Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 2be7273c

History | View | Annotate | Download (104.5 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
  return result
648

    
649

    
650
def GetBlockDevSizes(devices):
651
  """Return the size of the given block devices
652

653
  @type devices: list
654
  @param devices: list of block device nodes to query
655
  @rtype: dict
656
  @return:
657
    dictionary of all block devices under /dev (key). The value is their
658
    size in MiB.
659

660
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
661

662
  """
663
  DEV_PREFIX = "/dev/"
664
  blockdevs = {}
665

    
666
  for devpath in devices:
667
    if os.path.commonprefix([DEV_PREFIX, devpath]) != DEV_PREFIX:
668
      continue
669

    
670
    try:
671
      st = os.stat(devpath)
672
    except EnvironmentError, err:
673
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
674
      continue
675

    
676
    if stat.S_ISBLK(st.st_mode):
677
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
678
      if result.failed:
679
        # We don't want to fail, just do not list this device as available
680
        logging.warning("Cannot get size for block device %s", devpath)
681
        continue
682

    
683
      size = int(result.stdout) / (1024 * 1024)
684
      blockdevs[devpath] = size
685
  return blockdevs
686

    
687

    
688
def GetVolumeList(vg_names):
689
  """Compute list of logical volumes and their size.
690

691
  @type vg_names: list
692
  @param vg_names: the volume groups whose LVs we should list, or
693
      empty for all volume groups
694
  @rtype: dict
695
  @return:
696
      dictionary of all partions (key) with value being a tuple of
697
      their size (in MiB), inactive and online status::
698

699
        {'xenvg/test1': ('20.06', True, True)}
700

701
      in case of errors, a string is returned with the error
702
      details.
703

704
  """
705
  lvs = {}
706
  sep = '|'
707
  if not vg_names:
708
    vg_names = []
709
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
710
                         "--separator=%s" % sep,
711
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
712
  if result.failed:
713
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
714

    
715
  for line in result.stdout.splitlines():
716
    line = line.strip()
717
    match = _LVSLINE_REGEX.match(line)
718
    if not match:
719
      logging.error("Invalid line returned from lvs output: '%s'", line)
720
      continue
721
    vg_name, name, size, attr = match.groups()
722
    inactive = attr[4] == '-'
723
    online = attr[5] == 'o'
724
    virtual = attr[0] == 'v'
725
    if virtual:
726
      # we don't want to report such volumes as existing, since they
727
      # don't really hold data
728
      continue
729
    lvs[vg_name+"/"+name] = (size, inactive, online)
730

    
731
  return lvs
732

    
733

    
734
def ListVolumeGroups():
735
  """List the volume groups and their size.
736

737
  @rtype: dict
738
  @return: dictionary with keys volume name and values the
739
      size of the volume
740

741
  """
742
  return utils.ListVolumeGroups()
743

    
744

    
745
def NodeVolumes():
746
  """List all volumes on this node.
747

748
  @rtype: list
749
  @return:
750
    A list of dictionaries, each having four keys:
751
      - name: the logical volume name,
752
      - size: the size of the logical volume
753
      - dev: the physical device on which the LV lives
754
      - vg: the volume group to which it belongs
755

756
    In case of errors, we return an empty list and log the
757
    error.
758

759
    Note that since a logical volume can live on multiple physical
760
    volumes, the resulting list might include a logical volume
761
    multiple times.
762

763
  """
764
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
765
                         "--separator=|",
766
                         "--options=lv_name,lv_size,devices,vg_name"])
767
  if result.failed:
768
    _Fail("Failed to list logical volumes, lvs output: %s",
769
          result.output)
770

    
771
  def parse_dev(dev):
772
    return dev.split('(')[0]
773

    
774
  def handle_dev(dev):
775
    return [parse_dev(x) for x in dev.split(",")]
776

    
777
  def map_line(line):
778
    line = [v.strip() for v in line]
779
    return [{'name': line[0], 'size': line[1],
780
             'dev': dev, 'vg': line[3]} for dev in handle_dev(line[2])]
781

    
782
  all_devs = []
783
  for line in result.stdout.splitlines():
784
    if line.count('|') >= 3:
785
      all_devs.extend(map_line(line.split('|')))
786
    else:
787
      logging.warning("Strange line in the output from lvs: '%s'", line)
788
  return all_devs
789

    
790

    
791
def BridgesExist(bridges_list):
792
  """Check if a list of bridges exist on the current node.
793

794
  @rtype: boolean
795
  @return: C{True} if all of them exist, C{False} otherwise
796

797
  """
798
  missing = []
799
  for bridge in bridges_list:
800
    if not utils.BridgeExists(bridge):
801
      missing.append(bridge)
802

    
803
  if missing:
804
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
805

    
806

    
807
def GetInstanceList(hypervisor_list):
808
  """Provides a list of instances.
809

810
  @type hypervisor_list: list
811
  @param hypervisor_list: the list of hypervisors to query information
812

813
  @rtype: list
814
  @return: a list of all running instances on the current node
815
    - instance1.example.com
816
    - instance2.example.com
817

818
  """
819
  results = []
820
  for hname in hypervisor_list:
821
    try:
822
      names = hypervisor.GetHypervisor(hname).ListInstances()
823
      results.extend(names)
824
    except errors.HypervisorError, err:
825
      _Fail("Error enumerating instances (hypervisor %s): %s",
826
            hname, err, exc=True)
827

    
828
  return results
829

    
830

    
831
def GetInstanceInfo(instance, hname):
832
  """Gives back the information about an instance as a dictionary.
833

834
  @type instance: string
835
  @param instance: the instance name
836
  @type hname: string
837
  @param hname: the hypervisor type of the instance
838

839
  @rtype: dict
840
  @return: dictionary with the following keys:
841
      - memory: memory size of instance (int)
842
      - state: xen state of instance (string)
843
      - time: cpu time of instance (float)
844

845
  """
846
  output = {}
847

    
848
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
849
  if iinfo is not None:
850
    output['memory'] = iinfo[2]
851
    output['state'] = iinfo[4]
852
    output['time'] = iinfo[5]
853

    
854
  return output
855

    
856

    
857
def GetInstanceMigratable(instance):
858
  """Gives whether an instance can be migrated.
859

860
  @type instance: L{objects.Instance}
861
  @param instance: object representing the instance to be checked.
862

863
  @rtype: tuple
864
  @return: tuple of (result, description) where:
865
      - result: whether the instance can be migrated or not
866
      - description: a description of the issue, if relevant
867

868
  """
869
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
870
  iname = instance.name
871
  if iname not in hyper.ListInstances():
872
    _Fail("Instance %s is not running", iname)
873

    
874
  for idx in range(len(instance.disks)):
875
    link_name = _GetBlockDevSymlinkPath(iname, idx)
876
    if not os.path.islink(link_name):
877
      logging.warning("Instance %s is missing symlink %s for disk %d",
878
                      iname, link_name, idx)
879

    
880

    
881
def GetAllInstancesInfo(hypervisor_list):
882
  """Gather data about all instances.
883

884
  This is the equivalent of L{GetInstanceInfo}, except that it
885
  computes data for all instances at once, thus being faster if one
886
  needs data about more than one instance.
887

888
  @type hypervisor_list: list
889
  @param hypervisor_list: list of hypervisors to query for instance data
890

891
  @rtype: dict
892
  @return: dictionary of instance: data, with data having the following keys:
893
      - memory: memory size of instance (int)
894
      - state: xen state of instance (string)
895
      - time: cpu time of instance (float)
896
      - vcpus: the number of vcpus
897

898
  """
899
  output = {}
900

    
901
  for hname in hypervisor_list:
902
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
903
    if iinfo:
904
      for name, _, memory, vcpus, state, times in iinfo:
905
        value = {
906
          'memory': memory,
907
          'vcpus': vcpus,
908
          'state': state,
909
          'time': times,
910
          }
911
        if name in output:
912
          # we only check static parameters, like memory and vcpus,
913
          # and not state and time which can change between the
914
          # invocations of the different hypervisors
915
          for key in 'memory', 'vcpus':
916
            if value[key] != output[name][key]:
917
              _Fail("Instance %s is running twice"
918
                    " with different parameters", name)
919
        output[name] = value
920

    
921
  return output
922

    
923

    
924
def _InstanceLogName(kind, os_name, instance):
925
  """Compute the OS log filename for a given instance and operation.
926

927
  The instance name and os name are passed in as strings since not all
928
  operations have these as part of an instance object.
929

930
  @type kind: string
931
  @param kind: the operation type (e.g. add, import, etc.)
932
  @type os_name: string
933
  @param os_name: the os name
934
  @type instance: string
935
  @param instance: the name of the instance being imported/added/etc.
936

937
  """
938
  # TODO: Use tempfile.mkstemp to create unique filename
939
  base = ("%s-%s-%s-%s.log" %
940
          (kind, os_name, instance, utils.TimestampForFilename()))
941
  return utils.PathJoin(constants.LOG_OS_DIR, base)
942

    
943

    
944
def InstanceOsAdd(instance, reinstall, debug):
945
  """Add an OS to an instance.
946

947
  @type instance: L{objects.Instance}
948
  @param instance: Instance whose OS is to be installed
949
  @type reinstall: boolean
950
  @param reinstall: whether this is an instance reinstall
951
  @type debug: integer
952
  @param debug: debug level, passed to the OS scripts
953
  @rtype: None
954

955
  """
956
  inst_os = OSFromDisk(instance.os)
957

    
958
  create_env = OSEnvironment(instance, inst_os, debug)
959
  if reinstall:
960
    create_env['INSTANCE_REINSTALL'] = "1"
961

    
962
  logfile = _InstanceLogName("add", instance.os, instance.name)
963

    
964
  result = utils.RunCmd([inst_os.create_script], env=create_env,
965
                        cwd=inst_os.path, output=logfile,)
966
  if result.failed:
967
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
968
                  " output: %s", result.cmd, result.fail_reason, logfile,
969
                  result.output)
970
    lines = [utils.SafeEncode(val)
971
             for val in utils.TailFile(logfile, lines=20)]
972
    _Fail("OS create script failed (%s), last lines in the"
973
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
974

    
975

    
976
def RunRenameInstance(instance, old_name, debug):
977
  """Run the OS rename script for an instance.
978

979
  @type instance: L{objects.Instance}
980
  @param instance: Instance whose OS is to be installed
981
  @type old_name: string
982
  @param old_name: previous instance name
983
  @type debug: integer
984
  @param debug: debug level, passed to the OS scripts
985
  @rtype: boolean
986
  @return: the success of the operation
987

988
  """
989
  inst_os = OSFromDisk(instance.os)
990

    
991
  rename_env = OSEnvironment(instance, inst_os, debug)
992
  rename_env['OLD_INSTANCE_NAME'] = old_name
993

    
994
  logfile = _InstanceLogName("rename", instance.os,
995
                             "%s-%s" % (old_name, instance.name))
996

    
997
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
998
                        cwd=inst_os.path, output=logfile)
999

    
1000
  if result.failed:
1001
    logging.error("os create command '%s' returned error: %s output: %s",
1002
                  result.cmd, result.fail_reason, result.output)
1003
    lines = [utils.SafeEncode(val)
1004
             for val in utils.TailFile(logfile, lines=20)]
1005
    _Fail("OS rename script failed (%s), last lines in the"
1006
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1007

    
1008

    
1009
def _GetBlockDevSymlinkPath(instance_name, idx):
1010
  return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1011
                        (instance_name, constants.DISK_SEPARATOR, idx))
1012

    
1013

    
1014
def _SymlinkBlockDev(instance_name, device_path, idx):
1015
  """Set up symlinks to a instance's block device.
1016

1017
  This is an auxiliary function run when an instance is start (on the primary
1018
  node) or when an instance is migrated (on the target node).
1019

1020

1021
  @param instance_name: the name of the target instance
1022
  @param device_path: path of the physical block device, on the node
1023
  @param idx: the disk index
1024
  @return: absolute path to the disk's symlink
1025

1026
  """
1027
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1028
  try:
1029
    os.symlink(device_path, link_name)
1030
  except OSError, err:
1031
    if err.errno == errno.EEXIST:
1032
      if (not os.path.islink(link_name) or
1033
          os.readlink(link_name) != device_path):
1034
        os.remove(link_name)
1035
        os.symlink(device_path, link_name)
1036
    else:
1037
      raise
1038

    
1039
  return link_name
1040

    
1041

    
1042
def _RemoveBlockDevLinks(instance_name, disks):
1043
  """Remove the block device symlinks belonging to the given instance.
1044

1045
  """
1046
  for idx, _ in enumerate(disks):
1047
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1048
    if os.path.islink(link_name):
1049
      try:
1050
        os.remove(link_name)
1051
      except OSError:
1052
        logging.exception("Can't remove symlink '%s'", link_name)
1053

    
1054

    
1055
def _GatherAndLinkBlockDevs(instance):
1056
  """Set up an instance's block device(s).
1057

1058
  This is run on the primary node at instance startup. The block
1059
  devices must be already assembled.
1060

1061
  @type instance: L{objects.Instance}
1062
  @param instance: the instance whose disks we shoul assemble
1063
  @rtype: list
1064
  @return: list of (disk_object, device_path)
1065

1066
  """
1067
  block_devices = []
1068
  for idx, disk in enumerate(instance.disks):
1069
    device = _RecursiveFindBD(disk)
1070
    if device is None:
1071
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1072
                                    str(disk))
1073
    device.Open()
1074
    try:
1075
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1076
    except OSError, e:
1077
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1078
                                    e.strerror)
1079

    
1080
    block_devices.append((disk, link_name))
1081

    
1082
  return block_devices
1083

    
1084

    
1085
def StartInstance(instance):
1086
  """Start an instance.
1087

1088
  @type instance: L{objects.Instance}
1089
  @param instance: the instance object
1090
  @rtype: None
1091

1092
  """
1093
  running_instances = GetInstanceList([instance.hypervisor])
1094

    
1095
  if instance.name in running_instances:
1096
    logging.info("Instance %s already running, not starting", instance.name)
1097
    return
1098

    
1099
  try:
1100
    block_devices = _GatherAndLinkBlockDevs(instance)
1101
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1102
    hyper.StartInstance(instance, block_devices)
1103
  except errors.BlockDeviceError, err:
1104
    _Fail("Block device error: %s", err, exc=True)
1105
  except errors.HypervisorError, err:
1106
    _RemoveBlockDevLinks(instance.name, instance.disks)
1107
    _Fail("Hypervisor error: %s", err, exc=True)
1108

    
1109

    
1110
def InstanceShutdown(instance, timeout):
1111
  """Shut an instance down.
1112

1113
  @note: this functions uses polling with a hardcoded timeout.
1114

1115
  @type instance: L{objects.Instance}
1116
  @param instance: the instance object
1117
  @type timeout: integer
1118
  @param timeout: maximum timeout for soft shutdown
1119
  @rtype: None
1120

1121
  """
1122
  hv_name = instance.hypervisor
1123
  hyper = hypervisor.GetHypervisor(hv_name)
1124
  iname = instance.name
1125

    
1126
  if instance.name not in hyper.ListInstances():
1127
    logging.info("Instance %s not running, doing nothing", iname)
1128
    return
1129

    
1130
  class _TryShutdown:
1131
    def __init__(self):
1132
      self.tried_once = False
1133

    
1134
    def __call__(self):
1135
      if iname not in hyper.ListInstances():
1136
        return
1137

    
1138
      try:
1139
        hyper.StopInstance(instance, retry=self.tried_once)
1140
      except errors.HypervisorError, err:
1141
        if iname not in hyper.ListInstances():
1142
          # if the instance is no longer existing, consider this a
1143
          # success and go to cleanup
1144
          return
1145

    
1146
        _Fail("Failed to stop instance %s: %s", iname, err)
1147

    
1148
      self.tried_once = True
1149

    
1150
      raise utils.RetryAgain()
1151

    
1152
  try:
1153
    utils.Retry(_TryShutdown(), 5, timeout)
1154
  except utils.RetryTimeout:
1155
    # the shutdown did not succeed
1156
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1157

    
1158
    try:
1159
      hyper.StopInstance(instance, force=True)
1160
    except errors.HypervisorError, err:
1161
      if iname in hyper.ListInstances():
1162
        # only raise an error if the instance still exists, otherwise
1163
        # the error could simply be "instance ... unknown"!
1164
        _Fail("Failed to force stop instance %s: %s", iname, err)
1165

    
1166
    time.sleep(1)
1167

    
1168
    if iname in hyper.ListInstances():
1169
      _Fail("Could not shutdown instance %s even by destroy", iname)
1170

    
1171
  try:
1172
    hyper.CleanupInstance(instance.name)
1173
  except errors.HypervisorError, err:
1174
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1175

    
1176
  _RemoveBlockDevLinks(iname, instance.disks)
1177

    
1178

    
1179
def InstanceReboot(instance, reboot_type, shutdown_timeout):
1180
  """Reboot an instance.
1181

1182
  @type instance: L{objects.Instance}
1183
  @param instance: the instance object to reboot
1184
  @type reboot_type: str
1185
  @param reboot_type: the type of reboot, one the following
1186
    constants:
1187
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1188
        instance OS, do not recreate the VM
1189
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1190
        restart the VM (at the hypervisor level)
1191
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1192
        not accepted here, since that mode is handled differently, in
1193
        cmdlib, and translates into full stop and start of the
1194
        instance (instead of a call_instance_reboot RPC)
1195
  @type shutdown_timeout: integer
1196
  @param shutdown_timeout: maximum timeout for soft shutdown
1197
  @rtype: None
1198

1199
  """
1200
  running_instances = GetInstanceList([instance.hypervisor])
1201

    
1202
  if instance.name not in running_instances:
1203
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1204

    
1205
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1206
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1207
    try:
1208
      hyper.RebootInstance(instance)
1209
    except errors.HypervisorError, err:
1210
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1211
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1212
    try:
1213
      InstanceShutdown(instance, shutdown_timeout)
1214
      return StartInstance(instance)
1215
    except errors.HypervisorError, err:
1216
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1217
  else:
1218
    _Fail("Invalid reboot_type received: %s", reboot_type)
1219

    
1220

    
1221
def MigrationInfo(instance):
1222
  """Gather information about an instance to be migrated.
1223

1224
  @type instance: L{objects.Instance}
1225
  @param instance: the instance definition
1226

1227
  """
1228
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1229
  try:
1230
    info = hyper.MigrationInfo(instance)
1231
  except errors.HypervisorError, err:
1232
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1233
  return info
1234

    
1235

    
1236
def AcceptInstance(instance, info, target):
1237
  """Prepare the node to accept an instance.
1238

1239
  @type instance: L{objects.Instance}
1240
  @param instance: the instance definition
1241
  @type info: string/data (opaque)
1242
  @param info: migration information, from the source node
1243
  @type target: string
1244
  @param target: target host (usually ip), on this node
1245

1246
  """
1247
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1248
  try:
1249
    hyper.AcceptInstance(instance, info, target)
1250
  except errors.HypervisorError, err:
1251
    _Fail("Failed to accept instance: %s", err, exc=True)
1252

    
1253

    
1254
def FinalizeMigration(instance, info, success):
1255
  """Finalize any preparation to accept an instance.
1256

1257
  @type instance: L{objects.Instance}
1258
  @param instance: the instance definition
1259
  @type info: string/data (opaque)
1260
  @param info: migration information, from the source node
1261
  @type success: boolean
1262
  @param success: whether the migration was a success or a failure
1263

1264
  """
1265
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1266
  try:
1267
    hyper.FinalizeMigration(instance, info, success)
1268
  except errors.HypervisorError, err:
1269
    _Fail("Failed to finalize migration: %s", err, exc=True)
1270

    
1271

    
1272
def MigrateInstance(instance, target, live):
1273
  """Migrates an instance to another node.
1274

1275
  @type instance: L{objects.Instance}
1276
  @param instance: the instance definition
1277
  @type target: string
1278
  @param target: the target node name
1279
  @type live: boolean
1280
  @param live: whether the migration should be done live or not (the
1281
      interpretation of this parameter is left to the hypervisor)
1282
  @rtype: tuple
1283
  @return: a tuple of (success, msg) where:
1284
      - succes is a boolean denoting the success/failure of the operation
1285
      - msg is a string with details in case of failure
1286

1287
  """
1288
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1289

    
1290
  try:
1291
    hyper.MigrateInstance(instance, target, live)
1292
  except errors.HypervisorError, err:
1293
    _Fail("Failed to migrate instance: %s", err, exc=True)
1294

    
1295

    
1296
def BlockdevCreate(disk, size, owner, on_primary, info):
1297
  """Creates a block device for an instance.
1298

1299
  @type disk: L{objects.Disk}
1300
  @param disk: the object describing the disk we should create
1301
  @type size: int
1302
  @param size: the size of the physical underlying device, in MiB
1303
  @type owner: str
1304
  @param owner: the name of the instance for which disk is created,
1305
      used for device cache data
1306
  @type on_primary: boolean
1307
  @param on_primary:  indicates if it is the primary node or not
1308
  @type info: string
1309
  @param info: string that will be sent to the physical device
1310
      creation, used for example to set (LVM) tags on LVs
1311

1312
  @return: the new unique_id of the device (this can sometime be
1313
      computed only after creation), or None. On secondary nodes,
1314
      it's not required to return anything.
1315

1316
  """
1317
  # TODO: remove the obsolete 'size' argument
1318
  # pylint: disable-msg=W0613
1319
  clist = []
1320
  if disk.children:
1321
    for child in disk.children:
1322
      try:
1323
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1324
      except errors.BlockDeviceError, err:
1325
        _Fail("Can't assemble device %s: %s", child, err)
1326
      if on_primary or disk.AssembleOnSecondary():
1327
        # we need the children open in case the device itself has to
1328
        # be assembled
1329
        try:
1330
          # pylint: disable-msg=E1103
1331
          crdev.Open()
1332
        except errors.BlockDeviceError, err:
1333
          _Fail("Can't make child '%s' read-write: %s", child, err)
1334
      clist.append(crdev)
1335

    
1336
  try:
1337
    device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1338
  except errors.BlockDeviceError, err:
1339
    _Fail("Can't create block device: %s", err)
1340

    
1341
  if on_primary or disk.AssembleOnSecondary():
1342
    try:
1343
      device.Assemble()
1344
    except errors.BlockDeviceError, err:
1345
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1346
    device.SetSyncSpeed(constants.SYNC_SPEED)
1347
    if on_primary or disk.OpenOnSecondary():
1348
      try:
1349
        device.Open(force=True)
1350
      except errors.BlockDeviceError, err:
1351
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1352
    DevCacheManager.UpdateCache(device.dev_path, owner,
1353
                                on_primary, disk.iv_name)
1354

    
1355
  device.SetInfo(info)
1356

    
1357
  return device.unique_id
1358

    
1359

    
1360
def _WipeDevice(path, offset, size):
1361
  """This function actually wipes the device.
1362

1363
  @param path: The path to the device to wipe
1364
  @param offset: The offset in MiB in the file
1365
  @param size: The size in MiB to write
1366

1367
  """
1368
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1369
         "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1370
         "count=%d" % size]
1371
  result = utils.RunCmd(cmd)
1372

    
1373
  if result.failed:
1374
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1375
          result.fail_reason, result.output)
1376

    
1377

    
1378
def BlockdevWipe(disk, offset, size):
1379
  """Wipes a block device.
1380

1381
  @type disk: L{objects.Disk}
1382
  @param disk: the disk object we want to wipe
1383
  @type offset: int
1384
  @param offset: The offset in MiB in the file
1385
  @type size: int
1386
  @param size: The size in MiB to write
1387

1388
  """
1389
  try:
1390
    rdev = _RecursiveFindBD(disk)
1391
  except errors.BlockDeviceError:
1392
    rdev = None
1393

    
1394
  if not rdev:
1395
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1396

    
1397
  # Do cross verify some of the parameters
1398
  if offset > rdev.size:
1399
    _Fail("Offset is bigger than device size")
1400
  if (offset + size) > rdev.size:
1401
    _Fail("The provided offset and size to wipe is bigger than device size")
1402

    
1403
  _WipeDevice(rdev.dev_path, offset, size)
1404

    
1405

    
1406
def BlockdevPauseResumeSync(disks, pause):
1407
  """Pause or resume the sync of the block device.
1408

1409
  @type disks: list of L{objects.Disk}
1410
  @param disks: the disks object we want to pause/resume
1411
  @type pause: bool
1412
  @param pause: Wheater to pause or resume
1413

1414
  """
1415
  success = []
1416
  for disk in disks:
1417
    try:
1418
      rdev = _RecursiveFindBD(disk)
1419
    except errors.BlockDeviceError:
1420
      rdev = None
1421

    
1422
    if not rdev:
1423
      success.append((False, ("Cannot change sync for device %s:"
1424
                              " device not found" % disk.iv_name)))
1425
      continue
1426

    
1427
    result = rdev.PauseResumeSync(pause)
1428

    
1429
    if result:
1430
      success.append((result, None))
1431
    else:
1432
      if pause:
1433
        msg = "Pause"
1434
      else:
1435
        msg = "Resume"
1436
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1437

    
1438
  return success
1439

    
1440

    
1441
def BlockdevRemove(disk):
1442
  """Remove a block device.
1443

1444
  @note: This is intended to be called recursively.
1445

1446
  @type disk: L{objects.Disk}
1447
  @param disk: the disk object we should remove
1448
  @rtype: boolean
1449
  @return: the success of the operation
1450

1451
  """
1452
  msgs = []
1453
  try:
1454
    rdev = _RecursiveFindBD(disk)
1455
  except errors.BlockDeviceError, err:
1456
    # probably can't attach
1457
    logging.info("Can't attach to device %s in remove", disk)
1458
    rdev = None
1459
  if rdev is not None:
1460
    r_path = rdev.dev_path
1461
    try:
1462
      rdev.Remove()
1463
    except errors.BlockDeviceError, err:
1464
      msgs.append(str(err))
1465
    if not msgs:
1466
      DevCacheManager.RemoveCache(r_path)
1467

    
1468
  if disk.children:
1469
    for child in disk.children:
1470
      try:
1471
        BlockdevRemove(child)
1472
      except RPCFail, err:
1473
        msgs.append(str(err))
1474

    
1475
  if msgs:
1476
    _Fail("; ".join(msgs))
1477

    
1478

    
1479
def _RecursiveAssembleBD(disk, owner, as_primary):
1480
  """Activate a block device for an instance.
1481

1482
  This is run on the primary and secondary nodes for an instance.
1483

1484
  @note: this function is called recursively.
1485

1486
  @type disk: L{objects.Disk}
1487
  @param disk: the disk we try to assemble
1488
  @type owner: str
1489
  @param owner: the name of the instance which owns the disk
1490
  @type as_primary: boolean
1491
  @param as_primary: if we should make the block device
1492
      read/write
1493

1494
  @return: the assembled device or None (in case no device
1495
      was assembled)
1496
  @raise errors.BlockDeviceError: in case there is an error
1497
      during the activation of the children or the device
1498
      itself
1499

1500
  """
1501
  children = []
1502
  if disk.children:
1503
    mcn = disk.ChildrenNeeded()
1504
    if mcn == -1:
1505
      mcn = 0 # max number of Nones allowed
1506
    else:
1507
      mcn = len(disk.children) - mcn # max number of Nones
1508
    for chld_disk in disk.children:
1509
      try:
1510
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1511
      except errors.BlockDeviceError, err:
1512
        if children.count(None) >= mcn:
1513
          raise
1514
        cdev = None
1515
        logging.error("Error in child activation (but continuing): %s",
1516
                      str(err))
1517
      children.append(cdev)
1518

    
1519
  if as_primary or disk.AssembleOnSecondary():
1520
    r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1521
    r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1522
    result = r_dev
1523
    if as_primary or disk.OpenOnSecondary():
1524
      r_dev.Open()
1525
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1526
                                as_primary, disk.iv_name)
1527

    
1528
  else:
1529
    result = True
1530
  return result
1531

    
1532

    
1533
def BlockdevAssemble(disk, owner, as_primary, idx):
1534
  """Activate a block device for an instance.
1535

1536
  This is a wrapper over _RecursiveAssembleBD.
1537

1538
  @rtype: str or boolean
1539
  @return: a C{/dev/...} path for primary nodes, and
1540
      C{True} for secondary nodes
1541

1542
  """
1543
  try:
1544
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1545
    if isinstance(result, bdev.BlockDev):
1546
      # pylint: disable-msg=E1103
1547
      result = result.dev_path
1548
      if as_primary:
1549
        _SymlinkBlockDev(owner, result, idx)
1550
  except errors.BlockDeviceError, err:
1551
    _Fail("Error while assembling disk: %s", err, exc=True)
1552
  except OSError, err:
1553
    _Fail("Error while symlinking disk: %s", err, exc=True)
1554

    
1555
  return result
1556

    
1557

    
1558
def BlockdevShutdown(disk):
1559
  """Shut down a block device.
1560

1561
  First, if the device is assembled (Attach() is successful), then
1562
  the device is shutdown. Then the children of the device are
1563
  shutdown.
1564

1565
  This function is called recursively. Note that we don't cache the
1566
  children or such, as oppossed to assemble, shutdown of different
1567
  devices doesn't require that the upper device was active.
1568

1569
  @type disk: L{objects.Disk}
1570
  @param disk: the description of the disk we should
1571
      shutdown
1572
  @rtype: None
1573

1574
  """
1575
  msgs = []
1576
  r_dev = _RecursiveFindBD(disk)
1577
  if r_dev is not None:
1578
    r_path = r_dev.dev_path
1579
    try:
1580
      r_dev.Shutdown()
1581
      DevCacheManager.RemoveCache(r_path)
1582
    except errors.BlockDeviceError, err:
1583
      msgs.append(str(err))
1584

    
1585
  if disk.children:
1586
    for child in disk.children:
1587
      try:
1588
        BlockdevShutdown(child)
1589
      except RPCFail, err:
1590
        msgs.append(str(err))
1591

    
1592
  if msgs:
1593
    _Fail("; ".join(msgs))
1594

    
1595

    
1596
def BlockdevAddchildren(parent_cdev, new_cdevs):
1597
  """Extend a mirrored block device.
1598

1599
  @type parent_cdev: L{objects.Disk}
1600
  @param parent_cdev: the disk to which we should add children
1601
  @type new_cdevs: list of L{objects.Disk}
1602
  @param new_cdevs: the list of children which we should add
1603
  @rtype: None
1604

1605
  """
1606
  parent_bdev = _RecursiveFindBD(parent_cdev)
1607
  if parent_bdev is None:
1608
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1609
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1610
  if new_bdevs.count(None) > 0:
1611
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1612
  parent_bdev.AddChildren(new_bdevs)
1613

    
1614

    
1615
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1616
  """Shrink a mirrored block device.
1617

1618
  @type parent_cdev: L{objects.Disk}
1619
  @param parent_cdev: the disk from which we should remove children
1620
  @type new_cdevs: list of L{objects.Disk}
1621
  @param new_cdevs: the list of children which we should remove
1622
  @rtype: None
1623

1624
  """
1625
  parent_bdev = _RecursiveFindBD(parent_cdev)
1626
  if parent_bdev is None:
1627
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1628
  devs = []
1629
  for disk in new_cdevs:
1630
    rpath = disk.StaticDevPath()
1631
    if rpath is None:
1632
      bd = _RecursiveFindBD(disk)
1633
      if bd is None:
1634
        _Fail("Can't find device %s while removing children", disk)
1635
      else:
1636
        devs.append(bd.dev_path)
1637
    else:
1638
      if not utils.IsNormAbsPath(rpath):
1639
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1640
      devs.append(rpath)
1641
  parent_bdev.RemoveChildren(devs)
1642

    
1643

    
1644
def BlockdevGetmirrorstatus(disks):
1645
  """Get the mirroring status of a list of devices.
1646

1647
  @type disks: list of L{objects.Disk}
1648
  @param disks: the list of disks which we should query
1649
  @rtype: disk
1650
  @return: List of L{objects.BlockDevStatus}, one for each disk
1651
  @raise errors.BlockDeviceError: if any of the disks cannot be
1652
      found
1653

1654
  """
1655
  stats = []
1656
  for dsk in disks:
1657
    rbd = _RecursiveFindBD(dsk)
1658
    if rbd is None:
1659
      _Fail("Can't find device %s", dsk)
1660

    
1661
    stats.append(rbd.CombinedSyncStatus())
1662

    
1663
  return stats
1664

    
1665

    
1666
def BlockdevGetmirrorstatusMulti(disks):
1667
  """Get the mirroring status of a list of devices.
1668

1669
  @type disks: list of L{objects.Disk}
1670
  @param disks: the list of disks which we should query
1671
  @rtype: disk
1672
  @return: List of tuples, (bool, status), one for each disk; bool denotes
1673
    success/failure, status is L{objects.BlockDevStatus} on success, string
1674
    otherwise
1675

1676
  """
1677
  result = []
1678
  for disk in disks:
1679
    try:
1680
      rbd = _RecursiveFindBD(disk)
1681
      if rbd is None:
1682
        result.append((False, "Can't find device %s" % disk))
1683
        continue
1684

    
1685
      status = rbd.CombinedSyncStatus()
1686
    except errors.BlockDeviceError, err:
1687
      logging.exception("Error while getting disk status")
1688
      result.append((False, str(err)))
1689
    else:
1690
      result.append((True, status))
1691

    
1692
  assert len(disks) == len(result)
1693

    
1694
  return result
1695

    
1696

    
1697
def _RecursiveFindBD(disk):
1698
  """Check if a device is activated.
1699

1700
  If so, return information about the real device.
1701

1702
  @type disk: L{objects.Disk}
1703
  @param disk: the disk object we need to find
1704

1705
  @return: None if the device can't be found,
1706
      otherwise the device instance
1707

1708
  """
1709
  children = []
1710
  if disk.children:
1711
    for chdisk in disk.children:
1712
      children.append(_RecursiveFindBD(chdisk))
1713

    
1714
  return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1715

    
1716

    
1717
def _OpenRealBD(disk):
1718
  """Opens the underlying block device of a disk.
1719

1720
  @type disk: L{objects.Disk}
1721
  @param disk: the disk object we want to open
1722

1723
  """
1724
  real_disk = _RecursiveFindBD(disk)
1725
  if real_disk is None:
1726
    _Fail("Block device '%s' is not set up", disk)
1727

    
1728
  real_disk.Open()
1729

    
1730
  return real_disk
1731

    
1732

    
1733
def BlockdevFind(disk):
1734
  """Check if a device is activated.
1735

1736
  If it is, return information about the real device.
1737

1738
  @type disk: L{objects.Disk}
1739
  @param disk: the disk to find
1740
  @rtype: None or objects.BlockDevStatus
1741
  @return: None if the disk cannot be found, otherwise a the current
1742
           information
1743

1744
  """
1745
  try:
1746
    rbd = _RecursiveFindBD(disk)
1747
  except errors.BlockDeviceError, err:
1748
    _Fail("Failed to find device: %s", err, exc=True)
1749

    
1750
  if rbd is None:
1751
    return None
1752

    
1753
  return rbd.GetSyncStatus()
1754

    
1755

    
1756
def BlockdevGetsize(disks):
1757
  """Computes the size of the given disks.
1758

1759
  If a disk is not found, returns None instead.
1760

1761
  @type disks: list of L{objects.Disk}
1762
  @param disks: the list of disk to compute the size for
1763
  @rtype: list
1764
  @return: list with elements None if the disk cannot be found,
1765
      otherwise the size
1766

1767
  """
1768
  result = []
1769
  for cf in disks:
1770
    try:
1771
      rbd = _RecursiveFindBD(cf)
1772
    except errors.BlockDeviceError:
1773
      result.append(None)
1774
      continue
1775
    if rbd is None:
1776
      result.append(None)
1777
    else:
1778
      result.append(rbd.GetActualSize())
1779
  return result
1780

    
1781

    
1782
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1783
  """Export a block device to a remote node.
1784

1785
  @type disk: L{objects.Disk}
1786
  @param disk: the description of the disk to export
1787
  @type dest_node: str
1788
  @param dest_node: the destination node to export to
1789
  @type dest_path: str
1790
  @param dest_path: the destination path on the target node
1791
  @type cluster_name: str
1792
  @param cluster_name: the cluster name, needed for SSH hostalias
1793
  @rtype: None
1794

1795
  """
1796
  real_disk = _OpenRealBD(disk)
1797

    
1798
  # the block size on the read dd is 1MiB to match our units
1799
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1800
                               "dd if=%s bs=1048576 count=%s",
1801
                               real_disk.dev_path, str(disk.size))
1802

    
1803
  # we set here a smaller block size as, due to ssh buffering, more
1804
  # than 64-128k will mostly ignored; we use nocreat to fail if the
1805
  # device is not already there or we pass a wrong path; we use
1806
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1807
  # to not buffer too much memory; this means that at best, we flush
1808
  # every 64k, which will not be very fast
1809
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1810
                                " oflag=dsync", dest_path)
1811

    
1812
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1813
                                                   constants.GANETI_RUNAS,
1814
                                                   destcmd)
1815

    
1816
  # all commands have been checked, so we're safe to combine them
1817
  command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1818

    
1819
  result = utils.RunCmd(["bash", "-c", command])
1820

    
1821
  if result.failed:
1822
    _Fail("Disk copy command '%s' returned error: %s"
1823
          " output: %s", command, result.fail_reason, result.output)
1824

    
1825

    
1826
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1827
  """Write a file to the filesystem.
1828

1829
  This allows the master to overwrite(!) a file. It will only perform
1830
  the operation if the file belongs to a list of configuration files.
1831

1832
  @type file_name: str
1833
  @param file_name: the target file name
1834
  @type data: str
1835
  @param data: the new contents of the file
1836
  @type mode: int
1837
  @param mode: the mode to give the file (can be None)
1838
  @type uid: int
1839
  @param uid: the owner of the file (can be -1 for default)
1840
  @type gid: int
1841
  @param gid: the group of the file (can be -1 for default)
1842
  @type atime: float
1843
  @param atime: the atime to set on the file (can be None)
1844
  @type mtime: float
1845
  @param mtime: the mtime to set on the file (can be None)
1846
  @rtype: None
1847

1848
  """
1849
  if not os.path.isabs(file_name):
1850
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1851

    
1852
  if file_name not in _ALLOWED_UPLOAD_FILES:
1853
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1854
          file_name)
1855

    
1856
  raw_data = _Decompress(data)
1857

    
1858
  utils.SafeWriteFile(file_name, None,
1859
                      data=raw_data, mode=mode, uid=uid, gid=gid,
1860
                      atime=atime, mtime=mtime)
1861

    
1862

    
1863
def RunOob(oob_program, command, node, timeout):
1864
  """Executes oob_program with given command on given node.
1865

1866
  @param oob_program: The path to the executable oob_program
1867
  @param command: The command to invoke on oob_program
1868
  @param node: The node given as an argument to the program
1869
  @param timeout: Timeout after which we kill the oob program
1870

1871
  @return: stdout
1872
  @raise RPCFail: If execution fails for some reason
1873

1874
  """
1875
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
1876

    
1877
  if result.failed:
1878
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
1879
          result.fail_reason, result.output)
1880

    
1881
  return result.stdout
1882

    
1883

    
1884
def WriteSsconfFiles(values):
1885
  """Update all ssconf files.
1886

1887
  Wrapper around the SimpleStore.WriteFiles.
1888

1889
  """
1890
  ssconf.SimpleStore().WriteFiles(values)
1891

    
1892

    
1893
def _ErrnoOrStr(err):
1894
  """Format an EnvironmentError exception.
1895

1896
  If the L{err} argument has an errno attribute, it will be looked up
1897
  and converted into a textual C{E...} description. Otherwise the
1898
  string representation of the error will be returned.
1899

1900
  @type err: L{EnvironmentError}
1901
  @param err: the exception to format
1902

1903
  """
1904
  if hasattr(err, 'errno'):
1905
    detail = errno.errorcode[err.errno]
1906
  else:
1907
    detail = str(err)
1908
  return detail
1909

    
1910

    
1911
def _OSOndiskAPIVersion(os_dir):
1912
  """Compute and return the API version of a given OS.
1913

1914
  This function will try to read the API version of the OS residing in
1915
  the 'os_dir' directory.
1916

1917
  @type os_dir: str
1918
  @param os_dir: the directory in which we should look for the OS
1919
  @rtype: tuple
1920
  @return: tuple (status, data) with status denoting the validity and
1921
      data holding either the vaid versions or an error message
1922

1923
  """
1924
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1925

    
1926
  try:
1927
    st = os.stat(api_file)
1928
  except EnvironmentError, err:
1929
    return False, ("Required file '%s' not found under path %s: %s" %
1930
                   (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1931

    
1932
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1933
    return False, ("File '%s' in %s is not a regular file" %
1934
                   (constants.OS_API_FILE, os_dir))
1935

    
1936
  try:
1937
    api_versions = utils.ReadFile(api_file).splitlines()
1938
  except EnvironmentError, err:
1939
    return False, ("Error while reading the API version file at %s: %s" %
1940
                   (api_file, _ErrnoOrStr(err)))
1941

    
1942
  try:
1943
    api_versions = [int(version.strip()) for version in api_versions]
1944
  except (TypeError, ValueError), err:
1945
    return False, ("API version(s) can't be converted to integer: %s" %
1946
                   str(err))
1947

    
1948
  return True, api_versions
1949

    
1950

    
1951
def DiagnoseOS(top_dirs=None):
1952
  """Compute the validity for all OSes.
1953

1954
  @type top_dirs: list
1955
  @param top_dirs: the list of directories in which to
1956
      search (if not given defaults to
1957
      L{constants.OS_SEARCH_PATH})
1958
  @rtype: list of L{objects.OS}
1959
  @return: a list of tuples (name, path, status, diagnose, variants,
1960
      parameters, api_version) for all (potential) OSes under all
1961
      search paths, where:
1962
          - name is the (potential) OS name
1963
          - path is the full path to the OS
1964
          - status True/False is the validity of the OS
1965
          - diagnose is the error message for an invalid OS, otherwise empty
1966
          - variants is a list of supported OS variants, if any
1967
          - parameters is a list of (name, help) parameters, if any
1968
          - api_version is a list of support OS API versions
1969

1970
  """
1971
  if top_dirs is None:
1972
    top_dirs = constants.OS_SEARCH_PATH
1973

    
1974
  result = []
1975
  for dir_name in top_dirs:
1976
    if os.path.isdir(dir_name):
1977
      try:
1978
        f_names = utils.ListVisibleFiles(dir_name)
1979
      except EnvironmentError, err:
1980
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1981
        break
1982
      for name in f_names:
1983
        os_path = utils.PathJoin(dir_name, name)
1984
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1985
        if status:
1986
          diagnose = ""
1987
          variants = os_inst.supported_variants
1988
          parameters = os_inst.supported_parameters
1989
          api_versions = os_inst.api_versions
1990
        else:
1991
          diagnose = os_inst
1992
          variants = parameters = api_versions = []
1993
        result.append((name, os_path, status, diagnose, variants,
1994
                       parameters, api_versions))
1995

    
1996
  return result
1997

    
1998

    
1999
def _TryOSFromDisk(name, base_dir=None):
2000
  """Create an OS instance from disk.
2001

2002
  This function will return an OS instance if the given name is a
2003
  valid OS name.
2004

2005
  @type base_dir: string
2006
  @keyword base_dir: Base directory containing OS installations.
2007
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2008
  @rtype: tuple
2009
  @return: success and either the OS instance if we find a valid one,
2010
      or error message
2011

2012
  """
2013
  if base_dir is None:
2014
    os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2015
  else:
2016
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2017

    
2018
  if os_dir is None:
2019
    return False, "Directory for OS %s not found in search path" % name
2020

    
2021
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2022
  if not status:
2023
    # push the error up
2024
    return status, api_versions
2025

    
2026
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2027
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2028
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2029

    
2030
  # OS Files dictionary, we will populate it with the absolute path names
2031
  os_files = dict.fromkeys(constants.OS_SCRIPTS)
2032

    
2033
  if max(api_versions) >= constants.OS_API_V15:
2034
    os_files[constants.OS_VARIANTS_FILE] = ''
2035

    
2036
  if max(api_versions) >= constants.OS_API_V20:
2037
    os_files[constants.OS_PARAMETERS_FILE] = ''
2038
  else:
2039
    del os_files[constants.OS_SCRIPT_VERIFY]
2040

    
2041
  for filename in os_files:
2042
    os_files[filename] = utils.PathJoin(os_dir, filename)
2043

    
2044
    try:
2045
      st = os.stat(os_files[filename])
2046
    except EnvironmentError, err:
2047
      return False, ("File '%s' under path '%s' is missing (%s)" %
2048
                     (filename, os_dir, _ErrnoOrStr(err)))
2049

    
2050
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2051
      return False, ("File '%s' under path '%s' is not a regular file" %
2052
                     (filename, os_dir))
2053

    
2054
    if filename in constants.OS_SCRIPTS:
2055
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2056
        return False, ("File '%s' under path '%s' is not executable" %
2057
                       (filename, os_dir))
2058

    
2059
  variants = []
2060
  if constants.OS_VARIANTS_FILE in os_files:
2061
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2062
    try:
2063
      variants = utils.ReadFile(variants_file).splitlines()
2064
    except EnvironmentError, err:
2065
      return False, ("Error while reading the OS variants file at %s: %s" %
2066
                     (variants_file, _ErrnoOrStr(err)))
2067
    if not variants:
2068
      return False, ("No supported os variant found")
2069

    
2070
  parameters = []
2071
  if constants.OS_PARAMETERS_FILE in os_files:
2072
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2073
    try:
2074
      parameters = utils.ReadFile(parameters_file).splitlines()
2075
    except EnvironmentError, err:
2076
      return False, ("Error while reading the OS parameters file at %s: %s" %
2077
                     (parameters_file, _ErrnoOrStr(err)))
2078
    parameters = [v.split(None, 1) for v in parameters]
2079

    
2080
  os_obj = objects.OS(name=name, path=os_dir,
2081
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2082
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2083
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2084
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2085
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2086
                                                 None),
2087
                      supported_variants=variants,
2088
                      supported_parameters=parameters,
2089
                      api_versions=api_versions)
2090
  return True, os_obj
2091

    
2092

    
2093
def OSFromDisk(name, base_dir=None):
2094
  """Create an OS instance from disk.
2095

2096
  This function will return an OS instance if the given name is a
2097
  valid OS name. Otherwise, it will raise an appropriate
2098
  L{RPCFail} exception, detailing why this is not a valid OS.
2099

2100
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2101
  an exception but returns true/false status data.
2102

2103
  @type base_dir: string
2104
  @keyword base_dir: Base directory containing OS installations.
2105
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2106
  @rtype: L{objects.OS}
2107
  @return: the OS instance if we find a valid one
2108
  @raise RPCFail: if we don't find a valid OS
2109

2110
  """
2111
  name_only = objects.OS.GetName(name)
2112
  status, payload = _TryOSFromDisk(name_only, base_dir)
2113

    
2114
  if not status:
2115
    _Fail(payload)
2116

    
2117
  return payload
2118

    
2119

    
2120
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2121
  """Calculate the basic environment for an os script.
2122

2123
  @type os_name: str
2124
  @param os_name: full operating system name (including variant)
2125
  @type inst_os: L{objects.OS}
2126
  @param inst_os: operating system for which the environment is being built
2127
  @type os_params: dict
2128
  @param os_params: the OS parameters
2129
  @type debug: integer
2130
  @param debug: debug level (0 or 1, for OS Api 10)
2131
  @rtype: dict
2132
  @return: dict of environment variables
2133
  @raise errors.BlockDeviceError: if the block device
2134
      cannot be found
2135

2136
  """
2137
  result = {}
2138
  api_version = \
2139
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2140
  result['OS_API_VERSION'] = '%d' % api_version
2141
  result['OS_NAME'] = inst_os.name
2142
  result['DEBUG_LEVEL'] = '%d' % debug
2143

    
2144
  # OS variants
2145
  if api_version >= constants.OS_API_V15:
2146
    variant = objects.OS.GetVariant(os_name)
2147
    if not variant:
2148
      variant = inst_os.supported_variants[0]
2149
    result['OS_VARIANT'] = variant
2150

    
2151
  # OS params
2152
  for pname, pvalue in os_params.items():
2153
    result['OSP_%s' % pname.upper()] = pvalue
2154

    
2155
  return result
2156

    
2157

    
2158
def OSEnvironment(instance, inst_os, debug=0):
2159
  """Calculate the environment for an os script.
2160

2161
  @type instance: L{objects.Instance}
2162
  @param instance: target instance for the os script run
2163
  @type inst_os: L{objects.OS}
2164
  @param inst_os: operating system for which the environment is being built
2165
  @type debug: integer
2166
  @param debug: debug level (0 or 1, for OS Api 10)
2167
  @rtype: dict
2168
  @return: dict of environment variables
2169
  @raise errors.BlockDeviceError: if the block device
2170
      cannot be found
2171

2172
  """
2173
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2174

    
2175
  for attr in ["name", "os", "uuid", "ctime", "mtime"]:
2176
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2177

    
2178
  result['HYPERVISOR'] = instance.hypervisor
2179
  result['DISK_COUNT'] = '%d' % len(instance.disks)
2180
  result['NIC_COUNT'] = '%d' % len(instance.nics)
2181

    
2182
  # Disks
2183
  for idx, disk in enumerate(instance.disks):
2184
    real_disk = _OpenRealBD(disk)
2185
    result['DISK_%d_PATH' % idx] = real_disk.dev_path
2186
    result['DISK_%d_ACCESS' % idx] = disk.mode
2187
    if constants.HV_DISK_TYPE in instance.hvparams:
2188
      result['DISK_%d_FRONTEND_TYPE' % idx] = \
2189
        instance.hvparams[constants.HV_DISK_TYPE]
2190
    if disk.dev_type in constants.LDS_BLOCK:
2191
      result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
2192
    elif disk.dev_type == constants.LD_FILE:
2193
      result['DISK_%d_BACKEND_TYPE' % idx] = \
2194
        'file:%s' % disk.physical_id[0]
2195

    
2196
  # NICs
2197
  for idx, nic in enumerate(instance.nics):
2198
    result['NIC_%d_MAC' % idx] = nic.mac
2199
    if nic.ip:
2200
      result['NIC_%d_IP' % idx] = nic.ip
2201
    result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
2202
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2203
      result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
2204
    if nic.nicparams[constants.NIC_LINK]:
2205
      result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
2206
    if constants.HV_NIC_TYPE in instance.hvparams:
2207
      result['NIC_%d_FRONTEND_TYPE' % idx] = \
2208
        instance.hvparams[constants.HV_NIC_TYPE]
2209

    
2210
  # HV/BE params
2211
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2212
    for key, value in source.items():
2213
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2214

    
2215
  return result
2216

    
2217

    
2218
def BlockdevGrow(disk, amount):
2219
  """Grow a stack of block devices.
2220

2221
  This function is called recursively, with the childrens being the
2222
  first ones to resize.
2223

2224
  @type disk: L{objects.Disk}
2225
  @param disk: the disk to be grown
2226
  @rtype: (status, result)
2227
  @return: a tuple with the status of the operation
2228
      (True/False), and the errors message if status
2229
      is False
2230

2231
  """
2232
  r_dev = _RecursiveFindBD(disk)
2233
  if r_dev is None:
2234
    _Fail("Cannot find block device %s", disk)
2235

    
2236
  try:
2237
    r_dev.Grow(amount)
2238
  except errors.BlockDeviceError, err:
2239
    _Fail("Failed to grow block device: %s", err, exc=True)
2240

    
2241

    
2242
def BlockdevSnapshot(disk):
2243
  """Create a snapshot copy of a block device.
2244

2245
  This function is called recursively, and the snapshot is actually created
2246
  just for the leaf lvm backend device.
2247

2248
  @type disk: L{objects.Disk}
2249
  @param disk: the disk to be snapshotted
2250
  @rtype: string
2251
  @return: snapshot disk ID as (vg, lv)
2252

2253
  """
2254
  if disk.dev_type == constants.LD_DRBD8:
2255
    if not disk.children:
2256
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2257
            disk.unique_id)
2258
    return BlockdevSnapshot(disk.children[0])
2259
  elif disk.dev_type == constants.LD_LV:
2260
    r_dev = _RecursiveFindBD(disk)
2261
    if r_dev is not None:
2262
      # FIXME: choose a saner value for the snapshot size
2263
      # let's stay on the safe side and ask for the full size, for now
2264
      return r_dev.Snapshot(disk.size)
2265
    else:
2266
      _Fail("Cannot find block device %s", disk)
2267
  else:
2268
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2269
          disk.unique_id, disk.dev_type)
2270

    
2271

    
2272
def FinalizeExport(instance, snap_disks):
2273
  """Write out the export configuration information.
2274

2275
  @type instance: L{objects.Instance}
2276
  @param instance: the instance which we export, used for
2277
      saving configuration
2278
  @type snap_disks: list of L{objects.Disk}
2279
  @param snap_disks: list of snapshot block devices, which
2280
      will be used to get the actual name of the dump file
2281

2282
  @rtype: None
2283

2284
  """
2285
  destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2286
  finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2287

    
2288
  config = objects.SerializableConfigParser()
2289

    
2290
  config.add_section(constants.INISECT_EXP)
2291
  config.set(constants.INISECT_EXP, 'version', '0')
2292
  config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2293
  config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2294
  config.set(constants.INISECT_EXP, 'os', instance.os)
2295
  config.set(constants.INISECT_EXP, "compression", "none")
2296

    
2297
  config.add_section(constants.INISECT_INS)
2298
  config.set(constants.INISECT_INS, 'name', instance.name)
2299
  config.set(constants.INISECT_INS, 'memory', '%d' %
2300
             instance.beparams[constants.BE_MEMORY])
2301
  config.set(constants.INISECT_INS, 'vcpus', '%d' %
2302
             instance.beparams[constants.BE_VCPUS])
2303
  config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2304
  config.set(constants.INISECT_INS, 'hypervisor', instance.hypervisor)
2305

    
2306
  nic_total = 0
2307
  for nic_count, nic in enumerate(instance.nics):
2308
    nic_total += 1
2309
    config.set(constants.INISECT_INS, 'nic%d_mac' %
2310
               nic_count, '%s' % nic.mac)
2311
    config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2312
    for param in constants.NICS_PARAMETER_TYPES:
2313
      config.set(constants.INISECT_INS, 'nic%d_%s' % (nic_count, param),
2314
                 '%s' % nic.nicparams.get(param, None))
2315
  # TODO: redundant: on load can read nics until it doesn't exist
2316
  config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2317

    
2318
  disk_total = 0
2319
  for disk_count, disk in enumerate(snap_disks):
2320
    if disk:
2321
      disk_total += 1
2322
      config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2323
                 ('%s' % disk.iv_name))
2324
      config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2325
                 ('%s' % disk.physical_id[1]))
2326
      config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2327
                 ('%d' % disk.size))
2328

    
2329
  config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2330

    
2331
  # New-style hypervisor/backend parameters
2332

    
2333
  config.add_section(constants.INISECT_HYP)
2334
  for name, value in instance.hvparams.items():
2335
    if name not in constants.HVC_GLOBALS:
2336
      config.set(constants.INISECT_HYP, name, str(value))
2337

    
2338
  config.add_section(constants.INISECT_BEP)
2339
  for name, value in instance.beparams.items():
2340
    config.set(constants.INISECT_BEP, name, str(value))
2341

    
2342
  config.add_section(constants.INISECT_OSP)
2343
  for name, value in instance.osparams.items():
2344
    config.set(constants.INISECT_OSP, name, str(value))
2345

    
2346
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2347
                  data=config.Dumps())
2348
  shutil.rmtree(finaldestdir, ignore_errors=True)
2349
  shutil.move(destdir, finaldestdir)
2350

    
2351

    
2352
def ExportInfo(dest):
2353
  """Get export configuration information.
2354

2355
  @type dest: str
2356
  @param dest: directory containing the export
2357

2358
  @rtype: L{objects.SerializableConfigParser}
2359
  @return: a serializable config file containing the
2360
      export info
2361

2362
  """
2363
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2364

    
2365
  config = objects.SerializableConfigParser()
2366
  config.read(cff)
2367

    
2368
  if (not config.has_section(constants.INISECT_EXP) or
2369
      not config.has_section(constants.INISECT_INS)):
2370
    _Fail("Export info file doesn't have the required fields")
2371

    
2372
  return config.Dumps()
2373

    
2374

    
2375
def ListExports():
2376
  """Return a list of exports currently available on this machine.
2377

2378
  @rtype: list
2379
  @return: list of the exports
2380

2381
  """
2382
  if os.path.isdir(constants.EXPORT_DIR):
2383
    return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2384
  else:
2385
    _Fail("No exports directory")
2386

    
2387

    
2388
def RemoveExport(export):
2389
  """Remove an existing export from the node.
2390

2391
  @type export: str
2392
  @param export: the name of the export to remove
2393
  @rtype: None
2394

2395
  """
2396
  target = utils.PathJoin(constants.EXPORT_DIR, export)
2397

    
2398
  try:
2399
    shutil.rmtree(target)
2400
  except EnvironmentError, err:
2401
    _Fail("Error while removing the export: %s", err, exc=True)
2402

    
2403

    
2404
def BlockdevRename(devlist):
2405
  """Rename a list of block devices.
2406

2407
  @type devlist: list of tuples
2408
  @param devlist: list of tuples of the form  (disk,
2409
      new_logical_id, new_physical_id); disk is an
2410
      L{objects.Disk} object describing the current disk,
2411
      and new logical_id/physical_id is the name we
2412
      rename it to
2413
  @rtype: boolean
2414
  @return: True if all renames succeeded, False otherwise
2415

2416
  """
2417
  msgs = []
2418
  result = True
2419
  for disk, unique_id in devlist:
2420
    dev = _RecursiveFindBD(disk)
2421
    if dev is None:
2422
      msgs.append("Can't find device %s in rename" % str(disk))
2423
      result = False
2424
      continue
2425
    try:
2426
      old_rpath = dev.dev_path
2427
      dev.Rename(unique_id)
2428
      new_rpath = dev.dev_path
2429
      if old_rpath != new_rpath:
2430
        DevCacheManager.RemoveCache(old_rpath)
2431
        # FIXME: we should add the new cache information here, like:
2432
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2433
        # but we don't have the owner here - maybe parse from existing
2434
        # cache? for now, we only lose lvm data when we rename, which
2435
        # is less critical than DRBD or MD
2436
    except errors.BlockDeviceError, err:
2437
      msgs.append("Can't rename device '%s' to '%s': %s" %
2438
                  (dev, unique_id, err))
2439
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2440
      result = False
2441
  if not result:
2442
    _Fail("; ".join(msgs))
2443

    
2444

    
2445
def _TransformFileStorageDir(fs_dir):
2446
  """Checks whether given file_storage_dir is valid.
2447

2448
  Checks wheter the given fs_dir is within the cluster-wide default
2449
  file_storage_dir or the shared_file_storage_dir, which are stored in
2450
  SimpleStore. Only paths under those directories are allowed.
2451

2452
  @type fs_dir: str
2453
  @param fs_dir: the path to check
2454

2455
  @return: the normalized path if valid, None otherwise
2456

2457
  """
2458
  if not constants.ENABLE_FILE_STORAGE:
2459
    _Fail("File storage disabled at configure time")
2460
  cfg = _GetConfig()
2461
  fs_dir = os.path.normpath(fs_dir)
2462
  base_fstore = cfg.GetFileStorageDir()
2463
  base_shared = cfg.GetSharedFileStorageDir()
2464
  if ((os.path.commonprefix([fs_dir, base_fstore]) != base_fstore) and
2465
      (os.path.commonprefix([fs_dir, base_shared]) != base_shared)):
2466
    _Fail("File storage directory '%s' is not under base file"
2467
          " storage directory '%s' or shared storage directory '%s'",
2468
          fs_dir, base_fstore, base_shared)
2469
  return fs_dir
2470

    
2471

    
2472
def CreateFileStorageDir(file_storage_dir):
2473
  """Create file storage directory.
2474

2475
  @type file_storage_dir: str
2476
  @param file_storage_dir: directory to create
2477

2478
  @rtype: tuple
2479
  @return: tuple with first element a boolean indicating wheter dir
2480
      creation was successful or not
2481

2482
  """
2483
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2484
  if os.path.exists(file_storage_dir):
2485
    if not os.path.isdir(file_storage_dir):
2486
      _Fail("Specified storage dir '%s' is not a directory",
2487
            file_storage_dir)
2488
  else:
2489
    try:
2490
      os.makedirs(file_storage_dir, 0750)
2491
    except OSError, err:
2492
      _Fail("Cannot create file storage directory '%s': %s",
2493
            file_storage_dir, err, exc=True)
2494

    
2495

    
2496
def RemoveFileStorageDir(file_storage_dir):
2497
  """Remove file storage directory.
2498

2499
  Remove it only if it's empty. If not log an error and return.
2500

2501
  @type file_storage_dir: str
2502
  @param file_storage_dir: the directory we should cleanup
2503
  @rtype: tuple (success,)
2504
  @return: tuple of one element, C{success}, denoting
2505
      whether the operation was successful
2506

2507
  """
2508
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2509
  if os.path.exists(file_storage_dir):
2510
    if not os.path.isdir(file_storage_dir):
2511
      _Fail("Specified Storage directory '%s' is not a directory",
2512
            file_storage_dir)
2513
    # deletes dir only if empty, otherwise we want to fail the rpc call
2514
    try:
2515
      os.rmdir(file_storage_dir)
2516
    except OSError, err:
2517
      _Fail("Cannot remove file storage directory '%s': %s",
2518
            file_storage_dir, err)
2519

    
2520

    
2521
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2522
  """Rename the file storage directory.
2523

2524
  @type old_file_storage_dir: str
2525
  @param old_file_storage_dir: the current path
2526
  @type new_file_storage_dir: str
2527
  @param new_file_storage_dir: the name we should rename to
2528
  @rtype: tuple (success,)
2529
  @return: tuple of one element, C{success}, denoting
2530
      whether the operation was successful
2531

2532
  """
2533
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2534
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2535
  if not os.path.exists(new_file_storage_dir):
2536
    if os.path.isdir(old_file_storage_dir):
2537
      try:
2538
        os.rename(old_file_storage_dir, new_file_storage_dir)
2539
      except OSError, err:
2540
        _Fail("Cannot rename '%s' to '%s': %s",
2541
              old_file_storage_dir, new_file_storage_dir, err)
2542
    else:
2543
      _Fail("Specified storage dir '%s' is not a directory",
2544
            old_file_storage_dir)
2545
  else:
2546
    if os.path.exists(old_file_storage_dir):
2547
      _Fail("Cannot rename '%s' to '%s': both locations exist",
2548
            old_file_storage_dir, new_file_storage_dir)
2549

    
2550

    
2551
def _EnsureJobQueueFile(file_name):
2552
  """Checks whether the given filename is in the queue directory.
2553

2554
  @type file_name: str
2555
  @param file_name: the file name we should check
2556
  @rtype: None
2557
  @raises RPCFail: if the file is not valid
2558

2559
  """
2560
  queue_dir = os.path.normpath(constants.QUEUE_DIR)
2561
  result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2562

    
2563
  if not result:
2564
    _Fail("Passed job queue file '%s' does not belong to"
2565
          " the queue directory '%s'", file_name, queue_dir)
2566

    
2567

    
2568
def JobQueueUpdate(file_name, content):
2569
  """Updates a file in the queue directory.
2570

2571
  This is just a wrapper over L{utils.io.WriteFile}, with proper
2572
  checking.
2573

2574
  @type file_name: str
2575
  @param file_name: the job file name
2576
  @type content: str
2577
  @param content: the new job contents
2578
  @rtype: boolean
2579
  @return: the success of the operation
2580

2581
  """
2582
  _EnsureJobQueueFile(file_name)
2583
  getents = runtime.GetEnts()
2584

    
2585
  # Write and replace the file atomically
2586
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2587
                  gid=getents.masterd_gid)
2588

    
2589

    
2590
def JobQueueRename(old, new):
2591
  """Renames a job queue file.
2592

2593
  This is just a wrapper over os.rename with proper checking.
2594

2595
  @type old: str
2596
  @param old: the old (actual) file name
2597
  @type new: str
2598
  @param new: the desired file name
2599
  @rtype: tuple
2600
  @return: the success of the operation and payload
2601

2602
  """
2603
  _EnsureJobQueueFile(old)
2604
  _EnsureJobQueueFile(new)
2605

    
2606
  utils.RenameFile(old, new, mkdir=True)
2607

    
2608

    
2609
def BlockdevClose(instance_name, disks):
2610
  """Closes the given block devices.
2611

2612
  This means they will be switched to secondary mode (in case of
2613
  DRBD).
2614

2615
  @param instance_name: if the argument is not empty, the symlinks
2616
      of this instance will be removed
2617
  @type disks: list of L{objects.Disk}
2618
  @param disks: the list of disks to be closed
2619
  @rtype: tuple (success, message)
2620
  @return: a tuple of success and message, where success
2621
      indicates the succes of the operation, and message
2622
      which will contain the error details in case we
2623
      failed
2624

2625
  """
2626
  bdevs = []
2627
  for cf in disks:
2628
    rd = _RecursiveFindBD(cf)
2629
    if rd is None:
2630
      _Fail("Can't find device %s", cf)
2631
    bdevs.append(rd)
2632

    
2633
  msg = []
2634
  for rd in bdevs:
2635
    try:
2636
      rd.Close()
2637
    except errors.BlockDeviceError, err:
2638
      msg.append(str(err))
2639
  if msg:
2640
    _Fail("Can't make devices secondary: %s", ",".join(msg))
2641
  else:
2642
    if instance_name:
2643
      _RemoveBlockDevLinks(instance_name, disks)
2644

    
2645

    
2646
def ValidateHVParams(hvname, hvparams):
2647
  """Validates the given hypervisor parameters.
2648

2649
  @type hvname: string
2650
  @param hvname: the hypervisor name
2651
  @type hvparams: dict
2652
  @param hvparams: the hypervisor parameters to be validated
2653
  @rtype: None
2654

2655
  """
2656
  try:
2657
    hv_type = hypervisor.GetHypervisor(hvname)
2658
    hv_type.ValidateParameters(hvparams)
2659
  except errors.HypervisorError, err:
2660
    _Fail(str(err), log=False)
2661

    
2662

    
2663
def _CheckOSPList(os_obj, parameters):
2664
  """Check whether a list of parameters is supported by the OS.
2665

2666
  @type os_obj: L{objects.OS}
2667
  @param os_obj: OS object to check
2668
  @type parameters: list
2669
  @param parameters: the list of parameters to check
2670

2671
  """
2672
  supported = [v[0] for v in os_obj.supported_parameters]
2673
  delta = frozenset(parameters).difference(supported)
2674
  if delta:
2675
    _Fail("The following parameters are not supported"
2676
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2677

    
2678

    
2679
def ValidateOS(required, osname, checks, osparams):
2680
  """Validate the given OS' parameters.
2681

2682
  @type required: boolean
2683
  @param required: whether absence of the OS should translate into
2684
      failure or not
2685
  @type osname: string
2686
  @param osname: the OS to be validated
2687
  @type checks: list
2688
  @param checks: list of the checks to run (currently only 'parameters')
2689
  @type osparams: dict
2690
  @param osparams: dictionary with OS parameters
2691
  @rtype: boolean
2692
  @return: True if the validation passed, or False if the OS was not
2693
      found and L{required} was false
2694

2695
  """
2696
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2697
    _Fail("Unknown checks required for OS %s: %s", osname,
2698
          set(checks).difference(constants.OS_VALIDATE_CALLS))
2699

    
2700
  name_only = objects.OS.GetName(osname)
2701
  status, tbv = _TryOSFromDisk(name_only, None)
2702

    
2703
  if not status:
2704
    if required:
2705
      _Fail(tbv)
2706
    else:
2707
      return False
2708

    
2709
  if max(tbv.api_versions) < constants.OS_API_V20:
2710
    return True
2711

    
2712
  if constants.OS_VALIDATE_PARAMETERS in checks:
2713
    _CheckOSPList(tbv, osparams.keys())
2714

    
2715
  validate_env = OSCoreEnv(osname, tbv, osparams)
2716
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2717
                        cwd=tbv.path)
2718
  if result.failed:
2719
    logging.error("os validate command '%s' returned error: %s output: %s",
2720
                  result.cmd, result.fail_reason, result.output)
2721
    _Fail("OS validation script failed (%s), output: %s",
2722
          result.fail_reason, result.output, log=False)
2723

    
2724
  return True
2725

    
2726

    
2727
def DemoteFromMC():
2728
  """Demotes the current node from master candidate role.
2729

2730
  """
2731
  # try to ensure we're not the master by mistake
2732
  master, myself = ssconf.GetMasterAndMyself()
2733
  if master == myself:
2734
    _Fail("ssconf status shows I'm the master node, will not demote")
2735

    
2736
  result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2737
  if not result.failed:
2738
    _Fail("The master daemon is running, will not demote")
2739

    
2740
  try:
2741
    if os.path.isfile(constants.CLUSTER_CONF_FILE):
2742
      utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2743
  except EnvironmentError, err:
2744
    if err.errno != errno.ENOENT:
2745
      _Fail("Error while backing up cluster file: %s", err, exc=True)
2746

    
2747
  utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2748

    
2749

    
2750
def _GetX509Filenames(cryptodir, name):
2751
  """Returns the full paths for the private key and certificate.
2752

2753
  """
2754
  return (utils.PathJoin(cryptodir, name),
2755
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2756
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2757

    
2758

    
2759
def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2760
  """Creates a new X509 certificate for SSL/TLS.
2761

2762
  @type validity: int
2763
  @param validity: Validity in seconds
2764
  @rtype: tuple; (string, string)
2765
  @return: Certificate name and public part
2766

2767
  """
2768
  (key_pem, cert_pem) = \
2769
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2770
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
2771

    
2772
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
2773
                              prefix="x509-%s-" % utils.TimestampForFilename())
2774
  try:
2775
    name = os.path.basename(cert_dir)
2776
    assert len(name) > 5
2777

    
2778
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2779

    
2780
    utils.WriteFile(key_file, mode=0400, data=key_pem)
2781
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2782

    
2783
    # Never return private key as it shouldn't leave the node
2784
    return (name, cert_pem)
2785
  except Exception:
2786
    shutil.rmtree(cert_dir, ignore_errors=True)
2787
    raise
2788

    
2789

    
2790
def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2791
  """Removes a X509 certificate.
2792

2793
  @type name: string
2794
  @param name: Certificate name
2795

2796
  """
2797
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2798

    
2799
  utils.RemoveFile(key_file)
2800
  utils.RemoveFile(cert_file)
2801

    
2802
  try:
2803
    os.rmdir(cert_dir)
2804
  except EnvironmentError, err:
2805
    _Fail("Cannot remove certificate directory '%s': %s",
2806
          cert_dir, err)
2807

    
2808

    
2809
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2810
  """Returns the command for the requested input/output.
2811

2812
  @type instance: L{objects.Instance}
2813
  @param instance: The instance object
2814
  @param mode: Import/export mode
2815
  @param ieio: Input/output type
2816
  @param ieargs: Input/output arguments
2817

2818
  """
2819
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2820

    
2821
  env = None
2822
  prefix = None
2823
  suffix = None
2824
  exp_size = None
2825

    
2826
  if ieio == constants.IEIO_FILE:
2827
    (filename, ) = ieargs
2828

    
2829
    if not utils.IsNormAbsPath(filename):
2830
      _Fail("Path '%s' is not normalized or absolute", filename)
2831

    
2832
    directory = os.path.normpath(os.path.dirname(filename))
2833

    
2834
    if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2835
        constants.EXPORT_DIR):
2836
      _Fail("File '%s' is not under exports directory '%s'",
2837
            filename, constants.EXPORT_DIR)
2838

    
2839
    # Create directory
2840
    utils.Makedirs(directory, mode=0750)
2841

    
2842
    quoted_filename = utils.ShellQuote(filename)
2843

    
2844
    if mode == constants.IEM_IMPORT:
2845
      suffix = "> %s" % quoted_filename
2846
    elif mode == constants.IEM_EXPORT:
2847
      suffix = "< %s" % quoted_filename
2848

    
2849
      # Retrieve file size
2850
      try:
2851
        st = os.stat(filename)
2852
      except EnvironmentError, err:
2853
        logging.error("Can't stat(2) %s: %s", filename, err)
2854
      else:
2855
        exp_size = utils.BytesToMebibyte(st.st_size)
2856

    
2857
  elif ieio == constants.IEIO_RAW_DISK:
2858
    (disk, ) = ieargs
2859

    
2860
    real_disk = _OpenRealBD(disk)
2861

    
2862
    if mode == constants.IEM_IMPORT:
2863
      # we set here a smaller block size as, due to transport buffering, more
2864
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
2865
      # is not already there or we pass a wrong path; we use notrunc to no
2866
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2867
      # much memory; this means that at best, we flush every 64k, which will
2868
      # not be very fast
2869
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2870
                                    " bs=%s oflag=dsync"),
2871
                                    real_disk.dev_path,
2872
                                    str(64 * 1024))
2873

    
2874
    elif mode == constants.IEM_EXPORT:
2875
      # the block size on the read dd is 1MiB to match our units
2876
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2877
                                   real_disk.dev_path,
2878
                                   str(1024 * 1024), # 1 MB
2879
                                   str(disk.size))
2880
      exp_size = disk.size
2881

    
2882
  elif ieio == constants.IEIO_SCRIPT:
2883
    (disk, disk_index, ) = ieargs
2884

    
2885
    assert isinstance(disk_index, (int, long))
2886

    
2887
    real_disk = _OpenRealBD(disk)
2888

    
2889
    inst_os = OSFromDisk(instance.os)
2890
    env = OSEnvironment(instance, inst_os)
2891

    
2892
    if mode == constants.IEM_IMPORT:
2893
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2894
      env["IMPORT_INDEX"] = str(disk_index)
2895
      script = inst_os.import_script
2896

    
2897
    elif mode == constants.IEM_EXPORT:
2898
      env["EXPORT_DEVICE"] = real_disk.dev_path
2899
      env["EXPORT_INDEX"] = str(disk_index)
2900
      script = inst_os.export_script
2901

    
2902
    # TODO: Pass special environment only to script
2903
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2904

    
2905
    if mode == constants.IEM_IMPORT:
2906
      suffix = "| %s" % script_cmd
2907

    
2908
    elif mode == constants.IEM_EXPORT:
2909
      prefix = "%s |" % script_cmd
2910

    
2911
    # Let script predict size
2912
    exp_size = constants.IE_CUSTOM_SIZE
2913

    
2914
  else:
2915
    _Fail("Invalid %s I/O mode %r", mode, ieio)
2916

    
2917
  return (env, prefix, suffix, exp_size)
2918

    
2919

    
2920
def _CreateImportExportStatusDir(prefix):
2921
  """Creates status directory for import/export.
2922

2923
  """
2924
  return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2925
                          prefix=("%s-%s-" %
2926
                                  (prefix, utils.TimestampForFilename())))
2927

    
2928

    
2929
def StartImportExportDaemon(mode, opts, host, port, instance, ieio, ieioargs):
2930
  """Starts an import or export daemon.
2931

2932
  @param mode: Import/output mode
2933
  @type opts: L{objects.ImportExportOptions}
2934
  @param opts: Daemon options
2935
  @type host: string
2936
  @param host: Remote host for export (None for import)
2937
  @type port: int
2938
  @param port: Remote port for export (None for import)
2939
  @type instance: L{objects.Instance}
2940
  @param instance: Instance object
2941
  @param ieio: Input/output type
2942
  @param ieioargs: Input/output arguments
2943

2944
  """
2945
  if mode == constants.IEM_IMPORT:
2946
    prefix = "import"
2947

    
2948
    if not (host is None and port is None):
2949
      _Fail("Can not specify host or port on import")
2950

    
2951
  elif mode == constants.IEM_EXPORT:
2952
    prefix = "export"
2953

    
2954
    if host is None or port is None:
2955
      _Fail("Host and port must be specified for an export")
2956

    
2957
  else:
2958
    _Fail("Invalid mode %r", mode)
2959

    
2960
  if (opts.key_name is None) ^ (opts.ca_pem is None):
2961
    _Fail("Cluster certificate can only be used for both key and CA")
2962

    
2963
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
2964
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
2965

    
2966
  if opts.key_name is None:
2967
    # Use server.pem
2968
    key_path = constants.NODED_CERT_FILE
2969
    cert_path = constants.NODED_CERT_FILE
2970
    assert opts.ca_pem is None
2971
  else:
2972
    (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
2973
                                                 opts.key_name)
2974
    assert opts.ca_pem is not None
2975

    
2976
  for i in [key_path, cert_path]:
2977
    if not os.path.exists(i):
2978
      _Fail("File '%s' does not exist" % i)
2979

    
2980
  status_dir = _CreateImportExportStatusDir(prefix)
2981
  try:
2982
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
2983
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
2984
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
2985

    
2986
    if opts.ca_pem is None:
2987
      # Use server.pem
2988
      ca = utils.ReadFile(constants.NODED_CERT_FILE)
2989
    else:
2990
      ca = opts.ca_pem
2991

    
2992
    # Write CA file
2993
    utils.WriteFile(ca_file, data=ca, mode=0400)
2994

    
2995
    cmd = [
2996
      constants.IMPORT_EXPORT_DAEMON,
2997
      status_file, mode,
2998
      "--key=%s" % key_path,
2999
      "--cert=%s" % cert_path,
3000
      "--ca=%s" % ca_file,
3001
      ]
3002

    
3003
    if host:
3004
      cmd.append("--host=%s" % host)
3005

    
3006
    if port:
3007
      cmd.append("--port=%s" % port)
3008

    
3009
    if opts.ipv6:
3010
      cmd.append("--ipv6")
3011
    else:
3012
      cmd.append("--ipv4")
3013

    
3014
    if opts.compress:
3015
      cmd.append("--compress=%s" % opts.compress)
3016

    
3017
    if opts.magic:
3018
      cmd.append("--magic=%s" % opts.magic)
3019

    
3020
    if exp_size is not None:
3021
      cmd.append("--expected-size=%s" % exp_size)
3022

    
3023
    if cmd_prefix:
3024
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3025

    
3026
    if cmd_suffix:
3027
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3028

    
3029
    if mode == constants.IEM_EXPORT:
3030
      # Retry connection a few times when connecting to remote peer
3031
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3032
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3033
    elif opts.connect_timeout is not None:
3034
      assert mode == constants.IEM_IMPORT
3035
      # Overall timeout for establishing connection while listening
3036
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3037

    
3038
    logfile = _InstanceLogName(prefix, instance.os, instance.name)
3039

    
3040
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3041
    # support for receiving a file descriptor for output
3042
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3043
                      output=logfile)
3044

    
3045
    # The import/export name is simply the status directory name
3046
    return os.path.basename(status_dir)
3047

    
3048
  except Exception:
3049
    shutil.rmtree(status_dir, ignore_errors=True)
3050
    raise
3051

    
3052

    
3053
def GetImportExportStatus(names):
3054
  """Returns import/export daemon status.
3055

3056
  @type names: sequence
3057
  @param names: List of names
3058
  @rtype: List of dicts
3059
  @return: Returns a list of the state of each named import/export or None if a
3060
           status couldn't be read
3061

3062
  """
3063
  result = []
3064

    
3065
  for name in names:
3066
    status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3067
                                 _IES_STATUS_FILE)
3068

    
3069
    try:
3070
      data = utils.ReadFile(status_file)
3071
    except EnvironmentError, err:
3072
      if err.errno != errno.ENOENT:
3073
        raise
3074
      data = None
3075

    
3076
    if not data:
3077
      result.append(None)
3078
      continue
3079

    
3080
    result.append(serializer.LoadJson(data))
3081

    
3082
  return result
3083

    
3084

    
3085
def AbortImportExport(name):
3086
  """Sends SIGTERM to a running import/export daemon.
3087

3088
  """
3089
  logging.info("Abort import/export %s", name)
3090

    
3091
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3092
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3093

    
3094
  if pid:
3095
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3096
                 name, pid)
3097
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3098

    
3099

    
3100
def CleanupImportExport(name):
3101
  """Cleanup after an import or export.
3102

3103
  If the import/export daemon is still running it's killed. Afterwards the
3104
  whole status directory is removed.
3105

3106
  """
3107
  logging.info("Finalizing import/export %s", name)
3108

    
3109
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3110

    
3111
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3112

    
3113
  if pid:
3114
    logging.info("Import/export %s is still running with PID %s",
3115
                 name, pid)
3116
    utils.KillProcess(pid, waitpid=False)
3117

    
3118
  shutil.rmtree(status_dir, ignore_errors=True)
3119

    
3120

    
3121
def _FindDisks(nodes_ip, disks):
3122
  """Sets the physical ID on disks and returns the block devices.
3123

3124
  """
3125
  # set the correct physical ID
3126
  my_name = netutils.Hostname.GetSysName()
3127
  for cf in disks:
3128
    cf.SetPhysicalID(my_name, nodes_ip)
3129

    
3130
  bdevs = []
3131

    
3132
  for cf in disks:
3133
    rd = _RecursiveFindBD(cf)
3134
    if rd is None:
3135
      _Fail("Can't find device %s", cf)
3136
    bdevs.append(rd)
3137
  return bdevs
3138

    
3139

    
3140
def DrbdDisconnectNet(nodes_ip, disks):
3141
  """Disconnects the network on a list of drbd devices.
3142

3143
  """
3144
  bdevs = _FindDisks(nodes_ip, disks)
3145

    
3146
  # disconnect disks
3147
  for rd in bdevs:
3148
    try:
3149
      rd.DisconnectNet()
3150
    except errors.BlockDeviceError, err:
3151
      _Fail("Can't change network configuration to standalone mode: %s",
3152
            err, exc=True)
3153

    
3154

    
3155
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3156
  """Attaches the network on a list of drbd devices.
3157

3158
  """
3159
  bdevs = _FindDisks(nodes_ip, disks)
3160

    
3161
  if multimaster:
3162
    for idx, rd in enumerate(bdevs):
3163
      try:
3164
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3165
      except EnvironmentError, err:
3166
        _Fail("Can't create symlink: %s", err)
3167
  # reconnect disks, switch to new master configuration and if
3168
  # needed primary mode
3169
  for rd in bdevs:
3170
    try:
3171
      rd.AttachNet(multimaster)
3172
    except errors.BlockDeviceError, err:
3173
      _Fail("Can't change network configuration: %s", err)
3174

    
3175
  # wait until the disks are connected; we need to retry the re-attach
3176
  # if the device becomes standalone, as this might happen if the one
3177
  # node disconnects and reconnects in a different mode before the
3178
  # other node reconnects; in this case, one or both of the nodes will
3179
  # decide it has wrong configuration and switch to standalone
3180

    
3181
  def _Attach():
3182
    all_connected = True
3183

    
3184
    for rd in bdevs:
3185
      stats = rd.GetProcStatus()
3186

    
3187
      all_connected = (all_connected and
3188
                       (stats.is_connected or stats.is_in_resync))
3189

    
3190
      if stats.is_standalone:
3191
        # peer had different config info and this node became
3192
        # standalone, even though this should not happen with the
3193
        # new staged way of changing disk configs
3194
        try:
3195
          rd.AttachNet(multimaster)
3196
        except errors.BlockDeviceError, err:
3197
          _Fail("Can't change network configuration: %s", err)
3198

    
3199
    if not all_connected:
3200
      raise utils.RetryAgain()
3201

    
3202
  try:
3203
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3204
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3205
  except utils.RetryTimeout:
3206
    _Fail("Timeout in disk reconnecting")
3207

    
3208
  if multimaster:
3209
    # change to primary mode
3210
    for rd in bdevs:
3211
      try:
3212
        rd.Open()
3213
      except errors.BlockDeviceError, err:
3214
        _Fail("Can't change to primary mode: %s", err)
3215

    
3216

    
3217
def DrbdWaitSync(nodes_ip, disks):
3218
  """Wait until DRBDs have synchronized.
3219

3220
  """
3221
  def _helper(rd):
3222
    stats = rd.GetProcStatus()
3223
    if not (stats.is_connected or stats.is_in_resync):
3224
      raise utils.RetryAgain()
3225
    return stats
3226

    
3227
  bdevs = _FindDisks(nodes_ip, disks)
3228

    
3229
  min_resync = 100
3230
  alldone = True
3231
  for rd in bdevs:
3232
    try:
3233
      # poll each second for 15 seconds
3234
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3235
    except utils.RetryTimeout:
3236
      stats = rd.GetProcStatus()
3237
      # last check
3238
      if not (stats.is_connected or stats.is_in_resync):
3239
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3240
    alldone = alldone and (not stats.is_in_resync)
3241
    if stats.sync_percent is not None:
3242
      min_resync = min(min_resync, stats.sync_percent)
3243

    
3244
  return (alldone, min_resync)
3245

    
3246

    
3247
def GetDrbdUsermodeHelper():
3248
  """Returns DRBD usermode helper currently configured.
3249

3250
  """
3251
  try:
3252
    return bdev.BaseDRBD.GetUsermodeHelper()
3253
  except errors.BlockDeviceError, err:
3254
    _Fail(str(err))
3255

    
3256

    
3257
def PowercycleNode(hypervisor_type):
3258
  """Hard-powercycle the node.
3259

3260
  Because we need to return first, and schedule the powercycle in the
3261
  background, we won't be able to report failures nicely.
3262

3263
  """
3264
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3265
  try:
3266
    pid = os.fork()
3267
  except OSError:
3268
    # if we can't fork, we'll pretend that we're in the child process
3269
    pid = 0
3270
  if pid > 0:
3271
    return "Reboot scheduled in 5 seconds"
3272
  # ensure the child is running on ram
3273
  try:
3274
    utils.Mlockall()
3275
  except Exception: # pylint: disable-msg=W0703
3276
    pass
3277
  time.sleep(5)
3278
  hyper.PowercycleNode()
3279

    
3280

    
3281
class HooksRunner(object):
3282
  """Hook runner.
3283

3284
  This class is instantiated on the node side (ganeti-noded) and not
3285
  on the master side.
3286

3287
  """
3288
  def __init__(self, hooks_base_dir=None):
3289
    """Constructor for hooks runner.
3290

3291
    @type hooks_base_dir: str or None
3292
    @param hooks_base_dir: if not None, this overrides the
3293
        L{constants.HOOKS_BASE_DIR} (useful for unittests)
3294

3295
    """
3296
    if hooks_base_dir is None:
3297
      hooks_base_dir = constants.HOOKS_BASE_DIR
3298
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
3299
    # constant
3300
    self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3301

    
3302
  def RunHooks(self, hpath, phase, env):
3303
    """Run the scripts in the hooks directory.
3304

3305
    @type hpath: str
3306
    @param hpath: the path to the hooks directory which
3307
        holds the scripts
3308
    @type phase: str
3309
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
3310
        L{constants.HOOKS_PHASE_POST}
3311
    @type env: dict
3312
    @param env: dictionary with the environment for the hook
3313
    @rtype: list
3314
    @return: list of 3-element tuples:
3315
      - script path
3316
      - script result, either L{constants.HKR_SUCCESS} or
3317
        L{constants.HKR_FAIL}
3318
      - output of the script
3319

3320
    @raise errors.ProgrammerError: for invalid input
3321
        parameters
3322

3323
    """
3324
    if phase == constants.HOOKS_PHASE_PRE:
3325
      suffix = "pre"
3326
    elif phase == constants.HOOKS_PHASE_POST:
3327
      suffix = "post"
3328
    else:
3329
      _Fail("Unknown hooks phase '%s'", phase)
3330

    
3331

    
3332
    subdir = "%s-%s.d" % (hpath, suffix)
3333
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3334

    
3335
    results = []
3336

    
3337
    if not os.path.isdir(dir_name):
3338
      # for non-existing/non-dirs, we simply exit instead of logging a
3339
      # warning at every operation
3340
      return results
3341

    
3342
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3343

    
3344
    for (relname, relstatus, runresult)  in runparts_results:
3345
      if relstatus == constants.RUNPARTS_SKIP:
3346
        rrval = constants.HKR_SKIP
3347
        output = ""
3348
      elif relstatus == constants.RUNPARTS_ERR:
3349
        rrval = constants.HKR_FAIL
3350
        output = "Hook script execution error: %s" % runresult
3351
      elif relstatus == constants.RUNPARTS_RUN:
3352
        if runresult.failed:
3353
          rrval = constants.HKR_FAIL
3354
        else:
3355
          rrval = constants.HKR_SUCCESS
3356
        output = utils.SafeEncode(runresult.output.strip())
3357
      results.append(("%s/%s" % (subdir, relname), rrval, output))
3358

    
3359
    return results
3360

    
3361

    
3362
class IAllocatorRunner(object):
3363
  """IAllocator runner.
3364

3365
  This class is instantiated on the node side (ganeti-noded) and not on
3366
  the master side.
3367

3368
  """
3369
  @staticmethod
3370
  def Run(name, idata):
3371
    """Run an iallocator script.
3372

3373
    @type name: str
3374
    @param name: the iallocator script name
3375
    @type idata: str
3376
    @param idata: the allocator input data
3377

3378
    @rtype: tuple
3379
    @return: two element tuple of:
3380
       - status
3381
       - either error message or stdout of allocator (for success)
3382

3383
    """
3384
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3385
                                  os.path.isfile)
3386
    if alloc_script is None:
3387
      _Fail("iallocator module '%s' not found in the search path", name)
3388

    
3389
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3390
    try:
3391
      os.write(fd, idata)
3392
      os.close(fd)
3393
      result = utils.RunCmd([alloc_script, fin_name])
3394
      if result.failed:
3395
        _Fail("iallocator module '%s' failed: %s, output '%s'",
3396
              name, result.fail_reason, result.output)
3397
    finally:
3398
      os.unlink(fin_name)
3399

    
3400
    return result.stdout
3401

    
3402

    
3403
class DevCacheManager(object):
3404
  """Simple class for managing a cache of block device information.
3405

3406
  """
3407
  _DEV_PREFIX = "/dev/"
3408
  _ROOT_DIR = constants.BDEV_CACHE_DIR
3409

    
3410
  @classmethod
3411
  def _ConvertPath(cls, dev_path):
3412
    """Converts a /dev/name path to the cache file name.
3413

3414
    This replaces slashes with underscores and strips the /dev
3415
    prefix. It then returns the full path to the cache file.
3416

3417
    @type dev_path: str
3418
    @param dev_path: the C{/dev/} path name
3419
    @rtype: str
3420
    @return: the converted path name
3421

3422
    """
3423
    if dev_path.startswith(cls._DEV_PREFIX):
3424
      dev_path = dev_path[len(cls._DEV_PREFIX):]
3425
    dev_path = dev_path.replace("/", "_")
3426
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3427
    return fpath
3428

    
3429
  @classmethod
3430
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3431
    """Updates the cache information for a given device.
3432

3433
    @type dev_path: str
3434
    @param dev_path: the pathname of the device
3435
    @type owner: str
3436
    @param owner: the owner (instance name) of the device
3437
    @type on_primary: bool
3438
    @param on_primary: whether this is the primary
3439
        node nor not
3440
    @type iv_name: str
3441
    @param iv_name: the instance-visible name of the
3442
        device, as in objects.Disk.iv_name
3443

3444
    @rtype: None
3445

3446
    """
3447
    if dev_path is None:
3448
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
3449
      return
3450
    fpath = cls._ConvertPath(dev_path)
3451
    if on_primary:
3452
      state = "primary"
3453
    else:
3454
      state = "secondary"
3455
    if iv_name is None:
3456
      iv_name = "not_visible"
3457
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3458
    try:
3459
      utils.WriteFile(fpath, data=fdata)
3460
    except EnvironmentError, err:
3461
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3462

    
3463
  @classmethod
3464
  def RemoveCache(cls, dev_path):
3465
    """Remove data for a dev_path.
3466

3467
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
3468
    path name and logging.
3469

3470
    @type dev_path: str
3471
    @param dev_path: the pathname of the device
3472

3473
    @rtype: None
3474

3475
    """
3476
    if dev_path is None:
3477
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
3478
      return
3479
    fpath = cls._ConvertPath(dev_path)
3480
    try:
3481
      utils.RemoveFile(fpath)
3482
    except EnvironmentError, err:
3483
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)