Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 748c9884

History | View | Annotate | Download (106.6 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 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

    
64

    
65
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
66
_ALLOWED_CLEAN_DIRS = frozenset([
67
  constants.DATA_DIR,
68
  constants.JOB_QUEUE_ARCHIVE_DIR,
69
  constants.QUEUE_DIR,
70
  constants.CRYPTO_KEYS_DIR,
71
  ])
72
_MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
73
_X509_KEY_FILE = "key"
74
_X509_CERT_FILE = "cert"
75
_IES_STATUS_FILE = "status"
76
_IES_PID_FILE = "pid"
77
_IES_CA_FILE = "ca"
78

    
79
#: Valid LVS output line regex
80
_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
81

    
82

    
83
class RPCFail(Exception):
84
  """Class denoting RPC failure.
85

86
  Its argument is the error message.
87

88
  """
89

    
90

    
91
def _Fail(msg, *args, **kwargs):
92
  """Log an error and the raise an RPCFail exception.
93

94
  This exception is then handled specially in the ganeti daemon and
95
  turned into a 'failed' return type. As such, this function is a
96
  useful shortcut for logging the error and returning it to the master
97
  daemon.
98

99
  @type msg: string
100
  @param msg: the text of the exception
101
  @raise RPCFail
102

103
  """
104
  if args:
105
    msg = msg % args
106
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
107
    if "exc" in kwargs and kwargs["exc"]:
108
      logging.exception(msg)
109
    else:
110
      logging.error(msg)
111
  raise RPCFail(msg)
112

    
113

    
114
def _GetConfig():
115
  """Simple wrapper to return a SimpleStore.
116

117
  @rtype: L{ssconf.SimpleStore}
118
  @return: a SimpleStore instance
119

120
  """
121
  return ssconf.SimpleStore()
122

    
123

    
124
def _GetSshRunner(cluster_name):
125
  """Simple wrapper to return an SshRunner.
126

127
  @type cluster_name: str
128
  @param cluster_name: the cluster name, which is needed
129
      by the SshRunner constructor
130
  @rtype: L{ssh.SshRunner}
131
  @return: an SshRunner instance
132

133
  """
134
  return ssh.SshRunner(cluster_name)
135

    
136

    
137
def _Decompress(data):
138
  """Unpacks data compressed by the RPC client.
139

140
  @type data: list or tuple
141
  @param data: Data sent by RPC client
142
  @rtype: str
143
  @return: Decompressed data
144

145
  """
146
  assert isinstance(data, (list, tuple))
147
  assert len(data) == 2
148
  (encoding, content) = data
149
  if encoding == constants.RPC_ENCODING_NONE:
150
    return content
151
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
152
    return zlib.decompress(base64.b64decode(content))
153
  else:
154
    raise AssertionError("Unknown data encoding")
155

    
156

    
157
def _CleanDirectory(path, exclude=None):
158
  """Removes all regular files in a directory.
159

160
  @type path: str
161
  @param path: the directory to clean
162
  @type exclude: list
163
  @param exclude: list of files to be excluded, defaults
164
      to the empty list
165

166
  """
167
  if path not in _ALLOWED_CLEAN_DIRS:
168
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
169
          path)
170

    
171
  if not os.path.isdir(path):
172
    return
173
  if exclude is None:
174
    exclude = []
175
  else:
176
    # Normalize excluded paths
177
    exclude = [os.path.normpath(i) for i in exclude]
178

    
179
  for rel_name in utils.ListVisibleFiles(path):
180
    full_name = utils.PathJoin(path, rel_name)
181
    if full_name in exclude:
182
      continue
183
    if os.path.isfile(full_name) and not os.path.islink(full_name):
184
      utils.RemoveFile(full_name)
185

    
186

    
187
def _BuildUploadFileList():
188
  """Build the list of allowed upload files.
189

190
  This is abstracted so that it's built only once at module import time.
191

192
  """
193
  allowed_files = set([
194
    constants.CLUSTER_CONF_FILE,
195
    constants.ETC_HOSTS,
196
    constants.SSH_KNOWN_HOSTS_FILE,
197
    constants.VNC_PASSWORD_FILE,
198
    constants.RAPI_CERT_FILE,
199
    constants.RAPI_USERS_FILE,
200
    constants.CONFD_HMAC_KEY,
201
    constants.CLUSTER_DOMAIN_SECRET_FILE,
202
    ])
203

    
204
  for hv_name in constants.HYPER_TYPES:
205
    hv_class = hypervisor.GetHypervisorClass(hv_name)
206
    allowed_files.update(hv_class.GetAncillaryFiles())
207

    
208
  return frozenset(allowed_files)
209

    
210

    
211
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
212

    
213

    
214
def JobQueuePurge():
215
  """Removes job queue files and archived jobs.
216

217
  @rtype: tuple
218
  @return: True, None
219

220
  """
221
  _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
222
  _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
223

    
224

    
225
def GetMasterInfo():
226
  """Returns master information.
227

228
  This is an utility function to compute master information, either
229
  for consumption here or from the node daemon.
230

231
  @rtype: tuple
232
  @return: master_netdev, master_ip, master_name, primary_ip_family
233
  @raise RPCFail: in case of errors
234

235
  """
236
  try:
237
    cfg = _GetConfig()
238
    master_netdev = cfg.GetMasterNetdev()
239
    master_ip = cfg.GetMasterIP()
240
    master_node = cfg.GetMasterNode()
241
    primary_ip_family = cfg.GetPrimaryIPFamily()
242
  except errors.ConfigurationError, err:
243
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
244
  return (master_netdev, master_ip, master_node, primary_ip_family)
245

    
246

    
247
def StartMaster(start_daemons, no_voting):
248
  """Activate local node as master node.
249

250
  The function will either try activate the IP address of the master
251
  (unless someone else has it) or also start the master daemons, based
252
  on the start_daemons parameter.
253

254
  @type start_daemons: boolean
255
  @param start_daemons: whether to start the master daemons
256
      (ganeti-masterd and ganeti-rapi), or (if false) activate the
257
      master ip
258
  @type no_voting: boolean
259
  @param no_voting: whether to start ganeti-masterd without a node vote
260
      (if start_daemons is True), but still non-interactively
261
  @rtype: None
262

263
  """
264
  # GetMasterInfo will raise an exception if not able to return data
265
  master_netdev, master_ip, _, family = GetMasterInfo()
266

    
267
  err_msgs = []
268
  # either start the master and rapi daemons
269
  if start_daemons:
270
    if no_voting:
271
      masterd_args = "--no-voting --yes-do-it"
272
    else:
273
      masterd_args = ""
274

    
275
    env = {
276
      "EXTRA_MASTERD_ARGS": masterd_args,
277
      }
278

    
279
    result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
280
    if result.failed:
281
      msg = "Can't start Ganeti master: %s" % result.output
282
      logging.error(msg)
283
      err_msgs.append(msg)
284
  # or activate the IP
285
  else:
286
    if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
287
      if netutils.IPAddress.Own(master_ip):
288
        # we already have the ip:
289
        logging.debug("Master IP already configured, doing nothing")
290
      else:
291
        msg = "Someone else has the master ip, not activating"
292
        logging.error(msg)
293
        err_msgs.append(msg)
294
    else:
295
      ipcls = netutils.IP4Address
296
      if family == netutils.IP6Address.family:
297
        ipcls = netutils.IP6Address
298

    
299
      result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
300
                             "%s/%d" % (master_ip, ipcls.iplen),
301
                             "dev", master_netdev, "label",
302
                             "%s:0" % master_netdev])
303
      if result.failed:
304
        msg = "Can't activate master IP: %s" % result.output
305
        logging.error(msg)
306
        err_msgs.append(msg)
307

    
308
      # we ignore the exit code of the following cmds
309
      if ipcls == netutils.IP4Address:
310
        utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev, "-s",
311
                      master_ip, master_ip])
312
      elif ipcls == netutils.IP6Address:
313
        try:
314
          utils.RunCmd(["ndisc6", "-q", "-r 3", master_ip, master_netdev])
315
        except errors.OpExecError:
316
          # TODO: Better error reporting
317
          logging.warning("Can't execute ndisc6, please install if missing")
318

    
319
  if err_msgs:
320
    _Fail("; ".join(err_msgs))
321

    
322

    
323
def StopMaster(stop_daemons):
324
  """Deactivate this node as master.
325

326
  The function will always try to deactivate the IP address of the
327
  master. It will also stop the master daemons depending on the
328
  stop_daemons parameter.
329

330
  @type stop_daemons: boolean
331
  @param stop_daemons: whether to also stop the master daemons
332
      (ganeti-masterd and ganeti-rapi)
333
  @rtype: None
334

335
  """
336
  # TODO: log and report back to the caller the error failures; we
337
  # need to decide in which case we fail the RPC for this
338

    
339
  # GetMasterInfo will raise an exception if not able to return data
340
  master_netdev, master_ip, _, family = GetMasterInfo()
341

    
342
  ipcls = netutils.IP4Address
343
  if family == netutils.IP6Address.family:
344
    ipcls = netutils.IP6Address
345

    
346
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
347
                         "%s/%d" % (master_ip, ipcls.iplen),
348
                         "dev", master_netdev])
349
  if result.failed:
350
    logging.error("Can't remove the master IP, error: %s", result.output)
351
    # but otherwise ignore the failure
352

    
353
  if stop_daemons:
354
    result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
355
    if result.failed:
356
      logging.error("Could not stop Ganeti master, command %s had exitcode %s"
357
                    " and error %s",
358
                    result.cmd, result.exit_code, result.output)
359

    
360

    
361
def EtcHostsModify(mode, host, ip):
362
  """Modify a host entry in /etc/hosts.
363

364
  @param mode: The mode to operate. Either add or remove entry
365
  @param host: The host to operate on
366
  @param ip: The ip associated with the entry
367

368
  """
369
  if mode == constants.ETC_HOSTS_ADD:
370
    if not ip:
371
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
372
              " present")
373
    utils.AddHostToEtcHosts(host, ip)
374
  elif mode == constants.ETC_HOSTS_REMOVE:
375
    if ip:
376
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
377
              " parameter is present")
378
    utils.RemoveHostFromEtcHosts(host)
379
  else:
380
    RPCFail("Mode not supported")
381

    
382

    
383
def LeaveCluster(modify_ssh_setup):
384
  """Cleans up and remove the current node.
385

386
  This function cleans up and prepares the current node to be removed
387
  from the cluster.
388

389
  If processing is successful, then it raises an
390
  L{errors.QuitGanetiException} which is used as a special case to
391
  shutdown the node daemon.
392

393
  @param modify_ssh_setup: boolean
394

395
  """
396
  _CleanDirectory(constants.DATA_DIR)
397
  _CleanDirectory(constants.CRYPTO_KEYS_DIR)
398
  JobQueuePurge()
399

    
400
  if modify_ssh_setup:
401
    try:
402
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
403

    
404
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
405

    
406
      utils.RemoveFile(priv_key)
407
      utils.RemoveFile(pub_key)
408
    except errors.OpExecError:
409
      logging.exception("Error while processing ssh files")
410

    
411
  try:
412
    utils.RemoveFile(constants.CONFD_HMAC_KEY)
413
    utils.RemoveFile(constants.RAPI_CERT_FILE)
414
    utils.RemoveFile(constants.NODED_CERT_FILE)
415
  except: # pylint: disable=W0702
416
    logging.exception("Error while removing cluster secrets")
417

    
418
  result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
419
  if result.failed:
420
    logging.error("Command %s failed with exitcode %s and error %s",
421
                  result.cmd, result.exit_code, result.output)
422

    
423
  # Raise a custom exception (handled in ganeti-noded)
424
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
425

    
426

    
427
def GetNodeInfo(vgname, hypervisor_type):
428
  """Gives back a hash with different information about the node.
429

430
  @type vgname: C{string}
431
  @param vgname: the name of the volume group to ask for disk space information
432
  @type hypervisor_type: C{str}
433
  @param hypervisor_type: the name of the hypervisor to ask for
434
      memory information
435
  @rtype: C{dict}
436
  @return: dictionary with the following keys:
437
      - vg_size is the size of the configured volume group in MiB
438
      - vg_free is the free size of the volume group in MiB
439
      - memory_dom0 is the memory allocated for domain0 in MiB
440
      - memory_free is the currently available (free) ram in MiB
441
      - memory_total is the total number of ram in MiB
442

443
  """
444
  outputarray = {}
445

    
446
  if vgname is not None:
447
    vginfo = bdev.LogicalVolume.GetVGInfo([vgname])
448
    vg_free = vg_size = None
449
    if vginfo:
450
      vg_free = int(round(vginfo[0][0], 0))
451
      vg_size = int(round(vginfo[0][1], 0))
452
    outputarray["vg_size"] = vg_size
453
    outputarray["vg_free"] = vg_free
454

    
455
  if hypervisor_type is not None:
456
    hyper = hypervisor.GetHypervisor(hypervisor_type)
457
    hyp_info = hyper.GetNodeInfo()
458
    if hyp_info is not None:
459
      outputarray.update(hyp_info)
460

    
461
  outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
462

    
463
  return outputarray
464

    
465

    
466
def VerifyNode(what, cluster_name):
467
  """Verify the status of the local node.
468

469
  Based on the input L{what} parameter, various checks are done on the
470
  local node.
471

472
  If the I{filelist} key is present, this list of
473
  files is checksummed and the file/checksum pairs are returned.
474

475
  If the I{nodelist} key is present, we check that we have
476
  connectivity via ssh with the target nodes (and check the hostname
477
  report).
478

479
  If the I{node-net-test} key is present, we check that we have
480
  connectivity to the given nodes via both primary IP and, if
481
  applicable, secondary IPs.
482

483
  @type what: C{dict}
484
  @param what: a dictionary of things to check:
485
      - filelist: list of files for which to compute checksums
486
      - nodelist: list of nodes we should check ssh communication with
487
      - node-net-test: list of nodes we should check node daemon port
488
        connectivity with
489
      - hypervisor: list with hypervisors to run the verify for
490
  @rtype: dict
491
  @return: a dictionary with the same keys as the input dict, and
492
      values representing the result of the checks
493

494
  """
495
  result = {}
496
  my_name = netutils.Hostname.GetSysName()
497
  port = netutils.GetDaemonPort(constants.NODED)
498
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
499

    
500
  if constants.NV_HYPERVISOR in what and vm_capable:
501
    result[constants.NV_HYPERVISOR] = tmp = {}
502
    for hv_name in what[constants.NV_HYPERVISOR]:
503
      try:
504
        val = hypervisor.GetHypervisor(hv_name).Verify()
505
      except errors.HypervisorError, err:
506
        val = "Error while checking hypervisor: %s" % str(err)
507
      tmp[hv_name] = val
508

    
509
  if constants.NV_HVPARAMS in what and vm_capable:
510
    result[constants.NV_HVPARAMS] = tmp = []
511
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
512
      try:
513
        logging.info("Validating hv %s, %s", hv_name, hvparms)
514
        hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
515
      except errors.HypervisorError, err:
516
        tmp.append((source, hv_name, str(err)))
517

    
518
  if constants.NV_FILELIST in what:
519
    result[constants.NV_FILELIST] = utils.FingerprintFiles(
520
      what[constants.NV_FILELIST])
521

    
522
  if constants.NV_NODELIST in what:
523
    result[constants.NV_NODELIST] = tmp = {}
524
    random.shuffle(what[constants.NV_NODELIST])
525
    for node in what[constants.NV_NODELIST]:
526
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
527
      if not success:
528
        tmp[node] = message
529

    
530
  if constants.NV_NODENETTEST in what:
531
    result[constants.NV_NODENETTEST] = tmp = {}
532
    my_pip = my_sip = None
533
    for name, pip, sip in what[constants.NV_NODENETTEST]:
534
      if name == my_name:
535
        my_pip = pip
536
        my_sip = sip
537
        break
538
    if not my_pip:
539
      tmp[my_name] = ("Can't find my own primary/secondary IP"
540
                      " in the node list")
541
    else:
542
      for name, pip, sip in what[constants.NV_NODENETTEST]:
543
        fail = []
544
        if not netutils.TcpPing(pip, port, source=my_pip):
545
          fail.append("primary")
546
        if sip != pip:
547
          if not netutils.TcpPing(sip, port, source=my_sip):
548
            fail.append("secondary")
549
        if fail:
550
          tmp[name] = ("failure using the %s interface(s)" %
551
                       " and ".join(fail))
552

    
553
  if constants.NV_MASTERIP in what:
554
    # FIXME: add checks on incoming data structures (here and in the
555
    # rest of the function)
556
    master_name, master_ip = what[constants.NV_MASTERIP]
557
    if master_name == my_name:
558
      source = constants.IP4_ADDRESS_LOCALHOST
559
    else:
560
      source = None
561
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
562
                                                  source=source)
563

    
564
  if constants.NV_OOB_PATHS in what:
565
    result[constants.NV_OOB_PATHS] = tmp = []
566
    for path in what[constants.NV_OOB_PATHS]:
567
      try:
568
        st = os.stat(path)
569
      except OSError, err:
570
        tmp.append("error stating out of band helper: %s" % err)
571
      else:
572
        if stat.S_ISREG(st.st_mode):
573
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
574
            tmp.append(None)
575
          else:
576
            tmp.append("out of band helper %s is not executable" % path)
577
        else:
578
          tmp.append("out of band helper %s is not a file" % path)
579

    
580
  if constants.NV_LVLIST in what and vm_capable:
581
    try:
582
      val = GetVolumeList(utils.ListVolumeGroups().keys())
583
    except RPCFail, err:
584
      val = str(err)
585
    result[constants.NV_LVLIST] = val
586

    
587
  if constants.NV_INSTANCELIST in what and vm_capable:
588
    # GetInstanceList can fail
589
    try:
590
      val = GetInstanceList(what[constants.NV_INSTANCELIST])
591
    except RPCFail, err:
592
      val = str(err)
593
    result[constants.NV_INSTANCELIST] = val
594

    
595
  if constants.NV_VGLIST in what and vm_capable:
596
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
597

    
598
  if constants.NV_PVLIST in what and vm_capable:
599
    result[constants.NV_PVLIST] = \
600
      bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
601
                                   filter_allocatable=False)
602

    
603
  if constants.NV_VERSION in what:
604
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
605
                                    constants.RELEASE_VERSION)
606

    
607
  if constants.NV_HVINFO in what and vm_capable:
608
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
609
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
610

    
611
  if constants.NV_DRBDLIST in what and vm_capable:
612
    try:
613
      used_minors = bdev.DRBD8.GetUsedDevs().keys()
614
    except errors.BlockDeviceError, err:
615
      logging.warning("Can't get used minors list", exc_info=True)
616
      used_minors = str(err)
617
    result[constants.NV_DRBDLIST] = used_minors
618

    
619
  if constants.NV_DRBDHELPER in what and vm_capable:
620
    status = True
621
    try:
622
      payload = bdev.BaseDRBD.GetUsermodeHelper()
623
    except errors.BlockDeviceError, err:
624
      logging.error("Can't get DRBD usermode helper: %s", str(err))
625
      status = False
626
      payload = str(err)
627
    result[constants.NV_DRBDHELPER] = (status, payload)
628

    
629
  if constants.NV_NODESETUP in what:
630
    result[constants.NV_NODESETUP] = tmpr = []
631
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
632
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
633
                  " under /sys, missing required directories /sys/block"
634
                  " and /sys/class/net")
635
    if (not os.path.isdir("/proc/sys") or
636
        not os.path.isfile("/proc/sysrq-trigger")):
637
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
638
                  " under /proc, missing required directory /proc/sys and"
639
                  " the file /proc/sysrq-trigger")
640

    
641
  if constants.NV_TIME in what:
642
    result[constants.NV_TIME] = utils.SplitTime(time.time())
643

    
644
  if constants.NV_OSLIST in what and vm_capable:
645
    result[constants.NV_OSLIST] = DiagnoseOS()
646

    
647
  if constants.NV_BRIDGES in what and vm_capable:
648
    result[constants.NV_BRIDGES] = [bridge
649
                                    for bridge in what[constants.NV_BRIDGES]
650
                                    if not utils.BridgeExists(bridge)]
651
  return result
652

    
653

    
654
def GetBlockDevSizes(devices):
655
  """Return the size of the given block devices
656

657
  @type devices: list
658
  @param devices: list of block device nodes to query
659
  @rtype: dict
660
  @return:
661
    dictionary of all block devices under /dev (key). The value is their
662
    size in MiB.
663

664
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
665

666
  """
667
  DEV_PREFIX = "/dev/"
668
  blockdevs = {}
669

    
670
  for devpath in devices:
671
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
672
      continue
673

    
674
    try:
675
      st = os.stat(devpath)
676
    except EnvironmentError, err:
677
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
678
      continue
679

    
680
    if stat.S_ISBLK(st.st_mode):
681
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
682
      if result.failed:
683
        # We don't want to fail, just do not list this device as available
684
        logging.warning("Cannot get size for block device %s", devpath)
685
        continue
686

    
687
      size = int(result.stdout) / (1024 * 1024)
688
      blockdevs[devpath] = size
689
  return blockdevs
690

    
691

    
692
def GetVolumeList(vg_names):
693
  """Compute list of logical volumes and their size.
694

695
  @type vg_names: list
696
  @param vg_names: the volume groups whose LVs we should list, or
697
      empty for all volume groups
698
  @rtype: dict
699
  @return:
700
      dictionary of all partions (key) with value being a tuple of
701
      their size (in MiB), inactive and online status::
702

703
        {'xenvg/test1': ('20.06', True, True)}
704

705
      in case of errors, a string is returned with the error
706
      details.
707

708
  """
709
  lvs = {}
710
  sep = "|"
711
  if not vg_names:
712
    vg_names = []
713
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
714
                         "--separator=%s" % sep,
715
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
716
  if result.failed:
717
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
718

    
719
  for line in result.stdout.splitlines():
720
    line = line.strip()
721
    match = _LVSLINE_REGEX.match(line)
722
    if not match:
723
      logging.error("Invalid line returned from lvs output: '%s'", line)
724
      continue
725
    vg_name, name, size, attr = match.groups()
726
    inactive = attr[4] == "-"
727
    online = attr[5] == "o"
728
    virtual = attr[0] == "v"
729
    if virtual:
730
      # we don't want to report such volumes as existing, since they
731
      # don't really hold data
732
      continue
733
    lvs[vg_name + "/" + name] = (size, inactive, online)
734

    
735
  return lvs
736

    
737

    
738
def ListVolumeGroups():
739
  """List the volume groups and their size.
740

741
  @rtype: dict
742
  @return: dictionary with keys volume name and values the
743
      size of the volume
744

745
  """
746
  return utils.ListVolumeGroups()
747

    
748

    
749
def NodeVolumes():
750
  """List all volumes on this node.
751

752
  @rtype: list
753
  @return:
754
    A list of dictionaries, each having four keys:
755
      - name: the logical volume name,
756
      - size: the size of the logical volume
757
      - dev: the physical device on which the LV lives
758
      - vg: the volume group to which it belongs
759

760
    In case of errors, we return an empty list and log the
761
    error.
762

763
    Note that since a logical volume can live on multiple physical
764
    volumes, the resulting list might include a logical volume
765
    multiple times.
766

767
  """
768
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
769
                         "--separator=|",
770
                         "--options=lv_name,lv_size,devices,vg_name"])
771
  if result.failed:
772
    _Fail("Failed to list logical volumes, lvs output: %s",
773
          result.output)
774

    
775
  def parse_dev(dev):
776
    return dev.split("(")[0]
777

    
778
  def handle_dev(dev):
779
    return [parse_dev(x) for x in dev.split(",")]
780

    
781
  def map_line(line):
782
    line = [v.strip() for v in line]
783
    return [{"name": line[0], "size": line[1],
784
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
785

    
786
  all_devs = []
787
  for line in result.stdout.splitlines():
788
    if line.count("|") >= 3:
789
      all_devs.extend(map_line(line.split("|")))
790
    else:
791
      logging.warning("Strange line in the output from lvs: '%s'", line)
792
  return all_devs
793

    
794

    
795
def BridgesExist(bridges_list):
796
  """Check if a list of bridges exist on the current node.
797

798
  @rtype: boolean
799
  @return: C{True} if all of them exist, C{False} otherwise
800

801
  """
802
  missing = []
803
  for bridge in bridges_list:
804
    if not utils.BridgeExists(bridge):
805
      missing.append(bridge)
806

    
807
  if missing:
808
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
809

    
810

    
811
def GetInstanceList(hypervisor_list):
812
  """Provides a list of instances.
813

814
  @type hypervisor_list: list
815
  @param hypervisor_list: the list of hypervisors to query information
816

817
  @rtype: list
818
  @return: a list of all running instances on the current node
819
    - instance1.example.com
820
    - instance2.example.com
821

822
  """
823
  results = []
824
  for hname in hypervisor_list:
825
    try:
826
      names = hypervisor.GetHypervisor(hname).ListInstances()
827
      results.extend(names)
828
    except errors.HypervisorError, err:
829
      _Fail("Error enumerating instances (hypervisor %s): %s",
830
            hname, err, exc=True)
831

    
832
  return results
833

    
834

    
835
def GetInstanceInfo(instance, hname):
836
  """Gives back the information about an instance as a dictionary.
837

838
  @type instance: string
839
  @param instance: the instance name
840
  @type hname: string
841
  @param hname: the hypervisor type of the instance
842

843
  @rtype: dict
844
  @return: dictionary with the following keys:
845
      - memory: memory size of instance (int)
846
      - state: xen state of instance (string)
847
      - time: cpu time of instance (float)
848

849
  """
850
  output = {}
851

    
852
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
853
  if iinfo is not None:
854
    output["memory"] = iinfo[2]
855
    output["state"] = iinfo[4]
856
    output["time"] = iinfo[5]
857

    
858
  return output
859

    
860

    
861
def GetInstanceMigratable(instance):
862
  """Gives whether an instance can be migrated.
863

864
  @type instance: L{objects.Instance}
865
  @param instance: object representing the instance to be checked.
866

867
  @rtype: tuple
868
  @return: tuple of (result, description) where:
869
      - result: whether the instance can be migrated or not
870
      - description: a description of the issue, if relevant
871

872
  """
873
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
874
  iname = instance.name
875
  if iname not in hyper.ListInstances():
876
    _Fail("Instance %s is not running", iname)
877

    
878
  for idx in range(len(instance.disks)):
879
    link_name = _GetBlockDevSymlinkPath(iname, idx)
880
    if not os.path.islink(link_name):
881
      logging.warning("Instance %s is missing symlink %s for disk %d",
882
                      iname, link_name, idx)
883

    
884

    
885
def GetAllInstancesInfo(hypervisor_list):
886
  """Gather data about all instances.
887

888
  This is the equivalent of L{GetInstanceInfo}, except that it
889
  computes data for all instances at once, thus being faster if one
890
  needs data about more than one instance.
891

892
  @type hypervisor_list: list
893
  @param hypervisor_list: list of hypervisors to query for instance data
894

895
  @rtype: dict
896
  @return: dictionary of instance: data, with data having the following keys:
897
      - memory: memory size of instance (int)
898
      - state: xen state of instance (string)
899
      - time: cpu time of instance (float)
900
      - vcpus: the number of vcpus
901

902
  """
903
  output = {}
904

    
905
  for hname in hypervisor_list:
906
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
907
    if iinfo:
908
      for name, _, memory, vcpus, state, times in iinfo:
909
        value = {
910
          "memory": memory,
911
          "vcpus": vcpus,
912
          "state": state,
913
          "time": times,
914
          }
915
        if name in output:
916
          # we only check static parameters, like memory and vcpus,
917
          # and not state and time which can change between the
918
          # invocations of the different hypervisors
919
          for key in "memory", "vcpus":
920
            if value[key] != output[name][key]:
921
              _Fail("Instance %s is running twice"
922
                    " with different parameters", name)
923
        output[name] = value
924

    
925
  return output
926

    
927

    
928
def _InstanceLogName(kind, os_name, instance, component):
929
  """Compute the OS log filename for a given instance and operation.
930

931
  The instance name and os name are passed in as strings since not all
932
  operations have these as part of an instance object.
933

934
  @type kind: string
935
  @param kind: the operation type (e.g. add, import, etc.)
936
  @type os_name: string
937
  @param os_name: the os name
938
  @type instance: string
939
  @param instance: the name of the instance being imported/added/etc.
940
  @type component: string or None
941
  @param component: the name of the component of the instance being
942
      transferred
943

944
  """
945
  # TODO: Use tempfile.mkstemp to create unique filename
946
  if component:
947
    assert "/" not in component
948
    c_msg = "-%s" % component
949
  else:
950
    c_msg = ""
951
  base = ("%s-%s-%s%s-%s.log" %
952
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
953
  return utils.PathJoin(constants.LOG_OS_DIR, base)
954

    
955

    
956
def InstanceOsAdd(instance, reinstall, debug):
957
  """Add an OS to an instance.
958

959
  @type instance: L{objects.Instance}
960
  @param instance: Instance whose OS is to be installed
961
  @type reinstall: boolean
962
  @param reinstall: whether this is an instance reinstall
963
  @type debug: integer
964
  @param debug: debug level, passed to the OS scripts
965
  @rtype: None
966

967
  """
968
  inst_os = OSFromDisk(instance.os)
969

    
970
  create_env = OSEnvironment(instance, inst_os, debug)
971
  if reinstall:
972
    create_env["INSTANCE_REINSTALL"] = "1"
973

    
974
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
975

    
976
  result = utils.RunCmd([inst_os.create_script], env=create_env,
977
                        cwd=inst_os.path, output=logfile, reset_env=True)
978
  if result.failed:
979
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
980
                  " output: %s", result.cmd, result.fail_reason, logfile,
981
                  result.output)
982
    lines = [utils.SafeEncode(val)
983
             for val in utils.TailFile(logfile, lines=20)]
984
    _Fail("OS create script failed (%s), last lines in the"
985
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
986

    
987

    
988
def RunRenameInstance(instance, old_name, debug):
989
  """Run the OS rename script for an instance.
990

991
  @type instance: L{objects.Instance}
992
  @param instance: Instance whose OS is to be installed
993
  @type old_name: string
994
  @param old_name: previous instance name
995
  @type debug: integer
996
  @param debug: debug level, passed to the OS scripts
997
  @rtype: boolean
998
  @return: the success of the operation
999

1000
  """
1001
  inst_os = OSFromDisk(instance.os)
1002

    
1003
  rename_env = OSEnvironment(instance, inst_os, debug)
1004
  rename_env["OLD_INSTANCE_NAME"] = old_name
1005

    
1006
  logfile = _InstanceLogName("rename", instance.os,
1007
                             "%s-%s" % (old_name, instance.name), None)
1008

    
1009
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1010
                        cwd=inst_os.path, output=logfile, reset_env=True)
1011

    
1012
  if result.failed:
1013
    logging.error("os create command '%s' returned error: %s output: %s",
1014
                  result.cmd, result.fail_reason, result.output)
1015
    lines = [utils.SafeEncode(val)
1016
             for val in utils.TailFile(logfile, lines=20)]
1017
    _Fail("OS rename script failed (%s), last lines in the"
1018
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1019

    
1020

    
1021
def _GetBlockDevSymlinkPath(instance_name, idx):
1022
  return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1023
                        (instance_name, constants.DISK_SEPARATOR, idx))
1024

    
1025

    
1026
def _SymlinkBlockDev(instance_name, device_path, idx):
1027
  """Set up symlinks to a instance's block device.
1028

1029
  This is an auxiliary function run when an instance is start (on the primary
1030
  node) or when an instance is migrated (on the target node).
1031

1032

1033
  @param instance_name: the name of the target instance
1034
  @param device_path: path of the physical block device, on the node
1035
  @param idx: the disk index
1036
  @return: absolute path to the disk's symlink
1037

1038
  """
1039
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1040
  try:
1041
    os.symlink(device_path, link_name)
1042
  except OSError, err:
1043
    if err.errno == errno.EEXIST:
1044
      if (not os.path.islink(link_name) or
1045
          os.readlink(link_name) != device_path):
1046
        os.remove(link_name)
1047
        os.symlink(device_path, link_name)
1048
    else:
1049
      raise
1050

    
1051
  return link_name
1052

    
1053

    
1054
def _RemoveBlockDevLinks(instance_name, disks):
1055
  """Remove the block device symlinks belonging to the given instance.
1056

1057
  """
1058
  for idx, _ in enumerate(disks):
1059
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1060
    if os.path.islink(link_name):
1061
      try:
1062
        os.remove(link_name)
1063
      except OSError:
1064
        logging.exception("Can't remove symlink '%s'", link_name)
1065

    
1066

    
1067
def _GatherAndLinkBlockDevs(instance):
1068
  """Set up an instance's block device(s).
1069

1070
  This is run on the primary node at instance startup. The block
1071
  devices must be already assembled.
1072

1073
  @type instance: L{objects.Instance}
1074
  @param instance: the instance whose disks we shoul assemble
1075
  @rtype: list
1076
  @return: list of (disk_object, device_path)
1077

1078
  """
1079
  block_devices = []
1080
  for idx, disk in enumerate(instance.disks):
1081
    device = _RecursiveFindBD(disk)
1082
    if device is None:
1083
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1084
                                    str(disk))
1085
    device.Open()
1086
    try:
1087
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1088
    except OSError, e:
1089
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1090
                                    e.strerror)
1091

    
1092
    block_devices.append((disk, link_name))
1093

    
1094
  return block_devices
1095

    
1096

    
1097
def StartInstance(instance, startup_paused):
1098
  """Start an instance.
1099

1100
  @type instance: L{objects.Instance}
1101
  @param instance: the instance object
1102
  @type startup_paused: bool
1103
  @param instance: pause instance at startup?
1104
  @rtype: None
1105

1106
  """
1107
  running_instances = GetInstanceList([instance.hypervisor])
1108

    
1109
  if instance.name in running_instances:
1110
    logging.info("Instance %s already running, not starting", instance.name)
1111
    return
1112

    
1113
  try:
1114
    block_devices = _GatherAndLinkBlockDevs(instance)
1115
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1116
    hyper.StartInstance(instance, block_devices, startup_paused)
1117
  except errors.BlockDeviceError, err:
1118
    _Fail("Block device error: %s", err, exc=True)
1119
  except errors.HypervisorError, err:
1120
    _RemoveBlockDevLinks(instance.name, instance.disks)
1121
    _Fail("Hypervisor error: %s", err, exc=True)
1122

    
1123

    
1124
def InstanceShutdown(instance, timeout):
1125
  """Shut an instance down.
1126

1127
  @note: this functions uses polling with a hardcoded timeout.
1128

1129
  @type instance: L{objects.Instance}
1130
  @param instance: the instance object
1131
  @type timeout: integer
1132
  @param timeout: maximum timeout for soft shutdown
1133
  @rtype: None
1134

1135
  """
1136
  hv_name = instance.hypervisor
1137
  hyper = hypervisor.GetHypervisor(hv_name)
1138
  iname = instance.name
1139

    
1140
  if instance.name not in hyper.ListInstances():
1141
    logging.info("Instance %s not running, doing nothing", iname)
1142
    return
1143

    
1144
  class _TryShutdown:
1145
    def __init__(self):
1146
      self.tried_once = False
1147

    
1148
    def __call__(self):
1149
      if iname not in hyper.ListInstances():
1150
        return
1151

    
1152
      try:
1153
        hyper.StopInstance(instance, retry=self.tried_once)
1154
      except errors.HypervisorError, err:
1155
        if iname not in hyper.ListInstances():
1156
          # if the instance is no longer existing, consider this a
1157
          # success and go to cleanup
1158
          return
1159

    
1160
        _Fail("Failed to stop instance %s: %s", iname, err)
1161

    
1162
      self.tried_once = True
1163

    
1164
      raise utils.RetryAgain()
1165

    
1166
  try:
1167
    utils.Retry(_TryShutdown(), 5, timeout)
1168
  except utils.RetryTimeout:
1169
    # the shutdown did not succeed
1170
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1171

    
1172
    try:
1173
      hyper.StopInstance(instance, force=True)
1174
    except errors.HypervisorError, err:
1175
      if iname in hyper.ListInstances():
1176
        # only raise an error if the instance still exists, otherwise
1177
        # the error could simply be "instance ... unknown"!
1178
        _Fail("Failed to force stop instance %s: %s", iname, err)
1179

    
1180
    time.sleep(1)
1181

    
1182
    if iname in hyper.ListInstances():
1183
      _Fail("Could not shutdown instance %s even by destroy", iname)
1184

    
1185
  try:
1186
    hyper.CleanupInstance(instance.name)
1187
  except errors.HypervisorError, err:
1188
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1189

    
1190
  _RemoveBlockDevLinks(iname, instance.disks)
1191

    
1192

    
1193
def InstanceReboot(instance, reboot_type, shutdown_timeout):
1194
  """Reboot an instance.
1195

1196
  @type instance: L{objects.Instance}
1197
  @param instance: the instance object to reboot
1198
  @type reboot_type: str
1199
  @param reboot_type: the type of reboot, one the following
1200
    constants:
1201
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1202
        instance OS, do not recreate the VM
1203
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1204
        restart the VM (at the hypervisor level)
1205
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1206
        not accepted here, since that mode is handled differently, in
1207
        cmdlib, and translates into full stop and start of the
1208
        instance (instead of a call_instance_reboot RPC)
1209
  @type shutdown_timeout: integer
1210
  @param shutdown_timeout: maximum timeout for soft shutdown
1211
  @rtype: None
1212

1213
  """
1214
  running_instances = GetInstanceList([instance.hypervisor])
1215

    
1216
  if instance.name not in running_instances:
1217
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1218

    
1219
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1220
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1221
    try:
1222
      hyper.RebootInstance(instance)
1223
    except errors.HypervisorError, err:
1224
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1225
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1226
    try:
1227
      InstanceShutdown(instance, shutdown_timeout)
1228
      return StartInstance(instance, False)
1229
    except errors.HypervisorError, err:
1230
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1231
  else:
1232
    _Fail("Invalid reboot_type received: %s", reboot_type)
1233

    
1234

    
1235
def MigrationInfo(instance):
1236
  """Gather information about an instance to be migrated.
1237

1238
  @type instance: L{objects.Instance}
1239
  @param instance: the instance definition
1240

1241
  """
1242
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1243
  try:
1244
    info = hyper.MigrationInfo(instance)
1245
  except errors.HypervisorError, err:
1246
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1247
  return info
1248

    
1249

    
1250
def AcceptInstance(instance, info, target):
1251
  """Prepare the node to accept an instance.
1252

1253
  @type instance: L{objects.Instance}
1254
  @param instance: the instance definition
1255
  @type info: string/data (opaque)
1256
  @param info: migration information, from the source node
1257
  @type target: string
1258
  @param target: target host (usually ip), on this node
1259

1260
  """
1261
  # TODO: why is this required only for DTS_EXT_MIRROR?
1262
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1263
    # Create the symlinks, as the disks are not active
1264
    # in any way
1265
    try:
1266
      _GatherAndLinkBlockDevs(instance)
1267
    except errors.BlockDeviceError, err:
1268
      _Fail("Block device error: %s", err, exc=True)
1269

    
1270
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1271
  try:
1272
    hyper.AcceptInstance(instance, info, target)
1273
  except errors.HypervisorError, err:
1274
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1275
      _RemoveBlockDevLinks(instance.name, instance.disks)
1276
    _Fail("Failed to accept instance: %s", err, exc=True)
1277

    
1278

    
1279
def FinalizeMigration(instance, info, success):
1280
  """Finalize any preparation to accept an instance.
1281

1282
  @type instance: L{objects.Instance}
1283
  @param instance: the instance definition
1284
  @type info: string/data (opaque)
1285
  @param info: migration information, from the source node
1286
  @type success: boolean
1287
  @param success: whether the migration was a success or a failure
1288

1289
  """
1290
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1291
  try:
1292
    hyper.FinalizeMigration(instance, info, success)
1293
  except errors.HypervisorError, err:
1294
    _Fail("Failed to finalize migration: %s", err, exc=True)
1295

    
1296

    
1297
def MigrateInstance(instance, target, live):
1298
  """Migrates an instance to another node.
1299

1300
  @type instance: L{objects.Instance}
1301
  @param instance: the instance definition
1302
  @type target: string
1303
  @param target: the target node name
1304
  @type live: boolean
1305
  @param live: whether the migration should be done live or not (the
1306
      interpretation of this parameter is left to the hypervisor)
1307
  @rtype: tuple
1308
  @return: a tuple of (success, msg) where:
1309
      - succes is a boolean denoting the success/failure of the operation
1310
      - msg is a string with details in case of failure
1311

1312
  """
1313
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1314

    
1315
  try:
1316
    hyper.MigrateInstance(instance, target, live)
1317
  except errors.HypervisorError, err:
1318
    _Fail("Failed to migrate instance: %s", err, exc=True)
1319

    
1320

    
1321
def BlockdevCreate(disk, size, owner, on_primary, info):
1322
  """Creates a block device for an instance.
1323

1324
  @type disk: L{objects.Disk}
1325
  @param disk: the object describing the disk we should create
1326
  @type size: int
1327
  @param size: the size of the physical underlying device, in MiB
1328
  @type owner: str
1329
  @param owner: the name of the instance for which disk is created,
1330
      used for device cache data
1331
  @type on_primary: boolean
1332
  @param on_primary:  indicates if it is the primary node or not
1333
  @type info: string
1334
  @param info: string that will be sent to the physical device
1335
      creation, used for example to set (LVM) tags on LVs
1336

1337
  @return: the new unique_id of the device (this can sometime be
1338
      computed only after creation), or None. On secondary nodes,
1339
      it's not required to return anything.
1340

1341
  """
1342
  # TODO: remove the obsolete "size" argument
1343
  # pylint: disable=W0613
1344
  clist = []
1345
  if disk.children:
1346
    for child in disk.children:
1347
      try:
1348
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1349
      except errors.BlockDeviceError, err:
1350
        _Fail("Can't assemble device %s: %s", child, err)
1351
      if on_primary or disk.AssembleOnSecondary():
1352
        # we need the children open in case the device itself has to
1353
        # be assembled
1354
        try:
1355
          # pylint: disable=E1103
1356
          crdev.Open()
1357
        except errors.BlockDeviceError, err:
1358
          _Fail("Can't make child '%s' read-write: %s", child, err)
1359
      clist.append(crdev)
1360

    
1361
  try:
1362
    device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1363
  except errors.BlockDeviceError, err:
1364
    _Fail("Can't create block device: %s", err)
1365

    
1366
  if on_primary or disk.AssembleOnSecondary():
1367
    try:
1368
      device.Assemble()
1369
    except errors.BlockDeviceError, err:
1370
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1371
    device.SetSyncSpeed(constants.SYNC_SPEED)
1372
    if on_primary or disk.OpenOnSecondary():
1373
      try:
1374
        device.Open(force=True)
1375
      except errors.BlockDeviceError, err:
1376
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1377
    DevCacheManager.UpdateCache(device.dev_path, owner,
1378
                                on_primary, disk.iv_name)
1379

    
1380
  device.SetInfo(info)
1381

    
1382
  return device.unique_id
1383

    
1384

    
1385
def _WipeDevice(path, offset, size):
1386
  """This function actually wipes the device.
1387

1388
  @param path: The path to the device to wipe
1389
  @param offset: The offset in MiB in the file
1390
  @param size: The size in MiB to write
1391

1392
  """
1393
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1394
         "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1395
         "count=%d" % size]
1396
  result = utils.RunCmd(cmd)
1397

    
1398
  if result.failed:
1399
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1400
          result.fail_reason, result.output)
1401

    
1402

    
1403
def BlockdevWipe(disk, offset, size):
1404
  """Wipes a block device.
1405

1406
  @type disk: L{objects.Disk}
1407
  @param disk: the disk object we want to wipe
1408
  @type offset: int
1409
  @param offset: The offset in MiB in the file
1410
  @type size: int
1411
  @param size: The size in MiB to write
1412

1413
  """
1414
  try:
1415
    rdev = _RecursiveFindBD(disk)
1416
  except errors.BlockDeviceError:
1417
    rdev = None
1418

    
1419
  if not rdev:
1420
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1421

    
1422
  # Do cross verify some of the parameters
1423
  if offset > rdev.size:
1424
    _Fail("Offset is bigger than device size")
1425
  if (offset + size) > rdev.size:
1426
    _Fail("The provided offset and size to wipe is bigger than device size")
1427

    
1428
  _WipeDevice(rdev.dev_path, offset, size)
1429

    
1430

    
1431
def BlockdevPauseResumeSync(disks, pause):
1432
  """Pause or resume the sync of the block device.
1433

1434
  @type disks: list of L{objects.Disk}
1435
  @param disks: the disks object we want to pause/resume
1436
  @type pause: bool
1437
  @param pause: Wheater to pause or resume
1438

1439
  """
1440
  success = []
1441
  for disk in disks:
1442
    try:
1443
      rdev = _RecursiveFindBD(disk)
1444
    except errors.BlockDeviceError:
1445
      rdev = None
1446

    
1447
    if not rdev:
1448
      success.append((False, ("Cannot change sync for device %s:"
1449
                              " device not found" % disk.iv_name)))
1450
      continue
1451

    
1452
    result = rdev.PauseResumeSync(pause)
1453

    
1454
    if result:
1455
      success.append((result, None))
1456
    else:
1457
      if pause:
1458
        msg = "Pause"
1459
      else:
1460
        msg = "Resume"
1461
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1462

    
1463
  return success
1464

    
1465

    
1466
def BlockdevRemove(disk):
1467
  """Remove a block device.
1468

1469
  @note: This is intended to be called recursively.
1470

1471
  @type disk: L{objects.Disk}
1472
  @param disk: the disk object we should remove
1473
  @rtype: boolean
1474
  @return: the success of the operation
1475

1476
  """
1477
  msgs = []
1478
  try:
1479
    rdev = _RecursiveFindBD(disk)
1480
  except errors.BlockDeviceError, err:
1481
    # probably can't attach
1482
    logging.info("Can't attach to device %s in remove", disk)
1483
    rdev = None
1484
  if rdev is not None:
1485
    r_path = rdev.dev_path
1486
    try:
1487
      rdev.Remove()
1488
    except errors.BlockDeviceError, err:
1489
      msgs.append(str(err))
1490
    if not msgs:
1491
      DevCacheManager.RemoveCache(r_path)
1492

    
1493
  if disk.children:
1494
    for child in disk.children:
1495
      try:
1496
        BlockdevRemove(child)
1497
      except RPCFail, err:
1498
        msgs.append(str(err))
1499

    
1500
  if msgs:
1501
    _Fail("; ".join(msgs))
1502

    
1503

    
1504
def _RecursiveAssembleBD(disk, owner, as_primary):
1505
  """Activate a block device for an instance.
1506

1507
  This is run on the primary and secondary nodes for an instance.
1508

1509
  @note: this function is called recursively.
1510

1511
  @type disk: L{objects.Disk}
1512
  @param disk: the disk we try to assemble
1513
  @type owner: str
1514
  @param owner: the name of the instance which owns the disk
1515
  @type as_primary: boolean
1516
  @param as_primary: if we should make the block device
1517
      read/write
1518

1519
  @return: the assembled device or None (in case no device
1520
      was assembled)
1521
  @raise errors.BlockDeviceError: in case there is an error
1522
      during the activation of the children or the device
1523
      itself
1524

1525
  """
1526
  children = []
1527
  if disk.children:
1528
    mcn = disk.ChildrenNeeded()
1529
    if mcn == -1:
1530
      mcn = 0 # max number of Nones allowed
1531
    else:
1532
      mcn = len(disk.children) - mcn # max number of Nones
1533
    for chld_disk in disk.children:
1534
      try:
1535
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1536
      except errors.BlockDeviceError, err:
1537
        if children.count(None) >= mcn:
1538
          raise
1539
        cdev = None
1540
        logging.error("Error in child activation (but continuing): %s",
1541
                      str(err))
1542
      children.append(cdev)
1543

    
1544
  if as_primary or disk.AssembleOnSecondary():
1545
    r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1546
    r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1547
    result = r_dev
1548
    if as_primary or disk.OpenOnSecondary():
1549
      r_dev.Open()
1550
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1551
                                as_primary, disk.iv_name)
1552

    
1553
  else:
1554
    result = True
1555
  return result
1556

    
1557

    
1558
def BlockdevAssemble(disk, owner, as_primary, idx):
1559
  """Activate a block device for an instance.
1560

1561
  This is a wrapper over _RecursiveAssembleBD.
1562

1563
  @rtype: str or boolean
1564
  @return: a C{/dev/...} path for primary nodes, and
1565
      C{True} for secondary nodes
1566

1567
  """
1568
  try:
1569
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1570
    if isinstance(result, bdev.BlockDev):
1571
      # pylint: disable=E1103
1572
      result = result.dev_path
1573
      if as_primary:
1574
        _SymlinkBlockDev(owner, result, idx)
1575
  except errors.BlockDeviceError, err:
1576
    _Fail("Error while assembling disk: %s", err, exc=True)
1577
  except OSError, err:
1578
    _Fail("Error while symlinking disk: %s", err, exc=True)
1579

    
1580
  return result
1581

    
1582

    
1583
def BlockdevShutdown(disk):
1584
  """Shut down a block device.
1585

1586
  First, if the device is assembled (Attach() is successful), then
1587
  the device is shutdown. Then the children of the device are
1588
  shutdown.
1589

1590
  This function is called recursively. Note that we don't cache the
1591
  children or such, as oppossed to assemble, shutdown of different
1592
  devices doesn't require that the upper device was active.
1593

1594
  @type disk: L{objects.Disk}
1595
  @param disk: the description of the disk we should
1596
      shutdown
1597
  @rtype: None
1598

1599
  """
1600
  msgs = []
1601
  r_dev = _RecursiveFindBD(disk)
1602
  if r_dev is not None:
1603
    r_path = r_dev.dev_path
1604
    try:
1605
      r_dev.Shutdown()
1606
      DevCacheManager.RemoveCache(r_path)
1607
    except errors.BlockDeviceError, err:
1608
      msgs.append(str(err))
1609

    
1610
  if disk.children:
1611
    for child in disk.children:
1612
      try:
1613
        BlockdevShutdown(child)
1614
      except RPCFail, err:
1615
        msgs.append(str(err))
1616

    
1617
  if msgs:
1618
    _Fail("; ".join(msgs))
1619

    
1620

    
1621
def BlockdevAddchildren(parent_cdev, new_cdevs):
1622
  """Extend a mirrored block device.
1623

1624
  @type parent_cdev: L{objects.Disk}
1625
  @param parent_cdev: the disk to which we should add children
1626
  @type new_cdevs: list of L{objects.Disk}
1627
  @param new_cdevs: the list of children which we should add
1628
  @rtype: None
1629

1630
  """
1631
  parent_bdev = _RecursiveFindBD(parent_cdev)
1632
  if parent_bdev is None:
1633
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1634
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1635
  if new_bdevs.count(None) > 0:
1636
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1637
  parent_bdev.AddChildren(new_bdevs)
1638

    
1639

    
1640
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1641
  """Shrink a mirrored block device.
1642

1643
  @type parent_cdev: L{objects.Disk}
1644
  @param parent_cdev: the disk from which we should remove children
1645
  @type new_cdevs: list of L{objects.Disk}
1646
  @param new_cdevs: the list of children which we should remove
1647
  @rtype: None
1648

1649
  """
1650
  parent_bdev = _RecursiveFindBD(parent_cdev)
1651
  if parent_bdev is None:
1652
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1653
  devs = []
1654
  for disk in new_cdevs:
1655
    rpath = disk.StaticDevPath()
1656
    if rpath is None:
1657
      bd = _RecursiveFindBD(disk)
1658
      if bd is None:
1659
        _Fail("Can't find device %s while removing children", disk)
1660
      else:
1661
        devs.append(bd.dev_path)
1662
    else:
1663
      if not utils.IsNormAbsPath(rpath):
1664
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1665
      devs.append(rpath)
1666
  parent_bdev.RemoveChildren(devs)
1667

    
1668

    
1669
def BlockdevGetmirrorstatus(disks):
1670
  """Get the mirroring status of a list of devices.
1671

1672
  @type disks: list of L{objects.Disk}
1673
  @param disks: the list of disks which we should query
1674
  @rtype: disk
1675
  @return: List of L{objects.BlockDevStatus}, one for each disk
1676
  @raise errors.BlockDeviceError: if any of the disks cannot be
1677
      found
1678

1679
  """
1680
  stats = []
1681
  for dsk in disks:
1682
    rbd = _RecursiveFindBD(dsk)
1683
    if rbd is None:
1684
      _Fail("Can't find device %s", dsk)
1685

    
1686
    stats.append(rbd.CombinedSyncStatus())
1687

    
1688
  return stats
1689

    
1690

    
1691
def BlockdevGetmirrorstatusMulti(disks):
1692
  """Get the mirroring status of a list of devices.
1693

1694
  @type disks: list of L{objects.Disk}
1695
  @param disks: the list of disks which we should query
1696
  @rtype: disk
1697
  @return: List of tuples, (bool, status), one for each disk; bool denotes
1698
    success/failure, status is L{objects.BlockDevStatus} on success, string
1699
    otherwise
1700

1701
  """
1702
  result = []
1703
  for disk in disks:
1704
    try:
1705
      rbd = _RecursiveFindBD(disk)
1706
      if rbd is None:
1707
        result.append((False, "Can't find device %s" % disk))
1708
        continue
1709

    
1710
      status = rbd.CombinedSyncStatus()
1711
    except errors.BlockDeviceError, err:
1712
      logging.exception("Error while getting disk status")
1713
      result.append((False, str(err)))
1714
    else:
1715
      result.append((True, status))
1716

    
1717
  assert len(disks) == len(result)
1718

    
1719
  return result
1720

    
1721

    
1722
def _RecursiveFindBD(disk):
1723
  """Check if a device is activated.
1724

1725
  If so, return information about the real device.
1726

1727
  @type disk: L{objects.Disk}
1728
  @param disk: the disk object we need to find
1729

1730
  @return: None if the device can't be found,
1731
      otherwise the device instance
1732

1733
  """
1734
  children = []
1735
  if disk.children:
1736
    for chdisk in disk.children:
1737
      children.append(_RecursiveFindBD(chdisk))
1738

    
1739
  return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1740

    
1741

    
1742
def _OpenRealBD(disk):
1743
  """Opens the underlying block device of a disk.
1744

1745
  @type disk: L{objects.Disk}
1746
  @param disk: the disk object we want to open
1747

1748
  """
1749
  real_disk = _RecursiveFindBD(disk)
1750
  if real_disk is None:
1751
    _Fail("Block device '%s' is not set up", disk)
1752

    
1753
  real_disk.Open()
1754

    
1755
  return real_disk
1756

    
1757

    
1758
def BlockdevFind(disk):
1759
  """Check if a device is activated.
1760

1761
  If it is, return information about the real device.
1762

1763
  @type disk: L{objects.Disk}
1764
  @param disk: the disk to find
1765
  @rtype: None or objects.BlockDevStatus
1766
  @return: None if the disk cannot be found, otherwise a the current
1767
           information
1768

1769
  """
1770
  try:
1771
    rbd = _RecursiveFindBD(disk)
1772
  except errors.BlockDeviceError, err:
1773
    _Fail("Failed to find device: %s", err, exc=True)
1774

    
1775
  if rbd is None:
1776
    return None
1777

    
1778
  return rbd.GetSyncStatus()
1779

    
1780

    
1781
def BlockdevGetsize(disks):
1782
  """Computes the size of the given disks.
1783

1784
  If a disk is not found, returns None instead.
1785

1786
  @type disks: list of L{objects.Disk}
1787
  @param disks: the list of disk to compute the size for
1788
  @rtype: list
1789
  @return: list with elements None if the disk cannot be found,
1790
      otherwise the size
1791

1792
  """
1793
  result = []
1794
  for cf in disks:
1795
    try:
1796
      rbd = _RecursiveFindBD(cf)
1797
    except errors.BlockDeviceError:
1798
      result.append(None)
1799
      continue
1800
    if rbd is None:
1801
      result.append(None)
1802
    else:
1803
      result.append(rbd.GetActualSize())
1804
  return result
1805

    
1806

    
1807
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1808
  """Export a block device to a remote node.
1809

1810
  @type disk: L{objects.Disk}
1811
  @param disk: the description of the disk to export
1812
  @type dest_node: str
1813
  @param dest_node: the destination node to export to
1814
  @type dest_path: str
1815
  @param dest_path: the destination path on the target node
1816
  @type cluster_name: str
1817
  @param cluster_name: the cluster name, needed for SSH hostalias
1818
  @rtype: None
1819

1820
  """
1821
  real_disk = _OpenRealBD(disk)
1822

    
1823
  # the block size on the read dd is 1MiB to match our units
1824
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1825
                               "dd if=%s bs=1048576 count=%s",
1826
                               real_disk.dev_path, str(disk.size))
1827

    
1828
  # we set here a smaller block size as, due to ssh buffering, more
1829
  # than 64-128k will mostly ignored; we use nocreat to fail if the
1830
  # device is not already there or we pass a wrong path; we use
1831
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1832
  # to not buffer too much memory; this means that at best, we flush
1833
  # every 64k, which will not be very fast
1834
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1835
                                " oflag=dsync", dest_path)
1836

    
1837
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1838
                                                   constants.GANETI_RUNAS,
1839
                                                   destcmd)
1840

    
1841
  # all commands have been checked, so we're safe to combine them
1842
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1843

    
1844
  result = utils.RunCmd(["bash", "-c", command])
1845

    
1846
  if result.failed:
1847
    _Fail("Disk copy command '%s' returned error: %s"
1848
          " output: %s", command, result.fail_reason, result.output)
1849

    
1850

    
1851
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1852
  """Write a file to the filesystem.
1853

1854
  This allows the master to overwrite(!) a file. It will only perform
1855
  the operation if the file belongs to a list of configuration files.
1856

1857
  @type file_name: str
1858
  @param file_name: the target file name
1859
  @type data: str
1860
  @param data: the new contents of the file
1861
  @type mode: int
1862
  @param mode: the mode to give the file (can be None)
1863
  @type uid: string
1864
  @param uid: the owner of the file
1865
  @type gid: string
1866
  @param gid: the group of the file
1867
  @type atime: float
1868
  @param atime: the atime to set on the file (can be None)
1869
  @type mtime: float
1870
  @param mtime: the mtime to set on the file (can be None)
1871
  @rtype: None
1872

1873
  """
1874
  if not os.path.isabs(file_name):
1875
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1876

    
1877
  if file_name not in _ALLOWED_UPLOAD_FILES:
1878
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1879
          file_name)
1880

    
1881
  raw_data = _Decompress(data)
1882

    
1883
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
1884
    _Fail("Invalid username/groupname type")
1885

    
1886
  getents = runtime.GetEnts()
1887
  uid = getents.LookupUser(uid)
1888
  gid = getents.LookupGroup(gid)
1889

    
1890
  utils.SafeWriteFile(file_name, None,
1891
                      data=raw_data, mode=mode, uid=uid, gid=gid,
1892
                      atime=atime, mtime=mtime)
1893

    
1894

    
1895
def RunOob(oob_program, command, node, timeout):
1896
  """Executes oob_program with given command on given node.
1897

1898
  @param oob_program: The path to the executable oob_program
1899
  @param command: The command to invoke on oob_program
1900
  @param node: The node given as an argument to the program
1901
  @param timeout: Timeout after which we kill the oob program
1902

1903
  @return: stdout
1904
  @raise RPCFail: If execution fails for some reason
1905

1906
  """
1907
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
1908

    
1909
  if result.failed:
1910
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
1911
          result.fail_reason, result.output)
1912

    
1913
  return result.stdout
1914

    
1915

    
1916
def WriteSsconfFiles(values):
1917
  """Update all ssconf files.
1918

1919
  Wrapper around the SimpleStore.WriteFiles.
1920

1921
  """
1922
  ssconf.SimpleStore().WriteFiles(values)
1923

    
1924

    
1925
def _ErrnoOrStr(err):
1926
  """Format an EnvironmentError exception.
1927

1928
  If the L{err} argument has an errno attribute, it will be looked up
1929
  and converted into a textual C{E...} description. Otherwise the
1930
  string representation of the error will be returned.
1931

1932
  @type err: L{EnvironmentError}
1933
  @param err: the exception to format
1934

1935
  """
1936
  if hasattr(err, "errno"):
1937
    detail = errno.errorcode[err.errno]
1938
  else:
1939
    detail = str(err)
1940
  return detail
1941

    
1942

    
1943
def _OSOndiskAPIVersion(os_dir):
1944
  """Compute and return the API version of a given OS.
1945

1946
  This function will try to read the API version of the OS residing in
1947
  the 'os_dir' directory.
1948

1949
  @type os_dir: str
1950
  @param os_dir: the directory in which we should look for the OS
1951
  @rtype: tuple
1952
  @return: tuple (status, data) with status denoting the validity and
1953
      data holding either the vaid versions or an error message
1954

1955
  """
1956
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1957

    
1958
  try:
1959
    st = os.stat(api_file)
1960
  except EnvironmentError, err:
1961
    return False, ("Required file '%s' not found under path %s: %s" %
1962
                   (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1963

    
1964
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1965
    return False, ("File '%s' in %s is not a regular file" %
1966
                   (constants.OS_API_FILE, os_dir))
1967

    
1968
  try:
1969
    api_versions = utils.ReadFile(api_file).splitlines()
1970
  except EnvironmentError, err:
1971
    return False, ("Error while reading the API version file at %s: %s" %
1972
                   (api_file, _ErrnoOrStr(err)))
1973

    
1974
  try:
1975
    api_versions = [int(version.strip()) for version in api_versions]
1976
  except (TypeError, ValueError), err:
1977
    return False, ("API version(s) can't be converted to integer: %s" %
1978
                   str(err))
1979

    
1980
  return True, api_versions
1981

    
1982

    
1983
def DiagnoseOS(top_dirs=None):
1984
  """Compute the validity for all OSes.
1985

1986
  @type top_dirs: list
1987
  @param top_dirs: the list of directories in which to
1988
      search (if not given defaults to
1989
      L{constants.OS_SEARCH_PATH})
1990
  @rtype: list of L{objects.OS}
1991
  @return: a list of tuples (name, path, status, diagnose, variants,
1992
      parameters, api_version) for all (potential) OSes under all
1993
      search paths, where:
1994
          - name is the (potential) OS name
1995
          - path is the full path to the OS
1996
          - status True/False is the validity of the OS
1997
          - diagnose is the error message for an invalid OS, otherwise empty
1998
          - variants is a list of supported OS variants, if any
1999
          - parameters is a list of (name, help) parameters, if any
2000
          - api_version is a list of support OS API versions
2001

2002
  """
2003
  if top_dirs is None:
2004
    top_dirs = constants.OS_SEARCH_PATH
2005

    
2006
  result = []
2007
  for dir_name in top_dirs:
2008
    if os.path.isdir(dir_name):
2009
      try:
2010
        f_names = utils.ListVisibleFiles(dir_name)
2011
      except EnvironmentError, err:
2012
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2013
        break
2014
      for name in f_names:
2015
        os_path = utils.PathJoin(dir_name, name)
2016
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2017
        if status:
2018
          diagnose = ""
2019
          variants = os_inst.supported_variants
2020
          parameters = os_inst.supported_parameters
2021
          api_versions = os_inst.api_versions
2022
        else:
2023
          diagnose = os_inst
2024
          variants = parameters = api_versions = []
2025
        result.append((name, os_path, status, diagnose, variants,
2026
                       parameters, api_versions))
2027

    
2028
  return result
2029

    
2030

    
2031
def _TryOSFromDisk(name, base_dir=None):
2032
  """Create an OS instance from disk.
2033

2034
  This function will return an OS instance if the given name is a
2035
  valid OS name.
2036

2037
  @type base_dir: string
2038
  @keyword base_dir: Base directory containing OS installations.
2039
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2040
  @rtype: tuple
2041
  @return: success and either the OS instance if we find a valid one,
2042
      or error message
2043

2044
  """
2045
  if base_dir is None:
2046
    os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2047
  else:
2048
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2049

    
2050
  if os_dir is None:
2051
    return False, "Directory for OS %s not found in search path" % name
2052

    
2053
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2054
  if not status:
2055
    # push the error up
2056
    return status, api_versions
2057

    
2058
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2059
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2060
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2061

    
2062
  # OS Files dictionary, we will populate it with the absolute path
2063
  # names; if the value is True, then it is a required file, otherwise
2064
  # an optional one
2065
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2066

    
2067
  if max(api_versions) >= constants.OS_API_V15:
2068
    os_files[constants.OS_VARIANTS_FILE] = False
2069

    
2070
  if max(api_versions) >= constants.OS_API_V20:
2071
    os_files[constants.OS_PARAMETERS_FILE] = True
2072
  else:
2073
    del os_files[constants.OS_SCRIPT_VERIFY]
2074

    
2075
  for (filename, required) in os_files.items():
2076
    os_files[filename] = utils.PathJoin(os_dir, filename)
2077

    
2078
    try:
2079
      st = os.stat(os_files[filename])
2080
    except EnvironmentError, err:
2081
      if err.errno == errno.ENOENT and not required:
2082
        del os_files[filename]
2083
        continue
2084
      return False, ("File '%s' under path '%s' is missing (%s)" %
2085
                     (filename, os_dir, _ErrnoOrStr(err)))
2086

    
2087
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2088
      return False, ("File '%s' under path '%s' is not a regular file" %
2089
                     (filename, os_dir))
2090

    
2091
    if filename in constants.OS_SCRIPTS:
2092
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2093
        return False, ("File '%s' under path '%s' is not executable" %
2094
                       (filename, os_dir))
2095

    
2096
  variants = []
2097
  if constants.OS_VARIANTS_FILE in os_files:
2098
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2099
    try:
2100
      variants = utils.ReadFile(variants_file).splitlines()
2101
    except EnvironmentError, err:
2102
      # we accept missing files, but not other errors
2103
      if err.errno != errno.ENOENT:
2104
        return False, ("Error while reading the OS variants file at %s: %s" %
2105
                       (variants_file, _ErrnoOrStr(err)))
2106

    
2107
  parameters = []
2108
  if constants.OS_PARAMETERS_FILE in os_files:
2109
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2110
    try:
2111
      parameters = utils.ReadFile(parameters_file).splitlines()
2112
    except EnvironmentError, err:
2113
      return False, ("Error while reading the OS parameters file at %s: %s" %
2114
                     (parameters_file, _ErrnoOrStr(err)))
2115
    parameters = [v.split(None, 1) for v in parameters]
2116

    
2117
  os_obj = objects.OS(name=name, path=os_dir,
2118
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2119
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2120
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2121
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2122
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2123
                                                 None),
2124
                      supported_variants=variants,
2125
                      supported_parameters=parameters,
2126
                      api_versions=api_versions)
2127
  return True, os_obj
2128

    
2129

    
2130
def OSFromDisk(name, base_dir=None):
2131
  """Create an OS instance from disk.
2132

2133
  This function will return an OS instance if the given name is a
2134
  valid OS name. Otherwise, it will raise an appropriate
2135
  L{RPCFail} exception, detailing why this is not a valid OS.
2136

2137
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2138
  an exception but returns true/false status data.
2139

2140
  @type base_dir: string
2141
  @keyword base_dir: Base directory containing OS installations.
2142
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2143
  @rtype: L{objects.OS}
2144
  @return: the OS instance if we find a valid one
2145
  @raise RPCFail: if we don't find a valid OS
2146

2147
  """
2148
  name_only = objects.OS.GetName(name)
2149
  status, payload = _TryOSFromDisk(name_only, base_dir)
2150

    
2151
  if not status:
2152
    _Fail(payload)
2153

    
2154
  return payload
2155

    
2156

    
2157
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2158
  """Calculate the basic environment for an os script.
2159

2160
  @type os_name: str
2161
  @param os_name: full operating system name (including variant)
2162
  @type inst_os: L{objects.OS}
2163
  @param inst_os: operating system for which the environment is being built
2164
  @type os_params: dict
2165
  @param os_params: the OS parameters
2166
  @type debug: integer
2167
  @param debug: debug level (0 or 1, for OS Api 10)
2168
  @rtype: dict
2169
  @return: dict of environment variables
2170
  @raise errors.BlockDeviceError: if the block device
2171
      cannot be found
2172

2173
  """
2174
  result = {}
2175
  api_version = \
2176
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2177
  result["OS_API_VERSION"] = "%d" % api_version
2178
  result["OS_NAME"] = inst_os.name
2179
  result["DEBUG_LEVEL"] = "%d" % debug
2180

    
2181
  # OS variants
2182
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2183
    variant = objects.OS.GetVariant(os_name)
2184
    if not variant:
2185
      variant = inst_os.supported_variants[0]
2186
  else:
2187
    variant = ""
2188
  result["OS_VARIANT"] = variant
2189

    
2190
  # OS params
2191
  for pname, pvalue in os_params.items():
2192
    result["OSP_%s" % pname.upper()] = pvalue
2193

    
2194
  return result
2195

    
2196

    
2197
def OSEnvironment(instance, inst_os, debug=0):
2198
  """Calculate the environment for an os script.
2199

2200
  @type instance: L{objects.Instance}
2201
  @param instance: target instance for the os script run
2202
  @type inst_os: L{objects.OS}
2203
  @param inst_os: operating system for which the environment is being built
2204
  @type debug: integer
2205
  @param debug: debug level (0 or 1, for OS Api 10)
2206
  @rtype: dict
2207
  @return: dict of environment variables
2208
  @raise errors.BlockDeviceError: if the block device
2209
      cannot be found
2210

2211
  """
2212
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2213

    
2214
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2215
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2216

    
2217
  result["HYPERVISOR"] = instance.hypervisor
2218
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2219
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2220
  result["INSTANCE_SECONDARY_NODES"] = \
2221
      ("%s" % " ".join(instance.secondary_nodes))
2222

    
2223
  # Disks
2224
  for idx, disk in enumerate(instance.disks):
2225
    real_disk = _OpenRealBD(disk)
2226
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2227
    result["DISK_%d_ACCESS" % idx] = disk.mode
2228
    if constants.HV_DISK_TYPE in instance.hvparams:
2229
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2230
        instance.hvparams[constants.HV_DISK_TYPE]
2231
    if disk.dev_type in constants.LDS_BLOCK:
2232
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2233
    elif disk.dev_type == constants.LD_FILE:
2234
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2235
        "file:%s" % disk.physical_id[0]
2236

    
2237
  # NICs
2238
  for idx, nic in enumerate(instance.nics):
2239
    result["NIC_%d_MAC" % idx] = nic.mac
2240
    if nic.ip:
2241
      result["NIC_%d_IP" % idx] = nic.ip
2242
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2243
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2244
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2245
    if nic.nicparams[constants.NIC_LINK]:
2246
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2247
    if constants.HV_NIC_TYPE in instance.hvparams:
2248
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2249
        instance.hvparams[constants.HV_NIC_TYPE]
2250

    
2251
  # HV/BE params
2252
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2253
    for key, value in source.items():
2254
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2255

    
2256
  return result
2257

    
2258

    
2259
def BlockdevGrow(disk, amount, dryrun):
2260
  """Grow a stack of block devices.
2261

2262
  This function is called recursively, with the childrens being the
2263
  first ones to resize.
2264

2265
  @type disk: L{objects.Disk}
2266
  @param disk: the disk to be grown
2267
  @type amount: integer
2268
  @param amount: the amount (in mebibytes) to grow with
2269
  @type dryrun: boolean
2270
  @param dryrun: whether to execute the operation in simulation mode
2271
      only, without actually increasing the size
2272
  @rtype: (status, result)
2273
  @return: a tuple with the status of the operation (True/False), and
2274
      the errors message if status is False
2275

2276
  """
2277
  r_dev = _RecursiveFindBD(disk)
2278
  if r_dev is None:
2279
    _Fail("Cannot find block device %s", disk)
2280

    
2281
  try:
2282
    r_dev.Grow(amount, dryrun)
2283
  except errors.BlockDeviceError, err:
2284
    _Fail("Failed to grow block device: %s", err, exc=True)
2285

    
2286

    
2287
def BlockdevSnapshot(disk):
2288
  """Create a snapshot copy of a block device.
2289

2290
  This function is called recursively, and the snapshot is actually created
2291
  just for the leaf lvm backend device.
2292

2293
  @type disk: L{objects.Disk}
2294
  @param disk: the disk to be snapshotted
2295
  @rtype: string
2296
  @return: snapshot disk ID as (vg, lv)
2297

2298
  """
2299
  if disk.dev_type == constants.LD_DRBD8:
2300
    if not disk.children:
2301
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2302
            disk.unique_id)
2303
    return BlockdevSnapshot(disk.children[0])
2304
  elif disk.dev_type == constants.LD_LV:
2305
    r_dev = _RecursiveFindBD(disk)
2306
    if r_dev is not None:
2307
      # FIXME: choose a saner value for the snapshot size
2308
      # let's stay on the safe side and ask for the full size, for now
2309
      return r_dev.Snapshot(disk.size)
2310
    else:
2311
      _Fail("Cannot find block device %s", disk)
2312
  else:
2313
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2314
          disk.unique_id, disk.dev_type)
2315

    
2316

    
2317
def FinalizeExport(instance, snap_disks):
2318
  """Write out the export configuration information.
2319

2320
  @type instance: L{objects.Instance}
2321
  @param instance: the instance which we export, used for
2322
      saving configuration
2323
  @type snap_disks: list of L{objects.Disk}
2324
  @param snap_disks: list of snapshot block devices, which
2325
      will be used to get the actual name of the dump file
2326

2327
  @rtype: None
2328

2329
  """
2330
  destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2331
  finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2332

    
2333
  config = objects.SerializableConfigParser()
2334

    
2335
  config.add_section(constants.INISECT_EXP)
2336
  config.set(constants.INISECT_EXP, "version", "0")
2337
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2338
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
2339
  config.set(constants.INISECT_EXP, "os", instance.os)
2340
  config.set(constants.INISECT_EXP, "compression", "none")
2341

    
2342
  config.add_section(constants.INISECT_INS)
2343
  config.set(constants.INISECT_INS, "name", instance.name)
2344
  config.set(constants.INISECT_INS, "memory", "%d" %
2345
             instance.beparams[constants.BE_MEMORY])
2346
  config.set(constants.INISECT_INS, "vcpus", "%d" %
2347
             instance.beparams[constants.BE_VCPUS])
2348
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2349
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2350
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2351

    
2352
  nic_total = 0
2353
  for nic_count, nic in enumerate(instance.nics):
2354
    nic_total += 1
2355
    config.set(constants.INISECT_INS, "nic%d_mac" %
2356
               nic_count, "%s" % nic.mac)
2357
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2358
    for param in constants.NICS_PARAMETER_TYPES:
2359
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2360
                 "%s" % nic.nicparams.get(param, None))
2361
  # TODO: redundant: on load can read nics until it doesn't exist
2362
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2363

    
2364
  disk_total = 0
2365
  for disk_count, disk in enumerate(snap_disks):
2366
    if disk:
2367
      disk_total += 1
2368
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2369
                 ("%s" % disk.iv_name))
2370
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2371
                 ("%s" % disk.physical_id[1]))
2372
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2373
                 ("%d" % disk.size))
2374

    
2375
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2376

    
2377
  # New-style hypervisor/backend parameters
2378

    
2379
  config.add_section(constants.INISECT_HYP)
2380
  for name, value in instance.hvparams.items():
2381
    if name not in constants.HVC_GLOBALS:
2382
      config.set(constants.INISECT_HYP, name, str(value))
2383

    
2384
  config.add_section(constants.INISECT_BEP)
2385
  for name, value in instance.beparams.items():
2386
    config.set(constants.INISECT_BEP, name, str(value))
2387

    
2388
  config.add_section(constants.INISECT_OSP)
2389
  for name, value in instance.osparams.items():
2390
    config.set(constants.INISECT_OSP, name, str(value))
2391

    
2392
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2393
                  data=config.Dumps())
2394
  shutil.rmtree(finaldestdir, ignore_errors=True)
2395
  shutil.move(destdir, finaldestdir)
2396

    
2397

    
2398
def ExportInfo(dest):
2399
  """Get export configuration information.
2400

2401
  @type dest: str
2402
  @param dest: directory containing the export
2403

2404
  @rtype: L{objects.SerializableConfigParser}
2405
  @return: a serializable config file containing the
2406
      export info
2407

2408
  """
2409
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2410

    
2411
  config = objects.SerializableConfigParser()
2412
  config.read(cff)
2413

    
2414
  if (not config.has_section(constants.INISECT_EXP) or
2415
      not config.has_section(constants.INISECT_INS)):
2416
    _Fail("Export info file doesn't have the required fields")
2417

    
2418
  return config.Dumps()
2419

    
2420

    
2421
def ListExports():
2422
  """Return a list of exports currently available on this machine.
2423

2424
  @rtype: list
2425
  @return: list of the exports
2426

2427
  """
2428
  if os.path.isdir(constants.EXPORT_DIR):
2429
    return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2430
  else:
2431
    _Fail("No exports directory")
2432

    
2433

    
2434
def RemoveExport(export):
2435
  """Remove an existing export from the node.
2436

2437
  @type export: str
2438
  @param export: the name of the export to remove
2439
  @rtype: None
2440

2441
  """
2442
  target = utils.PathJoin(constants.EXPORT_DIR, export)
2443

    
2444
  try:
2445
    shutil.rmtree(target)
2446
  except EnvironmentError, err:
2447
    _Fail("Error while removing the export: %s", err, exc=True)
2448

    
2449

    
2450
def BlockdevRename(devlist):
2451
  """Rename a list of block devices.
2452

2453
  @type devlist: list of tuples
2454
  @param devlist: list of tuples of the form  (disk,
2455
      new_logical_id, new_physical_id); disk is an
2456
      L{objects.Disk} object describing the current disk,
2457
      and new logical_id/physical_id is the name we
2458
      rename it to
2459
  @rtype: boolean
2460
  @return: True if all renames succeeded, False otherwise
2461

2462
  """
2463
  msgs = []
2464
  result = True
2465
  for disk, unique_id in devlist:
2466
    dev = _RecursiveFindBD(disk)
2467
    if dev is None:
2468
      msgs.append("Can't find device %s in rename" % str(disk))
2469
      result = False
2470
      continue
2471
    try:
2472
      old_rpath = dev.dev_path
2473
      dev.Rename(unique_id)
2474
      new_rpath = dev.dev_path
2475
      if old_rpath != new_rpath:
2476
        DevCacheManager.RemoveCache(old_rpath)
2477
        # FIXME: we should add the new cache information here, like:
2478
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2479
        # but we don't have the owner here - maybe parse from existing
2480
        # cache? for now, we only lose lvm data when we rename, which
2481
        # is less critical than DRBD or MD
2482
    except errors.BlockDeviceError, err:
2483
      msgs.append("Can't rename device '%s' to '%s': %s" %
2484
                  (dev, unique_id, err))
2485
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2486
      result = False
2487
  if not result:
2488
    _Fail("; ".join(msgs))
2489

    
2490

    
2491
def _TransformFileStorageDir(fs_dir):
2492
  """Checks whether given file_storage_dir is valid.
2493

2494
  Checks wheter the given fs_dir is within the cluster-wide default
2495
  file_storage_dir or the shared_file_storage_dir, which are stored in
2496
  SimpleStore. Only paths under those directories are allowed.
2497

2498
  @type fs_dir: str
2499
  @param fs_dir: the path to check
2500

2501
  @return: the normalized path if valid, None otherwise
2502

2503
  """
2504
  if not constants.ENABLE_FILE_STORAGE:
2505
    _Fail("File storage disabled at configure time")
2506
  cfg = _GetConfig()
2507
  fs_dir = os.path.normpath(fs_dir)
2508
  base_fstore = cfg.GetFileStorageDir()
2509
  base_shared = cfg.GetSharedFileStorageDir()
2510
  if not (utils.IsBelowDir(base_fstore, fs_dir) or
2511
          utils.IsBelowDir(base_shared, fs_dir)):
2512
    _Fail("File storage directory '%s' is not under base file"
2513
          " storage directory '%s' or shared storage directory '%s'",
2514
          fs_dir, base_fstore, base_shared)
2515
  return fs_dir
2516

    
2517

    
2518
def CreateFileStorageDir(file_storage_dir):
2519
  """Create file storage directory.
2520

2521
  @type file_storage_dir: str
2522
  @param file_storage_dir: directory to create
2523

2524
  @rtype: tuple
2525
  @return: tuple with first element a boolean indicating wheter dir
2526
      creation was successful or not
2527

2528
  """
2529
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2530
  if os.path.exists(file_storage_dir):
2531
    if not os.path.isdir(file_storage_dir):
2532
      _Fail("Specified storage dir '%s' is not a directory",
2533
            file_storage_dir)
2534
  else:
2535
    try:
2536
      os.makedirs(file_storage_dir, 0750)
2537
    except OSError, err:
2538
      _Fail("Cannot create file storage directory '%s': %s",
2539
            file_storage_dir, err, exc=True)
2540

    
2541

    
2542
def RemoveFileStorageDir(file_storage_dir):
2543
  """Remove file storage directory.
2544

2545
  Remove it only if it's empty. If not log an error and return.
2546

2547
  @type file_storage_dir: str
2548
  @param file_storage_dir: the directory we should cleanup
2549
  @rtype: tuple (success,)
2550
  @return: tuple of one element, C{success}, denoting
2551
      whether the operation was successful
2552

2553
  """
2554
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2555
  if os.path.exists(file_storage_dir):
2556
    if not os.path.isdir(file_storage_dir):
2557
      _Fail("Specified Storage directory '%s' is not a directory",
2558
            file_storage_dir)
2559
    # deletes dir only if empty, otherwise we want to fail the rpc call
2560
    try:
2561
      os.rmdir(file_storage_dir)
2562
    except OSError, err:
2563
      _Fail("Cannot remove file storage directory '%s': %s",
2564
            file_storage_dir, err)
2565

    
2566

    
2567
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2568
  """Rename the file storage directory.
2569

2570
  @type old_file_storage_dir: str
2571
  @param old_file_storage_dir: the current path
2572
  @type new_file_storage_dir: str
2573
  @param new_file_storage_dir: the name we should rename to
2574
  @rtype: tuple (success,)
2575
  @return: tuple of one element, C{success}, denoting
2576
      whether the operation was successful
2577

2578
  """
2579
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2580
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2581
  if not os.path.exists(new_file_storage_dir):
2582
    if os.path.isdir(old_file_storage_dir):
2583
      try:
2584
        os.rename(old_file_storage_dir, new_file_storage_dir)
2585
      except OSError, err:
2586
        _Fail("Cannot rename '%s' to '%s': %s",
2587
              old_file_storage_dir, new_file_storage_dir, err)
2588
    else:
2589
      _Fail("Specified storage dir '%s' is not a directory",
2590
            old_file_storage_dir)
2591
  else:
2592
    if os.path.exists(old_file_storage_dir):
2593
      _Fail("Cannot rename '%s' to '%s': both locations exist",
2594
            old_file_storage_dir, new_file_storage_dir)
2595

    
2596

    
2597
def _EnsureJobQueueFile(file_name):
2598
  """Checks whether the given filename is in the queue directory.
2599

2600
  @type file_name: str
2601
  @param file_name: the file name we should check
2602
  @rtype: None
2603
  @raises RPCFail: if the file is not valid
2604

2605
  """
2606
  queue_dir = os.path.normpath(constants.QUEUE_DIR)
2607
  result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2608

    
2609
  if not result:
2610
    _Fail("Passed job queue file '%s' does not belong to"
2611
          " the queue directory '%s'", file_name, queue_dir)
2612

    
2613

    
2614
def JobQueueUpdate(file_name, content):
2615
  """Updates a file in the queue directory.
2616

2617
  This is just a wrapper over L{utils.io.WriteFile}, with proper
2618
  checking.
2619

2620
  @type file_name: str
2621
  @param file_name: the job file name
2622
  @type content: str
2623
  @param content: the new job contents
2624
  @rtype: boolean
2625
  @return: the success of the operation
2626

2627
  """
2628
  _EnsureJobQueueFile(file_name)
2629
  getents = runtime.GetEnts()
2630

    
2631
  # Write and replace the file atomically
2632
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2633
                  gid=getents.masterd_gid)
2634

    
2635

    
2636
def JobQueueRename(old, new):
2637
  """Renames a job queue file.
2638

2639
  This is just a wrapper over os.rename with proper checking.
2640

2641
  @type old: str
2642
  @param old: the old (actual) file name
2643
  @type new: str
2644
  @param new: the desired file name
2645
  @rtype: tuple
2646
  @return: the success of the operation and payload
2647

2648
  """
2649
  _EnsureJobQueueFile(old)
2650
  _EnsureJobQueueFile(new)
2651

    
2652
  utils.RenameFile(old, new, mkdir=True)
2653

    
2654

    
2655
def BlockdevClose(instance_name, disks):
2656
  """Closes the given block devices.
2657

2658
  This means they will be switched to secondary mode (in case of
2659
  DRBD).
2660

2661
  @param instance_name: if the argument is not empty, the symlinks
2662
      of this instance will be removed
2663
  @type disks: list of L{objects.Disk}
2664
  @param disks: the list of disks to be closed
2665
  @rtype: tuple (success, message)
2666
  @return: a tuple of success and message, where success
2667
      indicates the succes of the operation, and message
2668
      which will contain the error details in case we
2669
      failed
2670

2671
  """
2672
  bdevs = []
2673
  for cf in disks:
2674
    rd = _RecursiveFindBD(cf)
2675
    if rd is None:
2676
      _Fail("Can't find device %s", cf)
2677
    bdevs.append(rd)
2678

    
2679
  msg = []
2680
  for rd in bdevs:
2681
    try:
2682
      rd.Close()
2683
    except errors.BlockDeviceError, err:
2684
      msg.append(str(err))
2685
  if msg:
2686
    _Fail("Can't make devices secondary: %s", ",".join(msg))
2687
  else:
2688
    if instance_name:
2689
      _RemoveBlockDevLinks(instance_name, disks)
2690

    
2691

    
2692
def ValidateHVParams(hvname, hvparams):
2693
  """Validates the given hypervisor parameters.
2694

2695
  @type hvname: string
2696
  @param hvname: the hypervisor name
2697
  @type hvparams: dict
2698
  @param hvparams: the hypervisor parameters to be validated
2699
  @rtype: None
2700

2701
  """
2702
  try:
2703
    hv_type = hypervisor.GetHypervisor(hvname)
2704
    hv_type.ValidateParameters(hvparams)
2705
  except errors.HypervisorError, err:
2706
    _Fail(str(err), log=False)
2707

    
2708

    
2709
def _CheckOSPList(os_obj, parameters):
2710
  """Check whether a list of parameters is supported by the OS.
2711

2712
  @type os_obj: L{objects.OS}
2713
  @param os_obj: OS object to check
2714
  @type parameters: list
2715
  @param parameters: the list of parameters to check
2716

2717
  """
2718
  supported = [v[0] for v in os_obj.supported_parameters]
2719
  delta = frozenset(parameters).difference(supported)
2720
  if delta:
2721
    _Fail("The following parameters are not supported"
2722
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2723

    
2724

    
2725
def ValidateOS(required, osname, checks, osparams):
2726
  """Validate the given OS' parameters.
2727

2728
  @type required: boolean
2729
  @param required: whether absence of the OS should translate into
2730
      failure or not
2731
  @type osname: string
2732
  @param osname: the OS to be validated
2733
  @type checks: list
2734
  @param checks: list of the checks to run (currently only 'parameters')
2735
  @type osparams: dict
2736
  @param osparams: dictionary with OS parameters
2737
  @rtype: boolean
2738
  @return: True if the validation passed, or False if the OS was not
2739
      found and L{required} was false
2740

2741
  """
2742
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2743
    _Fail("Unknown checks required for OS %s: %s", osname,
2744
          set(checks).difference(constants.OS_VALIDATE_CALLS))
2745

    
2746
  name_only = objects.OS.GetName(osname)
2747
  status, tbv = _TryOSFromDisk(name_only, None)
2748

    
2749
  if not status:
2750
    if required:
2751
      _Fail(tbv)
2752
    else:
2753
      return False
2754

    
2755
  if max(tbv.api_versions) < constants.OS_API_V20:
2756
    return True
2757

    
2758
  if constants.OS_VALIDATE_PARAMETERS in checks:
2759
    _CheckOSPList(tbv, osparams.keys())
2760

    
2761
  validate_env = OSCoreEnv(osname, tbv, osparams)
2762
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2763
                        cwd=tbv.path, reset_env=True)
2764
  if result.failed:
2765
    logging.error("os validate command '%s' returned error: %s output: %s",
2766
                  result.cmd, result.fail_reason, result.output)
2767
    _Fail("OS validation script failed (%s), output: %s",
2768
          result.fail_reason, result.output, log=False)
2769

    
2770
  return True
2771

    
2772

    
2773
def DemoteFromMC():
2774
  """Demotes the current node from master candidate role.
2775

2776
  """
2777
  # try to ensure we're not the master by mistake
2778
  master, myself = ssconf.GetMasterAndMyself()
2779
  if master == myself:
2780
    _Fail("ssconf status shows I'm the master node, will not demote")
2781

    
2782
  result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2783
  if not result.failed:
2784
    _Fail("The master daemon is running, will not demote")
2785

    
2786
  try:
2787
    if os.path.isfile(constants.CLUSTER_CONF_FILE):
2788
      utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2789
  except EnvironmentError, err:
2790
    if err.errno != errno.ENOENT:
2791
      _Fail("Error while backing up cluster file: %s", err, exc=True)
2792

    
2793
  utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2794

    
2795

    
2796
def _GetX509Filenames(cryptodir, name):
2797
  """Returns the full paths for the private key and certificate.
2798

2799
  """
2800
  return (utils.PathJoin(cryptodir, name),
2801
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2802
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2803

    
2804

    
2805
def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2806
  """Creates a new X509 certificate for SSL/TLS.
2807

2808
  @type validity: int
2809
  @param validity: Validity in seconds
2810
  @rtype: tuple; (string, string)
2811
  @return: Certificate name and public part
2812

2813
  """
2814
  (key_pem, cert_pem) = \
2815
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2816
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
2817

    
2818
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
2819
                              prefix="x509-%s-" % utils.TimestampForFilename())
2820
  try:
2821
    name = os.path.basename(cert_dir)
2822
    assert len(name) > 5
2823

    
2824
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2825

    
2826
    utils.WriteFile(key_file, mode=0400, data=key_pem)
2827
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2828

    
2829
    # Never return private key as it shouldn't leave the node
2830
    return (name, cert_pem)
2831
  except Exception:
2832
    shutil.rmtree(cert_dir, ignore_errors=True)
2833
    raise
2834

    
2835

    
2836
def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2837
  """Removes a X509 certificate.
2838

2839
  @type name: string
2840
  @param name: Certificate name
2841

2842
  """
2843
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2844

    
2845
  utils.RemoveFile(key_file)
2846
  utils.RemoveFile(cert_file)
2847

    
2848
  try:
2849
    os.rmdir(cert_dir)
2850
  except EnvironmentError, err:
2851
    _Fail("Cannot remove certificate directory '%s': %s",
2852
          cert_dir, err)
2853

    
2854

    
2855
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2856
  """Returns the command for the requested input/output.
2857

2858
  @type instance: L{objects.Instance}
2859
  @param instance: The instance object
2860
  @param mode: Import/export mode
2861
  @param ieio: Input/output type
2862
  @param ieargs: Input/output arguments
2863

2864
  """
2865
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2866

    
2867
  env = None
2868
  prefix = None
2869
  suffix = None
2870
  exp_size = None
2871

    
2872
  if ieio == constants.IEIO_FILE:
2873
    (filename, ) = ieargs
2874

    
2875
    if not utils.IsNormAbsPath(filename):
2876
      _Fail("Path '%s' is not normalized or absolute", filename)
2877

    
2878
    real_filename = os.path.realpath(filename)
2879
    directory = os.path.dirname(real_filename)
2880

    
2881
    if utils.IsBelowDir(constants.EXPORT_DIR, real_filename):
2882
      _Fail("File '%s' is not under exports directory '%s': %s",
2883
            filename, constants.EXPORT_DIR, real_filename)
2884

    
2885
    # Create directory
2886
    utils.Makedirs(directory, mode=0750)
2887

    
2888
    quoted_filename = utils.ShellQuote(filename)
2889

    
2890
    if mode == constants.IEM_IMPORT:
2891
      suffix = "> %s" % quoted_filename
2892
    elif mode == constants.IEM_EXPORT:
2893
      suffix = "< %s" % quoted_filename
2894

    
2895
      # Retrieve file size
2896
      try:
2897
        st = os.stat(filename)
2898
      except EnvironmentError, err:
2899
        logging.error("Can't stat(2) %s: %s", filename, err)
2900
      else:
2901
        exp_size = utils.BytesToMebibyte(st.st_size)
2902

    
2903
  elif ieio == constants.IEIO_RAW_DISK:
2904
    (disk, ) = ieargs
2905

    
2906
    real_disk = _OpenRealBD(disk)
2907

    
2908
    if mode == constants.IEM_IMPORT:
2909
      # we set here a smaller block size as, due to transport buffering, more
2910
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
2911
      # is not already there or we pass a wrong path; we use notrunc to no
2912
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2913
      # much memory; this means that at best, we flush every 64k, which will
2914
      # not be very fast
2915
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2916
                                    " bs=%s oflag=dsync"),
2917
                                    real_disk.dev_path,
2918
                                    str(64 * 1024))
2919

    
2920
    elif mode == constants.IEM_EXPORT:
2921
      # the block size on the read dd is 1MiB to match our units
2922
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2923
                                   real_disk.dev_path,
2924
                                   str(1024 * 1024), # 1 MB
2925
                                   str(disk.size))
2926
      exp_size = disk.size
2927

    
2928
  elif ieio == constants.IEIO_SCRIPT:
2929
    (disk, disk_index, ) = ieargs
2930

    
2931
    assert isinstance(disk_index, (int, long))
2932

    
2933
    real_disk = _OpenRealBD(disk)
2934

    
2935
    inst_os = OSFromDisk(instance.os)
2936
    env = OSEnvironment(instance, inst_os)
2937

    
2938
    if mode == constants.IEM_IMPORT:
2939
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2940
      env["IMPORT_INDEX"] = str(disk_index)
2941
      script = inst_os.import_script
2942

    
2943
    elif mode == constants.IEM_EXPORT:
2944
      env["EXPORT_DEVICE"] = real_disk.dev_path
2945
      env["EXPORT_INDEX"] = str(disk_index)
2946
      script = inst_os.export_script
2947

    
2948
    # TODO: Pass special environment only to script
2949
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2950

    
2951
    if mode == constants.IEM_IMPORT:
2952
      suffix = "| %s" % script_cmd
2953

    
2954
    elif mode == constants.IEM_EXPORT:
2955
      prefix = "%s |" % script_cmd
2956

    
2957
    # Let script predict size
2958
    exp_size = constants.IE_CUSTOM_SIZE
2959

    
2960
  else:
2961
    _Fail("Invalid %s I/O mode %r", mode, ieio)
2962

    
2963
  return (env, prefix, suffix, exp_size)
2964

    
2965

    
2966
def _CreateImportExportStatusDir(prefix):
2967
  """Creates status directory for import/export.
2968

2969
  """
2970
  return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2971
                          prefix=("%s-%s-" %
2972
                                  (prefix, utils.TimestampForFilename())))
2973

    
2974

    
2975
def StartImportExportDaemon(mode, opts, host, port, instance, component,
2976
                            ieio, ieioargs):
2977
  """Starts an import or export daemon.
2978

2979
  @param mode: Import/output mode
2980
  @type opts: L{objects.ImportExportOptions}
2981
  @param opts: Daemon options
2982
  @type host: string
2983
  @param host: Remote host for export (None for import)
2984
  @type port: int
2985
  @param port: Remote port for export (None for import)
2986
  @type instance: L{objects.Instance}
2987
  @param instance: Instance object
2988
  @type component: string
2989
  @param component: which part of the instance is transferred now,
2990
      e.g. 'disk/0'
2991
  @param ieio: Input/output type
2992
  @param ieioargs: Input/output arguments
2993

2994
  """
2995
  if mode == constants.IEM_IMPORT:
2996
    prefix = "import"
2997

    
2998
    if not (host is None and port is None):
2999
      _Fail("Can not specify host or port on import")
3000

    
3001
  elif mode == constants.IEM_EXPORT:
3002
    prefix = "export"
3003

    
3004
    if host is None or port is None:
3005
      _Fail("Host and port must be specified for an export")
3006

    
3007
  else:
3008
    _Fail("Invalid mode %r", mode)
3009

    
3010
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3011
    _Fail("Cluster certificate can only be used for both key and CA")
3012

    
3013
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3014
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3015

    
3016
  if opts.key_name is None:
3017
    # Use server.pem
3018
    key_path = constants.NODED_CERT_FILE
3019
    cert_path = constants.NODED_CERT_FILE
3020
    assert opts.ca_pem is None
3021
  else:
3022
    (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
3023
                                                 opts.key_name)
3024
    assert opts.ca_pem is not None
3025

    
3026
  for i in [key_path, cert_path]:
3027
    if not os.path.exists(i):
3028
      _Fail("File '%s' does not exist" % i)
3029

    
3030
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3031
  try:
3032
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3033
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3034
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3035

    
3036
    if opts.ca_pem is None:
3037
      # Use server.pem
3038
      ca = utils.ReadFile(constants.NODED_CERT_FILE)
3039
    else:
3040
      ca = opts.ca_pem
3041

    
3042
    # Write CA file
3043
    utils.WriteFile(ca_file, data=ca, mode=0400)
3044

    
3045
    cmd = [
3046
      constants.IMPORT_EXPORT_DAEMON,
3047
      status_file, mode,
3048
      "--key=%s" % key_path,
3049
      "--cert=%s" % cert_path,
3050
      "--ca=%s" % ca_file,
3051
      ]
3052

    
3053
    if host:
3054
      cmd.append("--host=%s" % host)
3055

    
3056
    if port:
3057
      cmd.append("--port=%s" % port)
3058

    
3059
    if opts.ipv6:
3060
      cmd.append("--ipv6")
3061
    else:
3062
      cmd.append("--ipv4")
3063

    
3064
    if opts.compress:
3065
      cmd.append("--compress=%s" % opts.compress)
3066

    
3067
    if opts.magic:
3068
      cmd.append("--magic=%s" % opts.magic)
3069

    
3070
    if exp_size is not None:
3071
      cmd.append("--expected-size=%s" % exp_size)
3072

    
3073
    if cmd_prefix:
3074
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3075

    
3076
    if cmd_suffix:
3077
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3078

    
3079
    if mode == constants.IEM_EXPORT:
3080
      # Retry connection a few times when connecting to remote peer
3081
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3082
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3083
    elif opts.connect_timeout is not None:
3084
      assert mode == constants.IEM_IMPORT
3085
      # Overall timeout for establishing connection while listening
3086
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3087

    
3088
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3089

    
3090
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3091
    # support for receiving a file descriptor for output
3092
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3093
                      output=logfile)
3094

    
3095
    # The import/export name is simply the status directory name
3096
    return os.path.basename(status_dir)
3097

    
3098
  except Exception:
3099
    shutil.rmtree(status_dir, ignore_errors=True)
3100
    raise
3101

    
3102

    
3103
def GetImportExportStatus(names):
3104
  """Returns import/export daemon status.
3105

3106
  @type names: sequence
3107
  @param names: List of names
3108
  @rtype: List of dicts
3109
  @return: Returns a list of the state of each named import/export or None if a
3110
           status couldn't be read
3111

3112
  """
3113
  result = []
3114

    
3115
  for name in names:
3116
    status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3117
                                 _IES_STATUS_FILE)
3118

    
3119
    try:
3120
      data = utils.ReadFile(status_file)
3121
    except EnvironmentError, err:
3122
      if err.errno != errno.ENOENT:
3123
        raise
3124
      data = None
3125

    
3126
    if not data:
3127
      result.append(None)
3128
      continue
3129

    
3130
    result.append(serializer.LoadJson(data))
3131

    
3132
  return result
3133

    
3134

    
3135
def AbortImportExport(name):
3136
  """Sends SIGTERM to a running import/export daemon.
3137

3138
  """
3139
  logging.info("Abort import/export %s", name)
3140

    
3141
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3142
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3143

    
3144
  if pid:
3145
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3146
                 name, pid)
3147
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3148

    
3149

    
3150
def CleanupImportExport(name):
3151
  """Cleanup after an import or export.
3152

3153
  If the import/export daemon is still running it's killed. Afterwards the
3154
  whole status directory is removed.
3155

3156
  """
3157
  logging.info("Finalizing import/export %s", name)
3158

    
3159
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3160

    
3161
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3162

    
3163
  if pid:
3164
    logging.info("Import/export %s is still running with PID %s",
3165
                 name, pid)
3166
    utils.KillProcess(pid, waitpid=False)
3167

    
3168
  shutil.rmtree(status_dir, ignore_errors=True)
3169

    
3170

    
3171
def _FindDisks(nodes_ip, disks):
3172
  """Sets the physical ID on disks and returns the block devices.
3173

3174
  """
3175
  # set the correct physical ID
3176
  my_name = netutils.Hostname.GetSysName()
3177
  for cf in disks:
3178
    cf.SetPhysicalID(my_name, nodes_ip)
3179

    
3180
  bdevs = []
3181

    
3182
  for cf in disks:
3183
    rd = _RecursiveFindBD(cf)
3184
    if rd is None:
3185
      _Fail("Can't find device %s", cf)
3186
    bdevs.append(rd)
3187
  return bdevs
3188

    
3189

    
3190
def DrbdDisconnectNet(nodes_ip, disks):
3191
  """Disconnects the network on a list of drbd devices.
3192

3193
  """
3194
  bdevs = _FindDisks(nodes_ip, disks)
3195

    
3196
  # disconnect disks
3197
  for rd in bdevs:
3198
    try:
3199
      rd.DisconnectNet()
3200
    except errors.BlockDeviceError, err:
3201
      _Fail("Can't change network configuration to standalone mode: %s",
3202
            err, exc=True)
3203

    
3204

    
3205
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3206
  """Attaches the network on a list of drbd devices.
3207

3208
  """
3209
  bdevs = _FindDisks(nodes_ip, disks)
3210

    
3211
  if multimaster:
3212
    for idx, rd in enumerate(bdevs):
3213
      try:
3214
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3215
      except EnvironmentError, err:
3216
        _Fail("Can't create symlink: %s", err)
3217
  # reconnect disks, switch to new master configuration and if
3218
  # needed primary mode
3219
  for rd in bdevs:
3220
    try:
3221
      rd.AttachNet(multimaster)
3222
    except errors.BlockDeviceError, err:
3223
      _Fail("Can't change network configuration: %s", err)
3224

    
3225
  # wait until the disks are connected; we need to retry the re-attach
3226
  # if the device becomes standalone, as this might happen if the one
3227
  # node disconnects and reconnects in a different mode before the
3228
  # other node reconnects; in this case, one or both of the nodes will
3229
  # decide it has wrong configuration and switch to standalone
3230

    
3231
  def _Attach():
3232
    all_connected = True
3233

    
3234
    for rd in bdevs:
3235
      stats = rd.GetProcStatus()
3236

    
3237
      all_connected = (all_connected and
3238
                       (stats.is_connected or stats.is_in_resync))
3239

    
3240
      if stats.is_standalone:
3241
        # peer had different config info and this node became
3242
        # standalone, even though this should not happen with the
3243
        # new staged way of changing disk configs
3244
        try:
3245
          rd.AttachNet(multimaster)
3246
        except errors.BlockDeviceError, err:
3247
          _Fail("Can't change network configuration: %s", err)
3248

    
3249
    if not all_connected:
3250
      raise utils.RetryAgain()
3251

    
3252
  try:
3253
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3254
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3255
  except utils.RetryTimeout:
3256
    _Fail("Timeout in disk reconnecting")
3257

    
3258
  if multimaster:
3259
    # change to primary mode
3260
    for rd in bdevs:
3261
      try:
3262
        rd.Open()
3263
      except errors.BlockDeviceError, err:
3264
        _Fail("Can't change to primary mode: %s", err)
3265

    
3266

    
3267
def DrbdWaitSync(nodes_ip, disks):
3268
  """Wait until DRBDs have synchronized.
3269

3270
  """
3271
  def _helper(rd):
3272
    stats = rd.GetProcStatus()
3273
    if not (stats.is_connected or stats.is_in_resync):
3274
      raise utils.RetryAgain()
3275
    return stats
3276

    
3277
  bdevs = _FindDisks(nodes_ip, disks)
3278

    
3279
  min_resync = 100
3280
  alldone = True
3281
  for rd in bdevs:
3282
    try:
3283
      # poll each second for 15 seconds
3284
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3285
    except utils.RetryTimeout:
3286
      stats = rd.GetProcStatus()
3287
      # last check
3288
      if not (stats.is_connected or stats.is_in_resync):
3289
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3290
    alldone = alldone and (not stats.is_in_resync)
3291
    if stats.sync_percent is not None:
3292
      min_resync = min(min_resync, stats.sync_percent)
3293

    
3294
  return (alldone, min_resync)
3295

    
3296

    
3297
def GetDrbdUsermodeHelper():
3298
  """Returns DRBD usermode helper currently configured.
3299

3300
  """
3301
  try:
3302
    return bdev.BaseDRBD.GetUsermodeHelper()
3303
  except errors.BlockDeviceError, err:
3304
    _Fail(str(err))
3305

    
3306

    
3307
def PowercycleNode(hypervisor_type):
3308
  """Hard-powercycle the node.
3309

3310
  Because we need to return first, and schedule the powercycle in the
3311
  background, we won't be able to report failures nicely.
3312

3313
  """
3314
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3315
  try:
3316
    pid = os.fork()
3317
  except OSError:
3318
    # if we can't fork, we'll pretend that we're in the child process
3319
    pid = 0
3320
  if pid > 0:
3321
    return "Reboot scheduled in 5 seconds"
3322
  # ensure the child is running on ram
3323
  try:
3324
    utils.Mlockall()
3325
  except Exception: # pylint: disable=W0703
3326
    pass
3327
  time.sleep(5)
3328
  hyper.PowercycleNode()
3329

    
3330

    
3331
class HooksRunner(object):
3332
  """Hook runner.
3333

3334
  This class is instantiated on the node side (ganeti-noded) and not
3335
  on the master side.
3336

3337
  """
3338
  def __init__(self, hooks_base_dir=None):
3339
    """Constructor for hooks runner.
3340

3341
    @type hooks_base_dir: str or None
3342
    @param hooks_base_dir: if not None, this overrides the
3343
        L{constants.HOOKS_BASE_DIR} (useful for unittests)
3344

3345
    """
3346
    if hooks_base_dir is None:
3347
      hooks_base_dir = constants.HOOKS_BASE_DIR
3348
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
3349
    # constant
3350
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
3351

    
3352
  def RunHooks(self, hpath, phase, env):
3353
    """Run the scripts in the hooks directory.
3354

3355
    @type hpath: str
3356
    @param hpath: the path to the hooks directory which
3357
        holds the scripts
3358
    @type phase: str
3359
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
3360
        L{constants.HOOKS_PHASE_POST}
3361
    @type env: dict
3362
    @param env: dictionary with the environment for the hook
3363
    @rtype: list
3364
    @return: list of 3-element tuples:
3365
      - script path
3366
      - script result, either L{constants.HKR_SUCCESS} or
3367
        L{constants.HKR_FAIL}
3368
      - output of the script
3369

3370
    @raise errors.ProgrammerError: for invalid input
3371
        parameters
3372

3373
    """
3374
    if phase == constants.HOOKS_PHASE_PRE:
3375
      suffix = "pre"
3376
    elif phase == constants.HOOKS_PHASE_POST:
3377
      suffix = "post"
3378
    else:
3379
      _Fail("Unknown hooks phase '%s'", phase)
3380

    
3381
    subdir = "%s-%s.d" % (hpath, suffix)
3382
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3383

    
3384
    results = []
3385

    
3386
    if not os.path.isdir(dir_name):
3387
      # for non-existing/non-dirs, we simply exit instead of logging a
3388
      # warning at every operation
3389
      return results
3390

    
3391
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3392

    
3393
    for (relname, relstatus, runresult)  in runparts_results:
3394
      if relstatus == constants.RUNPARTS_SKIP:
3395
        rrval = constants.HKR_SKIP
3396
        output = ""
3397
      elif relstatus == constants.RUNPARTS_ERR:
3398
        rrval = constants.HKR_FAIL
3399
        output = "Hook script execution error: %s" % runresult
3400
      elif relstatus == constants.RUNPARTS_RUN:
3401
        if runresult.failed:
3402
          rrval = constants.HKR_FAIL
3403
        else:
3404
          rrval = constants.HKR_SUCCESS
3405
        output = utils.SafeEncode(runresult.output.strip())
3406
      results.append(("%s/%s" % (subdir, relname), rrval, output))
3407

    
3408
    return results
3409

    
3410

    
3411
class IAllocatorRunner(object):
3412
  """IAllocator runner.
3413

3414
  This class is instantiated on the node side (ganeti-noded) and not on
3415
  the master side.
3416

3417
  """
3418
  @staticmethod
3419
  def Run(name, idata):
3420
    """Run an iallocator script.
3421

3422
    @type name: str
3423
    @param name: the iallocator script name
3424
    @type idata: str
3425
    @param idata: the allocator input data
3426

3427
    @rtype: tuple
3428
    @return: two element tuple of:
3429
       - status
3430
       - either error message or stdout of allocator (for success)
3431

3432
    """
3433
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3434
                                  os.path.isfile)
3435
    if alloc_script is None:
3436
      _Fail("iallocator module '%s' not found in the search path", name)
3437

    
3438
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3439
    try:
3440
      os.write(fd, idata)
3441
      os.close(fd)
3442
      result = utils.RunCmd([alloc_script, fin_name])
3443
      if result.failed:
3444
        _Fail("iallocator module '%s' failed: %s, output '%s'",
3445
              name, result.fail_reason, result.output)
3446
    finally:
3447
      os.unlink(fin_name)
3448

    
3449
    return result.stdout
3450

    
3451

    
3452
class DevCacheManager(object):
3453
  """Simple class for managing a cache of block device information.
3454

3455
  """
3456
  _DEV_PREFIX = "/dev/"
3457
  _ROOT_DIR = constants.BDEV_CACHE_DIR
3458

    
3459
  @classmethod
3460
  def _ConvertPath(cls, dev_path):
3461
    """Converts a /dev/name path to the cache file name.
3462

3463
    This replaces slashes with underscores and strips the /dev
3464
    prefix. It then returns the full path to the cache file.
3465

3466
    @type dev_path: str
3467
    @param dev_path: the C{/dev/} path name
3468
    @rtype: str
3469
    @return: the converted path name
3470

3471
    """
3472
    if dev_path.startswith(cls._DEV_PREFIX):
3473
      dev_path = dev_path[len(cls._DEV_PREFIX):]
3474
    dev_path = dev_path.replace("/", "_")
3475
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3476
    return fpath
3477

    
3478
  @classmethod
3479
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3480
    """Updates the cache information for a given device.
3481

3482
    @type dev_path: str
3483
    @param dev_path: the pathname of the device
3484
    @type owner: str
3485
    @param owner: the owner (instance name) of the device
3486
    @type on_primary: bool
3487
    @param on_primary: whether this is the primary
3488
        node nor not
3489
    @type iv_name: str
3490
    @param iv_name: the instance-visible name of the
3491
        device, as in objects.Disk.iv_name
3492

3493
    @rtype: None
3494

3495
    """
3496
    if dev_path is None:
3497
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
3498
      return
3499
    fpath = cls._ConvertPath(dev_path)
3500
    if on_primary:
3501
      state = "primary"
3502
    else:
3503
      state = "secondary"
3504
    if iv_name is None:
3505
      iv_name = "not_visible"
3506
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3507
    try:
3508
      utils.WriteFile(fpath, data=fdata)
3509
    except EnvironmentError, err:
3510
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3511

    
3512
  @classmethod
3513
  def RemoveCache(cls, dev_path):
3514
    """Remove data for a dev_path.
3515

3516
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
3517
    path name and logging.
3518

3519
    @type dev_path: str
3520
    @param dev_path: the pathname of the device
3521

3522
    @rtype: None
3523

3524
    """
3525
    if dev_path is None:
3526
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
3527
      return
3528
    fpath = cls._ConvertPath(dev_path)
3529
    try:
3530
      utils.RemoveFile(fpath)
3531
    except EnvironmentError, err:
3532
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)