Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ ad00ee21

History | View | Annotate | Download (127.9 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc.
5
#
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
10
#
11
# This program is distributed in the hope that it will be useful, but
12
# WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
# General Public License for more details.
15
#
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19
# 02110-1301, USA.
20

    
21

    
22
"""Functions used by the node daemon
23

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

29
"""
30

    
31
# pylint: disable=E1103
32

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

    
37

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

    
52
from ganeti import errors
53
from ganeti import utils
54
from ganeti import ssh
55
from ganeti import hypervisor
56
from ganeti import constants
57
from ganeti 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 compat
64
from ganeti import pathutils
65
from ganeti import vcluster
66
from ganeti import ht
67
from ganeti import hooksmaster
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 restricted 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 restricted commands
97
_RCMD_INVALID_DELAY = 10
98

    
99
#: How long to wait to acquire lock for restricted 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 _GetInstReasonFilename(instance_name):
114
  """Path of the file containing the reason of the instance status change.
115

116
  @type instance_name: string
117
  @param instance_name: The name of the instance
118
  @rtype: string
119
  @return: The path of the file
120

121
  """
122
  return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
123

    
124

    
125
def _StoreInstReasonTrail(instance_name, trail):
126
  """Serialize a reason trail related to an instance change of state to file.
127

128
  The exact location of the file depends on the name of the instance and on
129
  the configuration of the Ganeti cluster defined at deploy time.
130

131
  @type instance_name: string
132
  @param instance_name: The name of the instance
133
  @rtype: None
134

135
  """
136
  json = serializer.DumpJson(trail)
137
  filename = _GetInstReasonFilename(instance_name)
138
  utils.WriteFile(filename, data=json)
139

    
140

    
141
def _Fail(msg, *args, **kwargs):
142
  """Log an error and the raise an RPCFail exception.
143

144
  This exception is then handled specially in the ganeti daemon and
145
  turned into a 'failed' return type. As such, this function is a
146
  useful shortcut for logging the error and returning it to the master
147
  daemon.
148

149
  @type msg: string
150
  @param msg: the text of the exception
151
  @raise RPCFail
152

153
  """
154
  if args:
155
    msg = msg % args
156
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
157
    if "exc" in kwargs and kwargs["exc"]:
158
      logging.exception(msg)
159
    else:
160
      logging.error(msg)
161
  raise RPCFail(msg)
162

    
163

    
164
def _GetConfig():
165
  """Simple wrapper to return a SimpleStore.
166

167
  @rtype: L{ssconf.SimpleStore}
168
  @return: a SimpleStore instance
169

170
  """
171
  return ssconf.SimpleStore()
172

    
173

    
174
def _GetSshRunner(cluster_name):
175
  """Simple wrapper to return an SshRunner.
176

177
  @type cluster_name: str
178
  @param cluster_name: the cluster name, which is needed
179
      by the SshRunner constructor
180
  @rtype: L{ssh.SshRunner}
181
  @return: an SshRunner instance
182

183
  """
184
  return ssh.SshRunner(cluster_name)
185

    
186

    
187
def _Decompress(data):
188
  """Unpacks data compressed by the RPC client.
189

190
  @type data: list or tuple
191
  @param data: Data sent by RPC client
192
  @rtype: str
193
  @return: Decompressed data
194

195
  """
196
  assert isinstance(data, (list, tuple))
197
  assert len(data) == 2
198
  (encoding, content) = data
199
  if encoding == constants.RPC_ENCODING_NONE:
200
    return content
201
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
202
    return zlib.decompress(base64.b64decode(content))
203
  else:
204
    raise AssertionError("Unknown data encoding")
205

    
206

    
207
def _CleanDirectory(path, exclude=None):
208
  """Removes all regular files in a directory.
209

210
  @type path: str
211
  @param path: the directory to clean
212
  @type exclude: list
213
  @param exclude: list of files to be excluded, defaults
214
      to the empty list
215

216
  """
217
  if path not in _ALLOWED_CLEAN_DIRS:
218
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
219
          path)
220

    
221
  if not os.path.isdir(path):
222
    return
223
  if exclude is None:
224
    exclude = []
225
  else:
226
    # Normalize excluded paths
227
    exclude = [os.path.normpath(i) for i in exclude]
228

    
229
  for rel_name in utils.ListVisibleFiles(path):
230
    full_name = utils.PathJoin(path, rel_name)
231
    if full_name in exclude:
232
      continue
233
    if os.path.isfile(full_name) and not os.path.islink(full_name):
234
      utils.RemoveFile(full_name)
235

    
236

    
237
def _BuildUploadFileList():
238
  """Build the list of allowed upload files.
239

240
  This is abstracted so that it's built only once at module import time.
241

242
  """
243
  allowed_files = set([
244
    pathutils.CLUSTER_CONF_FILE,
245
    pathutils.ETC_HOSTS,
246
    pathutils.SSH_KNOWN_HOSTS_FILE,
247
    pathutils.VNC_PASSWORD_FILE,
248
    pathutils.RAPI_CERT_FILE,
249
    pathutils.SPICE_CERT_FILE,
250
    pathutils.SPICE_CACERT_FILE,
251
    pathutils.RAPI_USERS_FILE,
252
    pathutils.CONFD_HMAC_KEY,
253
    pathutils.CLUSTER_DOMAIN_SECRET_FILE,
254
    ])
255

    
256
  for hv_name in constants.HYPER_TYPES:
257
    hv_class = hypervisor.GetHypervisorClass(hv_name)
258
    allowed_files.update(hv_class.GetAncillaryFiles()[0])
259

    
260
  assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
261
    "Allowed file storage paths should never be uploaded via RPC"
262

    
263
  return frozenset(allowed_files)
264

    
265

    
266
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
267

    
268

    
269
def JobQueuePurge():
270
  """Removes job queue files and archived jobs.
271

272
  @rtype: tuple
273
  @return: True, None
274

275
  """
276
  _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
277
  _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
278

    
279

    
280
def GetMasterInfo():
281
  """Returns master information.
282

283
  This is an utility function to compute master information, either
284
  for consumption here or from the node daemon.
285

286
  @rtype: tuple
287
  @return: master_netdev, master_ip, master_name, primary_ip_family,
288
    master_netmask
289
  @raise RPCFail: in case of errors
290

291
  """
292
  try:
293
    cfg = _GetConfig()
294
    master_netdev = cfg.GetMasterNetdev()
295
    master_ip = cfg.GetMasterIP()
296
    master_netmask = cfg.GetMasterNetmask()
297
    master_node = cfg.GetMasterNode()
298
    primary_ip_family = cfg.GetPrimaryIPFamily()
299
  except errors.ConfigurationError, err:
300
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
301
  return (master_netdev, master_ip, master_node, primary_ip_family,
302
          master_netmask)
303

    
304

    
305
def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
306
  """Decorator that runs hooks before and after the decorated function.
307

308
  @type hook_opcode: string
309
  @param hook_opcode: opcode of the hook
310
  @type hooks_path: string
311
  @param hooks_path: path of the hooks
312
  @type env_builder_fn: function
313
  @param env_builder_fn: function that returns a dictionary containing the
314
    environment variables for the hooks. Will get all the parameters of the
315
    decorated function.
316
  @raise RPCFail: in case of pre-hook failure
317

318
  """
319
  def decorator(fn):
320
    def wrapper(*args, **kwargs):
321
      _, myself = ssconf.GetMasterAndMyself()
322
      nodes = ([myself], [myself])  # these hooks run locally
323

    
324
      env_fn = compat.partial(env_builder_fn, *args, **kwargs)
325

    
326
      cfg = _GetConfig()
327
      hr = HooksRunner()
328
      hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
329
                                   hr.RunLocalHooks, None, env_fn,
330
                                   logging.warning, cfg.GetClusterName(),
331
                                   cfg.GetMasterNode())
332
      hm.RunPhase(constants.HOOKS_PHASE_PRE)
333
      result = fn(*args, **kwargs)
334
      hm.RunPhase(constants.HOOKS_PHASE_POST)
335

    
336
      return result
337
    return wrapper
338
  return decorator
339

    
340

    
341
def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
342
  """Builds environment variables for master IP hooks.
343

344
  @type master_params: L{objects.MasterNetworkParameters}
345
  @param master_params: network parameters of the master
346
  @type use_external_mip_script: boolean
347
  @param use_external_mip_script: whether to use an external master IP
348
    address setup script (unused, but necessary per the implementation of the
349
    _RunLocalHooks decorator)
350

351
  """
352
  # pylint: disable=W0613
353
  ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
354
  env = {
355
    "MASTER_NETDEV": master_params.netdev,
356
    "MASTER_IP": master_params.ip,
357
    "MASTER_NETMASK": str(master_params.netmask),
358
    "CLUSTER_IP_VERSION": str(ver),
359
  }
360

    
361
  return env
362

    
363

    
364
def _RunMasterSetupScript(master_params, action, use_external_mip_script):
365
  """Execute the master IP address setup script.
366

367
  @type master_params: L{objects.MasterNetworkParameters}
368
  @param master_params: network parameters of the master
369
  @type action: string
370
  @param action: action to pass to the script. Must be one of
371
    L{backend._MASTER_START} or L{backend._MASTER_STOP}
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 backend.RPCFail: if there are errors during the execution of the
376
    script
377

378
  """
379
  env = _BuildMasterIpEnv(master_params)
380

    
381
  if use_external_mip_script:
382
    setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
383
  else:
384
    setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
385

    
386
  result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
387

    
388
  if result.failed:
389
    _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
390
          (action, result.exit_code, result.output), log=True)
391

    
392

    
393
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
394
               _BuildMasterIpEnv)
395
def ActivateMasterIp(master_params, use_external_mip_script):
396
  """Activate the IP address of the master daemon.
397

398
  @type master_params: L{objects.MasterNetworkParameters}
399
  @param master_params: network parameters of the master
400
  @type use_external_mip_script: boolean
401
  @param use_external_mip_script: whether to use an external master IP
402
    address setup script
403
  @raise RPCFail: in case of errors during the IP startup
404

405
  """
406
  _RunMasterSetupScript(master_params, _MASTER_START,
407
                        use_external_mip_script)
408

    
409

    
410
def StartMasterDaemons(no_voting):
411
  """Activate local node as master node.
412

413
  The function will start the master daemons (ganeti-masterd and ganeti-rapi).
414

415
  @type no_voting: boolean
416
  @param no_voting: whether to start ganeti-masterd without a node vote
417
      but still non-interactively
418
  @rtype: None
419

420
  """
421

    
422
  if no_voting:
423
    masterd_args = "--no-voting --yes-do-it"
424
  else:
425
    masterd_args = ""
426

    
427
  env = {
428
    "EXTRA_MASTERD_ARGS": masterd_args,
429
    }
430

    
431
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
432
  if result.failed:
433
    msg = "Can't start Ganeti master: %s" % result.output
434
    logging.error(msg)
435
    _Fail(msg)
436

    
437

    
438
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
439
               _BuildMasterIpEnv)
440
def DeactivateMasterIp(master_params, use_external_mip_script):
441
  """Deactivate the master IP on this node.
442

443
  @type master_params: L{objects.MasterNetworkParameters}
444
  @param master_params: network parameters of the master
445
  @type use_external_mip_script: boolean
446
  @param use_external_mip_script: whether to use an external master IP
447
    address setup script
448
  @raise RPCFail: in case of errors during the IP turndown
449

450
  """
451
  _RunMasterSetupScript(master_params, _MASTER_STOP,
452
                        use_external_mip_script)
453

    
454

    
455
def StopMasterDaemons():
456
  """Stop the master daemons on this node.
457

458
  Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
459

460
  @rtype: None
461

462
  """
463
  # TODO: log and report back to the caller the error failures; we
464
  # need to decide in which case we fail the RPC for this
465

    
466
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
467
  if result.failed:
468
    logging.error("Could not stop Ganeti master, command %s had exitcode %s"
469
                  " and error %s",
470
                  result.cmd, result.exit_code, result.output)
471

    
472

    
473
def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
474
  """Change the netmask of the master IP.
475

476
  @param old_netmask: the old value of the netmask
477
  @param netmask: the new value of the netmask
478
  @param master_ip: the master IP
479
  @param master_netdev: the master network device
480

481
  """
482
  if old_netmask == netmask:
483
    return
484

    
485
  if not netutils.IPAddress.Own(master_ip):
486
    _Fail("The master IP address is not up, not attempting to change its"
487
          " netmask")
488

    
489
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
490
                         "%s/%s" % (master_ip, netmask),
491
                         "dev", master_netdev, "label",
492
                         "%s:0" % master_netdev])
493
  if result.failed:
494
    _Fail("Could not set the new netmask on the master IP address")
495

    
496
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
497
                         "%s/%s" % (master_ip, old_netmask),
498
                         "dev", master_netdev, "label",
499
                         "%s:0" % master_netdev])
500
  if result.failed:
501
    _Fail("Could not bring down the master IP address with the old netmask")
502

    
503

    
504
def EtcHostsModify(mode, host, ip):
505
  """Modify a host entry in /etc/hosts.
506

507
  @param mode: The mode to operate. Either add or remove entry
508
  @param host: The host to operate on
509
  @param ip: The ip associated with the entry
510

511
  """
512
  if mode == constants.ETC_HOSTS_ADD:
513
    if not ip:
514
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
515
              " present")
516
    utils.AddHostToEtcHosts(host, ip)
517
  elif mode == constants.ETC_HOSTS_REMOVE:
518
    if ip:
519
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
520
              " parameter is present")
521
    utils.RemoveHostFromEtcHosts(host)
522
  else:
523
    RPCFail("Mode not supported")
524

    
525

    
526
def LeaveCluster(modify_ssh_setup):
527
  """Cleans up and remove the current node.
528

529
  This function cleans up and prepares the current node to be removed
530
  from the cluster.
531

532
  If processing is successful, then it raises an
533
  L{errors.QuitGanetiException} which is used as a special case to
534
  shutdown the node daemon.
535

536
  @param modify_ssh_setup: boolean
537

538
  """
539
  _CleanDirectory(pathutils.DATA_DIR)
540
  _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
541
  JobQueuePurge()
542

    
543
  if modify_ssh_setup:
544
    try:
545
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
546

    
547
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
548

    
549
      utils.RemoveFile(priv_key)
550
      utils.RemoveFile(pub_key)
551
    except errors.OpExecError:
552
      logging.exception("Error while processing ssh files")
553

    
554
  try:
555
    utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
556
    utils.RemoveFile(pathutils.RAPI_CERT_FILE)
557
    utils.RemoveFile(pathutils.SPICE_CERT_FILE)
558
    utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
559
    utils.RemoveFile(pathutils.NODED_CERT_FILE)
560
  except: # pylint: disable=W0702
561
    logging.exception("Error while removing cluster secrets")
562

    
563
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
564
  if result.failed:
565
    logging.error("Command %s failed with exitcode %s and error %s",
566
                  result.cmd, result.exit_code, result.output)
567

    
568
  # Raise a custom exception (handled in ganeti-noded)
569
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
570

    
571

    
572
def _GetVgInfo(name, excl_stor):
573
  """Retrieves information about a LVM volume group.
574

575
  """
576
  # TODO: GetVGInfo supports returning information for multiple VGs at once
577
  vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
578
  if vginfo:
579
    vg_free = int(round(vginfo[0][0], 0))
580
    vg_size = int(round(vginfo[0][1], 0))
581
  else:
582
    vg_free = None
583
    vg_size = None
584

    
585
  return {
586
    "name": name,
587
    "vg_free": vg_free,
588
    "vg_size": vg_size,
589
    }
590

    
591

    
592
def _GetHvInfo(name):
593
  """Retrieves node information from a hypervisor.
594

595
  The information returned depends on the hypervisor. Common items:
596

597
    - vg_size is the size of the configured volume group in MiB
598
    - vg_free is the free size of the volume group in MiB
599
    - memory_dom0 is the memory allocated for domain0 in MiB
600
    - memory_free is the currently available (free) ram in MiB
601
    - memory_total is the total number of ram in MiB
602
    - hv_version: the hypervisor version, if available
603

604
  """
605
  return hypervisor.GetHypervisor(name).GetNodeInfo()
606

    
607

    
608
def _GetNamedNodeInfo(names, fn):
609
  """Calls C{fn} for all names in C{names} and returns a dictionary.
610

611
  @rtype: None or dict
612

613
  """
614
  if names is None:
615
    return None
616
  else:
617
    return map(fn, names)
618

    
619

    
620
def GetNodeInfo(vg_names, hv_names, excl_stor):
621
  """Gives back a hash with different information about the node.
622

623
  @type vg_names: list of string
624
  @param vg_names: Names of the volume groups to ask for disk space information
625
  @type hv_names: list of string
626
  @param hv_names: Names of the hypervisors to ask for node information
627
  @type excl_stor: boolean
628
  @param excl_stor: Whether exclusive_storage is active
629
  @rtype: tuple; (string, None/dict, None/dict)
630
  @return: Tuple containing boot ID, volume group information and hypervisor
631
    information
632

633
  """
634
  bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
635
  vg_info = _GetNamedNodeInfo(vg_names, (lambda vg: _GetVgInfo(vg, excl_stor)))
636
  hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo)
637

    
638
  return (bootid, vg_info, hv_info)
639

    
640

    
641
def _CheckExclusivePvs(pvi_list):
642
  """Check that PVs are not shared among LVs
643

644
  @type pvi_list: list of L{objects.LvmPvInfo} objects
645
  @param pvi_list: information about the PVs
646

647
  @rtype: list of tuples (string, list of strings)
648
  @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
649

650
  """
651
  res = []
652
  for pvi in pvi_list:
653
    if len(pvi.lv_list) > 1:
654
      res.append((pvi.name, pvi.lv_list))
655
  return res
656

    
657

    
658
def VerifyNode(what, cluster_name):
659
  """Verify the status of the local node.
660

661
  Based on the input L{what} parameter, various checks are done on the
662
  local node.
663

664
  If the I{filelist} key is present, this list of
665
  files is checksummed and the file/checksum pairs are returned.
666

667
  If the I{nodelist} key is present, we check that we have
668
  connectivity via ssh with the target nodes (and check the hostname
669
  report).
670

671
  If the I{node-net-test} key is present, we check that we have
672
  connectivity to the given nodes via both primary IP and, if
673
  applicable, secondary IPs.
674

675
  @type what: C{dict}
676
  @param what: a dictionary of things to check:
677
      - filelist: list of files for which to compute checksums
678
      - nodelist: list of nodes we should check ssh communication with
679
      - node-net-test: list of nodes we should check node daemon port
680
        connectivity with
681
      - hypervisor: list with hypervisors to run the verify for
682
  @rtype: dict
683
  @return: a dictionary with the same keys as the input dict, and
684
      values representing the result of the checks
685

686
  """
687
  result = {}
688
  my_name = netutils.Hostname.GetSysName()
689
  port = netutils.GetDaemonPort(constants.NODED)
690
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
691

    
692
  if constants.NV_HYPERVISOR in what and vm_capable:
693
    result[constants.NV_HYPERVISOR] = tmp = {}
694
    for hv_name in what[constants.NV_HYPERVISOR]:
695
      try:
696
        val = hypervisor.GetHypervisor(hv_name).Verify()
697
      except errors.HypervisorError, err:
698
        val = "Error while checking hypervisor: %s" % str(err)
699
      tmp[hv_name] = val
700

    
701
  if constants.NV_HVPARAMS in what and vm_capable:
702
    result[constants.NV_HVPARAMS] = tmp = []
703
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
704
      try:
705
        logging.info("Validating hv %s, %s", hv_name, hvparms)
706
        hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
707
      except errors.HypervisorError, err:
708
        tmp.append((source, hv_name, str(err)))
709

    
710
  if constants.NV_FILELIST in what:
711
    fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
712
                                              what[constants.NV_FILELIST]))
713
    result[constants.NV_FILELIST] = \
714
      dict((vcluster.MakeVirtualPath(key), value)
715
           for (key, value) in fingerprints.items())
716

    
717
  if constants.NV_NODELIST in what:
718
    (nodes, bynode) = what[constants.NV_NODELIST]
719

    
720
    # Add nodes from other groups (different for each node)
721
    try:
722
      nodes.extend(bynode[my_name])
723
    except KeyError:
724
      pass
725

    
726
    # Use a random order
727
    random.shuffle(nodes)
728

    
729
    # Try to contact all nodes
730
    val = {}
731
    for node in nodes:
732
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
733
      if not success:
734
        val[node] = message
735

    
736
    result[constants.NV_NODELIST] = val
737

    
738
  if constants.NV_NODENETTEST in what:
739
    result[constants.NV_NODENETTEST] = tmp = {}
740
    my_pip = my_sip = None
741
    for name, pip, sip in what[constants.NV_NODENETTEST]:
742
      if name == my_name:
743
        my_pip = pip
744
        my_sip = sip
745
        break
746
    if not my_pip:
747
      tmp[my_name] = ("Can't find my own primary/secondary IP"
748
                      " in the node list")
749
    else:
750
      for name, pip, sip in what[constants.NV_NODENETTEST]:
751
        fail = []
752
        if not netutils.TcpPing(pip, port, source=my_pip):
753
          fail.append("primary")
754
        if sip != pip:
755
          if not netutils.TcpPing(sip, port, source=my_sip):
756
            fail.append("secondary")
757
        if fail:
758
          tmp[name] = ("failure using the %s interface(s)" %
759
                       " and ".join(fail))
760

    
761
  if constants.NV_MASTERIP in what:
762
    # FIXME: add checks on incoming data structures (here and in the
763
    # rest of the function)
764
    master_name, master_ip = what[constants.NV_MASTERIP]
765
    if master_name == my_name:
766
      source = constants.IP4_ADDRESS_LOCALHOST
767
    else:
768
      source = None
769
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
770
                                                     source=source)
771

    
772
  if constants.NV_USERSCRIPTS in what:
773
    result[constants.NV_USERSCRIPTS] = \
774
      [script for script in what[constants.NV_USERSCRIPTS]
775
       if not utils.IsExecutable(script)]
776

    
777
  if constants.NV_OOB_PATHS in what:
778
    result[constants.NV_OOB_PATHS] = tmp = []
779
    for path in what[constants.NV_OOB_PATHS]:
780
      try:
781
        st = os.stat(path)
782
      except OSError, err:
783
        tmp.append("error stating out of band helper: %s" % err)
784
      else:
785
        if stat.S_ISREG(st.st_mode):
786
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
787
            tmp.append(None)
788
          else:
789
            tmp.append("out of band helper %s is not executable" % path)
790
        else:
791
          tmp.append("out of band helper %s is not a file" % path)
792

    
793
  if constants.NV_LVLIST in what and vm_capable:
794
    try:
795
      val = GetVolumeList(utils.ListVolumeGroups().keys())
796
    except RPCFail, err:
797
      val = str(err)
798
    result[constants.NV_LVLIST] = val
799

    
800
  if constants.NV_INSTANCELIST in what and vm_capable:
801
    # GetInstanceList can fail
802
    try:
803
      val = GetInstanceList(what[constants.NV_INSTANCELIST])
804
    except RPCFail, err:
805
      val = str(err)
806
    result[constants.NV_INSTANCELIST] = val
807

    
808
  if constants.NV_VGLIST in what and vm_capable:
809
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
810

    
811
  if constants.NV_PVLIST in what and vm_capable:
812
    check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
813
    val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
814
                                       filter_allocatable=False,
815
                                       include_lvs=check_exclusive_pvs)
816
    if check_exclusive_pvs:
817
      result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
818
      for pvi in val:
819
        # Avoid sending useless data on the wire
820
        pvi.lv_list = []
821
    result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
822

    
823
  if constants.NV_VERSION in what:
824
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
825
                                    constants.RELEASE_VERSION)
826

    
827
  if constants.NV_HVINFO in what and vm_capable:
828
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
829
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
830

    
831
  if constants.NV_DRBDLIST in what and vm_capable:
832
    try:
833
      used_minors = bdev.DRBD8.GetUsedDevs().keys()
834
    except errors.BlockDeviceError, err:
835
      logging.warning("Can't get used minors list", exc_info=True)
836
      used_minors = str(err)
837
    result[constants.NV_DRBDLIST] = used_minors
838

    
839
  if constants.NV_DRBDHELPER in what and vm_capable:
840
    status = True
841
    try:
842
      payload = bdev.BaseDRBD.GetUsermodeHelper()
843
    except errors.BlockDeviceError, err:
844
      logging.error("Can't get DRBD usermode helper: %s", str(err))
845
      status = False
846
      payload = str(err)
847
    result[constants.NV_DRBDHELPER] = (status, payload)
848

    
849
  if constants.NV_NODESETUP in what:
850
    result[constants.NV_NODESETUP] = tmpr = []
851
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
852
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
853
                  " under /sys, missing required directories /sys/block"
854
                  " and /sys/class/net")
855
    if (not os.path.isdir("/proc/sys") or
856
        not os.path.isfile("/proc/sysrq-trigger")):
857
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
858
                  " under /proc, missing required directory /proc/sys and"
859
                  " the file /proc/sysrq-trigger")
860

    
861
  if constants.NV_TIME in what:
862
    result[constants.NV_TIME] = utils.SplitTime(time.time())
863

    
864
  if constants.NV_OSLIST in what and vm_capable:
865
    result[constants.NV_OSLIST] = DiagnoseOS()
866

    
867
  if constants.NV_BRIDGES in what and vm_capable:
868
    result[constants.NV_BRIDGES] = [bridge
869
                                    for bridge in what[constants.NV_BRIDGES]
870
                                    if not utils.BridgeExists(bridge)]
871

    
872
  if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
873
    result[constants.NV_FILE_STORAGE_PATHS] = \
874
      bdev.ComputeWrongFileStoragePaths()
875

    
876
  return result
877

    
878

    
879
def GetBlockDevSizes(devices):
880
  """Return the size of the given block devices
881

882
  @type devices: list
883
  @param devices: list of block device nodes to query
884
  @rtype: dict
885
  @return:
886
    dictionary of all block devices under /dev (key). The value is their
887
    size in MiB.
888

889
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
890

891
  """
892
  DEV_PREFIX = "/dev/"
893
  blockdevs = {}
894

    
895
  for devpath in devices:
896
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
897
      continue
898

    
899
    try:
900
      st = os.stat(devpath)
901
    except EnvironmentError, err:
902
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
903
      continue
904

    
905
    if stat.S_ISBLK(st.st_mode):
906
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
907
      if result.failed:
908
        # We don't want to fail, just do not list this device as available
909
        logging.warning("Cannot get size for block device %s", devpath)
910
        continue
911

    
912
      size = int(result.stdout) / (1024 * 1024)
913
      blockdevs[devpath] = size
914
  return blockdevs
915

    
916

    
917
def GetVolumeList(vg_names):
918
  """Compute list of logical volumes and their size.
919

920
  @type vg_names: list
921
  @param vg_names: the volume groups whose LVs we should list, or
922
      empty for all volume groups
923
  @rtype: dict
924
  @return:
925
      dictionary of all partions (key) with value being a tuple of
926
      their size (in MiB), inactive and online status::
927

928
        {'xenvg/test1': ('20.06', True, True)}
929

930
      in case of errors, a string is returned with the error
931
      details.
932

933
  """
934
  lvs = {}
935
  sep = "|"
936
  if not vg_names:
937
    vg_names = []
938
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
939
                         "--separator=%s" % sep,
940
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
941
  if result.failed:
942
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
943

    
944
  for line in result.stdout.splitlines():
945
    line = line.strip()
946
    match = _LVSLINE_REGEX.match(line)
947
    if not match:
948
      logging.error("Invalid line returned from lvs output: '%s'", line)
949
      continue
950
    vg_name, name, size, attr = match.groups()
951
    inactive = attr[4] == "-"
952
    online = attr[5] == "o"
953
    virtual = attr[0] == "v"
954
    if virtual:
955
      # we don't want to report such volumes as existing, since they
956
      # don't really hold data
957
      continue
958
    lvs[vg_name + "/" + name] = (size, inactive, online)
959

    
960
  return lvs
961

    
962

    
963
def ListVolumeGroups():
964
  """List the volume groups and their size.
965

966
  @rtype: dict
967
  @return: dictionary with keys volume name and values the
968
      size of the volume
969

970
  """
971
  return utils.ListVolumeGroups()
972

    
973

    
974
def NodeVolumes():
975
  """List all volumes on this node.
976

977
  @rtype: list
978
  @return:
979
    A list of dictionaries, each having four keys:
980
      - name: the logical volume name,
981
      - size: the size of the logical volume
982
      - dev: the physical device on which the LV lives
983
      - vg: the volume group to which it belongs
984

985
    In case of errors, we return an empty list and log the
986
    error.
987

988
    Note that since a logical volume can live on multiple physical
989
    volumes, the resulting list might include a logical volume
990
    multiple times.
991

992
  """
993
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
994
                         "--separator=|",
995
                         "--options=lv_name,lv_size,devices,vg_name"])
996
  if result.failed:
997
    _Fail("Failed to list logical volumes, lvs output: %s",
998
          result.output)
999

    
1000
  def parse_dev(dev):
1001
    return dev.split("(")[0]
1002

    
1003
  def handle_dev(dev):
1004
    return [parse_dev(x) for x in dev.split(",")]
1005

    
1006
  def map_line(line):
1007
    line = [v.strip() for v in line]
1008
    return [{"name": line[0], "size": line[1],
1009
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1010

    
1011
  all_devs = []
1012
  for line in result.stdout.splitlines():
1013
    if line.count("|") >= 3:
1014
      all_devs.extend(map_line(line.split("|")))
1015
    else:
1016
      logging.warning("Strange line in the output from lvs: '%s'", line)
1017
  return all_devs
1018

    
1019

    
1020
def BridgesExist(bridges_list):
1021
  """Check if a list of bridges exist on the current node.
1022

1023
  @rtype: boolean
1024
  @return: C{True} if all of them exist, C{False} otherwise
1025

1026
  """
1027
  missing = []
1028
  for bridge in bridges_list:
1029
    if not utils.BridgeExists(bridge):
1030
      missing.append(bridge)
1031

    
1032
  if missing:
1033
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
1034

    
1035

    
1036
def GetInstanceList(hypervisor_list):
1037
  """Provides a list of instances.
1038

1039
  @type hypervisor_list: list
1040
  @param hypervisor_list: the list of hypervisors to query information
1041

1042
  @rtype: list
1043
  @return: a list of all running instances on the current node
1044
    - instance1.example.com
1045
    - instance2.example.com
1046

1047
  """
1048
  results = []
1049
  for hname in hypervisor_list:
1050
    try:
1051
      names = hypervisor.GetHypervisor(hname).ListInstances()
1052
      results.extend(names)
1053
    except errors.HypervisorError, err:
1054
      _Fail("Error enumerating instances (hypervisor %s): %s",
1055
            hname, err, exc=True)
1056

    
1057
  return results
1058

    
1059

    
1060
def GetInstanceInfo(instance, hname):
1061
  """Gives back the information about an instance as a dictionary.
1062

1063
  @type instance: string
1064
  @param instance: the instance name
1065
  @type hname: string
1066
  @param hname: the hypervisor type of the instance
1067

1068
  @rtype: dict
1069
  @return: dictionary with the following keys:
1070
      - memory: memory size of instance (int)
1071
      - state: xen state of instance (string)
1072
      - time: cpu time of instance (float)
1073
      - vcpus: the number of vcpus (int)
1074

1075
  """
1076
  output = {}
1077

    
1078
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
1079
  if iinfo is not None:
1080
    output["memory"] = iinfo[2]
1081
    output["vcpus"] = iinfo[3]
1082
    output["state"] = iinfo[4]
1083
    output["time"] = iinfo[5]
1084

    
1085
  return output
1086

    
1087

    
1088
def GetInstanceMigratable(instance):
1089
  """Gives whether an instance can be migrated.
1090

1091
  @type instance: L{objects.Instance}
1092
  @param instance: object representing the instance to be checked.
1093

1094
  @rtype: tuple
1095
  @return: tuple of (result, description) where:
1096
      - result: whether the instance can be migrated or not
1097
      - description: a description of the issue, if relevant
1098

1099
  """
1100
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1101
  iname = instance.name
1102
  if iname not in hyper.ListInstances():
1103
    _Fail("Instance %s is not running", iname)
1104

    
1105
  for idx in range(len(instance.disks)):
1106
    link_name = _GetBlockDevSymlinkPath(iname, idx)
1107
    if not os.path.islink(link_name):
1108
      logging.warning("Instance %s is missing symlink %s for disk %d",
1109
                      iname, link_name, idx)
1110

    
1111

    
1112
def GetAllInstancesInfo(hypervisor_list):
1113
  """Gather data about all instances.
1114

1115
  This is the equivalent of L{GetInstanceInfo}, except that it
1116
  computes data for all instances at once, thus being faster if one
1117
  needs data about more than one instance.
1118

1119
  @type hypervisor_list: list
1120
  @param hypervisor_list: list of hypervisors to query for instance data
1121

1122
  @rtype: dict
1123
  @return: dictionary of instance: data, with data having the following keys:
1124
      - memory: memory size of instance (int)
1125
      - state: xen state of instance (string)
1126
      - time: cpu time of instance (float)
1127
      - vcpus: the number of vcpus
1128

1129
  """
1130
  output = {}
1131

    
1132
  for hname in hypervisor_list:
1133
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
1134
    if iinfo:
1135
      for name, _, memory, vcpus, state, times in iinfo:
1136
        value = {
1137
          "memory": memory,
1138
          "vcpus": vcpus,
1139
          "state": state,
1140
          "time": times,
1141
          }
1142
        if name in output:
1143
          # we only check static parameters, like memory and vcpus,
1144
          # and not state and time which can change between the
1145
          # invocations of the different hypervisors
1146
          for key in "memory", "vcpus":
1147
            if value[key] != output[name][key]:
1148
              _Fail("Instance %s is running twice"
1149
                    " with different parameters", name)
1150
        output[name] = value
1151

    
1152
  return output
1153

    
1154

    
1155
def _InstanceLogName(kind, os_name, instance, component):
1156
  """Compute the OS log filename for a given instance and operation.
1157

1158
  The instance name and os name are passed in as strings since not all
1159
  operations have these as part of an instance object.
1160

1161
  @type kind: string
1162
  @param kind: the operation type (e.g. add, import, etc.)
1163
  @type os_name: string
1164
  @param os_name: the os name
1165
  @type instance: string
1166
  @param instance: the name of the instance being imported/added/etc.
1167
  @type component: string or None
1168
  @param component: the name of the component of the instance being
1169
      transferred
1170

1171
  """
1172
  # TODO: Use tempfile.mkstemp to create unique filename
1173
  if component:
1174
    assert "/" not in component
1175
    c_msg = "-%s" % component
1176
  else:
1177
    c_msg = ""
1178
  base = ("%s-%s-%s%s-%s.log" %
1179
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1180
  return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1181

    
1182

    
1183
def InstanceOsAdd(instance, reinstall, debug):
1184
  """Add an OS to an instance.
1185

1186
  @type instance: L{objects.Instance}
1187
  @param instance: Instance whose OS is to be installed
1188
  @type reinstall: boolean
1189
  @param reinstall: whether this is an instance reinstall
1190
  @type debug: integer
1191
  @param debug: debug level, passed to the OS scripts
1192
  @rtype: None
1193

1194
  """
1195
  inst_os = OSFromDisk(instance.os)
1196

    
1197
  create_env = OSEnvironment(instance, inst_os, debug)
1198
  if reinstall:
1199
    create_env["INSTANCE_REINSTALL"] = "1"
1200

    
1201
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
1202

    
1203
  result = utils.RunCmd([inst_os.create_script], env=create_env,
1204
                        cwd=inst_os.path, output=logfile, reset_env=True)
1205
  if result.failed:
1206
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
1207
                  " output: %s", result.cmd, result.fail_reason, logfile,
1208
                  result.output)
1209
    lines = [utils.SafeEncode(val)
1210
             for val in utils.TailFile(logfile, lines=20)]
1211
    _Fail("OS create script failed (%s), last lines in the"
1212
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1213

    
1214

    
1215
def RunRenameInstance(instance, old_name, debug):
1216
  """Run the OS rename script for an instance.
1217

1218
  @type instance: L{objects.Instance}
1219
  @param instance: Instance whose OS is to be installed
1220
  @type old_name: string
1221
  @param old_name: previous instance name
1222
  @type debug: integer
1223
  @param debug: debug level, passed to the OS scripts
1224
  @rtype: boolean
1225
  @return: the success of the operation
1226

1227
  """
1228
  inst_os = OSFromDisk(instance.os)
1229

    
1230
  rename_env = OSEnvironment(instance, inst_os, debug)
1231
  rename_env["OLD_INSTANCE_NAME"] = old_name
1232

    
1233
  logfile = _InstanceLogName("rename", instance.os,
1234
                             "%s-%s" % (old_name, instance.name), None)
1235

    
1236
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1237
                        cwd=inst_os.path, output=logfile, reset_env=True)
1238

    
1239
  if result.failed:
1240
    logging.error("os create command '%s' returned error: %s output: %s",
1241
                  result.cmd, result.fail_reason, result.output)
1242
    lines = [utils.SafeEncode(val)
1243
             for val in utils.TailFile(logfile, lines=20)]
1244
    _Fail("OS rename script failed (%s), last lines in the"
1245
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1246

    
1247

    
1248
def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1249
  """Returns symlink path for block device.
1250

1251
  """
1252
  if _dir is None:
1253
    _dir = pathutils.DISK_LINKS_DIR
1254

    
1255
  return utils.PathJoin(_dir,
1256
                        ("%s%s%s" %
1257
                         (instance_name, constants.DISK_SEPARATOR, idx)))
1258

    
1259

    
1260
def _SymlinkBlockDev(instance_name, device_path, idx):
1261
  """Set up symlinks to a instance's block device.
1262

1263
  This is an auxiliary function run when an instance is start (on the primary
1264
  node) or when an instance is migrated (on the target node).
1265

1266

1267
  @param instance_name: the name of the target instance
1268
  @param device_path: path of the physical block device, on the node
1269
  @param idx: the disk index
1270
  @return: absolute path to the disk's symlink
1271

1272
  """
1273
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1274
  try:
1275
    os.symlink(device_path, link_name)
1276
  except OSError, err:
1277
    if err.errno == errno.EEXIST:
1278
      if (not os.path.islink(link_name) or
1279
          os.readlink(link_name) != device_path):
1280
        os.remove(link_name)
1281
        os.symlink(device_path, link_name)
1282
    else:
1283
      raise
1284

    
1285
  return link_name
1286

    
1287

    
1288
def _RemoveBlockDevLinks(instance_name, disks):
1289
  """Remove the block device symlinks belonging to the given instance.
1290

1291
  """
1292
  for idx, _ in enumerate(disks):
1293
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1294
    if os.path.islink(link_name):
1295
      try:
1296
        os.remove(link_name)
1297
      except OSError:
1298
        logging.exception("Can't remove symlink '%s'", link_name)
1299

    
1300

    
1301
def _GatherAndLinkBlockDevs(instance):
1302
  """Set up an instance's block device(s).
1303

1304
  This is run on the primary node at instance startup. The block
1305
  devices must be already assembled.
1306

1307
  @type instance: L{objects.Instance}
1308
  @param instance: the instance whose disks we shoul assemble
1309
  @rtype: list
1310
  @return: list of (disk_object, device_path)
1311

1312
  """
1313
  block_devices = []
1314
  for idx, disk in enumerate(instance.disks):
1315
    device = _RecursiveFindBD(disk)
1316
    if device is None:
1317
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1318
                                    str(disk))
1319
    device.Open()
1320
    try:
1321
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1322
    except OSError, e:
1323
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1324
                                    e.strerror)
1325

    
1326
    block_devices.append((disk, link_name))
1327

    
1328
  return block_devices
1329

    
1330

    
1331
def StartInstance(instance, startup_paused, reason, store_reason=True):
1332
  """Start an instance.
1333

1334
  @type instance: L{objects.Instance}
1335
  @param instance: the instance object
1336
  @type startup_paused: bool
1337
  @param instance: pause instance at startup?
1338
  @type reason: list of reasons
1339
  @param reason: the reason trail for this startup
1340
  @type store_reason: boolean
1341
  @param store_reason: whether to store the shutdown reason trail on file
1342
  @rtype: None
1343

1344
  """
1345
  running_instances = GetInstanceList([instance.hypervisor])
1346

    
1347
  if instance.name in running_instances:
1348
    logging.info("Instance %s already running, not starting", instance.name)
1349
    return
1350

    
1351
  try:
1352
    block_devices = _GatherAndLinkBlockDevs(instance)
1353
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1354
    hyper.StartInstance(instance, block_devices, startup_paused)
1355
    if store_reason:
1356
      _StoreInstReasonTrail(instance.name, reason)
1357
  except errors.BlockDeviceError, err:
1358
    _Fail("Block device error: %s", err, exc=True)
1359
  except errors.HypervisorError, err:
1360
    _RemoveBlockDevLinks(instance.name, instance.disks)
1361
    _Fail("Hypervisor error: %s", err, exc=True)
1362

    
1363

    
1364
def InstanceShutdown(instance, timeout, reason, store_reason=True):
1365
  """Shut an instance down.
1366

1367
  @note: this functions uses polling with a hardcoded timeout.
1368

1369
  @type instance: L{objects.Instance}
1370
  @param instance: the instance object
1371
  @type timeout: integer
1372
  @param timeout: maximum timeout for soft shutdown
1373
  @type reason: list of reasons
1374
  @param reason: the reason trail for this shutdown
1375
  @type store_reason: boolean
1376
  @param store_reason: whether to store the shutdown reason trail on file
1377
  @rtype: None
1378

1379
  """
1380
  hv_name = instance.hypervisor
1381
  hyper = hypervisor.GetHypervisor(hv_name)
1382
  iname = instance.name
1383

    
1384
  if instance.name not in hyper.ListInstances():
1385
    logging.info("Instance %s not running, doing nothing", iname)
1386
    return
1387

    
1388
  class _TryShutdown:
1389
    def __init__(self):
1390
      self.tried_once = False
1391

    
1392
    def __call__(self):
1393
      if iname not in hyper.ListInstances():
1394
        return
1395

    
1396
      try:
1397
        hyper.StopInstance(instance, retry=self.tried_once)
1398
        if store_reason:
1399
          _StoreInstReasonTrail(instance.name, reason)
1400
      except errors.HypervisorError, err:
1401
        if iname not in hyper.ListInstances():
1402
          # if the instance is no longer existing, consider this a
1403
          # success and go to cleanup
1404
          return
1405

    
1406
        _Fail("Failed to stop instance %s: %s", iname, err)
1407

    
1408
      self.tried_once = True
1409

    
1410
      raise utils.RetryAgain()
1411

    
1412
  try:
1413
    utils.Retry(_TryShutdown(), 5, timeout)
1414
  except utils.RetryTimeout:
1415
    # the shutdown did not succeed
1416
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1417

    
1418
    try:
1419
      hyper.StopInstance(instance, force=True)
1420
    except errors.HypervisorError, err:
1421
      if iname in hyper.ListInstances():
1422
        # only raise an error if the instance still exists, otherwise
1423
        # the error could simply be "instance ... unknown"!
1424
        _Fail("Failed to force stop instance %s: %s", iname, err)
1425

    
1426
    time.sleep(1)
1427

    
1428
    if iname in hyper.ListInstances():
1429
      _Fail("Could not shutdown instance %s even by destroy", iname)
1430

    
1431
  try:
1432
    hyper.CleanupInstance(instance.name)
1433
  except errors.HypervisorError, err:
1434
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1435

    
1436
  _RemoveBlockDevLinks(iname, instance.disks)
1437

    
1438

    
1439
def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1440
  """Reboot an instance.
1441

1442
  @type instance: L{objects.Instance}
1443
  @param instance: the instance object to reboot
1444
  @type reboot_type: str
1445
  @param reboot_type: the type of reboot, one the following
1446
    constants:
1447
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1448
        instance OS, do not recreate the VM
1449
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1450
        restart the VM (at the hypervisor level)
1451
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1452
        not accepted here, since that mode is handled differently, in
1453
        cmdlib, and translates into full stop and start of the
1454
        instance (instead of a call_instance_reboot RPC)
1455
  @type shutdown_timeout: integer
1456
  @param shutdown_timeout: maximum timeout for soft shutdown
1457
  @type reason: list of reasons
1458
  @param reason: the reason trail for this reboot
1459
  @rtype: None
1460

1461
  """
1462
  running_instances = GetInstanceList([instance.hypervisor])
1463

    
1464
  if instance.name not in running_instances:
1465
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1466

    
1467
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1468
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1469
    try:
1470
      hyper.RebootInstance(instance)
1471
    except errors.HypervisorError, err:
1472
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1473
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1474
    try:
1475
      InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1476
      result = StartInstance(instance, False, reason, store_reason=False)
1477
      _StoreInstReasonTrail(instance.name, reason)
1478
      return result
1479
    except errors.HypervisorError, err:
1480
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1481
  else:
1482
    _Fail("Invalid reboot_type received: %s", reboot_type)
1483

    
1484

    
1485
def InstanceBalloonMemory(instance, memory):
1486
  """Resize an instance's memory.
1487

1488
  @type instance: L{objects.Instance}
1489
  @param instance: the instance object
1490
  @type memory: int
1491
  @param memory: new memory amount in MB
1492
  @rtype: None
1493

1494
  """
1495
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1496
  running = hyper.ListInstances()
1497
  if instance.name not in running:
1498
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1499
    return
1500
  try:
1501
    hyper.BalloonInstanceMemory(instance, memory)
1502
  except errors.HypervisorError, err:
1503
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1504

    
1505

    
1506
def MigrationInfo(instance):
1507
  """Gather information about an instance to be migrated.
1508

1509
  @type instance: L{objects.Instance}
1510
  @param instance: the instance definition
1511

1512
  """
1513
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1514
  try:
1515
    info = hyper.MigrationInfo(instance)
1516
  except errors.HypervisorError, err:
1517
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1518
  return info
1519

    
1520

    
1521
def AcceptInstance(instance, info, target):
1522
  """Prepare the node to accept an instance.
1523

1524
  @type instance: L{objects.Instance}
1525
  @param instance: the instance definition
1526
  @type info: string/data (opaque)
1527
  @param info: migration information, from the source node
1528
  @type target: string
1529
  @param target: target host (usually ip), on this node
1530

1531
  """
1532
  # TODO: why is this required only for DTS_EXT_MIRROR?
1533
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1534
    # Create the symlinks, as the disks are not active
1535
    # in any way
1536
    try:
1537
      _GatherAndLinkBlockDevs(instance)
1538
    except errors.BlockDeviceError, err:
1539
      _Fail("Block device error: %s", err, exc=True)
1540

    
1541
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1542
  try:
1543
    hyper.AcceptInstance(instance, info, target)
1544
  except errors.HypervisorError, err:
1545
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1546
      _RemoveBlockDevLinks(instance.name, instance.disks)
1547
    _Fail("Failed to accept instance: %s", err, exc=True)
1548

    
1549

    
1550
def FinalizeMigrationDst(instance, info, success):
1551
  """Finalize any preparation to accept an instance.
1552

1553
  @type instance: L{objects.Instance}
1554
  @param instance: the instance definition
1555
  @type info: string/data (opaque)
1556
  @param info: migration information, from the source node
1557
  @type success: boolean
1558
  @param success: whether the migration was a success or a failure
1559

1560
  """
1561
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1562
  try:
1563
    hyper.FinalizeMigrationDst(instance, info, success)
1564
  except errors.HypervisorError, err:
1565
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1566

    
1567

    
1568
def MigrateInstance(instance, target, live):
1569
  """Migrates an instance to another node.
1570

1571
  @type instance: L{objects.Instance}
1572
  @param instance: the instance definition
1573
  @type target: string
1574
  @param target: the target node name
1575
  @type live: boolean
1576
  @param live: whether the migration should be done live or not (the
1577
      interpretation of this parameter is left to the hypervisor)
1578
  @raise RPCFail: if migration fails for some reason
1579

1580
  """
1581
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1582

    
1583
  try:
1584
    hyper.MigrateInstance(instance, target, live)
1585
  except errors.HypervisorError, err:
1586
    _Fail("Failed to migrate instance: %s", err, exc=True)
1587

    
1588

    
1589
def FinalizeMigrationSource(instance, success, live):
1590
  """Finalize the instance migration on the source node.
1591

1592
  @type instance: L{objects.Instance}
1593
  @param instance: the instance definition of the migrated instance
1594
  @type success: bool
1595
  @param success: whether the migration succeeded or not
1596
  @type live: bool
1597
  @param live: whether the user requested a live migration or not
1598
  @raise RPCFail: If the execution fails for some reason
1599

1600
  """
1601
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1602

    
1603
  try:
1604
    hyper.FinalizeMigrationSource(instance, success, live)
1605
  except Exception, err:  # pylint: disable=W0703
1606
    _Fail("Failed to finalize the migration on the source node: %s", err,
1607
          exc=True)
1608

    
1609

    
1610
def GetMigrationStatus(instance):
1611
  """Get the migration status
1612

1613
  @type instance: L{objects.Instance}
1614
  @param instance: the instance that is being migrated
1615
  @rtype: L{objects.MigrationStatus}
1616
  @return: the status of the current migration (one of
1617
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1618
           progress info that can be retrieved from the hypervisor
1619
  @raise RPCFail: If the migration status cannot be retrieved
1620

1621
  """
1622
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1623
  try:
1624
    return hyper.GetMigrationStatus(instance)
1625
  except Exception, err:  # pylint: disable=W0703
1626
    _Fail("Failed to get migration status: %s", err, exc=True)
1627

    
1628

    
1629
def HotplugDevice(instance, action, dev_type, device, extra, seq):
1630
  """Hotplug a device
1631

1632
  Hotplug is currently supported only for KVM Hypervisor.
1633
  @type instance: L{objects.Instance}
1634
  @param instance: the instance to which we hotplug a device
1635
  @type action: string
1636
  @param action: the hotplug action to perform
1637
  @type dev_type: string
1638
  @param dev_type: the device type to hotplug
1639
  @type device: either L{objects.NIC} or L{objects.Disk}
1640
  @param device: the device object to hotplug
1641
  @type extra: string
1642
  @param extra: extra info used by hotplug code (e.g. disk link)
1643
  @type seq: int
1644
  @param seq: the index of the device from master perspective
1645
  @raise RPCFail: in case instance does not have KVM hypervisor
1646

1647
  """
1648
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1649
  try:
1650
    hyper.VerifyHotplugSupport(instance, action, dev_type)
1651
  except errors.HotplugError, err:
1652
    _Fail("Hotplug is not supported: %s", err)
1653

    
1654
  if action == constants.HOTPLUG_ACTION_ADD:
1655
    fn = hyper.HotAddDevice
1656
  elif action == constants.HOTPLUG_ACTION_REMOVE:
1657
    fn = hyper.HotDelDevice
1658
  elif action == constants.HOTPLUG_ACTION_MODIFY:
1659
    fn = hyper.HotModDevice
1660
  else:
1661
    assert action in constants.HOTPLUG_ALL_ACTIONS
1662
  # This will raise an exception if hotplug is no supported for this case
1663
  return fn(instance, dev_type, device, extra, seq)
1664

    
1665

    
1666
def HotplugSupported(instance):
1667
  """Checks if hotplug is generally supported.
1668

1669
  """
1670
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1671
  try:
1672
    hyper.HotplugSupported(instance)
1673
  except errors.HotplugError, err:
1674
    _Fail("Hotplug is not supported: %s", err)
1675

    
1676

    
1677
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1678
  """Creates a block device for an instance.
1679

1680
  @type disk: L{objects.Disk}
1681
  @param disk: the object describing the disk we should create
1682
  @type size: int
1683
  @param size: the size of the physical underlying device, in MiB
1684
  @type owner: str
1685
  @param owner: the name of the instance for which disk is created,
1686
      used for device cache data
1687
  @type on_primary: boolean
1688
  @param on_primary:  indicates if it is the primary node or not
1689
  @type info: string
1690
  @param info: string that will be sent to the physical device
1691
      creation, used for example to set (LVM) tags on LVs
1692
  @type excl_stor: boolean
1693
  @param excl_stor: Whether exclusive_storage is active
1694

1695
  @return: the new unique_id of the device (this can sometime be
1696
      computed only after creation), or None. On secondary nodes,
1697
      it's not required to return anything.
1698

1699
  """
1700
  # TODO: remove the obsolete "size" argument
1701
  # pylint: disable=W0613
1702
  clist = []
1703
  if disk.children:
1704
    for child in disk.children:
1705
      try:
1706
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1707
      except errors.BlockDeviceError, err:
1708
        _Fail("Can't assemble device %s: %s", child, err)
1709
      if on_primary or disk.AssembleOnSecondary():
1710
        # we need the children open in case the device itself has to
1711
        # be assembled
1712
        try:
1713
          # pylint: disable=E1103
1714
          crdev.Open()
1715
        except errors.BlockDeviceError, err:
1716
          _Fail("Can't make child '%s' read-write: %s", child, err)
1717
      clist.append(crdev)
1718

    
1719
  try:
1720
    device = bdev.Create(disk, clist, excl_stor)
1721
  except errors.BlockDeviceError, err:
1722
    _Fail("Can't create block device: %s", err)
1723

    
1724
  if on_primary or disk.AssembleOnSecondary():
1725
    try:
1726
      device.Assemble()
1727
    except errors.BlockDeviceError, err:
1728
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1729
    if on_primary or disk.OpenOnSecondary():
1730
      try:
1731
        device.Open(force=True)
1732
      except errors.BlockDeviceError, err:
1733
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1734
    DevCacheManager.UpdateCache(device.dev_path, owner,
1735
                                on_primary, disk.iv_name)
1736

    
1737
  device.SetInfo(info)
1738

    
1739
  return device.unique_id
1740

    
1741

    
1742
def _WipeDevice(path, offset, size):
1743
  """This function actually wipes the device.
1744

1745
  @param path: The path to the device to wipe
1746
  @param offset: The offset in MiB in the file
1747
  @param size: The size in MiB to write
1748

1749
  """
1750
  # Internal sizes are always in Mebibytes; if the following "dd" command
1751
  # should use a different block size the offset and size given to this
1752
  # function must be adjusted accordingly before being passed to "dd".
1753
  block_size = 1024 * 1024
1754

    
1755
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1756
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1757
         "count=%d" % size]
1758
  result = utils.RunCmd(cmd)
1759

    
1760
  if result.failed:
1761
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1762
          result.fail_reason, result.output)
1763

    
1764

    
1765
def BlockdevWipe(disk, offset, size):
1766
  """Wipes a block device.
1767

1768
  @type disk: L{objects.Disk}
1769
  @param disk: the disk object we want to wipe
1770
  @type offset: int
1771
  @param offset: The offset in MiB in the file
1772
  @type size: int
1773
  @param size: The size in MiB to write
1774

1775
  """
1776
  try:
1777
    rdev = _RecursiveFindBD(disk)
1778
  except errors.BlockDeviceError:
1779
    rdev = None
1780

    
1781
  if not rdev:
1782
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1783

    
1784
  # Do cross verify some of the parameters
1785
  if offset < 0:
1786
    _Fail("Negative offset")
1787
  if size < 0:
1788
    _Fail("Negative size")
1789
  if offset > rdev.size:
1790
    _Fail("Offset is bigger than device size")
1791
  if (offset + size) > rdev.size:
1792
    _Fail("The provided offset and size to wipe is bigger than device size")
1793

    
1794
  _WipeDevice(rdev.dev_path, offset, size)
1795

    
1796

    
1797
def BlockdevPauseResumeSync(disks, pause):
1798
  """Pause or resume the sync of the block device.
1799

1800
  @type disks: list of L{objects.Disk}
1801
  @param disks: the disks object we want to pause/resume
1802
  @type pause: bool
1803
  @param pause: Wheater to pause or resume
1804

1805
  """
1806
  success = []
1807
  for disk in disks:
1808
    try:
1809
      rdev = _RecursiveFindBD(disk)
1810
    except errors.BlockDeviceError:
1811
      rdev = None
1812

    
1813
    if not rdev:
1814
      success.append((False, ("Cannot change sync for device %s:"
1815
                              " device not found" % disk.iv_name)))
1816
      continue
1817

    
1818
    result = rdev.PauseResumeSync(pause)
1819

    
1820
    if result:
1821
      success.append((result, None))
1822
    else:
1823
      if pause:
1824
        msg = "Pause"
1825
      else:
1826
        msg = "Resume"
1827
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1828

    
1829
  return success
1830

    
1831

    
1832
def BlockdevRemove(disk):
1833
  """Remove a block device.
1834

1835
  @note: This is intended to be called recursively.
1836

1837
  @type disk: L{objects.Disk}
1838
  @param disk: the disk object we should remove
1839
  @rtype: boolean
1840
  @return: the success of the operation
1841

1842
  """
1843
  msgs = []
1844
  try:
1845
    rdev = _RecursiveFindBD(disk)
1846
  except errors.BlockDeviceError, err:
1847
    # probably can't attach
1848
    logging.info("Can't attach to device %s in remove", disk)
1849
    rdev = None
1850
  if rdev is not None:
1851
    r_path = rdev.dev_path
1852
    try:
1853
      rdev.Remove()
1854
    except errors.BlockDeviceError, err:
1855
      msgs.append(str(err))
1856
    if not msgs:
1857
      DevCacheManager.RemoveCache(r_path)
1858

    
1859
  if disk.children:
1860
    for child in disk.children:
1861
      try:
1862
        BlockdevRemove(child)
1863
      except RPCFail, err:
1864
        msgs.append(str(err))
1865

    
1866
  if msgs:
1867
    _Fail("; ".join(msgs))
1868

    
1869

    
1870
def _RecursiveAssembleBD(disk, owner, as_primary):
1871
  """Activate a block device for an instance.
1872

1873
  This is run on the primary and secondary nodes for an instance.
1874

1875
  @note: this function is called recursively.
1876

1877
  @type disk: L{objects.Disk}
1878
  @param disk: the disk we try to assemble
1879
  @type owner: str
1880
  @param owner: the name of the instance which owns the disk
1881
  @type as_primary: boolean
1882
  @param as_primary: if we should make the block device
1883
      read/write
1884

1885
  @return: the assembled device or None (in case no device
1886
      was assembled)
1887
  @raise errors.BlockDeviceError: in case there is an error
1888
      during the activation of the children or the device
1889
      itself
1890

1891
  """
1892
  children = []
1893
  if disk.children:
1894
    mcn = disk.ChildrenNeeded()
1895
    if mcn == -1:
1896
      mcn = 0 # max number of Nones allowed
1897
    else:
1898
      mcn = len(disk.children) - mcn # max number of Nones
1899
    for chld_disk in disk.children:
1900
      try:
1901
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1902
      except errors.BlockDeviceError, err:
1903
        if children.count(None) >= mcn:
1904
          raise
1905
        cdev = None
1906
        logging.error("Error in child activation (but continuing): %s",
1907
                      str(err))
1908
      children.append(cdev)
1909

    
1910
  if as_primary or disk.AssembleOnSecondary():
1911
    r_dev = bdev.Assemble(disk, children)
1912
    result = r_dev
1913
    if as_primary or disk.OpenOnSecondary():
1914
      r_dev.Open()
1915
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1916
                                as_primary, disk.iv_name)
1917

    
1918
  else:
1919
    result = True
1920
  return result
1921

    
1922

    
1923
def BlockdevAssemble(disk, owner, as_primary, idx):
1924
  """Activate a block device for an instance.
1925

1926
  This is a wrapper over _RecursiveAssembleBD.
1927

1928
  @rtype: str or boolean
1929
  @return: a C{/dev/...} path for primary nodes, and
1930
      C{True} for secondary nodes
1931

1932
  """
1933
  try:
1934
    device = _RecursiveAssembleBD(disk, owner, as_primary)
1935
    if isinstance(device, bdev.BlockDev):
1936
      dev_path = device.dev_path
1937
      link_name = None
1938
      if as_primary:
1939
        link_name = _SymlinkBlockDev(owner, dev_path, idx)
1940
  except errors.BlockDeviceError, err:
1941
    _Fail("Error while assembling disk: %s", err, exc=True)
1942
  except OSError, err:
1943
    _Fail("Error while symlinking disk: %s", err, exc=True)
1944

    
1945
  return dev_path, link_name
1946

    
1947

    
1948
def BlockdevShutdown(disk):
1949
  """Shut down a block device.
1950

1951
  First, if the device is assembled (Attach() is successful), then
1952
  the device is shutdown. Then the children of the device are
1953
  shutdown.
1954

1955
  This function is called recursively. Note that we don't cache the
1956
  children or such, as oppossed to assemble, shutdown of different
1957
  devices doesn't require that the upper device was active.
1958

1959
  @type disk: L{objects.Disk}
1960
  @param disk: the description of the disk we should
1961
      shutdown
1962
  @rtype: None
1963

1964
  """
1965
  msgs = []
1966
  r_dev = _RecursiveFindBD(disk)
1967
  if r_dev is not None:
1968
    r_path = r_dev.dev_path
1969
    try:
1970
      r_dev.Shutdown()
1971
      DevCacheManager.RemoveCache(r_path)
1972
    except errors.BlockDeviceError, err:
1973
      msgs.append(str(err))
1974

    
1975
  if disk.children:
1976
    for child in disk.children:
1977
      try:
1978
        BlockdevShutdown(child)
1979
      except RPCFail, err:
1980
        msgs.append(str(err))
1981

    
1982
  if msgs:
1983
    _Fail("; ".join(msgs))
1984

    
1985

    
1986
def BlockdevAddchildren(parent_cdev, new_cdevs):
1987
  """Extend a mirrored block device.
1988

1989
  @type parent_cdev: L{objects.Disk}
1990
  @param parent_cdev: the disk to which we should add children
1991
  @type new_cdevs: list of L{objects.Disk}
1992
  @param new_cdevs: the list of children which we should add
1993
  @rtype: None
1994

1995
  """
1996
  parent_bdev = _RecursiveFindBD(parent_cdev)
1997
  if parent_bdev is None:
1998
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1999
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2000
  if new_bdevs.count(None) > 0:
2001
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2002
  parent_bdev.AddChildren(new_bdevs)
2003

    
2004

    
2005
def BlockdevRemovechildren(parent_cdev, new_cdevs):
2006
  """Shrink a mirrored block device.
2007

2008
  @type parent_cdev: L{objects.Disk}
2009
  @param parent_cdev: the disk from which we should remove children
2010
  @type new_cdevs: list of L{objects.Disk}
2011
  @param new_cdevs: the list of children which we should remove
2012
  @rtype: None
2013

2014
  """
2015
  parent_bdev = _RecursiveFindBD(parent_cdev)
2016
  if parent_bdev is None:
2017
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2018
  devs = []
2019
  for disk in new_cdevs:
2020
    rpath = disk.StaticDevPath()
2021
    if rpath is None:
2022
      bd = _RecursiveFindBD(disk)
2023
      if bd is None:
2024
        _Fail("Can't find device %s while removing children", disk)
2025
      else:
2026
        devs.append(bd.dev_path)
2027
    else:
2028
      if not utils.IsNormAbsPath(rpath):
2029
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2030
      devs.append(rpath)
2031
  parent_bdev.RemoveChildren(devs)
2032

    
2033

    
2034
def BlockdevGetmirrorstatus(disks):
2035
  """Get the mirroring status of a list of devices.
2036

2037
  @type disks: list of L{objects.Disk}
2038
  @param disks: the list of disks which we should query
2039
  @rtype: disk
2040
  @return: List of L{objects.BlockDevStatus}, one for each disk
2041
  @raise errors.BlockDeviceError: if any of the disks cannot be
2042
      found
2043

2044
  """
2045
  stats = []
2046
  for dsk in disks:
2047
    rbd = _RecursiveFindBD(dsk)
2048
    if rbd is None:
2049
      _Fail("Can't find device %s", dsk)
2050

    
2051
    stats.append(rbd.CombinedSyncStatus())
2052

    
2053
  return stats
2054

    
2055

    
2056
def BlockdevGetmirrorstatusMulti(disks):
2057
  """Get the mirroring status of a list of devices.
2058

2059
  @type disks: list of L{objects.Disk}
2060
  @param disks: the list of disks which we should query
2061
  @rtype: disk
2062
  @return: List of tuples, (bool, status), one for each disk; bool denotes
2063
    success/failure, status is L{objects.BlockDevStatus} on success, string
2064
    otherwise
2065

2066
  """
2067
  result = []
2068
  for disk in disks:
2069
    try:
2070
      rbd = _RecursiveFindBD(disk)
2071
      if rbd is None:
2072
        result.append((False, "Can't find device %s" % disk))
2073
        continue
2074

    
2075
      status = rbd.CombinedSyncStatus()
2076
    except errors.BlockDeviceError, err:
2077
      logging.exception("Error while getting disk status")
2078
      result.append((False, str(err)))
2079
    else:
2080
      result.append((True, status))
2081

    
2082
  assert len(disks) == len(result)
2083

    
2084
  return result
2085

    
2086

    
2087
def _RecursiveFindBD(disk):
2088
  """Check if a device is activated.
2089

2090
  If so, return information about the real device.
2091

2092
  @type disk: L{objects.Disk}
2093
  @param disk: the disk object we need to find
2094

2095
  @return: None if the device can't be found,
2096
      otherwise the device instance
2097

2098
  """
2099
  children = []
2100
  if disk.children:
2101
    for chdisk in disk.children:
2102
      children.append(_RecursiveFindBD(chdisk))
2103

    
2104
  return bdev.FindDevice(disk, children)
2105

    
2106

    
2107
def _OpenRealBD(disk):
2108
  """Opens the underlying block device of a disk.
2109

2110
  @type disk: L{objects.Disk}
2111
  @param disk: the disk object we want to open
2112

2113
  """
2114
  real_disk = _RecursiveFindBD(disk)
2115
  if real_disk is None:
2116
    _Fail("Block device '%s' is not set up", disk)
2117

    
2118
  real_disk.Open()
2119

    
2120
  return real_disk
2121

    
2122

    
2123
def BlockdevFind(disk):
2124
  """Check if a device is activated.
2125

2126
  If it is, return information about the real device.
2127

2128
  @type disk: L{objects.Disk}
2129
  @param disk: the disk to find
2130
  @rtype: None or objects.BlockDevStatus
2131
  @return: None if the disk cannot be found, otherwise a the current
2132
           information
2133

2134
  """
2135
  try:
2136
    rbd = _RecursiveFindBD(disk)
2137
  except errors.BlockDeviceError, err:
2138
    _Fail("Failed to find device: %s", err, exc=True)
2139

    
2140
  if rbd is None:
2141
    return None
2142

    
2143
  return rbd.GetSyncStatus()
2144

    
2145

    
2146
def BlockdevGetsize(disks):
2147
  """Computes the size of the given disks.
2148

2149
  If a disk is not found, returns None instead.
2150

2151
  @type disks: list of L{objects.Disk}
2152
  @param disks: the list of disk to compute the size for
2153
  @rtype: list
2154
  @return: list with elements None if the disk cannot be found,
2155
      otherwise the size
2156

2157
  """
2158
  result = []
2159
  for cf in disks:
2160
    try:
2161
      rbd = _RecursiveFindBD(cf)
2162
    except errors.BlockDeviceError:
2163
      result.append(None)
2164
      continue
2165
    if rbd is None:
2166
      result.append(None)
2167
    else:
2168
      result.append(rbd.GetActualSize())
2169
  return result
2170

    
2171

    
2172
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2173
  """Export a block device to a remote node.
2174

2175
  @type disk: L{objects.Disk}
2176
  @param disk: the description of the disk to export
2177
  @type dest_node: str
2178
  @param dest_node: the destination node to export to
2179
  @type dest_path: str
2180
  @param dest_path: the destination path on the target node
2181
  @type cluster_name: str
2182
  @param cluster_name: the cluster name, needed for SSH hostalias
2183
  @rtype: None
2184

2185
  """
2186
  real_disk = _OpenRealBD(disk)
2187

    
2188
  # the block size on the read dd is 1MiB to match our units
2189
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2190
                               "dd if=%s bs=1048576 count=%s",
2191
                               real_disk.dev_path, str(disk.size))
2192

    
2193
  # we set here a smaller block size as, due to ssh buffering, more
2194
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2195
  # device is not already there or we pass a wrong path; we use
2196
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2197
  # to not buffer too much memory; this means that at best, we flush
2198
  # every 64k, which will not be very fast
2199
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2200
                                " oflag=dsync", dest_path)
2201

    
2202
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2203
                                                   constants.SSH_LOGIN_USER,
2204
                                                   destcmd)
2205

    
2206
  # all commands have been checked, so we're safe to combine them
2207
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2208

    
2209
  result = utils.RunCmd(["bash", "-c", command])
2210

    
2211
  if result.failed:
2212
    _Fail("Disk copy command '%s' returned error: %s"
2213
          " output: %s", command, result.fail_reason, result.output)
2214

    
2215

    
2216
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2217
  """Write a file to the filesystem.
2218

2219
  This allows the master to overwrite(!) a file. It will only perform
2220
  the operation if the file belongs to a list of configuration files.
2221

2222
  @type file_name: str
2223
  @param file_name: the target file name
2224
  @type data: str
2225
  @param data: the new contents of the file
2226
  @type mode: int
2227
  @param mode: the mode to give the file (can be None)
2228
  @type uid: string
2229
  @param uid: the owner of the file
2230
  @type gid: string
2231
  @param gid: the group of the file
2232
  @type atime: float
2233
  @param atime: the atime to set on the file (can be None)
2234
  @type mtime: float
2235
  @param mtime: the mtime to set on the file (can be None)
2236
  @rtype: None
2237

2238
  """
2239
  file_name = vcluster.LocalizeVirtualPath(file_name)
2240

    
2241
  if not os.path.isabs(file_name):
2242
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2243

    
2244
  if file_name not in _ALLOWED_UPLOAD_FILES:
2245
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2246
          file_name)
2247

    
2248
  raw_data = _Decompress(data)
2249

    
2250
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2251
    _Fail("Invalid username/groupname type")
2252

    
2253
  getents = runtime.GetEnts()
2254
  uid = getents.LookupUser(uid)
2255
  gid = getents.LookupGroup(gid)
2256

    
2257
  utils.SafeWriteFile(file_name, None,
2258
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2259
                      atime=atime, mtime=mtime)
2260

    
2261

    
2262
def RunOob(oob_program, command, node, timeout):
2263
  """Executes oob_program with given command on given node.
2264

2265
  @param oob_program: The path to the executable oob_program
2266
  @param command: The command to invoke on oob_program
2267
  @param node: The node given as an argument to the program
2268
  @param timeout: Timeout after which we kill the oob program
2269

2270
  @return: stdout
2271
  @raise RPCFail: If execution fails for some reason
2272

2273
  """
2274
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2275

    
2276
  if result.failed:
2277
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2278
          result.fail_reason, result.output)
2279

    
2280
  return result.stdout
2281

    
2282

    
2283
def _OSOndiskAPIVersion(os_dir):
2284
  """Compute and return the API version of a given OS.
2285

2286
  This function will try to read the API version of the OS residing in
2287
  the 'os_dir' directory.
2288

2289
  @type os_dir: str
2290
  @param os_dir: the directory in which we should look for the OS
2291
  @rtype: tuple
2292
  @return: tuple (status, data) with status denoting the validity and
2293
      data holding either the vaid versions or an error message
2294

2295
  """
2296
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2297

    
2298
  try:
2299
    st = os.stat(api_file)
2300
  except EnvironmentError, err:
2301
    return False, ("Required file '%s' not found under path %s: %s" %
2302
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2303

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

    
2308
  try:
2309
    api_versions = utils.ReadFile(api_file).splitlines()
2310
  except EnvironmentError, err:
2311
    return False, ("Error while reading the API version file at %s: %s" %
2312
                   (api_file, utils.ErrnoOrStr(err)))
2313

    
2314
  try:
2315
    api_versions = [int(version.strip()) for version in api_versions]
2316
  except (TypeError, ValueError), err:
2317
    return False, ("API version(s) can't be converted to integer: %s" %
2318
                   str(err))
2319

    
2320
  return True, api_versions
2321

    
2322

    
2323
def DiagnoseOS(top_dirs=None):
2324
  """Compute the validity for all OSes.
2325

2326
  @type top_dirs: list
2327
  @param top_dirs: the list of directories in which to
2328
      search (if not given defaults to
2329
      L{pathutils.OS_SEARCH_PATH})
2330
  @rtype: list of L{objects.OS}
2331
  @return: a list of tuples (name, path, status, diagnose, variants,
2332
      parameters, api_version) for all (potential) OSes under all
2333
      search paths, where:
2334
          - name is the (potential) OS name
2335
          - path is the full path to the OS
2336
          - status True/False is the validity of the OS
2337
          - diagnose is the error message for an invalid OS, otherwise empty
2338
          - variants is a list of supported OS variants, if any
2339
          - parameters is a list of (name, help) parameters, if any
2340
          - api_version is a list of support OS API versions
2341

2342
  """
2343
  if top_dirs is None:
2344
    top_dirs = pathutils.OS_SEARCH_PATH
2345

    
2346
  result = []
2347
  for dir_name in top_dirs:
2348
    if os.path.isdir(dir_name):
2349
      try:
2350
        f_names = utils.ListVisibleFiles(dir_name)
2351
      except EnvironmentError, err:
2352
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2353
        break
2354
      for name in f_names:
2355
        os_path = utils.PathJoin(dir_name, name)
2356
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2357
        if status:
2358
          diagnose = ""
2359
          variants = os_inst.supported_variants
2360
          parameters = os_inst.supported_parameters
2361
          api_versions = os_inst.api_versions
2362
        else:
2363
          diagnose = os_inst
2364
          variants = parameters = api_versions = []
2365
        result.append((name, os_path, status, diagnose, variants,
2366
                       parameters, api_versions))
2367

    
2368
  return result
2369

    
2370

    
2371
def _TryOSFromDisk(name, base_dir=None):
2372
  """Create an OS instance from disk.
2373

2374
  This function will return an OS instance if the given name is a
2375
  valid OS name.
2376

2377
  @type base_dir: string
2378
  @keyword base_dir: Base directory containing OS installations.
2379
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2380
  @rtype: tuple
2381
  @return: success and either the OS instance if we find a valid one,
2382
      or error message
2383

2384
  """
2385
  if base_dir is None:
2386
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2387
  else:
2388
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2389

    
2390
  if os_dir is None:
2391
    return False, "Directory for OS %s not found in search path" % name
2392

    
2393
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2394
  if not status:
2395
    # push the error up
2396
    return status, api_versions
2397

    
2398
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2399
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2400
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2401

    
2402
  # OS Files dictionary, we will populate it with the absolute path
2403
  # names; if the value is True, then it is a required file, otherwise
2404
  # an optional one
2405
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2406

    
2407
  if max(api_versions) >= constants.OS_API_V15:
2408
    os_files[constants.OS_VARIANTS_FILE] = False
2409

    
2410
  if max(api_versions) >= constants.OS_API_V20:
2411
    os_files[constants.OS_PARAMETERS_FILE] = True
2412
  else:
2413
    del os_files[constants.OS_SCRIPT_VERIFY]
2414

    
2415
  for (filename, required) in os_files.items():
2416
    os_files[filename] = utils.PathJoin(os_dir, filename)
2417

    
2418
    try:
2419
      st = os.stat(os_files[filename])
2420
    except EnvironmentError, err:
2421
      if err.errno == errno.ENOENT and not required:
2422
        del os_files[filename]
2423
        continue
2424
      return False, ("File '%s' under path '%s' is missing (%s)" %
2425
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2426

    
2427
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2428
      return False, ("File '%s' under path '%s' is not a regular file" %
2429
                     (filename, os_dir))
2430

    
2431
    if filename in constants.OS_SCRIPTS:
2432
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2433
        return False, ("File '%s' under path '%s' is not executable" %
2434
                       (filename, os_dir))
2435

    
2436
  variants = []
2437
  if constants.OS_VARIANTS_FILE in os_files:
2438
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2439
    try:
2440
      variants = \
2441
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2442
    except EnvironmentError, err:
2443
      # we accept missing files, but not other errors
2444
      if err.errno != errno.ENOENT:
2445
        return False, ("Error while reading the OS variants file at %s: %s" %
2446
                       (variants_file, utils.ErrnoOrStr(err)))
2447

    
2448
  parameters = []
2449
  if constants.OS_PARAMETERS_FILE in os_files:
2450
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2451
    try:
2452
      parameters = utils.ReadFile(parameters_file).splitlines()
2453
    except EnvironmentError, err:
2454
      return False, ("Error while reading the OS parameters file at %s: %s" %
2455
                     (parameters_file, utils.ErrnoOrStr(err)))
2456
    parameters = [v.split(None, 1) for v in parameters]
2457

    
2458
  os_obj = objects.OS(name=name, path=os_dir,
2459
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2460
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2461
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2462
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2463
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2464
                                                 None),
2465
                      supported_variants=variants,
2466
                      supported_parameters=parameters,
2467
                      api_versions=api_versions)
2468
  return True, os_obj
2469

    
2470

    
2471
def OSFromDisk(name, base_dir=None):
2472
  """Create an OS instance from disk.
2473

2474
  This function will return an OS instance if the given name is a
2475
  valid OS name. Otherwise, it will raise an appropriate
2476
  L{RPCFail} exception, detailing why this is not a valid OS.
2477

2478
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2479
  an exception but returns true/false status data.
2480

2481
  @type base_dir: string
2482
  @keyword base_dir: Base directory containing OS installations.
2483
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2484
  @rtype: L{objects.OS}
2485
  @return: the OS instance if we find a valid one
2486
  @raise RPCFail: if we don't find a valid OS
2487

2488
  """
2489
  name_only = objects.OS.GetName(name)
2490
  status, payload = _TryOSFromDisk(name_only, base_dir)
2491

    
2492
  if not status:
2493
    _Fail(payload)
2494

    
2495
  return payload
2496

    
2497

    
2498
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2499
  """Calculate the basic environment for an os script.
2500

2501
  @type os_name: str
2502
  @param os_name: full operating system name (including variant)
2503
  @type inst_os: L{objects.OS}
2504
  @param inst_os: operating system for which the environment is being built
2505
  @type os_params: dict
2506
  @param os_params: the OS parameters
2507
  @type debug: integer
2508
  @param debug: debug level (0 or 1, for OS Api 10)
2509
  @rtype: dict
2510
  @return: dict of environment variables
2511
  @raise errors.BlockDeviceError: if the block device
2512
      cannot be found
2513

2514
  """
2515
  result = {}
2516
  api_version = \
2517
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2518
  result["OS_API_VERSION"] = "%d" % api_version
2519
  result["OS_NAME"] = inst_os.name
2520
  result["DEBUG_LEVEL"] = "%d" % debug
2521

    
2522
  # OS variants
2523
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2524
    variant = objects.OS.GetVariant(os_name)
2525
    if not variant:
2526
      variant = inst_os.supported_variants[0]
2527
  else:
2528
    variant = ""
2529
  result["OS_VARIANT"] = variant
2530

    
2531
  # OS params
2532
  for pname, pvalue in os_params.items():
2533
    result["OSP_%s" % pname.upper()] = pvalue
2534

    
2535
  # Set a default path otherwise programs called by OS scripts (or
2536
  # even hooks called from OS scripts) might break, and we don't want
2537
  # to have each script require setting a PATH variable
2538
  result["PATH"] = constants.HOOKS_PATH
2539

    
2540
  return result
2541

    
2542

    
2543
def OSEnvironment(instance, inst_os, debug=0):
2544
  """Calculate the environment for an os script.
2545

2546
  @type instance: L{objects.Instance}
2547
  @param instance: target instance for the os script run
2548
  @type inst_os: L{objects.OS}
2549
  @param inst_os: operating system for which the environment is being built
2550
  @type debug: integer
2551
  @param debug: debug level (0 or 1, for OS Api 10)
2552
  @rtype: dict
2553
  @return: dict of environment variables
2554
  @raise errors.BlockDeviceError: if the block device
2555
      cannot be found
2556

2557
  """
2558
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2559

    
2560
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2561
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2562

    
2563
  result["HYPERVISOR"] = instance.hypervisor
2564
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2565
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2566
  result["INSTANCE_SECONDARY_NODES"] = \
2567
      ("%s" % " ".join(instance.secondary_nodes))
2568

    
2569
  # Disks
2570
  for idx, disk in enumerate(instance.disks):
2571
    real_disk = _OpenRealBD(disk)
2572
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2573
    result["DISK_%d_ACCESS" % idx] = disk.mode
2574
    result["DISK_%d_UUID" % idx] = disk.uuid
2575
    if disk.name:
2576
      result["DISK_%d_NAME" % idx] = disk.name
2577
    if constants.HV_DISK_TYPE in instance.hvparams:
2578
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2579
        instance.hvparams[constants.HV_DISK_TYPE]
2580
    if disk.dev_type in constants.LDS_BLOCK:
2581
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2582
    elif disk.dev_type == constants.LD_FILE:
2583
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2584
        "file:%s" % disk.physical_id[0]
2585

    
2586
  # NICs
2587
  for idx, nic in enumerate(instance.nics):
2588
    result["NIC_%d_MAC" % idx] = nic.mac
2589
    result["NIC_%d_UUID" % idx] = nic.uuid
2590
    if nic.name:
2591
      result["NIC_%d_NAME" % idx] = nic.name
2592
    if nic.ip:
2593
      result["NIC_%d_IP" % idx] = nic.ip
2594
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2595
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2596
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2597
    if nic.nicparams[constants.NIC_LINK]:
2598
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2599
    if nic.netinfo:
2600
      nobj = objects.Network.FromDict(nic.netinfo)
2601
      result.update(nobj.HooksDict("NIC_%d_" % idx))
2602
    if constants.HV_NIC_TYPE in instance.hvparams:
2603
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2604
        instance.hvparams[constants.HV_NIC_TYPE]
2605

    
2606
  # HV/BE params
2607
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2608
    for key, value in source.items():
2609
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2610

    
2611
  return result
2612

    
2613

    
2614
def DiagnoseExtStorage(top_dirs=None):
2615
  """Compute the validity for all ExtStorage Providers.
2616

2617
  @type top_dirs: list
2618
  @param top_dirs: the list of directories in which to
2619
      search (if not given defaults to
2620
      L{pathutils.ES_SEARCH_PATH})
2621
  @rtype: list of L{objects.ExtStorage}
2622
  @return: a list of tuples (name, path, status, diagnose, parameters)
2623
      for all (potential) ExtStorage Providers under all
2624
      search paths, where:
2625
          - name is the (potential) ExtStorage Provider
2626
          - path is the full path to the ExtStorage Provider
2627
          - status True/False is the validity of the ExtStorage Provider
2628
          - diagnose is the error message for an invalid ExtStorage Provider,
2629
            otherwise empty
2630
          - parameters is a list of (name, help) parameters, if any
2631

2632
  """
2633
  if top_dirs is None:
2634
    top_dirs = pathutils.ES_SEARCH_PATH
2635

    
2636
  result = []
2637
  for dir_name in top_dirs:
2638
    if os.path.isdir(dir_name):
2639
      try:
2640
        f_names = utils.ListVisibleFiles(dir_name)
2641
      except EnvironmentError, err:
2642
        logging.exception("Can't list the ExtStorage directory %s: %s",
2643
                          dir_name, err)
2644
        break
2645
      for name in f_names:
2646
        es_path = utils.PathJoin(dir_name, name)
2647
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2648
        if status:
2649
          diagnose = ""
2650
          parameters = es_inst.supported_parameters
2651
        else:
2652
          diagnose = es_inst
2653
          parameters = []
2654
        result.append((name, es_path, status, diagnose, parameters))
2655

    
2656
  return result
2657

    
2658

    
2659
def BlockdevGrow(disk, amount, dryrun, backingstore):
2660
  """Grow a stack of block devices.
2661

2662
  This function is called recursively, with the childrens being the
2663
  first ones to resize.
2664

2665
  @type disk: L{objects.Disk}
2666
  @param disk: the disk to be grown
2667
  @type amount: integer
2668
  @param amount: the amount (in mebibytes) to grow with
2669
  @type dryrun: boolean
2670
  @param dryrun: whether to execute the operation in simulation mode
2671
      only, without actually increasing the size
2672
  @param backingstore: whether to execute the operation on backing storage
2673
      only, or on "logical" storage only; e.g. DRBD is logical storage,
2674
      whereas LVM, file, RBD are backing storage
2675
  @rtype: (status, result)
2676
  @return: a tuple with the status of the operation (True/False), and
2677
      the errors message if status is False
2678

2679
  """
2680
  r_dev = _RecursiveFindBD(disk)
2681
  if r_dev is None:
2682
    _Fail("Cannot find block device %s", disk)
2683

    
2684
  try:
2685
    r_dev.Grow(amount, dryrun, backingstore)
2686
  except errors.BlockDeviceError, err:
2687
    _Fail("Failed to grow block device: %s", err, exc=True)
2688

    
2689

    
2690
def BlockdevSnapshot(disk):
2691
  """Create a snapshot copy of a block device.
2692

2693
  This function is called recursively, and the snapshot is actually created
2694
  just for the leaf lvm backend device.
2695

2696
  @type disk: L{objects.Disk}
2697
  @param disk: the disk to be snapshotted
2698
  @rtype: string
2699
  @return: snapshot disk ID as (vg, lv)
2700

2701
  """
2702
  if disk.dev_type == constants.LD_DRBD8:
2703
    if not disk.children:
2704
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2705
            disk.unique_id)
2706
    return BlockdevSnapshot(disk.children[0])
2707
  elif disk.dev_type == constants.LD_LV:
2708
    r_dev = _RecursiveFindBD(disk)
2709
    if r_dev is not None:
2710
      # FIXME: choose a saner value for the snapshot size
2711
      # let's stay on the safe side and ask for the full size, for now
2712
      return r_dev.Snapshot(disk.size)
2713
    else:
2714
      _Fail("Cannot find block device %s", disk)
2715
  else:
2716
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2717
          disk.unique_id, disk.dev_type)
2718

    
2719

    
2720
def BlockdevSetInfo(disk, info):
2721
  """Sets 'metadata' information on block devices.
2722

2723
  This function sets 'info' metadata on block devices. Initial
2724
  information is set at device creation; this function should be used
2725
  for example after renames.
2726

2727
  @type disk: L{objects.Disk}
2728
  @param disk: the disk to be grown
2729
  @type info: string
2730
  @param info: new 'info' metadata
2731
  @rtype: (status, result)
2732
  @return: a tuple with the status of the operation (True/False), and
2733
      the errors message if status is False
2734

2735
  """
2736
  r_dev = _RecursiveFindBD(disk)
2737
  if r_dev is None:
2738
    _Fail("Cannot find block device %s", disk)
2739

    
2740
  try:
2741
    r_dev.SetInfo(info)
2742
  except errors.BlockDeviceError, err:
2743
    _Fail("Failed to set information on block device: %s", err, exc=True)
2744

    
2745

    
2746
def FinalizeExport(instance, snap_disks):
2747
  """Write out the export configuration information.
2748

2749
  @type instance: L{objects.Instance}
2750
  @param instance: the instance which we export, used for
2751
      saving configuration
2752
  @type snap_disks: list of L{objects.Disk}
2753
  @param snap_disks: list of snapshot block devices, which
2754
      will be used to get the actual name of the dump file
2755

2756
  @rtype: None
2757

2758
  """
2759
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2760
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2761

    
2762
  config = objects.SerializableConfigParser()
2763

    
2764
  config.add_section(constants.INISECT_EXP)
2765
  config.set(constants.INISECT_EXP, "version", "0")
2766
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2767
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
2768
  config.set(constants.INISECT_EXP, "os", instance.os)
2769
  config.set(constants.INISECT_EXP, "compression", "none")
2770

    
2771
  config.add_section(constants.INISECT_INS)
2772
  config.set(constants.INISECT_INS, "name", instance.name)
2773
  config.set(constants.INISECT_INS, "maxmem", "%d" %
2774
             instance.beparams[constants.BE_MAXMEM])
2775
  config.set(constants.INISECT_INS, "minmem", "%d" %
2776
             instance.beparams[constants.BE_MINMEM])
2777
  # "memory" is deprecated, but useful for exporting to old ganeti versions
2778
  config.set(constants.INISECT_INS, "memory", "%d" %
2779
             instance.beparams[constants.BE_MAXMEM])
2780
  config.set(constants.INISECT_INS, "vcpus", "%d" %
2781
             instance.beparams[constants.BE_VCPUS])
2782
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2783
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2784
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2785

    
2786
  nic_total = 0
2787
  for nic_count, nic in enumerate(instance.nics):
2788
    nic_total += 1
2789
    config.set(constants.INISECT_INS, "nic%d_mac" %
2790
               nic_count, "%s" % nic.mac)
2791
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2792
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2793
               "%s" % nic.network)
2794
    for param in constants.NICS_PARAMETER_TYPES:
2795
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2796
                 "%s" % nic.nicparams.get(param, None))
2797
  # TODO: redundant: on load can read nics until it doesn't exist
2798
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2799

    
2800
  disk_total = 0
2801
  for disk_count, disk in enumerate(snap_disks):
2802
    if disk:
2803
      disk_total += 1
2804
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2805
                 ("%s" % disk.iv_name))
2806
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2807
                 ("%s" % disk.physical_id[1]))
2808
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2809
                 ("%d" % disk.size))
2810

    
2811
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2812

    
2813
  # New-style hypervisor/backend parameters
2814

    
2815
  config.add_section(constants.INISECT_HYP)
2816
  for name, value in instance.hvparams.items():
2817
    if name not in constants.HVC_GLOBALS:
2818
      config.set(constants.INISECT_HYP, name, str(value))
2819

    
2820
  config.add_section(constants.INISECT_BEP)
2821
  for name, value in instance.beparams.items():
2822
    config.set(constants.INISECT_BEP, name, str(value))
2823

    
2824
  config.add_section(constants.INISECT_OSP)
2825
  for name, value in instance.osparams.items():
2826
    config.set(constants.INISECT_OSP, name, str(value))
2827

    
2828
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2829
                  data=config.Dumps())
2830
  shutil.rmtree(finaldestdir, ignore_errors=True)
2831
  shutil.move(destdir, finaldestdir)
2832

    
2833

    
2834
def ExportInfo(dest):
2835
  """Get export configuration information.
2836

2837
  @type dest: str
2838
  @param dest: directory containing the export
2839

2840
  @rtype: L{objects.SerializableConfigParser}
2841
  @return: a serializable config file containing the
2842
      export info
2843

2844
  """
2845
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2846

    
2847
  config = objects.SerializableConfigParser()
2848
  config.read(cff)
2849

    
2850
  if (not config.has_section(constants.INISECT_EXP) or
2851
      not config.has_section(constants.INISECT_INS)):
2852
    _Fail("Export info file doesn't have the required fields")
2853

    
2854
  return config.Dumps()
2855

    
2856

    
2857
def ListExports():
2858
  """Return a list of exports currently available on this machine.
2859

2860
  @rtype: list
2861
  @return: list of the exports
2862

2863
  """
2864
  if os.path.isdir(pathutils.EXPORT_DIR):
2865
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
2866
  else:
2867
    _Fail("No exports directory")
2868

    
2869

    
2870
def RemoveExport(export):
2871
  """Remove an existing export from the node.
2872

2873
  @type export: str
2874
  @param export: the name of the export to remove
2875
  @rtype: None
2876

2877
  """
2878
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
2879

    
2880
  try:
2881
    shutil.rmtree(target)
2882
  except EnvironmentError, err:
2883
    _Fail("Error while removing the export: %s", err, exc=True)
2884

    
2885

    
2886
def BlockdevRename(devlist):
2887
  """Rename a list of block devices.
2888

2889
  @type devlist: list of tuples
2890
  @param devlist: list of tuples of the form  (disk,
2891
      new_logical_id, new_physical_id); disk is an
2892
      L{objects.Disk} object describing the current disk,
2893
      and new logical_id/physical_id is the name we
2894
      rename it to
2895
  @rtype: boolean
2896
  @return: True if all renames succeeded, False otherwise
2897

2898
  """
2899
  msgs = []
2900
  result = True
2901
  for disk, unique_id in devlist:
2902
    dev = _RecursiveFindBD(disk)
2903
    if dev is None:
2904
      msgs.append("Can't find device %s in rename" % str(disk))
2905
      result = False
2906
      continue
2907
    try:
2908
      old_rpath = dev.dev_path
2909
      dev.Rename(unique_id)
2910
      new_rpath = dev.dev_path
2911
      if old_rpath != new_rpath:
2912
        DevCacheManager.RemoveCache(old_rpath)
2913
        # FIXME: we should add the new cache information here, like:
2914
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2915
        # but we don't have the owner here - maybe parse from existing
2916
        # cache? for now, we only lose lvm data when we rename, which
2917
        # is less critical than DRBD or MD
2918
    except errors.BlockDeviceError, err:
2919
      msgs.append("Can't rename device '%s' to '%s': %s" %
2920
                  (dev, unique_id, err))
2921
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2922
      result = False
2923
  if not result:
2924
    _Fail("; ".join(msgs))
2925

    
2926

    
2927
def _TransformFileStorageDir(fs_dir):
2928
  """Checks whether given file_storage_dir is valid.
2929

2930
  Checks wheter the given fs_dir is within the cluster-wide default
2931
  file_storage_dir or the shared_file_storage_dir, which are stored in
2932
  SimpleStore. Only paths under those directories are allowed.
2933

2934
  @type fs_dir: str
2935
  @param fs_dir: the path to check
2936

2937
  @return: the normalized path if valid, None otherwise
2938

2939
  """
2940
  if not (constants.ENABLE_FILE_STORAGE or
2941
          constants.ENABLE_SHARED_FILE_STORAGE):
2942
    _Fail("File storage disabled at configure time")
2943

    
2944
  bdev.CheckFileStoragePath(fs_dir)
2945

    
2946
  return os.path.normpath(fs_dir)
2947

    
2948

    
2949
def CreateFileStorageDir(file_storage_dir):
2950
  """Create file storage directory.
2951

2952
  @type file_storage_dir: str
2953
  @param file_storage_dir: directory to create
2954

2955
  @rtype: tuple
2956
  @return: tuple with first element a boolean indicating wheter dir
2957
      creation was successful or not
2958

2959
  """
2960
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2961
  if os.path.exists(file_storage_dir):
2962
    if not os.path.isdir(file_storage_dir):
2963
      _Fail("Specified storage dir '%s' is not a directory",
2964
            file_storage_dir)
2965
  else:
2966
    try:
2967
      os.makedirs(file_storage_dir, 0750)
2968
    except OSError, err:
2969
      _Fail("Cannot create file storage directory '%s': %s",
2970
            file_storage_dir, err, exc=True)
2971

    
2972

    
2973
def RemoveFileStorageDir(file_storage_dir):
2974
  """Remove file storage directory.
2975

2976
  Remove it only if it's empty. If not log an error and return.
2977

2978
  @type file_storage_dir: str
2979
  @param file_storage_dir: the directory we should cleanup
2980
  @rtype: tuple (success,)
2981
  @return: tuple of one element, C{success}, denoting
2982
      whether the operation was successful
2983

2984
  """
2985
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2986
  if os.path.exists(file_storage_dir):
2987
    if not os.path.isdir(file_storage_dir):
2988
      _Fail("Specified Storage directory '%s' is not a directory",
2989
            file_storage_dir)
2990
    # deletes dir only if empty, otherwise we want to fail the rpc call
2991
    try:
2992
      os.rmdir(file_storage_dir)
2993
    except OSError, err:
2994
      _Fail("Cannot remove file storage directory '%s': %s",
2995
            file_storage_dir, err)
2996

    
2997

    
2998
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2999
  """Rename the file storage directory.
3000

3001
  @type old_file_storage_dir: str
3002
  @param old_file_storage_dir: the current path
3003
  @type new_file_storage_dir: str
3004
  @param new_file_storage_dir: the name we should rename to
3005
  @rtype: tuple (success,)
3006
  @return: tuple of one element, C{success}, denoting
3007
      whether the operation was successful
3008

3009
  """
3010
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3011
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3012
  if not os.path.exists(new_file_storage_dir):
3013
    if os.path.isdir(old_file_storage_dir):
3014
      try:
3015
        os.rename(old_file_storage_dir, new_file_storage_dir)
3016
      except OSError, err:
3017
        _Fail("Cannot rename '%s' to '%s': %s",
3018
              old_file_storage_dir, new_file_storage_dir, err)
3019
    else:
3020
      _Fail("Specified storage dir '%s' is not a directory",
3021
            old_file_storage_dir)
3022
  else:
3023
    if os.path.exists(old_file_storage_dir):
3024
      _Fail("Cannot rename '%s' to '%s': both locations exist",
3025
            old_file_storage_dir, new_file_storage_dir)
3026

    
3027

    
3028
def _EnsureJobQueueFile(file_name):
3029
  """Checks whether the given filename is in the queue directory.
3030

3031
  @type file_name: str
3032
  @param file_name: the file name we should check
3033
  @rtype: None
3034
  @raises RPCFail: if the file is not valid
3035

3036
  """
3037
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3038
    _Fail("Passed job queue file '%s' does not belong to"
3039
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3040

    
3041

    
3042
def JobQueueUpdate(file_name, content):
3043
  """Updates a file in the queue directory.
3044

3045
  This is just a wrapper over L{utils.io.WriteFile}, with proper
3046
  checking.
3047

3048
  @type file_name: str
3049
  @param file_name: the job file name
3050
  @type content: str
3051
  @param content: the new job contents
3052
  @rtype: boolean
3053
  @return: the success of the operation
3054

3055
  """
3056
  file_name = vcluster.LocalizeVirtualPath(file_name)
3057

    
3058
  _EnsureJobQueueFile(file_name)
3059
  getents = runtime.GetEnts()
3060

    
3061
  # Write and replace the file atomically
3062
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3063
                  gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3064

    
3065

    
3066
def JobQueueRename(old, new):
3067
  """Renames a job queue file.
3068

3069
  This is just a wrapper over os.rename with proper checking.
3070

3071
  @type old: str
3072
  @param old: the old (actual) file name
3073
  @type new: str
3074
  @param new: the desired file name
3075
  @rtype: tuple
3076
  @return: the success of the operation and payload
3077

3078
  """
3079
  old = vcluster.LocalizeVirtualPath(old)
3080
  new = vcluster.LocalizeVirtualPath(new)
3081

    
3082
  _EnsureJobQueueFile(old)
3083
  _EnsureJobQueueFile(new)
3084

    
3085
  getents = runtime.GetEnts()
3086

    
3087
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3088
                   dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3089

    
3090

    
3091
def BlockdevClose(instance_name, disks):
3092
  """Closes the given block devices.
3093

3094
  This means they will be switched to secondary mode (in case of
3095
  DRBD).
3096

3097
  @param instance_name: if the argument is not empty, the symlinks
3098
      of this instance will be removed
3099
  @type disks: list of L{objects.Disk}
3100
  @param disks: the list of disks to be closed
3101
  @rtype: tuple (success, message)
3102
  @return: a tuple of success and message, where success
3103
      indicates the succes of the operation, and message
3104
      which will contain the error details in case we
3105
      failed
3106

3107
  """
3108
  bdevs = []
3109
  for cf in disks:
3110
    rd = _RecursiveFindBD(cf)
3111
    if rd is None:
3112
      _Fail("Can't find device %s", cf)
3113
    bdevs.append(rd)
3114

    
3115
  msg = []
3116
  for rd in bdevs:
3117
    try:
3118
      rd.Close()
3119
    except errors.BlockDeviceError, err:
3120
      msg.append(str(err))
3121
  if msg:
3122
    _Fail("Can't make devices secondary: %s", ",".join(msg))
3123
  else:
3124
    if instance_name:
3125
      _RemoveBlockDevLinks(instance_name, disks)
3126

    
3127

    
3128
def ValidateHVParams(hvname, hvparams):
3129
  """Validates the given hypervisor parameters.
3130

3131
  @type hvname: string
3132
  @param hvname: the hypervisor name
3133
  @type hvparams: dict
3134
  @param hvparams: the hypervisor parameters to be validated
3135
  @rtype: None
3136

3137
  """
3138
  try:
3139
    hv_type = hypervisor.GetHypervisor(hvname)
3140
    hv_type.ValidateParameters(hvparams)
3141
  except errors.HypervisorError, err:
3142
    _Fail(str(err), log=False)
3143

    
3144

    
3145
def _CheckOSPList(os_obj, parameters):
3146
  """Check whether a list of parameters is supported by the OS.
3147

3148
  @type os_obj: L{objects.OS}
3149
  @param os_obj: OS object to check
3150
  @type parameters: list
3151
  @param parameters: the list of parameters to check
3152

3153
  """
3154
  supported = [v[0] for v in os_obj.supported_parameters]
3155
  delta = frozenset(parameters).difference(supported)
3156
  if delta:
3157
    _Fail("The following parameters are not supported"
3158
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3159

    
3160

    
3161
def ValidateOS(required, osname, checks, osparams):
3162
  """Validate the given OS' parameters.
3163

3164
  @type required: boolean
3165
  @param required: whether absence of the OS should translate into
3166
      failure or not
3167
  @type osname: string
3168
  @param osname: the OS to be validated
3169
  @type checks: list
3170
  @param checks: list of the checks to run (currently only 'parameters')
3171
  @type osparams: dict
3172
  @param osparams: dictionary with OS parameters
3173
  @rtype: boolean
3174
  @return: True if the validation passed, or False if the OS was not
3175
      found and L{required} was false
3176

3177
  """
3178
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3179
    _Fail("Unknown checks required for OS %s: %s", osname,
3180
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3181

    
3182
  name_only = objects.OS.GetName(osname)
3183
  status, tbv = _TryOSFromDisk(name_only, None)
3184

    
3185
  if not status:
3186
    if required:
3187
      _Fail(tbv)
3188
    else:
3189
      return False
3190

    
3191
  if max(tbv.api_versions) < constants.OS_API_V20:
3192
    return True
3193

    
3194
  if constants.OS_VALIDATE_PARAMETERS in checks:
3195
    _CheckOSPList(tbv, osparams.keys())
3196

    
3197
  validate_env = OSCoreEnv(osname, tbv, osparams)
3198
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3199
                        cwd=tbv.path, reset_env=True)
3200
  if result.failed:
3201
    logging.error("os validate command '%s' returned error: %s output: %s",
3202
                  result.cmd, result.fail_reason, result.output)
3203
    _Fail("OS validation script failed (%s), output: %s",
3204
          result.fail_reason, result.output, log=False)
3205

    
3206
  return True
3207

    
3208

    
3209
def DemoteFromMC():
3210
  """Demotes the current node from master candidate role.
3211

3212
  """
3213
  # try to ensure we're not the master by mistake
3214
  master, myself = ssconf.GetMasterAndMyself()
3215
  if master == myself:
3216
    _Fail("ssconf status shows I'm the master node, will not demote")
3217

    
3218
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3219
  if not result.failed:
3220
    _Fail("The master daemon is running, will not demote")
3221

    
3222
  try:
3223
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3224
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3225
  except EnvironmentError, err:
3226
    if err.errno != errno.ENOENT:
3227
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3228

    
3229
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3230

    
3231

    
3232
def _GetX509Filenames(cryptodir, name):
3233
  """Returns the full paths for the private key and certificate.
3234

3235
  """
3236
  return (utils.PathJoin(cryptodir, name),
3237
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3238
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3239

    
3240

    
3241
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3242
  """Creates a new X509 certificate for SSL/TLS.
3243

3244
  @type validity: int
3245
  @param validity: Validity in seconds
3246
  @rtype: tuple; (string, string)
3247
  @return: Certificate name and public part
3248

3249
  """
3250
  (key_pem, cert_pem) = \
3251
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3252
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3253

    
3254
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3255
                              prefix="x509-%s-" % utils.TimestampForFilename())
3256
  try:
3257
    name = os.path.basename(cert_dir)
3258
    assert len(name) > 5
3259

    
3260
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3261

    
3262
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3263
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3264

    
3265
    # Never return private key as it shouldn't leave the node
3266
    return (name, cert_pem)
3267
  except Exception:
3268
    shutil.rmtree(cert_dir, ignore_errors=True)
3269
    raise
3270

    
3271

    
3272
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3273
  """Removes a X509 certificate.
3274

3275
  @type name: string
3276
  @param name: Certificate name
3277

3278
  """
3279
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3280

    
3281
  utils.RemoveFile(key_file)
3282
  utils.RemoveFile(cert_file)
3283

    
3284
  try:
3285
    os.rmdir(cert_dir)
3286
  except EnvironmentError, err:
3287
    _Fail("Cannot remove certificate directory '%s': %s",
3288
          cert_dir, err)
3289

    
3290

    
3291
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3292
  """Returns the command for the requested input/output.
3293

3294
  @type instance: L{objects.Instance}
3295
  @param instance: The instance object
3296
  @param mode: Import/export mode
3297
  @param ieio: Input/output type
3298
  @param ieargs: Input/output arguments
3299

3300
  """
3301
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3302

    
3303
  env = None
3304
  prefix = None
3305
  suffix = None
3306
  exp_size = None
3307

    
3308
  if ieio == constants.IEIO_FILE:
3309
    (filename, ) = ieargs
3310

    
3311
    if not utils.IsNormAbsPath(filename):
3312
      _Fail("Path '%s' is not normalized or absolute", filename)
3313

    
3314
    real_filename = os.path.realpath(filename)
3315
    directory = os.path.dirname(real_filename)
3316

    
3317
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3318
      _Fail("File '%s' is not under exports directory '%s': %s",
3319
            filename, pathutils.EXPORT_DIR, real_filename)
3320

    
3321
    # Create directory
3322
    utils.Makedirs(directory, mode=0750)
3323

    
3324
    quoted_filename = utils.ShellQuote(filename)
3325

    
3326
    if mode == constants.IEM_IMPORT:
3327
      suffix = "> %s" % quoted_filename
3328
    elif mode == constants.IEM_EXPORT:
3329
      suffix = "< %s" % quoted_filename
3330

    
3331
      # Retrieve file size
3332
      try:
3333
        st = os.stat(filename)
3334
      except EnvironmentError, err:
3335
        logging.error("Can't stat(2) %s: %s", filename, err)
3336
      else:
3337
        exp_size = utils.BytesToMebibyte(st.st_size)
3338

    
3339
  elif ieio == constants.IEIO_RAW_DISK:
3340
    (disk, ) = ieargs
3341

    
3342
    real_disk = _OpenRealBD(disk)
3343

    
3344
    if mode == constants.IEM_IMPORT:
3345
      # we set here a smaller block size as, due to transport buffering, more
3346
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3347
      # is not already there or we pass a wrong path; we use notrunc to no
3348
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3349
      # much memory; this means that at best, we flush every 64k, which will
3350
      # not be very fast
3351
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3352
                                    " bs=%s oflag=dsync"),
3353
                                    real_disk.dev_path,
3354
                                    str(64 * 1024))
3355

    
3356
    elif mode == constants.IEM_EXPORT:
3357
      # the block size on the read dd is 1MiB to match our units
3358
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3359
                                   real_disk.dev_path,
3360
                                   str(1024 * 1024), # 1 MB
3361
                                   str(disk.size))
3362
      exp_size = disk.size
3363

    
3364
  elif ieio == constants.IEIO_SCRIPT:
3365
    (disk, disk_index, ) = ieargs
3366

    
3367
    assert isinstance(disk_index, (int, long))
3368

    
3369
    real_disk = _OpenRealBD(disk)
3370

    
3371
    inst_os = OSFromDisk(instance.os)
3372
    env = OSEnvironment(instance, inst_os)
3373

    
3374
    if mode == constants.IEM_IMPORT:
3375
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3376
      env["IMPORT_INDEX"] = str(disk_index)
3377
      script = inst_os.import_script
3378

    
3379
    elif mode == constants.IEM_EXPORT:
3380
      env["EXPORT_DEVICE"] = real_disk.dev_path
3381
      env["EXPORT_INDEX"] = str(disk_index)
3382
      script = inst_os.export_script
3383

    
3384
    # TODO: Pass special environment only to script
3385
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3386

    
3387
    if mode == constants.IEM_IMPORT:
3388
      suffix = "| %s" % script_cmd
3389

    
3390
    elif mode == constants.IEM_EXPORT:
3391
      prefix = "%s |" % script_cmd
3392

    
3393
    # Let script predict size
3394
    exp_size = constants.IE_CUSTOM_SIZE
3395

    
3396
  else:
3397
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3398

    
3399
  return (env, prefix, suffix, exp_size)
3400

    
3401

    
3402
def _CreateImportExportStatusDir(prefix):
3403
  """Creates status directory for import/export.
3404

3405
  """
3406
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3407
                          prefix=("%s-%s-" %
3408
                                  (prefix, utils.TimestampForFilename())))
3409

    
3410

    
3411
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3412
                            ieio, ieioargs):
3413
  """Starts an import or export daemon.
3414

3415
  @param mode: Import/output mode
3416
  @type opts: L{objects.ImportExportOptions}
3417
  @param opts: Daemon options
3418
  @type host: string
3419
  @param host: Remote host for export (None for import)
3420
  @type port: int
3421
  @param port: Remote port for export (None for import)
3422
  @type instance: L{objects.Instance}
3423
  @param instance: Instance object
3424
  @type component: string
3425
  @param component: which part of the instance is transferred now,
3426
      e.g. 'disk/0'
3427
  @param ieio: Input/output type
3428
  @param ieioargs: Input/output arguments
3429

3430
  """
3431
  if mode == constants.IEM_IMPORT:
3432
    prefix = "import"
3433

    
3434
    if not (host is None and port is None):
3435
      _Fail("Can not specify host or port on import")
3436

    
3437
  elif mode == constants.IEM_EXPORT:
3438
    prefix = "export"
3439

    
3440
    if host is None or port is None:
3441
      _Fail("Host and port must be specified for an export")
3442

    
3443
  else:
3444
    _Fail("Invalid mode %r", mode)
3445

    
3446
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3447
    _Fail("Cluster certificate can only be used for both key and CA")
3448

    
3449
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3450
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3451

    
3452
  if opts.key_name is None:
3453
    # Use server.pem
3454
    key_path = pathutils.NODED_CERT_FILE
3455
    cert_path = pathutils.NODED_CERT_FILE
3456
    assert opts.ca_pem is None
3457
  else:
3458
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3459
                                                 opts.key_name)
3460
    assert opts.ca_pem is not None
3461

    
3462
  for i in [key_path, cert_path]:
3463
    if not os.path.exists(i):
3464
      _Fail("File '%s' does not exist" % i)
3465

    
3466
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3467
  try:
3468
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3469
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3470
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3471

    
3472
    if opts.ca_pem is None:
3473
      # Use server.pem
3474
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3475
    else:
3476
      ca = opts.ca_pem
3477

    
3478
    # Write CA file
3479
    utils.WriteFile(ca_file, data=ca, mode=0400)
3480

    
3481
    cmd = [
3482
      pathutils.IMPORT_EXPORT_DAEMON,
3483
      status_file, mode,
3484
      "--key=%s" % key_path,
3485
      "--cert=%s" % cert_path,
3486
      "--ca=%s" % ca_file,
3487
      ]
3488

    
3489
    if host:
3490
      cmd.append("--host=%s" % host)
3491

    
3492
    if port:
3493
      cmd.append("--port=%s" % port)
3494

    
3495
    if opts.ipv6:
3496
      cmd.append("--ipv6")
3497
    else:
3498
      cmd.append("--ipv4")
3499

    
3500
    if opts.compress:
3501
      cmd.append("--compress=%s" % opts.compress)
3502

    
3503
    if opts.magic:
3504
      cmd.append("--magic=%s" % opts.magic)
3505

    
3506
    if exp_size is not None:
3507
      cmd.append("--expected-size=%s" % exp_size)
3508

    
3509
    if cmd_prefix:
3510
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3511

    
3512
    if cmd_suffix:
3513
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3514

    
3515
    if mode == constants.IEM_EXPORT:
3516
      # Retry connection a few times when connecting to remote peer
3517
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3518
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3519
    elif opts.connect_timeout is not None:
3520
      assert mode == constants.IEM_IMPORT
3521
      # Overall timeout for establishing connection while listening
3522
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3523

    
3524
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3525

    
3526
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3527
    # support for receiving a file descriptor for output
3528
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3529
                      output=logfile)
3530

    
3531
    # The import/export name is simply the status directory name
3532
    return os.path.basename(status_dir)
3533

    
3534
  except Exception:
3535
    shutil.rmtree(status_dir, ignore_errors=True)
3536
    raise
3537

    
3538

    
3539
def GetImportExportStatus(names):
3540
  """Returns import/export daemon status.
3541

3542
  @type names: sequence
3543
  @param names: List of names
3544
  @rtype: List of dicts
3545
  @return: Returns a list of the state of each named import/export or None if a
3546
           status couldn't be read
3547

3548
  """
3549
  result = []
3550

    
3551
  for name in names:
3552
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3553
                                 _IES_STATUS_FILE)
3554

    
3555
    try:
3556
      data = utils.ReadFile(status_file)
3557
    except EnvironmentError, err:
3558
      if err.errno != errno.ENOENT:
3559
        raise
3560
      data = None
3561

    
3562
    if not data:
3563
      result.append(None)
3564
      continue
3565

    
3566
    result.append(serializer.LoadJson(data))
3567

    
3568
  return result
3569

    
3570

    
3571
def AbortImportExport(name):
3572
  """Sends SIGTERM to a running import/export daemon.
3573

3574
  """
3575
  logging.info("Abort import/export %s", name)
3576

    
3577
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3578
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3579

    
3580
  if pid:
3581
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3582
                 name, pid)
3583
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3584

    
3585

    
3586
def CleanupImportExport(name):
3587
  """Cleanup after an import or export.
3588

3589
  If the import/export daemon is still running it's killed. Afterwards the
3590
  whole status directory is removed.
3591

3592
  """
3593
  logging.info("Finalizing import/export %s", name)
3594

    
3595
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3596

    
3597
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3598

    
3599
  if pid:
3600
    logging.info("Import/export %s is still running with PID %s",
3601
                 name, pid)
3602
    utils.KillProcess(pid, waitpid=False)
3603

    
3604
  shutil.rmtree(status_dir, ignore_errors=True)
3605

    
3606

    
3607
def _FindDisks(nodes_ip, disks):
3608
  """Sets the physical ID on disks and returns the block devices.
3609

3610
  """
3611
  # set the correct physical ID
3612
  my_name = netutils.Hostname.GetSysName()
3613
  for cf in disks:
3614
    cf.SetPhysicalID(my_name, nodes_ip)
3615

    
3616
  bdevs = []
3617

    
3618
  for cf in disks:
3619
    rd = _RecursiveFindBD(cf)
3620
    if rd is None:
3621
      _Fail("Can't find device %s", cf)
3622
    bdevs.append(rd)
3623
  return bdevs
3624

    
3625

    
3626
def DrbdDisconnectNet(nodes_ip, disks):
3627
  """Disconnects the network on a list of drbd devices.
3628

3629
  """
3630
  bdevs = _FindDisks(nodes_ip, disks)
3631

    
3632
  # disconnect disks
3633
  for rd in bdevs:
3634
    try:
3635
      rd.DisconnectNet()
3636
    except errors.BlockDeviceError, err:
3637
      _Fail("Can't change network configuration to standalone mode: %s",
3638
            err, exc=True)
3639

    
3640

    
3641
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3642
  """Attaches the network on a list of drbd devices.
3643

3644
  """
3645
  bdevs = _FindDisks(nodes_ip, disks)
3646

    
3647
  if multimaster:
3648
    for idx, rd in enumerate(bdevs):
3649
      try:
3650
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3651
      except EnvironmentError, err:
3652
        _Fail("Can't create symlink: %s", err)
3653
  # reconnect disks, switch to new master configuration and if
3654
  # needed primary mode
3655
  for rd in bdevs:
3656
    try:
3657
      rd.AttachNet(multimaster)
3658
    except errors.BlockDeviceError, err:
3659
      _Fail("Can't change network configuration: %s", err)
3660

    
3661
  # wait until the disks are connected; we need to retry the re-attach
3662
  # if the device becomes standalone, as this might happen if the one
3663
  # node disconnects and reconnects in a different mode before the
3664
  # other node reconnects; in this case, one or both of the nodes will
3665
  # decide it has wrong configuration and switch to standalone
3666

    
3667
  def _Attach():
3668
    all_connected = True
3669

    
3670
    for rd in bdevs:
3671
      stats = rd.GetProcStatus()
3672

    
3673
      if multimaster:
3674
        # In the multimaster case we have to wait explicitly until
3675
        # the resource is Connected and UpToDate/UpToDate, because
3676
        # we promote *both nodes* to primary directly afterwards.
3677
        # Being in resync is not enough, since there is a race during which we
3678
        # may promote a node with an Outdated disk to primary, effectively
3679
        # tearing down the connection.
3680
        all_connected = (all_connected and
3681
                         stats.is_connected and
3682
                         stats.is_disk_uptodate and
3683
                         stats.peer_disk_uptodate)
3684
      else:
3685
        all_connected = (all_connected and
3686
                         (stats.is_connected or stats.is_in_resync))
3687

    
3688
      if stats.is_standalone:
3689
        # peer had different config info and this node became
3690
        # standalone, even though this should not happen with the
3691
        # new staged way of changing disk configs
3692
        try:
3693
          rd.AttachNet(multimaster)
3694
        except errors.BlockDeviceError, err:
3695
          _Fail("Can't change network configuration: %s", err)
3696

    
3697
    if not all_connected:
3698
      raise utils.RetryAgain()
3699

    
3700
  try:
3701
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3702
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3703
  except utils.RetryTimeout:
3704
    _Fail("Timeout in disk reconnecting")
3705

    
3706
  if multimaster:
3707
    # change to primary mode
3708
    for rd in bdevs:
3709
      try:
3710
        rd.Open()
3711
      except errors.BlockDeviceError, err:
3712
        _Fail("Can't change to primary mode: %s", err)
3713

    
3714

    
3715
def DrbdWaitSync(nodes_ip, disks):
3716
  """Wait until DRBDs have synchronized.
3717

3718
  """
3719
  def _helper(rd):
3720
    stats = rd.GetProcStatus()
3721
    if not (stats.is_connected or stats.is_in_resync):
3722
      raise utils.RetryAgain()
3723
    return stats
3724

    
3725
  bdevs = _FindDisks(nodes_ip, disks)
3726

    
3727
  min_resync = 100
3728
  alldone = True
3729
  for rd in bdevs:
3730
    try:
3731
      # poll each second for 15 seconds
3732
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3733
    except utils.RetryTimeout:
3734
      stats = rd.GetProcStatus()
3735
      # last check
3736
      if not (stats.is_connected or stats.is_in_resync):
3737
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3738
    alldone = alldone and (not stats.is_in_resync)
3739
    if stats.sync_percent is not None:
3740
      min_resync = min(min_resync, stats.sync_percent)
3741

    
3742
  return (alldone, min_resync)
3743

    
3744

    
3745
def GetDrbdUsermodeHelper():
3746
  """Returns DRBD usermode helper currently configured.
3747

3748
  """
3749
  try:
3750
    return bdev.BaseDRBD.GetUsermodeHelper()
3751
  except errors.BlockDeviceError, err:
3752
    _Fail(str(err))
3753

    
3754

    
3755
def PowercycleNode(hypervisor_type):
3756
  """Hard-powercycle the node.
3757

3758
  Because we need to return first, and schedule the powercycle in the
3759
  background, we won't be able to report failures nicely.
3760

3761
  """
3762
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3763
  try:
3764
    pid = os.fork()
3765
  except OSError:
3766
    # if we can't fork, we'll pretend that we're in the child process
3767
    pid = 0
3768
  if pid > 0:
3769
    return "Reboot scheduled in 5 seconds"
3770
  # ensure the child is running on ram
3771
  try:
3772
    utils.Mlockall()
3773
  except Exception: # pylint: disable=W0703
3774
    pass
3775
  time.sleep(5)
3776
  hyper.PowercycleNode()
3777

    
3778

    
3779
def _VerifyRestrictedCmdName(cmd):
3780
  """Verifies a restricted command name.
3781

3782
  @type cmd: string
3783
  @param cmd: Command name
3784
  @rtype: tuple; (boolean, string or None)
3785
  @return: The tuple's first element is the status; if C{False}, the second
3786
    element is an error message string, otherwise it's C{None}
3787

3788
  """
3789
  if not cmd.strip():
3790
    return (False, "Missing command name")
3791

    
3792
  if os.path.basename(cmd) != cmd:
3793
    return (False, "Invalid command name")
3794

    
3795
  if not constants.EXT_PLUGIN_MASK.match(cmd):
3796
    return (False, "Command name contains forbidden characters")
3797

    
3798
  return (True, None)
3799

    
3800

    
3801
def _CommonRestrictedCmdCheck(path, owner):
3802
  """Common checks for restricted command file system directories and files.
3803

3804
  @type path: string
3805
  @param path: Path to check
3806
  @param owner: C{None} or tuple containing UID and GID
3807
  @rtype: tuple; (boolean, string or C{os.stat} result)
3808
  @return: The tuple's first element is the status; if C{False}, the second
3809
    element is an error message string, otherwise it's the result of C{os.stat}
3810

3811
  """
3812
  if owner is None:
3813
    # Default to root as owner
3814
    owner = (0, 0)
3815

    
3816
  try:
3817
    st = os.stat(path)
3818
  except EnvironmentError, err:
3819
    return (False, "Can't stat(2) '%s': %s" % (path, err))
3820

    
3821
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
3822
    return (False, "Permissions on '%s' are too permissive" % path)
3823

    
3824
  if (st.st_uid, st.st_gid) != owner:
3825
    (owner_uid, owner_gid) = owner
3826
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
3827

    
3828
  return (True, st)
3829

    
3830

    
3831
def _VerifyRestrictedCmdDirectory(path, _owner=None):
3832
  """Verifies restricted command directory.
3833

3834
  @type path: string
3835
  @param path: Path to check
3836
  @rtype: tuple; (boolean, string or None)
3837
  @return: The tuple's first element is the status; if C{False}, the second
3838
    element is an error message string, otherwise it's C{None}
3839

3840
  """
3841
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
3842

    
3843
  if not status:
3844
    return (False, value)
3845

    
3846
  if not stat.S_ISDIR(value.st_mode):
3847
    return (False, "Path '%s' is not a directory" % path)
3848

    
3849
  return (True, None)
3850

    
3851

    
3852
def _VerifyRestrictedCmd(path, cmd, _owner=None):
3853
  """Verifies a whole restricted command and returns its executable filename.
3854

3855
  @type path: string
3856
  @param path: Directory containing restricted commands
3857
  @type cmd: string
3858
  @param cmd: Command name
3859
  @rtype: tuple; (boolean, string)
3860
  @return: The tuple's first element is the status; if C{False}, the second
3861
    element is an error message string, otherwise the second element is the
3862
    absolute path to the executable
3863

3864
  """
3865
  executable = utils.PathJoin(path, cmd)
3866

    
3867
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
3868

    
3869
  if not status:
3870
    return (False, msg)
3871

    
3872
  if not utils.IsExecutable(executable):
3873
    return (False, "access(2) thinks '%s' can't be executed" % executable)
3874

    
3875
  return (True, executable)
3876

    
3877

    
3878
def _PrepareRestrictedCmd(path, cmd,
3879
                          _verify_dir=_VerifyRestrictedCmdDirectory,
3880
                          _verify_name=_VerifyRestrictedCmdName,
3881
                          _verify_cmd=_VerifyRestrictedCmd):
3882
  """Performs a number of tests on a restricted command.
3883

3884
  @type path: string
3885
  @param path: Directory containing restricted commands
3886
  @type cmd: string
3887
  @param cmd: Command name
3888
  @return: Same as L{_VerifyRestrictedCmd}
3889

3890
  """
3891
  # Verify the directory first
3892
  (status, msg) = _verify_dir(path)
3893
  if status:
3894
    # Check command if everything was alright
3895
    (status, msg) = _verify_name(cmd)
3896

    
3897
  if not status:
3898
    return (False, msg)
3899

    
3900
  # Check actual executable
3901
  return _verify_cmd(path, cmd)
3902

    
3903

    
3904
def RunRestrictedCmd(cmd,
3905
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
3906
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
3907
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
3908
                     _sleep_fn=time.sleep,
3909
                     _prepare_fn=_PrepareRestrictedCmd,
3910
                     _runcmd_fn=utils.RunCmd,
3911
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
3912
  """Executes a restricted command after performing strict tests.
3913

3914
  @type cmd: string
3915
  @param cmd: Command name
3916
  @rtype: string
3917
  @return: Command output
3918
  @raise RPCFail: In case of an error
3919

3920
  """
3921
  logging.info("Preparing to run restricted command '%s'", cmd)
3922

    
3923
  if not _enabled:
3924
    _Fail("Restricted commands disabled at configure time")
3925

    
3926
  lock = None
3927
  try:
3928
    cmdresult = None
3929
    try:
3930
      lock = utils.FileLock.Open(_lock_file)
3931
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
3932

    
3933
      (status, value) = _prepare_fn(_path, cmd)
3934

    
3935
      if status:
3936
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
3937
                               postfork_fn=lambda _: lock.Unlock())
3938
      else:
3939
        logging.error(value)
3940
    except Exception: # pylint: disable=W0703
3941
      # Keep original error in log
3942
      logging.exception("Caught exception")
3943

    
3944
    if cmdresult is None:
3945
      logging.info("Sleeping for %0.1f seconds before returning",
3946
                   _RCMD_INVALID_DELAY)
3947
      _sleep_fn(_RCMD_INVALID_DELAY)
3948

    
3949
      # Do not include original error message in returned error
3950
      _Fail("Executing command '%s' failed" % cmd)
3951
    elif cmdresult.failed or cmdresult.fail_reason:
3952
      _Fail("Restricted command '%s' failed: %s; output: %s",
3953
            cmd, cmdresult.fail_reason, cmdresult.output)
3954
    else:
3955
      return cmdresult.output
3956
  finally:
3957
    if lock is not None:
3958
      # Release lock at last
3959
      lock.Close()
3960
      lock = None
3961

    
3962

    
3963
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
3964
  """Creates or removes the watcher pause file.
3965

3966
  @type until: None or number
3967
  @param until: Unix timestamp saying until when the watcher shouldn't run
3968

3969
  """
3970
  if until is None:
3971
    logging.info("Received request to no longer pause watcher")
3972
    utils.RemoveFile(_filename)
3973
  else:
3974
    logging.info("Received request to pause watcher until %s", until)
3975

    
3976
    if not ht.TNumber(until):
3977
      _Fail("Duration must be numeric")
3978

    
3979
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
3980

    
3981

    
3982
class HooksRunner(object):
3983
  """Hook runner.
3984

3985
  This class is instantiated on the node side (ganeti-noded) and not
3986
  on the master side.
3987

3988
  """
3989
  def __init__(self, hooks_base_dir=None):
3990
    """Constructor for hooks runner.
3991

3992
    @type hooks_base_dir: str or None
3993
    @param hooks_base_dir: if not None, this overrides the
3994
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
3995

3996
    """
3997
    if hooks_base_dir is None:
3998
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
3999
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
4000
    # constant
4001
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4002

    
4003
  def RunLocalHooks(self, node_list, hpath, phase, env):
4004
    """Check that the hooks will be run only locally and then run them.
4005

4006
    """
4007
    assert len(node_list) == 1
4008
    node = node_list[0]
4009
    _, myself = ssconf.GetMasterAndMyself()
4010
    assert node == myself
4011

    
4012
    results = self.RunHooks(hpath, phase, env)
4013

    
4014
    # Return values in the form expected by HooksMaster
4015
    return {node: (None, False, results)}
4016

    
4017
  def RunHooks(self, hpath, phase, env):
4018
    """Run the scripts in the hooks directory.
4019

4020
    @type hpath: str
4021
    @param hpath: the path to the hooks directory which
4022
        holds the scripts
4023
    @type phase: str
4024
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
4025
        L{constants.HOOKS_PHASE_POST}
4026
    @type env: dict
4027
    @param env: dictionary with the environment for the hook
4028
    @rtype: list
4029
    @return: list of 3-element tuples:
4030
      - script path
4031
      - script result, either L{constants.HKR_SUCCESS} or
4032
        L{constants.HKR_FAIL}
4033
      - output of the script
4034

4035
    @raise errors.ProgrammerError: for invalid input
4036
        parameters
4037

4038
    """
4039
    if phase == constants.HOOKS_PHASE_PRE:
4040
      suffix = "pre"
4041
    elif phase == constants.HOOKS_PHASE_POST:
4042
      suffix = "post"
4043
    else:
4044
      _Fail("Unknown hooks phase '%s'", phase)
4045

    
4046
    subdir = "%s-%s.d" % (hpath, suffix)
4047
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4048

    
4049
    results = []
4050

    
4051
    if not os.path.isdir(dir_name):
4052
      # for non-existing/non-dirs, we simply exit instead of logging a
4053
      # warning at every operation
4054
      return results
4055

    
4056
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4057

    
4058
    for (relname, relstatus, runresult) in runparts_results:
4059
      if relstatus == constants.RUNPARTS_SKIP:
4060
        rrval = constants.HKR_SKIP
4061
        output = ""
4062
      elif relstatus == constants.RUNPARTS_ERR:
4063
        rrval = constants.HKR_FAIL
4064
        output = "Hook script execution error: %s" % runresult
4065
      elif relstatus == constants.RUNPARTS_RUN:
4066
        if runresult.failed:
4067
          rrval = constants.HKR_FAIL
4068
        else:
4069
          rrval = constants.HKR_SUCCESS
4070
        output = utils.SafeEncode(runresult.output.strip())
4071
      results.append(("%s/%s" % (subdir, relname), rrval, output))
4072

    
4073
    return results
4074

    
4075

    
4076
class IAllocatorRunner(object):
4077
  """IAllocator runner.
4078

4079
  This class is instantiated on the node side (ganeti-noded) and not on
4080
  the master side.
4081

4082
  """
4083
  @staticmethod
4084
  def Run(name, idata):
4085
    """Run an iallocator script.
4086

4087
    @type name: str
4088
    @param name: the iallocator script name
4089
    @type idata: str
4090
    @param idata: the allocator input data
4091

4092
    @rtype: tuple
4093
    @return: two element tuple of:
4094
       - status
4095
       - either error message or stdout of allocator (for success)
4096

4097
    """
4098
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4099
                                  os.path.isfile)
4100
    if alloc_script is None:
4101
      _Fail("iallocator module '%s' not found in the search path", name)
4102

    
4103
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4104
    try:
4105
      os.write(fd, idata)
4106
      os.close(fd)
4107
      result = utils.RunCmd([alloc_script, fin_name])
4108
      if result.failed:
4109
        _Fail("iallocator module '%s' failed: %s, output '%s'",
4110
              name, result.fail_reason, result.output)
4111
    finally:
4112
      os.unlink(fin_name)
4113

    
4114
    return result.stdout
4115

    
4116

    
4117
class DevCacheManager(object):
4118
  """Simple class for managing a cache of block device information.
4119

4120
  """
4121
  _DEV_PREFIX = "/dev/"
4122
  _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4123

    
4124
  @classmethod
4125
  def _ConvertPath(cls, dev_path):
4126
    """Converts a /dev/name path to the cache file name.
4127

4128
    This replaces slashes with underscores and strips the /dev
4129
    prefix. It then returns the full path to the cache file.
4130

4131
    @type dev_path: str
4132
    @param dev_path: the C{/dev/} path name
4133
    @rtype: str
4134
    @return: the converted path name
4135

4136
    """
4137
    if dev_path.startswith(cls._DEV_PREFIX):
4138
      dev_path = dev_path[len(cls._DEV_PREFIX):]
4139
    dev_path = dev_path.replace("/", "_")
4140
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4141
    return fpath
4142

    
4143
  @classmethod
4144
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4145
    """Updates the cache information for a given device.
4146

4147
    @type dev_path: str
4148
    @param dev_path: the pathname of the device
4149
    @type owner: str
4150
    @param owner: the owner (instance name) of the device
4151
    @type on_primary: bool
4152
    @param on_primary: whether this is the primary
4153
        node nor not
4154
    @type iv_name: str
4155
    @param iv_name: the instance-visible name of the
4156
        device, as in objects.Disk.iv_name
4157

4158
    @rtype: None
4159

4160
    """
4161
    if dev_path is None:
4162
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
4163
      return
4164
    fpath = cls._ConvertPath(dev_path)
4165
    if on_primary:
4166
      state = "primary"
4167
    else:
4168
      state = "secondary"
4169
    if iv_name is None:
4170
      iv_name = "not_visible"
4171
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4172
    try:
4173
      utils.WriteFile(fpath, data=fdata)
4174
    except EnvironmentError, err:
4175
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4176

    
4177
  @classmethod
4178
  def RemoveCache(cls, dev_path):
4179
    """Remove data for a dev_path.
4180

4181
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
4182
    path name and logging.
4183

4184
    @type dev_path: str
4185
    @param dev_path: the pathname of the device
4186

4187
    @rtype: None
4188

4189
    """
4190
    if dev_path is None:
4191
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
4192
      return
4193
    fpath = cls._ConvertPath(dev_path)
4194
    try:
4195
      utils.RemoveFile(fpath)
4196
    except EnvironmentError, err:
4197
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)