Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ fbb2c636

History | View | Annotate | Download (105.8 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-msg=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(["ip", "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(["ip", "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-msg=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 os.path.commonprefix([DEV_PREFIX, devpath]) != DEV_PREFIX:
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):
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

941
  """
942
  # TODO: Use tempfile.mkstemp to create unique filename
943
  base = ("%s-%s-%s-%s.log" %
944
          (kind, os_name, instance, utils.TimestampForFilename()))
945
  return utils.PathJoin(constants.LOG_OS_DIR, base)
946

    
947

    
948
def InstanceOsAdd(instance, reinstall, debug):
949
  """Add an OS to an instance.
950

951
  @type instance: L{objects.Instance}
952
  @param instance: Instance whose OS is to be installed
953
  @type reinstall: boolean
954
  @param reinstall: whether this is an instance reinstall
955
  @type debug: integer
956
  @param debug: debug level, passed to the OS scripts
957
  @rtype: None
958

959
  """
960
  inst_os = OSFromDisk(instance.os)
961

    
962
  create_env = OSEnvironment(instance, inst_os, debug)
963
  if reinstall:
964
    create_env['INSTANCE_REINSTALL'] = "1"
965

    
966
  logfile = _InstanceLogName("add", instance.os, instance.name)
967

    
968
  result = utils.RunCmd([inst_os.create_script], env=create_env,
969
                        cwd=inst_os.path, output=logfile,)
970
  if result.failed:
971
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
972
                  " output: %s", result.cmd, result.fail_reason, logfile,
973
                  result.output)
974
    lines = [utils.SafeEncode(val)
975
             for val in utils.TailFile(logfile, lines=20)]
976
    _Fail("OS create script failed (%s), last lines in the"
977
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
978

    
979

    
980
def RunRenameInstance(instance, old_name, debug):
981
  """Run the OS rename script for an instance.
982

983
  @type instance: L{objects.Instance}
984
  @param instance: Instance whose OS is to be installed
985
  @type old_name: string
986
  @param old_name: previous instance name
987
  @type debug: integer
988
  @param debug: debug level, passed to the OS scripts
989
  @rtype: boolean
990
  @return: the success of the operation
991

992
  """
993
  inst_os = OSFromDisk(instance.os)
994

    
995
  rename_env = OSEnvironment(instance, inst_os, debug)
996
  rename_env['OLD_INSTANCE_NAME'] = old_name
997

    
998
  logfile = _InstanceLogName("rename", instance.os,
999
                             "%s-%s" % (old_name, instance.name))
1000

    
1001
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1002
                        cwd=inst_os.path, output=logfile)
1003

    
1004
  if result.failed:
1005
    logging.error("os create command '%s' returned error: %s output: %s",
1006
                  result.cmd, result.fail_reason, result.output)
1007
    lines = [utils.SafeEncode(val)
1008
             for val in utils.TailFile(logfile, lines=20)]
1009
    _Fail("OS rename script failed (%s), last lines in the"
1010
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1011

    
1012

    
1013
def _GetBlockDevSymlinkPath(instance_name, idx):
1014
  return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1015
                        (instance_name, constants.DISK_SEPARATOR, idx))
1016

    
1017

    
1018
def _SymlinkBlockDev(instance_name, device_path, idx):
1019
  """Set up symlinks to a instance's block device.
1020

1021
  This is an auxiliary function run when an instance is start (on the primary
1022
  node) or when an instance is migrated (on the target node).
1023

1024

1025
  @param instance_name: the name of the target instance
1026
  @param device_path: path of the physical block device, on the node
1027
  @param idx: the disk index
1028
  @return: absolute path to the disk's symlink
1029

1030
  """
1031
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1032
  try:
1033
    os.symlink(device_path, link_name)
1034
  except OSError, err:
1035
    if err.errno == errno.EEXIST:
1036
      if (not os.path.islink(link_name) or
1037
          os.readlink(link_name) != device_path):
1038
        os.remove(link_name)
1039
        os.symlink(device_path, link_name)
1040
    else:
1041
      raise
1042

    
1043
  return link_name
1044

    
1045

    
1046
def _RemoveBlockDevLinks(instance_name, disks):
1047
  """Remove the block device symlinks belonging to the given instance.
1048

1049
  """
1050
  for idx, _ in enumerate(disks):
1051
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1052
    if os.path.islink(link_name):
1053
      try:
1054
        os.remove(link_name)
1055
      except OSError:
1056
        logging.exception("Can't remove symlink '%s'", link_name)
1057

    
1058

    
1059
def _GatherAndLinkBlockDevs(instance):
1060
  """Set up an instance's block device(s).
1061

1062
  This is run on the primary node at instance startup. The block
1063
  devices must be already assembled.
1064

1065
  @type instance: L{objects.Instance}
1066
  @param instance: the instance whose disks we shoul assemble
1067
  @rtype: list
1068
  @return: list of (disk_object, device_path)
1069

1070
  """
1071
  block_devices = []
1072
  for idx, disk in enumerate(instance.disks):
1073
    device = _RecursiveFindBD(disk)
1074
    if device is None:
1075
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1076
                                    str(disk))
1077
    device.Open()
1078
    try:
1079
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1080
    except OSError, e:
1081
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1082
                                    e.strerror)
1083

    
1084
    block_devices.append((disk, link_name))
1085

    
1086
  return block_devices
1087

    
1088

    
1089
def StartInstance(instance):
1090
  """Start an instance.
1091

1092
  @type instance: L{objects.Instance}
1093
  @param instance: the instance object
1094
  @rtype: None
1095

1096
  """
1097
  running_instances = GetInstanceList([instance.hypervisor])
1098

    
1099
  if instance.name in running_instances:
1100
    logging.info("Instance %s already running, not starting", instance.name)
1101
    return
1102

    
1103
  try:
1104
    block_devices = _GatherAndLinkBlockDevs(instance)
1105
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1106
    hyper.StartInstance(instance, block_devices)
1107
  except errors.BlockDeviceError, err:
1108
    _Fail("Block device error: %s", err, exc=True)
1109
  except errors.HypervisorError, err:
1110
    _RemoveBlockDevLinks(instance.name, instance.disks)
1111
    _Fail("Hypervisor error: %s", err, exc=True)
1112

    
1113

    
1114
def InstanceShutdown(instance, timeout):
1115
  """Shut an instance down.
1116

1117
  @note: this functions uses polling with a hardcoded timeout.
1118

1119
  @type instance: L{objects.Instance}
1120
  @param instance: the instance object
1121
  @type timeout: integer
1122
  @param timeout: maximum timeout for soft shutdown
1123
  @rtype: None
1124

1125
  """
1126
  hv_name = instance.hypervisor
1127
  hyper = hypervisor.GetHypervisor(hv_name)
1128
  iname = instance.name
1129

    
1130
  if instance.name not in hyper.ListInstances():
1131
    logging.info("Instance %s not running, doing nothing", iname)
1132
    return
1133

    
1134
  class _TryShutdown:
1135
    def __init__(self):
1136
      self.tried_once = False
1137

    
1138
    def __call__(self):
1139
      if iname not in hyper.ListInstances():
1140
        return
1141

    
1142
      try:
1143
        hyper.StopInstance(instance, retry=self.tried_once)
1144
      except errors.HypervisorError, err:
1145
        if iname not in hyper.ListInstances():
1146
          # if the instance is no longer existing, consider this a
1147
          # success and go to cleanup
1148
          return
1149

    
1150
        _Fail("Failed to stop instance %s: %s", iname, err)
1151

    
1152
      self.tried_once = True
1153

    
1154
      raise utils.RetryAgain()
1155

    
1156
  try:
1157
    utils.Retry(_TryShutdown(), 5, timeout)
1158
  except utils.RetryTimeout:
1159
    # the shutdown did not succeed
1160
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1161

    
1162
    try:
1163
      hyper.StopInstance(instance, force=True)
1164
    except errors.HypervisorError, err:
1165
      if iname in hyper.ListInstances():
1166
        # only raise an error if the instance still exists, otherwise
1167
        # the error could simply be "instance ... unknown"!
1168
        _Fail("Failed to force stop instance %s: %s", iname, err)
1169

    
1170
    time.sleep(1)
1171

    
1172
    if iname in hyper.ListInstances():
1173
      _Fail("Could not shutdown instance %s even by destroy", iname)
1174

    
1175
  try:
1176
    hyper.CleanupInstance(instance.name)
1177
  except errors.HypervisorError, err:
1178
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1179

    
1180
  _RemoveBlockDevLinks(iname, instance.disks)
1181

    
1182

    
1183
def InstanceReboot(instance, reboot_type, shutdown_timeout):
1184
  """Reboot an instance.
1185

1186
  @type instance: L{objects.Instance}
1187
  @param instance: the instance object to reboot
1188
  @type reboot_type: str
1189
  @param reboot_type: the type of reboot, one the following
1190
    constants:
1191
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1192
        instance OS, do not recreate the VM
1193
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1194
        restart the VM (at the hypervisor level)
1195
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1196
        not accepted here, since that mode is handled differently, in
1197
        cmdlib, and translates into full stop and start of the
1198
        instance (instead of a call_instance_reboot RPC)
1199
  @type shutdown_timeout: integer
1200
  @param shutdown_timeout: maximum timeout for soft shutdown
1201
  @rtype: None
1202

1203
  """
1204
  running_instances = GetInstanceList([instance.hypervisor])
1205

    
1206
  if instance.name not in running_instances:
1207
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1208

    
1209
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1210
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1211
    try:
1212
      hyper.RebootInstance(instance)
1213
    except errors.HypervisorError, err:
1214
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1215
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1216
    try:
1217
      InstanceShutdown(instance, shutdown_timeout)
1218
      return StartInstance(instance)
1219
    except errors.HypervisorError, err:
1220
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1221
  else:
1222
    _Fail("Invalid reboot_type received: %s", reboot_type)
1223

    
1224

    
1225
def MigrationInfo(instance):
1226
  """Gather information about an instance to be migrated.
1227

1228
  @type instance: L{objects.Instance}
1229
  @param instance: the instance definition
1230

1231
  """
1232
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1233
  try:
1234
    info = hyper.MigrationInfo(instance)
1235
  except errors.HypervisorError, err:
1236
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1237
  return info
1238

    
1239

    
1240
def AcceptInstance(instance, info, target):
1241
  """Prepare the node to accept an instance.
1242

1243
  @type instance: L{objects.Instance}
1244
  @param instance: the instance definition
1245
  @type info: string/data (opaque)
1246
  @param info: migration information, from the source node
1247
  @type target: string
1248
  @param target: target host (usually ip), on this node
1249

1250
  """
1251
  # TODO: why is this required only for DTS_EXT_MIRROR?
1252
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1253
    # Create the symlinks, as the disks are not active
1254
    # in any way
1255
    try:
1256
      _GatherAndLinkBlockDevs(instance)
1257
    except errors.BlockDeviceError, err:
1258
      _Fail("Block device error: %s", err, exc=True)
1259

    
1260
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1261
  try:
1262
    hyper.AcceptInstance(instance, info, target)
1263
  except errors.HypervisorError, err:
1264
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1265
      _RemoveBlockDevLinks(instance.name, instance.disks)
1266
    _Fail("Failed to accept instance: %s", err, exc=True)
1267

    
1268

    
1269
def FinalizeMigration(instance, info, success):
1270
  """Finalize any preparation to accept an instance.
1271

1272
  @type instance: L{objects.Instance}
1273
  @param instance: the instance definition
1274
  @type info: string/data (opaque)
1275
  @param info: migration information, from the source node
1276
  @type success: boolean
1277
  @param success: whether the migration was a success or a failure
1278

1279
  """
1280
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1281
  try:
1282
    hyper.FinalizeMigration(instance, info, success)
1283
  except errors.HypervisorError, err:
1284
    _Fail("Failed to finalize migration: %s", err, exc=True)
1285

    
1286

    
1287
def MigrateInstance(instance, target, live):
1288
  """Migrates an instance to another node.
1289

1290
  @type instance: L{objects.Instance}
1291
  @param instance: the instance definition
1292
  @type target: string
1293
  @param target: the target node name
1294
  @type live: boolean
1295
  @param live: whether the migration should be done live or not (the
1296
      interpretation of this parameter is left to the hypervisor)
1297
  @rtype: tuple
1298
  @return: a tuple of (success, msg) where:
1299
      - succes is a boolean denoting the success/failure of the operation
1300
      - msg is a string with details in case of failure
1301

1302
  """
1303
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1304

    
1305
  try:
1306
    hyper.MigrateInstance(instance, target, live)
1307
  except errors.HypervisorError, err:
1308
    _Fail("Failed to migrate instance: %s", err, exc=True)
1309

    
1310

    
1311
def BlockdevCreate(disk, size, owner, on_primary, info):
1312
  """Creates a block device for an instance.
1313

1314
  @type disk: L{objects.Disk}
1315
  @param disk: the object describing the disk we should create
1316
  @type size: int
1317
  @param size: the size of the physical underlying device, in MiB
1318
  @type owner: str
1319
  @param owner: the name of the instance for which disk is created,
1320
      used for device cache data
1321
  @type on_primary: boolean
1322
  @param on_primary:  indicates if it is the primary node or not
1323
  @type info: string
1324
  @param info: string that will be sent to the physical device
1325
      creation, used for example to set (LVM) tags on LVs
1326

1327
  @return: the new unique_id of the device (this can sometime be
1328
      computed only after creation), or None. On secondary nodes,
1329
      it's not required to return anything.
1330

1331
  """
1332
  # TODO: remove the obsolete 'size' argument
1333
  # pylint: disable-msg=W0613
1334
  clist = []
1335
  if disk.children:
1336
    for child in disk.children:
1337
      try:
1338
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1339
      except errors.BlockDeviceError, err:
1340
        _Fail("Can't assemble device %s: %s", child, err)
1341
      if on_primary or disk.AssembleOnSecondary():
1342
        # we need the children open in case the device itself has to
1343
        # be assembled
1344
        try:
1345
          # pylint: disable-msg=E1103
1346
          crdev.Open()
1347
        except errors.BlockDeviceError, err:
1348
          _Fail("Can't make child '%s' read-write: %s", child, err)
1349
      clist.append(crdev)
1350

    
1351
  try:
1352
    device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1353
  except errors.BlockDeviceError, err:
1354
    _Fail("Can't create block device: %s", err)
1355

    
1356
  if on_primary or disk.AssembleOnSecondary():
1357
    try:
1358
      device.Assemble()
1359
    except errors.BlockDeviceError, err:
1360
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1361
    device.SetSyncSpeed(constants.SYNC_SPEED)
1362
    if on_primary or disk.OpenOnSecondary():
1363
      try:
1364
        device.Open(force=True)
1365
      except errors.BlockDeviceError, err:
1366
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1367
    DevCacheManager.UpdateCache(device.dev_path, owner,
1368
                                on_primary, disk.iv_name)
1369

    
1370
  device.SetInfo(info)
1371

    
1372
  return device.unique_id
1373

    
1374

    
1375
def _WipeDevice(path, offset, size):
1376
  """This function actually wipes the device.
1377

1378
  @param path: The path to the device to wipe
1379
  @param offset: The offset in MiB in the file
1380
  @param size: The size in MiB to write
1381

1382
  """
1383
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1384
         "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1385
         "count=%d" % size]
1386
  result = utils.RunCmd(cmd)
1387

    
1388
  if result.failed:
1389
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1390
          result.fail_reason, result.output)
1391

    
1392

    
1393
def BlockdevWipe(disk, offset, size):
1394
  """Wipes a block device.
1395

1396
  @type disk: L{objects.Disk}
1397
  @param disk: the disk object we want to wipe
1398
  @type offset: int
1399
  @param offset: The offset in MiB in the file
1400
  @type size: int
1401
  @param size: The size in MiB to write
1402

1403
  """
1404
  try:
1405
    rdev = _RecursiveFindBD(disk)
1406
  except errors.BlockDeviceError:
1407
    rdev = None
1408

    
1409
  if not rdev:
1410
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1411

    
1412
  # Do cross verify some of the parameters
1413
  if offset > rdev.size:
1414
    _Fail("Offset is bigger than device size")
1415
  if (offset + size) > rdev.size:
1416
    _Fail("The provided offset and size to wipe is bigger than device size")
1417

    
1418
  _WipeDevice(rdev.dev_path, offset, size)
1419

    
1420

    
1421
def BlockdevPauseResumeSync(disks, pause):
1422
  """Pause or resume the sync of the block device.
1423

1424
  @type disks: list of L{objects.Disk}
1425
  @param disks: the disks object we want to pause/resume
1426
  @type pause: bool
1427
  @param pause: Wheater to pause or resume
1428

1429
  """
1430
  success = []
1431
  for disk in disks:
1432
    try:
1433
      rdev = _RecursiveFindBD(disk)
1434
    except errors.BlockDeviceError:
1435
      rdev = None
1436

    
1437
    if not rdev:
1438
      success.append((False, ("Cannot change sync for device %s:"
1439
                              " device not found" % disk.iv_name)))
1440
      continue
1441

    
1442
    result = rdev.PauseResumeSync(pause)
1443

    
1444
    if result:
1445
      success.append((result, None))
1446
    else:
1447
      if pause:
1448
        msg = "Pause"
1449
      else:
1450
        msg = "Resume"
1451
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1452

    
1453
  return success
1454

    
1455

    
1456
def BlockdevRemove(disk):
1457
  """Remove a block device.
1458

1459
  @note: This is intended to be called recursively.
1460

1461
  @type disk: L{objects.Disk}
1462
  @param disk: the disk object we should remove
1463
  @rtype: boolean
1464
  @return: the success of the operation
1465

1466
  """
1467
  msgs = []
1468
  try:
1469
    rdev = _RecursiveFindBD(disk)
1470
  except errors.BlockDeviceError, err:
1471
    # probably can't attach
1472
    logging.info("Can't attach to device %s in remove", disk)
1473
    rdev = None
1474
  if rdev is not None:
1475
    r_path = rdev.dev_path
1476
    try:
1477
      rdev.Remove()
1478
    except errors.BlockDeviceError, err:
1479
      msgs.append(str(err))
1480
    if not msgs:
1481
      DevCacheManager.RemoveCache(r_path)
1482

    
1483
  if disk.children:
1484
    for child in disk.children:
1485
      try:
1486
        BlockdevRemove(child)
1487
      except RPCFail, err:
1488
        msgs.append(str(err))
1489

    
1490
  if msgs:
1491
    _Fail("; ".join(msgs))
1492

    
1493

    
1494
def _RecursiveAssembleBD(disk, owner, as_primary):
1495
  """Activate a block device for an instance.
1496

1497
  This is run on the primary and secondary nodes for an instance.
1498

1499
  @note: this function is called recursively.
1500

1501
  @type disk: L{objects.Disk}
1502
  @param disk: the disk we try to assemble
1503
  @type owner: str
1504
  @param owner: the name of the instance which owns the disk
1505
  @type as_primary: boolean
1506
  @param as_primary: if we should make the block device
1507
      read/write
1508

1509
  @return: the assembled device or None (in case no device
1510
      was assembled)
1511
  @raise errors.BlockDeviceError: in case there is an error
1512
      during the activation of the children or the device
1513
      itself
1514

1515
  """
1516
  children = []
1517
  if disk.children:
1518
    mcn = disk.ChildrenNeeded()
1519
    if mcn == -1:
1520
      mcn = 0 # max number of Nones allowed
1521
    else:
1522
      mcn = len(disk.children) - mcn # max number of Nones
1523
    for chld_disk in disk.children:
1524
      try:
1525
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1526
      except errors.BlockDeviceError, err:
1527
        if children.count(None) >= mcn:
1528
          raise
1529
        cdev = None
1530
        logging.error("Error in child activation (but continuing): %s",
1531
                      str(err))
1532
      children.append(cdev)
1533

    
1534
  if as_primary or disk.AssembleOnSecondary():
1535
    r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1536
    r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1537
    result = r_dev
1538
    if as_primary or disk.OpenOnSecondary():
1539
      r_dev.Open()
1540
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1541
                                as_primary, disk.iv_name)
1542

    
1543
  else:
1544
    result = True
1545
  return result
1546

    
1547

    
1548
def BlockdevAssemble(disk, owner, as_primary, idx):
1549
  """Activate a block device for an instance.
1550

1551
  This is a wrapper over _RecursiveAssembleBD.
1552

1553
  @rtype: str or boolean
1554
  @return: a C{/dev/...} path for primary nodes, and
1555
      C{True} for secondary nodes
1556

1557
  """
1558
  try:
1559
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1560
    if isinstance(result, bdev.BlockDev):
1561
      # pylint: disable-msg=E1103
1562
      result = result.dev_path
1563
      if as_primary:
1564
        _SymlinkBlockDev(owner, result, idx)
1565
  except errors.BlockDeviceError, err:
1566
    _Fail("Error while assembling disk: %s", err, exc=True)
1567
  except OSError, err:
1568
    _Fail("Error while symlinking disk: %s", err, exc=True)
1569

    
1570
  return result
1571

    
1572

    
1573
def BlockdevShutdown(disk):
1574
  """Shut down a block device.
1575

1576
  First, if the device is assembled (Attach() is successful), then
1577
  the device is shutdown. Then the children of the device are
1578
  shutdown.
1579

1580
  This function is called recursively. Note that we don't cache the
1581
  children or such, as oppossed to assemble, shutdown of different
1582
  devices doesn't require that the upper device was active.
1583

1584
  @type disk: L{objects.Disk}
1585
  @param disk: the description of the disk we should
1586
      shutdown
1587
  @rtype: None
1588

1589
  """
1590
  msgs = []
1591
  r_dev = _RecursiveFindBD(disk)
1592
  if r_dev is not None:
1593
    r_path = r_dev.dev_path
1594
    try:
1595
      r_dev.Shutdown()
1596
      DevCacheManager.RemoveCache(r_path)
1597
    except errors.BlockDeviceError, err:
1598
      msgs.append(str(err))
1599

    
1600
  if disk.children:
1601
    for child in disk.children:
1602
      try:
1603
        BlockdevShutdown(child)
1604
      except RPCFail, err:
1605
        msgs.append(str(err))
1606

    
1607
  if msgs:
1608
    _Fail("; ".join(msgs))
1609

    
1610

    
1611
def BlockdevAddchildren(parent_cdev, new_cdevs):
1612
  """Extend a mirrored block device.
1613

1614
  @type parent_cdev: L{objects.Disk}
1615
  @param parent_cdev: the disk to which we should add children
1616
  @type new_cdevs: list of L{objects.Disk}
1617
  @param new_cdevs: the list of children which we should add
1618
  @rtype: None
1619

1620
  """
1621
  parent_bdev = _RecursiveFindBD(parent_cdev)
1622
  if parent_bdev is None:
1623
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1624
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1625
  if new_bdevs.count(None) > 0:
1626
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1627
  parent_bdev.AddChildren(new_bdevs)
1628

    
1629

    
1630
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1631
  """Shrink a mirrored block device.
1632

1633
  @type parent_cdev: L{objects.Disk}
1634
  @param parent_cdev: the disk from which we should remove children
1635
  @type new_cdevs: list of L{objects.Disk}
1636
  @param new_cdevs: the list of children which we should remove
1637
  @rtype: None
1638

1639
  """
1640
  parent_bdev = _RecursiveFindBD(parent_cdev)
1641
  if parent_bdev is None:
1642
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1643
  devs = []
1644
  for disk in new_cdevs:
1645
    rpath = disk.StaticDevPath()
1646
    if rpath is None:
1647
      bd = _RecursiveFindBD(disk)
1648
      if bd is None:
1649
        _Fail("Can't find device %s while removing children", disk)
1650
      else:
1651
        devs.append(bd.dev_path)
1652
    else:
1653
      if not utils.IsNormAbsPath(rpath):
1654
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1655
      devs.append(rpath)
1656
  parent_bdev.RemoveChildren(devs)
1657

    
1658

    
1659
def BlockdevGetmirrorstatus(disks):
1660
  """Get the mirroring status of a list of devices.
1661

1662
  @type disks: list of L{objects.Disk}
1663
  @param disks: the list of disks which we should query
1664
  @rtype: disk
1665
  @return: List of L{objects.BlockDevStatus}, one for each disk
1666
  @raise errors.BlockDeviceError: if any of the disks cannot be
1667
      found
1668

1669
  """
1670
  stats = []
1671
  for dsk in disks:
1672
    rbd = _RecursiveFindBD(dsk)
1673
    if rbd is None:
1674
      _Fail("Can't find device %s", dsk)
1675

    
1676
    stats.append(rbd.CombinedSyncStatus())
1677

    
1678
  return stats
1679

    
1680

    
1681
def BlockdevGetmirrorstatusMulti(disks):
1682
  """Get the mirroring status of a list of devices.
1683

1684
  @type disks: list of L{objects.Disk}
1685
  @param disks: the list of disks which we should query
1686
  @rtype: disk
1687
  @return: List of tuples, (bool, status), one for each disk; bool denotes
1688
    success/failure, status is L{objects.BlockDevStatus} on success, string
1689
    otherwise
1690

1691
  """
1692
  result = []
1693
  for disk in disks:
1694
    try:
1695
      rbd = _RecursiveFindBD(disk)
1696
      if rbd is None:
1697
        result.append((False, "Can't find device %s" % disk))
1698
        continue
1699

    
1700
      status = rbd.CombinedSyncStatus()
1701
    except errors.BlockDeviceError, err:
1702
      logging.exception("Error while getting disk status")
1703
      result.append((False, str(err)))
1704
    else:
1705
      result.append((True, status))
1706

    
1707
  assert len(disks) == len(result)
1708

    
1709
  return result
1710

    
1711

    
1712
def _RecursiveFindBD(disk):
1713
  """Check if a device is activated.
1714

1715
  If so, return information about the real device.
1716

1717
  @type disk: L{objects.Disk}
1718
  @param disk: the disk object we need to find
1719

1720
  @return: None if the device can't be found,
1721
      otherwise the device instance
1722

1723
  """
1724
  children = []
1725
  if disk.children:
1726
    for chdisk in disk.children:
1727
      children.append(_RecursiveFindBD(chdisk))
1728

    
1729
  return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1730

    
1731

    
1732
def _OpenRealBD(disk):
1733
  """Opens the underlying block device of a disk.
1734

1735
  @type disk: L{objects.Disk}
1736
  @param disk: the disk object we want to open
1737

1738
  """
1739
  real_disk = _RecursiveFindBD(disk)
1740
  if real_disk is None:
1741
    _Fail("Block device '%s' is not set up", disk)
1742

    
1743
  real_disk.Open()
1744

    
1745
  return real_disk
1746

    
1747

    
1748
def BlockdevFind(disk):
1749
  """Check if a device is activated.
1750

1751
  If it is, return information about the real device.
1752

1753
  @type disk: L{objects.Disk}
1754
  @param disk: the disk to find
1755
  @rtype: None or objects.BlockDevStatus
1756
  @return: None if the disk cannot be found, otherwise a the current
1757
           information
1758

1759
  """
1760
  try:
1761
    rbd = _RecursiveFindBD(disk)
1762
  except errors.BlockDeviceError, err:
1763
    _Fail("Failed to find device: %s", err, exc=True)
1764

    
1765
  if rbd is None:
1766
    return None
1767

    
1768
  return rbd.GetSyncStatus()
1769

    
1770

    
1771
def BlockdevGetsize(disks):
1772
  """Computes the size of the given disks.
1773

1774
  If a disk is not found, returns None instead.
1775

1776
  @type disks: list of L{objects.Disk}
1777
  @param disks: the list of disk to compute the size for
1778
  @rtype: list
1779
  @return: list with elements None if the disk cannot be found,
1780
      otherwise the size
1781

1782
  """
1783
  result = []
1784
  for cf in disks:
1785
    try:
1786
      rbd = _RecursiveFindBD(cf)
1787
    except errors.BlockDeviceError:
1788
      result.append(None)
1789
      continue
1790
    if rbd is None:
1791
      result.append(None)
1792
    else:
1793
      result.append(rbd.GetActualSize())
1794
  return result
1795

    
1796

    
1797
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1798
  """Export a block device to a remote node.
1799

1800
  @type disk: L{objects.Disk}
1801
  @param disk: the description of the disk to export
1802
  @type dest_node: str
1803
  @param dest_node: the destination node to export to
1804
  @type dest_path: str
1805
  @param dest_path: the destination path on the target node
1806
  @type cluster_name: str
1807
  @param cluster_name: the cluster name, needed for SSH hostalias
1808
  @rtype: None
1809

1810
  """
1811
  real_disk = _OpenRealBD(disk)
1812

    
1813
  # the block size on the read dd is 1MiB to match our units
1814
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1815
                               "dd if=%s bs=1048576 count=%s",
1816
                               real_disk.dev_path, str(disk.size))
1817

    
1818
  # we set here a smaller block size as, due to ssh buffering, more
1819
  # than 64-128k will mostly ignored; we use nocreat to fail if the
1820
  # device is not already there or we pass a wrong path; we use
1821
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1822
  # to not buffer too much memory; this means that at best, we flush
1823
  # every 64k, which will not be very fast
1824
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1825
                                " oflag=dsync", dest_path)
1826

    
1827
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1828
                                                   constants.GANETI_RUNAS,
1829
                                                   destcmd)
1830

    
1831
  # all commands have been checked, so we're safe to combine them
1832
  command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1833

    
1834
  result = utils.RunCmd(["bash", "-c", command])
1835

    
1836
  if result.failed:
1837
    _Fail("Disk copy command '%s' returned error: %s"
1838
          " output: %s", command, result.fail_reason, result.output)
1839

    
1840

    
1841
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1842
  """Write a file to the filesystem.
1843

1844
  This allows the master to overwrite(!) a file. It will only perform
1845
  the operation if the file belongs to a list of configuration files.
1846

1847
  @type file_name: str
1848
  @param file_name: the target file name
1849
  @type data: str
1850
  @param data: the new contents of the file
1851
  @type mode: int
1852
  @param mode: the mode to give the file (can be None)
1853
  @type uid: string
1854
  @param uid: the owner of the file
1855
  @type gid: string
1856
  @param gid: the group of the file
1857
  @type atime: float
1858
  @param atime: the atime to set on the file (can be None)
1859
  @type mtime: float
1860
  @param mtime: the mtime to set on the file (can be None)
1861
  @rtype: None
1862

1863
  """
1864
  if not os.path.isabs(file_name):
1865
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1866

    
1867
  if file_name not in _ALLOWED_UPLOAD_FILES:
1868
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1869
          file_name)
1870

    
1871
  raw_data = _Decompress(data)
1872

    
1873
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
1874
    _Fail("Invalid username/groupname type")
1875

    
1876
  getents = runtime.GetEnts()
1877
  uid = getents.LookupUser(uid)
1878
  gid = getents.LookupGroup(gid)
1879

    
1880
  utils.SafeWriteFile(file_name, None,
1881
                      data=raw_data, mode=mode, uid=uid, gid=gid,
1882
                      atime=atime, mtime=mtime)
1883

    
1884

    
1885
def RunOob(oob_program, command, node, timeout):
1886
  """Executes oob_program with given command on given node.
1887

1888
  @param oob_program: The path to the executable oob_program
1889
  @param command: The command to invoke on oob_program
1890
  @param node: The node given as an argument to the program
1891
  @param timeout: Timeout after which we kill the oob program
1892

1893
  @return: stdout
1894
  @raise RPCFail: If execution fails for some reason
1895

1896
  """
1897
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
1898

    
1899
  if result.failed:
1900
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
1901
          result.fail_reason, result.output)
1902

    
1903
  return result.stdout
1904

    
1905

    
1906
def WriteSsconfFiles(values):
1907
  """Update all ssconf files.
1908

1909
  Wrapper around the SimpleStore.WriteFiles.
1910

1911
  """
1912
  ssconf.SimpleStore().WriteFiles(values)
1913

    
1914

    
1915
def _ErrnoOrStr(err):
1916
  """Format an EnvironmentError exception.
1917

1918
  If the L{err} argument has an errno attribute, it will be looked up
1919
  and converted into a textual C{E...} description. Otherwise the
1920
  string representation of the error will be returned.
1921

1922
  @type err: L{EnvironmentError}
1923
  @param err: the exception to format
1924

1925
  """
1926
  if hasattr(err, 'errno'):
1927
    detail = errno.errorcode[err.errno]
1928
  else:
1929
    detail = str(err)
1930
  return detail
1931

    
1932

    
1933
def _OSOndiskAPIVersion(os_dir):
1934
  """Compute and return the API version of a given OS.
1935

1936
  This function will try to read the API version of the OS residing in
1937
  the 'os_dir' directory.
1938

1939
  @type os_dir: str
1940
  @param os_dir: the directory in which we should look for the OS
1941
  @rtype: tuple
1942
  @return: tuple (status, data) with status denoting the validity and
1943
      data holding either the vaid versions or an error message
1944

1945
  """
1946
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1947

    
1948
  try:
1949
    st = os.stat(api_file)
1950
  except EnvironmentError, err:
1951
    return False, ("Required file '%s' not found under path %s: %s" %
1952
                   (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1953

    
1954
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1955
    return False, ("File '%s' in %s is not a regular file" %
1956
                   (constants.OS_API_FILE, os_dir))
1957

    
1958
  try:
1959
    api_versions = utils.ReadFile(api_file).splitlines()
1960
  except EnvironmentError, err:
1961
    return False, ("Error while reading the API version file at %s: %s" %
1962
                   (api_file, _ErrnoOrStr(err)))
1963

    
1964
  try:
1965
    api_versions = [int(version.strip()) for version in api_versions]
1966
  except (TypeError, ValueError), err:
1967
    return False, ("API version(s) can't be converted to integer: %s" %
1968
                   str(err))
1969

    
1970
  return True, api_versions
1971

    
1972

    
1973
def DiagnoseOS(top_dirs=None):
1974
  """Compute the validity for all OSes.
1975

1976
  @type top_dirs: list
1977
  @param top_dirs: the list of directories in which to
1978
      search (if not given defaults to
1979
      L{constants.OS_SEARCH_PATH})
1980
  @rtype: list of L{objects.OS}
1981
  @return: a list of tuples (name, path, status, diagnose, variants,
1982
      parameters, api_version) for all (potential) OSes under all
1983
      search paths, where:
1984
          - name is the (potential) OS name
1985
          - path is the full path to the OS
1986
          - status True/False is the validity of the OS
1987
          - diagnose is the error message for an invalid OS, otherwise empty
1988
          - variants is a list of supported OS variants, if any
1989
          - parameters is a list of (name, help) parameters, if any
1990
          - api_version is a list of support OS API versions
1991

1992
  """
1993
  if top_dirs is None:
1994
    top_dirs = constants.OS_SEARCH_PATH
1995

    
1996
  result = []
1997
  for dir_name in top_dirs:
1998
    if os.path.isdir(dir_name):
1999
      try:
2000
        f_names = utils.ListVisibleFiles(dir_name)
2001
      except EnvironmentError, err:
2002
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2003
        break
2004
      for name in f_names:
2005
        os_path = utils.PathJoin(dir_name, name)
2006
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2007
        if status:
2008
          diagnose = ""
2009
          variants = os_inst.supported_variants
2010
          parameters = os_inst.supported_parameters
2011
          api_versions = os_inst.api_versions
2012
        else:
2013
          diagnose = os_inst
2014
          variants = parameters = api_versions = []
2015
        result.append((name, os_path, status, diagnose, variants,
2016
                       parameters, api_versions))
2017

    
2018
  return result
2019

    
2020

    
2021
def _TryOSFromDisk(name, base_dir=None):
2022
  """Create an OS instance from disk.
2023

2024
  This function will return an OS instance if the given name is a
2025
  valid OS name.
2026

2027
  @type base_dir: string
2028
  @keyword base_dir: Base directory containing OS installations.
2029
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2030
  @rtype: tuple
2031
  @return: success and either the OS instance if we find a valid one,
2032
      or error message
2033

2034
  """
2035
  if base_dir is None:
2036
    os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2037
  else:
2038
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2039

    
2040
  if os_dir is None:
2041
    return False, "Directory for OS %s not found in search path" % name
2042

    
2043
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2044
  if not status:
2045
    # push the error up
2046
    return status, api_versions
2047

    
2048
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2049
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2050
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2051

    
2052
  # OS Files dictionary, we will populate it with the absolute path names
2053
  os_files = dict.fromkeys(constants.OS_SCRIPTS)
2054

    
2055
  if max(api_versions) >= constants.OS_API_V15:
2056
    os_files[constants.OS_VARIANTS_FILE] = ''
2057

    
2058
  if max(api_versions) >= constants.OS_API_V20:
2059
    os_files[constants.OS_PARAMETERS_FILE] = ''
2060
  else:
2061
    del os_files[constants.OS_SCRIPT_VERIFY]
2062

    
2063
  for filename in os_files:
2064
    os_files[filename] = utils.PathJoin(os_dir, filename)
2065

    
2066
    try:
2067
      st = os.stat(os_files[filename])
2068
    except EnvironmentError, err:
2069
      return False, ("File '%s' under path '%s' is missing (%s)" %
2070
                     (filename, os_dir, _ErrnoOrStr(err)))
2071

    
2072
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2073
      return False, ("File '%s' under path '%s' is not a regular file" %
2074
                     (filename, os_dir))
2075

    
2076
    if filename in constants.OS_SCRIPTS:
2077
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2078
        return False, ("File '%s' under path '%s' is not executable" %
2079
                       (filename, os_dir))
2080

    
2081
  variants = []
2082
  if constants.OS_VARIANTS_FILE in os_files:
2083
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2084
    try:
2085
      variants = utils.ReadFile(variants_file).splitlines()
2086
    except EnvironmentError, err:
2087
      return False, ("Error while reading the OS variants file at %s: %s" %
2088
                     (variants_file, _ErrnoOrStr(err)))
2089
    if not variants:
2090
      return False, ("No supported os variant found")
2091

    
2092
  parameters = []
2093
  if constants.OS_PARAMETERS_FILE in os_files:
2094
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2095
    try:
2096
      parameters = utils.ReadFile(parameters_file).splitlines()
2097
    except EnvironmentError, err:
2098
      return False, ("Error while reading the OS parameters file at %s: %s" %
2099
                     (parameters_file, _ErrnoOrStr(err)))
2100
    parameters = [v.split(None, 1) for v in parameters]
2101

    
2102
  os_obj = objects.OS(name=name, path=os_dir,
2103
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2104
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2105
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2106
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2107
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2108
                                                 None),
2109
                      supported_variants=variants,
2110
                      supported_parameters=parameters,
2111
                      api_versions=api_versions)
2112
  return True, os_obj
2113

    
2114

    
2115
def OSFromDisk(name, base_dir=None):
2116
  """Create an OS instance from disk.
2117

2118
  This function will return an OS instance if the given name is a
2119
  valid OS name. Otherwise, it will raise an appropriate
2120
  L{RPCFail} exception, detailing why this is not a valid OS.
2121

2122
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2123
  an exception but returns true/false status data.
2124

2125
  @type base_dir: string
2126
  @keyword base_dir: Base directory containing OS installations.
2127
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2128
  @rtype: L{objects.OS}
2129
  @return: the OS instance if we find a valid one
2130
  @raise RPCFail: if we don't find a valid OS
2131

2132
  """
2133
  name_only = objects.OS.GetName(name)
2134
  status, payload = _TryOSFromDisk(name_only, base_dir)
2135

    
2136
  if not status:
2137
    _Fail(payload)
2138

    
2139
  return payload
2140

    
2141

    
2142
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2143
  """Calculate the basic environment for an os script.
2144

2145
  @type os_name: str
2146
  @param os_name: full operating system name (including variant)
2147
  @type inst_os: L{objects.OS}
2148
  @param inst_os: operating system for which the environment is being built
2149
  @type os_params: dict
2150
  @param os_params: the OS parameters
2151
  @type debug: integer
2152
  @param debug: debug level (0 or 1, for OS Api 10)
2153
  @rtype: dict
2154
  @return: dict of environment variables
2155
  @raise errors.BlockDeviceError: if the block device
2156
      cannot be found
2157

2158
  """
2159
  result = {}
2160
  api_version = \
2161
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2162
  result['OS_API_VERSION'] = '%d' % api_version
2163
  result['OS_NAME'] = inst_os.name
2164
  result['DEBUG_LEVEL'] = '%d' % debug
2165

    
2166
  # OS variants
2167
  if api_version >= constants.OS_API_V15:
2168
    variant = objects.OS.GetVariant(os_name)
2169
    if not variant:
2170
      variant = inst_os.supported_variants[0]
2171
    result['OS_VARIANT'] = variant
2172

    
2173
  # OS params
2174
  for pname, pvalue in os_params.items():
2175
    result['OSP_%s' % pname.upper()] = pvalue
2176

    
2177
  return result
2178

    
2179

    
2180
def OSEnvironment(instance, inst_os, debug=0):
2181
  """Calculate the environment for an os script.
2182

2183
  @type instance: L{objects.Instance}
2184
  @param instance: target instance for the os script run
2185
  @type inst_os: L{objects.OS}
2186
  @param inst_os: operating system for which the environment is being built
2187
  @type debug: integer
2188
  @param debug: debug level (0 or 1, for OS Api 10)
2189
  @rtype: dict
2190
  @return: dict of environment variables
2191
  @raise errors.BlockDeviceError: if the block device
2192
      cannot be found
2193

2194
  """
2195
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2196

    
2197
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2198
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2199

    
2200
  result['HYPERVISOR'] = instance.hypervisor
2201
  result['DISK_COUNT'] = '%d' % len(instance.disks)
2202
  result['NIC_COUNT'] = '%d' % len(instance.nics)
2203
  result['INSTANCE_SECONDARY_NODES'] = \
2204
      ('%s' % " ".join(instance.secondary_nodes))
2205

    
2206
  # Disks
2207
  for idx, disk in enumerate(instance.disks):
2208
    real_disk = _OpenRealBD(disk)
2209
    result['DISK_%d_PATH' % idx] = real_disk.dev_path
2210
    result['DISK_%d_ACCESS' % idx] = disk.mode
2211
    if constants.HV_DISK_TYPE in instance.hvparams:
2212
      result['DISK_%d_FRONTEND_TYPE' % idx] = \
2213
        instance.hvparams[constants.HV_DISK_TYPE]
2214
    if disk.dev_type in constants.LDS_BLOCK:
2215
      result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
2216
    elif disk.dev_type == constants.LD_FILE:
2217
      result['DISK_%d_BACKEND_TYPE' % idx] = \
2218
        'file:%s' % disk.physical_id[0]
2219

    
2220
  # NICs
2221
  for idx, nic in enumerate(instance.nics):
2222
    result['NIC_%d_MAC' % idx] = nic.mac
2223
    if nic.ip:
2224
      result['NIC_%d_IP' % idx] = nic.ip
2225
    result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
2226
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2227
      result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
2228
    if nic.nicparams[constants.NIC_LINK]:
2229
      result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
2230
    if constants.HV_NIC_TYPE in instance.hvparams:
2231
      result['NIC_%d_FRONTEND_TYPE' % idx] = \
2232
        instance.hvparams[constants.HV_NIC_TYPE]
2233

    
2234
  # HV/BE params
2235
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2236
    for key, value in source.items():
2237
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2238

    
2239
  return result
2240

    
2241

    
2242
def BlockdevGrow(disk, amount, dryrun):
2243
  """Grow a stack of block devices.
2244

2245
  This function is called recursively, with the childrens being the
2246
  first ones to resize.
2247

2248
  @type disk: L{objects.Disk}
2249
  @param disk: the disk to be grown
2250
  @type amount: integer
2251
  @param amount: the amount (in mebibytes) to grow with
2252
  @type dryrun: boolean
2253
  @param dryrun: whether to execute the operation in simulation mode
2254
      only, without actually increasing the size
2255
  @rtype: (status, result)
2256
  @return: a tuple with the status of the operation (True/False), and
2257
      the errors message if status is False
2258

2259
  """
2260
  r_dev = _RecursiveFindBD(disk)
2261
  if r_dev is None:
2262
    _Fail("Cannot find block device %s", disk)
2263

    
2264
  try:
2265
    r_dev.Grow(amount, dryrun)
2266
  except errors.BlockDeviceError, err:
2267
    _Fail("Failed to grow block device: %s", err, exc=True)
2268

    
2269

    
2270
def BlockdevSnapshot(disk):
2271
  """Create a snapshot copy of a block device.
2272

2273
  This function is called recursively, and the snapshot is actually created
2274
  just for the leaf lvm backend device.
2275

2276
  @type disk: L{objects.Disk}
2277
  @param disk: the disk to be snapshotted
2278
  @rtype: string
2279
  @return: snapshot disk ID as (vg, lv)
2280

2281
  """
2282
  if disk.dev_type == constants.LD_DRBD8:
2283
    if not disk.children:
2284
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2285
            disk.unique_id)
2286
    return BlockdevSnapshot(disk.children[0])
2287
  elif disk.dev_type == constants.LD_LV:
2288
    r_dev = _RecursiveFindBD(disk)
2289
    if r_dev is not None:
2290
      # FIXME: choose a saner value for the snapshot size
2291
      # let's stay on the safe side and ask for the full size, for now
2292
      return r_dev.Snapshot(disk.size)
2293
    else:
2294
      _Fail("Cannot find block device %s", disk)
2295
  else:
2296
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2297
          disk.unique_id, disk.dev_type)
2298

    
2299

    
2300
def FinalizeExport(instance, snap_disks):
2301
  """Write out the export configuration information.
2302

2303
  @type instance: L{objects.Instance}
2304
  @param instance: the instance which we export, used for
2305
      saving configuration
2306
  @type snap_disks: list of L{objects.Disk}
2307
  @param snap_disks: list of snapshot block devices, which
2308
      will be used to get the actual name of the dump file
2309

2310
  @rtype: None
2311

2312
  """
2313
  destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2314
  finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2315

    
2316
  config = objects.SerializableConfigParser()
2317

    
2318
  config.add_section(constants.INISECT_EXP)
2319
  config.set(constants.INISECT_EXP, 'version', '0')
2320
  config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2321
  config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2322
  config.set(constants.INISECT_EXP, 'os', instance.os)
2323
  config.set(constants.INISECT_EXP, "compression", "none")
2324

    
2325
  config.add_section(constants.INISECT_INS)
2326
  config.set(constants.INISECT_INS, 'name', instance.name)
2327
  config.set(constants.INISECT_INS, 'memory', '%d' %
2328
             instance.beparams[constants.BE_MEMORY])
2329
  config.set(constants.INISECT_INS, 'vcpus', '%d' %
2330
             instance.beparams[constants.BE_VCPUS])
2331
  config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2332
  config.set(constants.INISECT_INS, 'hypervisor', instance.hypervisor)
2333
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2334

    
2335
  nic_total = 0
2336
  for nic_count, nic in enumerate(instance.nics):
2337
    nic_total += 1
2338
    config.set(constants.INISECT_INS, 'nic%d_mac' %
2339
               nic_count, '%s' % nic.mac)
2340
    config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2341
    for param in constants.NICS_PARAMETER_TYPES:
2342
      config.set(constants.INISECT_INS, 'nic%d_%s' % (nic_count, param),
2343
                 '%s' % nic.nicparams.get(param, None))
2344
  # TODO: redundant: on load can read nics until it doesn't exist
2345
  config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2346

    
2347
  disk_total = 0
2348
  for disk_count, disk in enumerate(snap_disks):
2349
    if disk:
2350
      disk_total += 1
2351
      config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2352
                 ('%s' % disk.iv_name))
2353
      config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2354
                 ('%s' % disk.physical_id[1]))
2355
      config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2356
                 ('%d' % disk.size))
2357

    
2358
  config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2359

    
2360
  # New-style hypervisor/backend parameters
2361

    
2362
  config.add_section(constants.INISECT_HYP)
2363
  for name, value in instance.hvparams.items():
2364
    if name not in constants.HVC_GLOBALS:
2365
      config.set(constants.INISECT_HYP, name, str(value))
2366

    
2367
  config.add_section(constants.INISECT_BEP)
2368
  for name, value in instance.beparams.items():
2369
    config.set(constants.INISECT_BEP, name, str(value))
2370

    
2371
  config.add_section(constants.INISECT_OSP)
2372
  for name, value in instance.osparams.items():
2373
    config.set(constants.INISECT_OSP, name, str(value))
2374

    
2375
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2376
                  data=config.Dumps())
2377
  shutil.rmtree(finaldestdir, ignore_errors=True)
2378
  shutil.move(destdir, finaldestdir)
2379

    
2380

    
2381
def ExportInfo(dest):
2382
  """Get export configuration information.
2383

2384
  @type dest: str
2385
  @param dest: directory containing the export
2386

2387
  @rtype: L{objects.SerializableConfigParser}
2388
  @return: a serializable config file containing the
2389
      export info
2390

2391
  """
2392
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2393

    
2394
  config = objects.SerializableConfigParser()
2395
  config.read(cff)
2396

    
2397
  if (not config.has_section(constants.INISECT_EXP) or
2398
      not config.has_section(constants.INISECT_INS)):
2399
    _Fail("Export info file doesn't have the required fields")
2400

    
2401
  return config.Dumps()
2402

    
2403

    
2404
def ListExports():
2405
  """Return a list of exports currently available on this machine.
2406

2407
  @rtype: list
2408
  @return: list of the exports
2409

2410
  """
2411
  if os.path.isdir(constants.EXPORT_DIR):
2412
    return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2413
  else:
2414
    _Fail("No exports directory")
2415

    
2416

    
2417
def RemoveExport(export):
2418
  """Remove an existing export from the node.
2419

2420
  @type export: str
2421
  @param export: the name of the export to remove
2422
  @rtype: None
2423

2424
  """
2425
  target = utils.PathJoin(constants.EXPORT_DIR, export)
2426

    
2427
  try:
2428
    shutil.rmtree(target)
2429
  except EnvironmentError, err:
2430
    _Fail("Error while removing the export: %s", err, exc=True)
2431

    
2432

    
2433
def BlockdevRename(devlist):
2434
  """Rename a list of block devices.
2435

2436
  @type devlist: list of tuples
2437
  @param devlist: list of tuples of the form  (disk,
2438
      new_logical_id, new_physical_id); disk is an
2439
      L{objects.Disk} object describing the current disk,
2440
      and new logical_id/physical_id is the name we
2441
      rename it to
2442
  @rtype: boolean
2443
  @return: True if all renames succeeded, False otherwise
2444

2445
  """
2446
  msgs = []
2447
  result = True
2448
  for disk, unique_id in devlist:
2449
    dev = _RecursiveFindBD(disk)
2450
    if dev is None:
2451
      msgs.append("Can't find device %s in rename" % str(disk))
2452
      result = False
2453
      continue
2454
    try:
2455
      old_rpath = dev.dev_path
2456
      dev.Rename(unique_id)
2457
      new_rpath = dev.dev_path
2458
      if old_rpath != new_rpath:
2459
        DevCacheManager.RemoveCache(old_rpath)
2460
        # FIXME: we should add the new cache information here, like:
2461
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2462
        # but we don't have the owner here - maybe parse from existing
2463
        # cache? for now, we only lose lvm data when we rename, which
2464
        # is less critical than DRBD or MD
2465
    except errors.BlockDeviceError, err:
2466
      msgs.append("Can't rename device '%s' to '%s': %s" %
2467
                  (dev, unique_id, err))
2468
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2469
      result = False
2470
  if not result:
2471
    _Fail("; ".join(msgs))
2472

    
2473

    
2474
def _TransformFileStorageDir(fs_dir):
2475
  """Checks whether given file_storage_dir is valid.
2476

2477
  Checks wheter the given fs_dir is within the cluster-wide default
2478
  file_storage_dir or the shared_file_storage_dir, which are stored in
2479
  SimpleStore. Only paths under those directories are allowed.
2480

2481
  @type fs_dir: str
2482
  @param fs_dir: the path to check
2483

2484
  @return: the normalized path if valid, None otherwise
2485

2486
  """
2487
  if not constants.ENABLE_FILE_STORAGE:
2488
    _Fail("File storage disabled at configure time")
2489
  cfg = _GetConfig()
2490
  fs_dir = os.path.normpath(fs_dir)
2491
  base_fstore = cfg.GetFileStorageDir()
2492
  base_shared = cfg.GetSharedFileStorageDir()
2493
  if ((os.path.commonprefix([fs_dir, base_fstore]) != base_fstore) and
2494
      (os.path.commonprefix([fs_dir, base_shared]) != base_shared)):
2495
    _Fail("File storage directory '%s' is not under base file"
2496
          " storage directory '%s' or shared storage directory '%s'",
2497
          fs_dir, base_fstore, base_shared)
2498
  return fs_dir
2499

    
2500

    
2501
def CreateFileStorageDir(file_storage_dir):
2502
  """Create file storage directory.
2503

2504
  @type file_storage_dir: str
2505
  @param file_storage_dir: directory to create
2506

2507
  @rtype: tuple
2508
  @return: tuple with first element a boolean indicating wheter dir
2509
      creation was successful or not
2510

2511
  """
2512
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2513
  if os.path.exists(file_storage_dir):
2514
    if not os.path.isdir(file_storage_dir):
2515
      _Fail("Specified storage dir '%s' is not a directory",
2516
            file_storage_dir)
2517
  else:
2518
    try:
2519
      os.makedirs(file_storage_dir, 0750)
2520
    except OSError, err:
2521
      _Fail("Cannot create file storage directory '%s': %s",
2522
            file_storage_dir, err, exc=True)
2523

    
2524

    
2525
def RemoveFileStorageDir(file_storage_dir):
2526
  """Remove file storage directory.
2527

2528
  Remove it only if it's empty. If not log an error and return.
2529

2530
  @type file_storage_dir: str
2531
  @param file_storage_dir: the directory we should cleanup
2532
  @rtype: tuple (success,)
2533
  @return: tuple of one element, C{success}, denoting
2534
      whether the operation was successful
2535

2536
  """
2537
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2538
  if os.path.exists(file_storage_dir):
2539
    if not os.path.isdir(file_storage_dir):
2540
      _Fail("Specified Storage directory '%s' is not a directory",
2541
            file_storage_dir)
2542
    # deletes dir only if empty, otherwise we want to fail the rpc call
2543
    try:
2544
      os.rmdir(file_storage_dir)
2545
    except OSError, err:
2546
      _Fail("Cannot remove file storage directory '%s': %s",
2547
            file_storage_dir, err)
2548

    
2549

    
2550
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2551
  """Rename the file storage directory.
2552

2553
  @type old_file_storage_dir: str
2554
  @param old_file_storage_dir: the current path
2555
  @type new_file_storage_dir: str
2556
  @param new_file_storage_dir: the name we should rename to
2557
  @rtype: tuple (success,)
2558
  @return: tuple of one element, C{success}, denoting
2559
      whether the operation was successful
2560

2561
  """
2562
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2563
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2564
  if not os.path.exists(new_file_storage_dir):
2565
    if os.path.isdir(old_file_storage_dir):
2566
      try:
2567
        os.rename(old_file_storage_dir, new_file_storage_dir)
2568
      except OSError, err:
2569
        _Fail("Cannot rename '%s' to '%s': %s",
2570
              old_file_storage_dir, new_file_storage_dir, err)
2571
    else:
2572
      _Fail("Specified storage dir '%s' is not a directory",
2573
            old_file_storage_dir)
2574
  else:
2575
    if os.path.exists(old_file_storage_dir):
2576
      _Fail("Cannot rename '%s' to '%s': both locations exist",
2577
            old_file_storage_dir, new_file_storage_dir)
2578

    
2579

    
2580
def _EnsureJobQueueFile(file_name):
2581
  """Checks whether the given filename is in the queue directory.
2582

2583
  @type file_name: str
2584
  @param file_name: the file name we should check
2585
  @rtype: None
2586
  @raises RPCFail: if the file is not valid
2587

2588
  """
2589
  queue_dir = os.path.normpath(constants.QUEUE_DIR)
2590
  result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2591

    
2592
  if not result:
2593
    _Fail("Passed job queue file '%s' does not belong to"
2594
          " the queue directory '%s'", file_name, queue_dir)
2595

    
2596

    
2597
def JobQueueUpdate(file_name, content):
2598
  """Updates a file in the queue directory.
2599

2600
  This is just a wrapper over L{utils.io.WriteFile}, with proper
2601
  checking.
2602

2603
  @type file_name: str
2604
  @param file_name: the job file name
2605
  @type content: str
2606
  @param content: the new job contents
2607
  @rtype: boolean
2608
  @return: the success of the operation
2609

2610
  """
2611
  _EnsureJobQueueFile(file_name)
2612
  getents = runtime.GetEnts()
2613

    
2614
  # Write and replace the file atomically
2615
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2616
                  gid=getents.masterd_gid)
2617

    
2618

    
2619
def JobQueueRename(old, new):
2620
  """Renames a job queue file.
2621

2622
  This is just a wrapper over os.rename with proper checking.
2623

2624
  @type old: str
2625
  @param old: the old (actual) file name
2626
  @type new: str
2627
  @param new: the desired file name
2628
  @rtype: tuple
2629
  @return: the success of the operation and payload
2630

2631
  """
2632
  _EnsureJobQueueFile(old)
2633
  _EnsureJobQueueFile(new)
2634

    
2635
  utils.RenameFile(old, new, mkdir=True)
2636

    
2637

    
2638
def BlockdevClose(instance_name, disks):
2639
  """Closes the given block devices.
2640

2641
  This means they will be switched to secondary mode (in case of
2642
  DRBD).
2643

2644
  @param instance_name: if the argument is not empty, the symlinks
2645
      of this instance will be removed
2646
  @type disks: list of L{objects.Disk}
2647
  @param disks: the list of disks to be closed
2648
  @rtype: tuple (success, message)
2649
  @return: a tuple of success and message, where success
2650
      indicates the succes of the operation, and message
2651
      which will contain the error details in case we
2652
      failed
2653

2654
  """
2655
  bdevs = []
2656
  for cf in disks:
2657
    rd = _RecursiveFindBD(cf)
2658
    if rd is None:
2659
      _Fail("Can't find device %s", cf)
2660
    bdevs.append(rd)
2661

    
2662
  msg = []
2663
  for rd in bdevs:
2664
    try:
2665
      rd.Close()
2666
    except errors.BlockDeviceError, err:
2667
      msg.append(str(err))
2668
  if msg:
2669
    _Fail("Can't make devices secondary: %s", ",".join(msg))
2670
  else:
2671
    if instance_name:
2672
      _RemoveBlockDevLinks(instance_name, disks)
2673

    
2674

    
2675
def ValidateHVParams(hvname, hvparams):
2676
  """Validates the given hypervisor parameters.
2677

2678
  @type hvname: string
2679
  @param hvname: the hypervisor name
2680
  @type hvparams: dict
2681
  @param hvparams: the hypervisor parameters to be validated
2682
  @rtype: None
2683

2684
  """
2685
  try:
2686
    hv_type = hypervisor.GetHypervisor(hvname)
2687
    hv_type.ValidateParameters(hvparams)
2688
  except errors.HypervisorError, err:
2689
    _Fail(str(err), log=False)
2690

    
2691

    
2692
def _CheckOSPList(os_obj, parameters):
2693
  """Check whether a list of parameters is supported by the OS.
2694

2695
  @type os_obj: L{objects.OS}
2696
  @param os_obj: OS object to check
2697
  @type parameters: list
2698
  @param parameters: the list of parameters to check
2699

2700
  """
2701
  supported = [v[0] for v in os_obj.supported_parameters]
2702
  delta = frozenset(parameters).difference(supported)
2703
  if delta:
2704
    _Fail("The following parameters are not supported"
2705
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2706

    
2707

    
2708
def ValidateOS(required, osname, checks, osparams):
2709
  """Validate the given OS' parameters.
2710

2711
  @type required: boolean
2712
  @param required: whether absence of the OS should translate into
2713
      failure or not
2714
  @type osname: string
2715
  @param osname: the OS to be validated
2716
  @type checks: list
2717
  @param checks: list of the checks to run (currently only 'parameters')
2718
  @type osparams: dict
2719
  @param osparams: dictionary with OS parameters
2720
  @rtype: boolean
2721
  @return: True if the validation passed, or False if the OS was not
2722
      found and L{required} was false
2723

2724
  """
2725
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2726
    _Fail("Unknown checks required for OS %s: %s", osname,
2727
          set(checks).difference(constants.OS_VALIDATE_CALLS))
2728

    
2729
  name_only = objects.OS.GetName(osname)
2730
  status, tbv = _TryOSFromDisk(name_only, None)
2731

    
2732
  if not status:
2733
    if required:
2734
      _Fail(tbv)
2735
    else:
2736
      return False
2737

    
2738
  if max(tbv.api_versions) < constants.OS_API_V20:
2739
    return True
2740

    
2741
  if constants.OS_VALIDATE_PARAMETERS in checks:
2742
    _CheckOSPList(tbv, osparams.keys())
2743

    
2744
  validate_env = OSCoreEnv(osname, tbv, osparams)
2745
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2746
                        cwd=tbv.path)
2747
  if result.failed:
2748
    logging.error("os validate command '%s' returned error: %s output: %s",
2749
                  result.cmd, result.fail_reason, result.output)
2750
    _Fail("OS validation script failed (%s), output: %s",
2751
          result.fail_reason, result.output, log=False)
2752

    
2753
  return True
2754

    
2755

    
2756
def DemoteFromMC():
2757
  """Demotes the current node from master candidate role.
2758

2759
  """
2760
  # try to ensure we're not the master by mistake
2761
  master, myself = ssconf.GetMasterAndMyself()
2762
  if master == myself:
2763
    _Fail("ssconf status shows I'm the master node, will not demote")
2764

    
2765
  result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2766
  if not result.failed:
2767
    _Fail("The master daemon is running, will not demote")
2768

    
2769
  try:
2770
    if os.path.isfile(constants.CLUSTER_CONF_FILE):
2771
      utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2772
  except EnvironmentError, err:
2773
    if err.errno != errno.ENOENT:
2774
      _Fail("Error while backing up cluster file: %s", err, exc=True)
2775

    
2776
  utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2777

    
2778

    
2779
def _GetX509Filenames(cryptodir, name):
2780
  """Returns the full paths for the private key and certificate.
2781

2782
  """
2783
  return (utils.PathJoin(cryptodir, name),
2784
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2785
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2786

    
2787

    
2788
def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2789
  """Creates a new X509 certificate for SSL/TLS.
2790

2791
  @type validity: int
2792
  @param validity: Validity in seconds
2793
  @rtype: tuple; (string, string)
2794
  @return: Certificate name and public part
2795

2796
  """
2797
  (key_pem, cert_pem) = \
2798
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2799
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
2800

    
2801
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
2802
                              prefix="x509-%s-" % utils.TimestampForFilename())
2803
  try:
2804
    name = os.path.basename(cert_dir)
2805
    assert len(name) > 5
2806

    
2807
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2808

    
2809
    utils.WriteFile(key_file, mode=0400, data=key_pem)
2810
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2811

    
2812
    # Never return private key as it shouldn't leave the node
2813
    return (name, cert_pem)
2814
  except Exception:
2815
    shutil.rmtree(cert_dir, ignore_errors=True)
2816
    raise
2817

    
2818

    
2819
def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2820
  """Removes a X509 certificate.
2821

2822
  @type name: string
2823
  @param name: Certificate name
2824

2825
  """
2826
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2827

    
2828
  utils.RemoveFile(key_file)
2829
  utils.RemoveFile(cert_file)
2830

    
2831
  try:
2832
    os.rmdir(cert_dir)
2833
  except EnvironmentError, err:
2834
    _Fail("Cannot remove certificate directory '%s': %s",
2835
          cert_dir, err)
2836

    
2837

    
2838
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2839
  """Returns the command for the requested input/output.
2840

2841
  @type instance: L{objects.Instance}
2842
  @param instance: The instance object
2843
  @param mode: Import/export mode
2844
  @param ieio: Input/output type
2845
  @param ieargs: Input/output arguments
2846

2847
  """
2848
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2849

    
2850
  env = None
2851
  prefix = None
2852
  suffix = None
2853
  exp_size = None
2854

    
2855
  if ieio == constants.IEIO_FILE:
2856
    (filename, ) = ieargs
2857

    
2858
    if not utils.IsNormAbsPath(filename):
2859
      _Fail("Path '%s' is not normalized or absolute", filename)
2860

    
2861
    directory = os.path.normpath(os.path.dirname(filename))
2862

    
2863
    if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2864
        constants.EXPORT_DIR):
2865
      _Fail("File '%s' is not under exports directory '%s'",
2866
            filename, constants.EXPORT_DIR)
2867

    
2868
    # Create directory
2869
    utils.Makedirs(directory, mode=0750)
2870

    
2871
    quoted_filename = utils.ShellQuote(filename)
2872

    
2873
    if mode == constants.IEM_IMPORT:
2874
      suffix = "> %s" % quoted_filename
2875
    elif mode == constants.IEM_EXPORT:
2876
      suffix = "< %s" % quoted_filename
2877

    
2878
      # Retrieve file size
2879
      try:
2880
        st = os.stat(filename)
2881
      except EnvironmentError, err:
2882
        logging.error("Can't stat(2) %s: %s", filename, err)
2883
      else:
2884
        exp_size = utils.BytesToMebibyte(st.st_size)
2885

    
2886
  elif ieio == constants.IEIO_RAW_DISK:
2887
    (disk, ) = ieargs
2888

    
2889
    real_disk = _OpenRealBD(disk)
2890

    
2891
    if mode == constants.IEM_IMPORT:
2892
      # we set here a smaller block size as, due to transport buffering, more
2893
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
2894
      # is not already there or we pass a wrong path; we use notrunc to no
2895
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2896
      # much memory; this means that at best, we flush every 64k, which will
2897
      # not be very fast
2898
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2899
                                    " bs=%s oflag=dsync"),
2900
                                    real_disk.dev_path,
2901
                                    str(64 * 1024))
2902

    
2903
    elif mode == constants.IEM_EXPORT:
2904
      # the block size on the read dd is 1MiB to match our units
2905
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2906
                                   real_disk.dev_path,
2907
                                   str(1024 * 1024), # 1 MB
2908
                                   str(disk.size))
2909
      exp_size = disk.size
2910

    
2911
  elif ieio == constants.IEIO_SCRIPT:
2912
    (disk, disk_index, ) = ieargs
2913

    
2914
    assert isinstance(disk_index, (int, long))
2915

    
2916
    real_disk = _OpenRealBD(disk)
2917

    
2918
    inst_os = OSFromDisk(instance.os)
2919
    env = OSEnvironment(instance, inst_os)
2920

    
2921
    if mode == constants.IEM_IMPORT:
2922
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2923
      env["IMPORT_INDEX"] = str(disk_index)
2924
      script = inst_os.import_script
2925

    
2926
    elif mode == constants.IEM_EXPORT:
2927
      env["EXPORT_DEVICE"] = real_disk.dev_path
2928
      env["EXPORT_INDEX"] = str(disk_index)
2929
      script = inst_os.export_script
2930

    
2931
    # TODO: Pass special environment only to script
2932
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2933

    
2934
    if mode == constants.IEM_IMPORT:
2935
      suffix = "| %s" % script_cmd
2936

    
2937
    elif mode == constants.IEM_EXPORT:
2938
      prefix = "%s |" % script_cmd
2939

    
2940
    # Let script predict size
2941
    exp_size = constants.IE_CUSTOM_SIZE
2942

    
2943
  else:
2944
    _Fail("Invalid %s I/O mode %r", mode, ieio)
2945

    
2946
  return (env, prefix, suffix, exp_size)
2947

    
2948

    
2949
def _CreateImportExportStatusDir(prefix):
2950
  """Creates status directory for import/export.
2951

2952
  """
2953
  return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2954
                          prefix=("%s-%s-" %
2955
                                  (prefix, utils.TimestampForFilename())))
2956

    
2957

    
2958
def StartImportExportDaemon(mode, opts, host, port, instance, ieio, ieioargs):
2959
  """Starts an import or export daemon.
2960

2961
  @param mode: Import/output mode
2962
  @type opts: L{objects.ImportExportOptions}
2963
  @param opts: Daemon options
2964
  @type host: string
2965
  @param host: Remote host for export (None for import)
2966
  @type port: int
2967
  @param port: Remote port for export (None for import)
2968
  @type instance: L{objects.Instance}
2969
  @param instance: Instance object
2970
  @param ieio: Input/output type
2971
  @param ieioargs: Input/output arguments
2972

2973
  """
2974
  if mode == constants.IEM_IMPORT:
2975
    prefix = "import"
2976

    
2977
    if not (host is None and port is None):
2978
      _Fail("Can not specify host or port on import")
2979

    
2980
  elif mode == constants.IEM_EXPORT:
2981
    prefix = "export"
2982

    
2983
    if host is None or port is None:
2984
      _Fail("Host and port must be specified for an export")
2985

    
2986
  else:
2987
    _Fail("Invalid mode %r", mode)
2988

    
2989
  if (opts.key_name is None) ^ (opts.ca_pem is None):
2990
    _Fail("Cluster certificate can only be used for both key and CA")
2991

    
2992
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
2993
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
2994

    
2995
  if opts.key_name is None:
2996
    # Use server.pem
2997
    key_path = constants.NODED_CERT_FILE
2998
    cert_path = constants.NODED_CERT_FILE
2999
    assert opts.ca_pem is None
3000
  else:
3001
    (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
3002
                                                 opts.key_name)
3003
    assert opts.ca_pem is not None
3004

    
3005
  for i in [key_path, cert_path]:
3006
    if not os.path.exists(i):
3007
      _Fail("File '%s' does not exist" % i)
3008

    
3009
  status_dir = _CreateImportExportStatusDir(prefix)
3010
  try:
3011
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3012
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3013
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3014

    
3015
    if opts.ca_pem is None:
3016
      # Use server.pem
3017
      ca = utils.ReadFile(constants.NODED_CERT_FILE)
3018
    else:
3019
      ca = opts.ca_pem
3020

    
3021
    # Write CA file
3022
    utils.WriteFile(ca_file, data=ca, mode=0400)
3023

    
3024
    cmd = [
3025
      constants.IMPORT_EXPORT_DAEMON,
3026
      status_file, mode,
3027
      "--key=%s" % key_path,
3028
      "--cert=%s" % cert_path,
3029
      "--ca=%s" % ca_file,
3030
      ]
3031

    
3032
    if host:
3033
      cmd.append("--host=%s" % host)
3034

    
3035
    if port:
3036
      cmd.append("--port=%s" % port)
3037

    
3038
    if opts.ipv6:
3039
      cmd.append("--ipv6")
3040
    else:
3041
      cmd.append("--ipv4")
3042

    
3043
    if opts.compress:
3044
      cmd.append("--compress=%s" % opts.compress)
3045

    
3046
    if opts.magic:
3047
      cmd.append("--magic=%s" % opts.magic)
3048

    
3049
    if exp_size is not None:
3050
      cmd.append("--expected-size=%s" % exp_size)
3051

    
3052
    if cmd_prefix:
3053
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3054

    
3055
    if cmd_suffix:
3056
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3057

    
3058
    if mode == constants.IEM_EXPORT:
3059
      # Retry connection a few times when connecting to remote peer
3060
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3061
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3062
    elif opts.connect_timeout is not None:
3063
      assert mode == constants.IEM_IMPORT
3064
      # Overall timeout for establishing connection while listening
3065
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3066

    
3067
    logfile = _InstanceLogName(prefix, instance.os, instance.name)
3068

    
3069
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3070
    # support for receiving a file descriptor for output
3071
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3072
                      output=logfile)
3073

    
3074
    # The import/export name is simply the status directory name
3075
    return os.path.basename(status_dir)
3076

    
3077
  except Exception:
3078
    shutil.rmtree(status_dir, ignore_errors=True)
3079
    raise
3080

    
3081

    
3082
def GetImportExportStatus(names):
3083
  """Returns import/export daemon status.
3084

3085
  @type names: sequence
3086
  @param names: List of names
3087
  @rtype: List of dicts
3088
  @return: Returns a list of the state of each named import/export or None if a
3089
           status couldn't be read
3090

3091
  """
3092
  result = []
3093

    
3094
  for name in names:
3095
    status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3096
                                 _IES_STATUS_FILE)
3097

    
3098
    try:
3099
      data = utils.ReadFile(status_file)
3100
    except EnvironmentError, err:
3101
      if err.errno != errno.ENOENT:
3102
        raise
3103
      data = None
3104

    
3105
    if not data:
3106
      result.append(None)
3107
      continue
3108

    
3109
    result.append(serializer.LoadJson(data))
3110

    
3111
  return result
3112

    
3113

    
3114
def AbortImportExport(name):
3115
  """Sends SIGTERM to a running import/export daemon.
3116

3117
  """
3118
  logging.info("Abort import/export %s", name)
3119

    
3120
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3121
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3122

    
3123
  if pid:
3124
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3125
                 name, pid)
3126
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3127

    
3128

    
3129
def CleanupImportExport(name):
3130
  """Cleanup after an import or export.
3131

3132
  If the import/export daemon is still running it's killed. Afterwards the
3133
  whole status directory is removed.
3134

3135
  """
3136
  logging.info("Finalizing import/export %s", name)
3137

    
3138
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3139

    
3140
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3141

    
3142
  if pid:
3143
    logging.info("Import/export %s is still running with PID %s",
3144
                 name, pid)
3145
    utils.KillProcess(pid, waitpid=False)
3146

    
3147
  shutil.rmtree(status_dir, ignore_errors=True)
3148

    
3149

    
3150
def _FindDisks(nodes_ip, disks):
3151
  """Sets the physical ID on disks and returns the block devices.
3152

3153
  """
3154
  # set the correct physical ID
3155
  my_name = netutils.Hostname.GetSysName()
3156
  for cf in disks:
3157
    cf.SetPhysicalID(my_name, nodes_ip)
3158

    
3159
  bdevs = []
3160

    
3161
  for cf in disks:
3162
    rd = _RecursiveFindBD(cf)
3163
    if rd is None:
3164
      _Fail("Can't find device %s", cf)
3165
    bdevs.append(rd)
3166
  return bdevs
3167

    
3168

    
3169
def DrbdDisconnectNet(nodes_ip, disks):
3170
  """Disconnects the network on a list of drbd devices.
3171

3172
  """
3173
  bdevs = _FindDisks(nodes_ip, disks)
3174

    
3175
  # disconnect disks
3176
  for rd in bdevs:
3177
    try:
3178
      rd.DisconnectNet()
3179
    except errors.BlockDeviceError, err:
3180
      _Fail("Can't change network configuration to standalone mode: %s",
3181
            err, exc=True)
3182

    
3183

    
3184
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3185
  """Attaches the network on a list of drbd devices.
3186

3187
  """
3188
  bdevs = _FindDisks(nodes_ip, disks)
3189

    
3190
  if multimaster:
3191
    for idx, rd in enumerate(bdevs):
3192
      try:
3193
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3194
      except EnvironmentError, err:
3195
        _Fail("Can't create symlink: %s", err)
3196
  # reconnect disks, switch to new master configuration and if
3197
  # needed primary mode
3198
  for rd in bdevs:
3199
    try:
3200
      rd.AttachNet(multimaster)
3201
    except errors.BlockDeviceError, err:
3202
      _Fail("Can't change network configuration: %s", err)
3203

    
3204
  # wait until the disks are connected; we need to retry the re-attach
3205
  # if the device becomes standalone, as this might happen if the one
3206
  # node disconnects and reconnects in a different mode before the
3207
  # other node reconnects; in this case, one or both of the nodes will
3208
  # decide it has wrong configuration and switch to standalone
3209

    
3210
  def _Attach():
3211
    all_connected = True
3212

    
3213
    for rd in bdevs:
3214
      stats = rd.GetProcStatus()
3215

    
3216
      all_connected = (all_connected and
3217
                       (stats.is_connected or stats.is_in_resync))
3218

    
3219
      if stats.is_standalone:
3220
        # peer had different config info and this node became
3221
        # standalone, even though this should not happen with the
3222
        # new staged way of changing disk configs
3223
        try:
3224
          rd.AttachNet(multimaster)
3225
        except errors.BlockDeviceError, err:
3226
          _Fail("Can't change network configuration: %s", err)
3227

    
3228
    if not all_connected:
3229
      raise utils.RetryAgain()
3230

    
3231
  try:
3232
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3233
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3234
  except utils.RetryTimeout:
3235
    _Fail("Timeout in disk reconnecting")
3236

    
3237
  if multimaster:
3238
    # change to primary mode
3239
    for rd in bdevs:
3240
      try:
3241
        rd.Open()
3242
      except errors.BlockDeviceError, err:
3243
        _Fail("Can't change to primary mode: %s", err)
3244

    
3245

    
3246
def DrbdWaitSync(nodes_ip, disks):
3247
  """Wait until DRBDs have synchronized.
3248

3249
  """
3250
  def _helper(rd):
3251
    stats = rd.GetProcStatus()
3252
    if not (stats.is_connected or stats.is_in_resync):
3253
      raise utils.RetryAgain()
3254
    return stats
3255

    
3256
  bdevs = _FindDisks(nodes_ip, disks)
3257

    
3258
  min_resync = 100
3259
  alldone = True
3260
  for rd in bdevs:
3261
    try:
3262
      # poll each second for 15 seconds
3263
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3264
    except utils.RetryTimeout:
3265
      stats = rd.GetProcStatus()
3266
      # last check
3267
      if not (stats.is_connected or stats.is_in_resync):
3268
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3269
    alldone = alldone and (not stats.is_in_resync)
3270
    if stats.sync_percent is not None:
3271
      min_resync = min(min_resync, stats.sync_percent)
3272

    
3273
  return (alldone, min_resync)
3274

    
3275

    
3276
def GetDrbdUsermodeHelper():
3277
  """Returns DRBD usermode helper currently configured.
3278

3279
  """
3280
  try:
3281
    return bdev.BaseDRBD.GetUsermodeHelper()
3282
  except errors.BlockDeviceError, err:
3283
    _Fail(str(err))
3284

    
3285

    
3286
def PowercycleNode(hypervisor_type):
3287
  """Hard-powercycle the node.
3288

3289
  Because we need to return first, and schedule the powercycle in the
3290
  background, we won't be able to report failures nicely.
3291

3292
  """
3293
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3294
  try:
3295
    pid = os.fork()
3296
  except OSError:
3297
    # if we can't fork, we'll pretend that we're in the child process
3298
    pid = 0
3299
  if pid > 0:
3300
    return "Reboot scheduled in 5 seconds"
3301
  # ensure the child is running on ram
3302
  try:
3303
    utils.Mlockall()
3304
  except Exception: # pylint: disable-msg=W0703
3305
    pass
3306
  time.sleep(5)
3307
  hyper.PowercycleNode()
3308

    
3309

    
3310
class HooksRunner(object):
3311
  """Hook runner.
3312

3313
  This class is instantiated on the node side (ganeti-noded) and not
3314
  on the master side.
3315

3316
  """
3317
  def __init__(self, hooks_base_dir=None):
3318
    """Constructor for hooks runner.
3319

3320
    @type hooks_base_dir: str or None
3321
    @param hooks_base_dir: if not None, this overrides the
3322
        L{constants.HOOKS_BASE_DIR} (useful for unittests)
3323

3324
    """
3325
    if hooks_base_dir is None:
3326
      hooks_base_dir = constants.HOOKS_BASE_DIR
3327
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
3328
    # constant
3329
    self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3330

    
3331
  def RunHooks(self, hpath, phase, env):
3332
    """Run the scripts in the hooks directory.
3333

3334
    @type hpath: str
3335
    @param hpath: the path to the hooks directory which
3336
        holds the scripts
3337
    @type phase: str
3338
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
3339
        L{constants.HOOKS_PHASE_POST}
3340
    @type env: dict
3341
    @param env: dictionary with the environment for the hook
3342
    @rtype: list
3343
    @return: list of 3-element tuples:
3344
      - script path
3345
      - script result, either L{constants.HKR_SUCCESS} or
3346
        L{constants.HKR_FAIL}
3347
      - output of the script
3348

3349
    @raise errors.ProgrammerError: for invalid input
3350
        parameters
3351

3352
    """
3353
    if phase == constants.HOOKS_PHASE_PRE:
3354
      suffix = "pre"
3355
    elif phase == constants.HOOKS_PHASE_POST:
3356
      suffix = "post"
3357
    else:
3358
      _Fail("Unknown hooks phase '%s'", phase)
3359

    
3360

    
3361
    subdir = "%s-%s.d" % (hpath, suffix)
3362
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3363

    
3364
    results = []
3365

    
3366
    if not os.path.isdir(dir_name):
3367
      # for non-existing/non-dirs, we simply exit instead of logging a
3368
      # warning at every operation
3369
      return results
3370

    
3371
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3372

    
3373
    for (relname, relstatus, runresult)  in runparts_results:
3374
      if relstatus == constants.RUNPARTS_SKIP:
3375
        rrval = constants.HKR_SKIP
3376
        output = ""
3377
      elif relstatus == constants.RUNPARTS_ERR:
3378
        rrval = constants.HKR_FAIL
3379
        output = "Hook script execution error: %s" % runresult
3380
      elif relstatus == constants.RUNPARTS_RUN:
3381
        if runresult.failed:
3382
          rrval = constants.HKR_FAIL
3383
        else:
3384
          rrval = constants.HKR_SUCCESS
3385
        output = utils.SafeEncode(runresult.output.strip())
3386
      results.append(("%s/%s" % (subdir, relname), rrval, output))
3387

    
3388
    return results
3389

    
3390

    
3391
class IAllocatorRunner(object):
3392
  """IAllocator runner.
3393

3394
  This class is instantiated on the node side (ganeti-noded) and not on
3395
  the master side.
3396

3397
  """
3398
  @staticmethod
3399
  def Run(name, idata):
3400
    """Run an iallocator script.
3401

3402
    @type name: str
3403
    @param name: the iallocator script name
3404
    @type idata: str
3405
    @param idata: the allocator input data
3406

3407
    @rtype: tuple
3408
    @return: two element tuple of:
3409
       - status
3410
       - either error message or stdout of allocator (for success)
3411

3412
    """
3413
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3414
                                  os.path.isfile)
3415
    if alloc_script is None:
3416
      _Fail("iallocator module '%s' not found in the search path", name)
3417

    
3418
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3419
    try:
3420
      os.write(fd, idata)
3421
      os.close(fd)
3422
      result = utils.RunCmd([alloc_script, fin_name])
3423
      if result.failed:
3424
        _Fail("iallocator module '%s' failed: %s, output '%s'",
3425
              name, result.fail_reason, result.output)
3426
    finally:
3427
      os.unlink(fin_name)
3428

    
3429
    return result.stdout
3430

    
3431

    
3432
class DevCacheManager(object):
3433
  """Simple class for managing a cache of block device information.
3434

3435
  """
3436
  _DEV_PREFIX = "/dev/"
3437
  _ROOT_DIR = constants.BDEV_CACHE_DIR
3438

    
3439
  @classmethod
3440
  def _ConvertPath(cls, dev_path):
3441
    """Converts a /dev/name path to the cache file name.
3442

3443
    This replaces slashes with underscores and strips the /dev
3444
    prefix. It then returns the full path to the cache file.
3445

3446
    @type dev_path: str
3447
    @param dev_path: the C{/dev/} path name
3448
    @rtype: str
3449
    @return: the converted path name
3450

3451
    """
3452
    if dev_path.startswith(cls._DEV_PREFIX):
3453
      dev_path = dev_path[len(cls._DEV_PREFIX):]
3454
    dev_path = dev_path.replace("/", "_")
3455
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3456
    return fpath
3457

    
3458
  @classmethod
3459
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3460
    """Updates the cache information for a given device.
3461

3462
    @type dev_path: str
3463
    @param dev_path: the pathname of the device
3464
    @type owner: str
3465
    @param owner: the owner (instance name) of the device
3466
    @type on_primary: bool
3467
    @param on_primary: whether this is the primary
3468
        node nor not
3469
    @type iv_name: str
3470
    @param iv_name: the instance-visible name of the
3471
        device, as in objects.Disk.iv_name
3472

3473
    @rtype: None
3474

3475
    """
3476
    if dev_path is None:
3477
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
3478
      return
3479
    fpath = cls._ConvertPath(dev_path)
3480
    if on_primary:
3481
      state = "primary"
3482
    else:
3483
      state = "secondary"
3484
    if iv_name is None:
3485
      iv_name = "not_visible"
3486
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3487
    try:
3488
      utils.WriteFile(fpath, data=fdata)
3489
    except EnvironmentError, err:
3490
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3491

    
3492
  @classmethod
3493
  def RemoveCache(cls, dev_path):
3494
    """Remove data for a dev_path.
3495

3496
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
3497
    path name and logging.
3498

3499
    @type dev_path: str
3500
    @param dev_path: the pathname of the device
3501

3502
    @rtype: None
3503

3504
    """
3505
    if dev_path is None:
3506
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
3507
      return
3508
    fpath = cls._ConvertPath(dev_path)
3509
    try:
3510
      utils.RemoveFile(fpath)
3511
    except EnvironmentError, err:
3512
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)