Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 64c7b383

History | View | Annotate | Download (106.9 kB)

1
#
2
#
3

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

    
21

    
22
"""Functions used by the node daemon
23

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

29
"""
30

    
31
# pylint: disable=E1103
32

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

    
37

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

    
52
from ganeti import errors
53
from ganeti import utils
54
from ganeti import ssh
55
from ganeti import hypervisor
56
from ganeti import constants
57
from ganeti import bdev
58
from ganeti import objects
59
from ganeti import ssconf
60
from ganeti import serializer
61
from ganeti import netutils
62
from ganeti import runtime
63

    
64

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

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

    
82

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

86
  Its argument is the error message.
87

88
  """
89

    
90

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

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

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

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

    
113

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

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

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

    
123

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

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

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

    
136

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

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

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

    
156

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

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

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

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

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

    
186

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

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

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

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

    
208
  return frozenset(allowed_files)
209

    
210

    
211
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
212

    
213

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

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

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

    
224

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

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

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

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

    
246

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

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

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

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

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

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

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

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

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

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

    
322

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

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

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

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

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

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

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

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

    
360

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

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

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

    
382

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

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

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

393
  @param modify_ssh_setup: boolean
394

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

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

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

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

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

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

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

    
426

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

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

443
  """
444
  outputarray = {}
445

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

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

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

    
463
  return outputarray
464

    
465

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

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

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

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

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

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

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

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

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

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

    
522
  if constants.NV_NODELIST in what:
523
    (nodes, bynode) = what[constants.NV_NODELIST]
524

    
525
    # Add nodes from other groups (different for each node)
526
    try:
527
      nodes.extend(bynode[my_name])
528
    except KeyError:
529
      pass
530

    
531
    # Use a random order
532
    random.shuffle(nodes)
533

    
534
    # Try to contact all nodes
535
    val = {}
536
    for node in nodes:
537
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
538
      if not success:
539
        val[node] = message
540

    
541
    result[constants.NV_NODELIST] = val
542

    
543
  if constants.NV_NODENETTEST in what:
544
    result[constants.NV_NODENETTEST] = tmp = {}
545
    my_pip = my_sip = None
546
    for name, pip, sip in what[constants.NV_NODENETTEST]:
547
      if name == my_name:
548
        my_pip = pip
549
        my_sip = sip
550
        break
551
    if not my_pip:
552
      tmp[my_name] = ("Can't find my own primary/secondary IP"
553
                      " in the node list")
554
    else:
555
      for name, pip, sip in what[constants.NV_NODENETTEST]:
556
        fail = []
557
        if not netutils.TcpPing(pip, port, source=my_pip):
558
          fail.append("primary")
559
        if sip != pip:
560
          if not netutils.TcpPing(sip, port, source=my_sip):
561
            fail.append("secondary")
562
        if fail:
563
          tmp[name] = ("failure using the %s interface(s)" %
564
                       " and ".join(fail))
565

    
566
  if constants.NV_MASTERIP in what:
567
    # FIXME: add checks on incoming data structures (here and in the
568
    # rest of the function)
569
    master_name, master_ip = what[constants.NV_MASTERIP]
570
    if master_name == my_name:
571
      source = constants.IP4_ADDRESS_LOCALHOST
572
    else:
573
      source = None
574
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
575
                                                  source=source)
576

    
577
  if constants.NV_OOB_PATHS in what:
578
    result[constants.NV_OOB_PATHS] = tmp = []
579
    for path in what[constants.NV_OOB_PATHS]:
580
      try:
581
        st = os.stat(path)
582
      except OSError, err:
583
        tmp.append("error stating out of band helper: %s" % err)
584
      else:
585
        if stat.S_ISREG(st.st_mode):
586
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
587
            tmp.append(None)
588
          else:
589
            tmp.append("out of band helper %s is not executable" % path)
590
        else:
591
          tmp.append("out of band helper %s is not a file" % path)
592

    
593
  if constants.NV_LVLIST in what and vm_capable:
594
    try:
595
      val = GetVolumeList(utils.ListVolumeGroups().keys())
596
    except RPCFail, err:
597
      val = str(err)
598
    result[constants.NV_LVLIST] = val
599

    
600
  if constants.NV_INSTANCELIST in what and vm_capable:
601
    # GetInstanceList can fail
602
    try:
603
      val = GetInstanceList(what[constants.NV_INSTANCELIST])
604
    except RPCFail, err:
605
      val = str(err)
606
    result[constants.NV_INSTANCELIST] = val
607

    
608
  if constants.NV_VGLIST in what and vm_capable:
609
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
610

    
611
  if constants.NV_PVLIST in what and vm_capable:
612
    result[constants.NV_PVLIST] = \
613
      bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
614
                                   filter_allocatable=False)
615

    
616
  if constants.NV_VERSION in what:
617
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
618
                                    constants.RELEASE_VERSION)
619

    
620
  if constants.NV_HVINFO in what and vm_capable:
621
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
622
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
623

    
624
  if constants.NV_DRBDLIST in what and vm_capable:
625
    try:
626
      used_minors = bdev.DRBD8.GetUsedDevs().keys()
627
    except errors.BlockDeviceError, err:
628
      logging.warning("Can't get used minors list", exc_info=True)
629
      used_minors = str(err)
630
    result[constants.NV_DRBDLIST] = used_minors
631

    
632
  if constants.NV_DRBDHELPER in what and vm_capable:
633
    status = True
634
    try:
635
      payload = bdev.BaseDRBD.GetUsermodeHelper()
636
    except errors.BlockDeviceError, err:
637
      logging.error("Can't get DRBD usermode helper: %s", str(err))
638
      status = False
639
      payload = str(err)
640
    result[constants.NV_DRBDHELPER] = (status, payload)
641

    
642
  if constants.NV_NODESETUP in what:
643
    result[constants.NV_NODESETUP] = tmpr = []
644
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
645
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
646
                  " under /sys, missing required directories /sys/block"
647
                  " and /sys/class/net")
648
    if (not os.path.isdir("/proc/sys") or
649
        not os.path.isfile("/proc/sysrq-trigger")):
650
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
651
                  " under /proc, missing required directory /proc/sys and"
652
                  " the file /proc/sysrq-trigger")
653

    
654
  if constants.NV_TIME in what:
655
    result[constants.NV_TIME] = utils.SplitTime(time.time())
656

    
657
  if constants.NV_OSLIST in what and vm_capable:
658
    result[constants.NV_OSLIST] = DiagnoseOS()
659

    
660
  if constants.NV_BRIDGES in what and vm_capable:
661
    result[constants.NV_BRIDGES] = [bridge
662
                                    for bridge in what[constants.NV_BRIDGES]
663
                                    if not utils.BridgeExists(bridge)]
664
  return result
665

    
666

    
667
def GetBlockDevSizes(devices):
668
  """Return the size of the given block devices
669

670
  @type devices: list
671
  @param devices: list of block device nodes to query
672
  @rtype: dict
673
  @return:
674
    dictionary of all block devices under /dev (key). The value is their
675
    size in MiB.
676

677
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
678

679
  """
680
  DEV_PREFIX = "/dev/"
681
  blockdevs = {}
682

    
683
  for devpath in devices:
684
    if os.path.commonprefix([DEV_PREFIX, devpath]) != DEV_PREFIX:
685
      continue
686

    
687
    try:
688
      st = os.stat(devpath)
689
    except EnvironmentError, err:
690
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
691
      continue
692

    
693
    if stat.S_ISBLK(st.st_mode):
694
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
695
      if result.failed:
696
        # We don't want to fail, just do not list this device as available
697
        logging.warning("Cannot get size for block device %s", devpath)
698
        continue
699

    
700
      size = int(result.stdout) / (1024 * 1024)
701
      blockdevs[devpath] = size
702
  return blockdevs
703

    
704

    
705
def GetVolumeList(vg_names):
706
  """Compute list of logical volumes and their size.
707

708
  @type vg_names: list
709
  @param vg_names: the volume groups whose LVs we should list, or
710
      empty for all volume groups
711
  @rtype: dict
712
  @return:
713
      dictionary of all partions (key) with value being a tuple of
714
      their size (in MiB), inactive and online status::
715

716
        {'xenvg/test1': ('20.06', True, True)}
717

718
      in case of errors, a string is returned with the error
719
      details.
720

721
  """
722
  lvs = {}
723
  sep = "|"
724
  if not vg_names:
725
    vg_names = []
726
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
727
                         "--separator=%s" % sep,
728
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
729
  if result.failed:
730
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
731

    
732
  for line in result.stdout.splitlines():
733
    line = line.strip()
734
    match = _LVSLINE_REGEX.match(line)
735
    if not match:
736
      logging.error("Invalid line returned from lvs output: '%s'", line)
737
      continue
738
    vg_name, name, size, attr = match.groups()
739
    inactive = attr[4] == "-"
740
    online = attr[5] == "o"
741
    virtual = attr[0] == "v"
742
    if virtual:
743
      # we don't want to report such volumes as existing, since they
744
      # don't really hold data
745
      continue
746
    lvs[vg_name + "/" + name] = (size, inactive, online)
747

    
748
  return lvs
749

    
750

    
751
def ListVolumeGroups():
752
  """List the volume groups and their size.
753

754
  @rtype: dict
755
  @return: dictionary with keys volume name and values the
756
      size of the volume
757

758
  """
759
  return utils.ListVolumeGroups()
760

    
761

    
762
def NodeVolumes():
763
  """List all volumes on this node.
764

765
  @rtype: list
766
  @return:
767
    A list of dictionaries, each having four keys:
768
      - name: the logical volume name,
769
      - size: the size of the logical volume
770
      - dev: the physical device on which the LV lives
771
      - vg: the volume group to which it belongs
772

773
    In case of errors, we return an empty list and log the
774
    error.
775

776
    Note that since a logical volume can live on multiple physical
777
    volumes, the resulting list might include a logical volume
778
    multiple times.
779

780
  """
781
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
782
                         "--separator=|",
783
                         "--options=lv_name,lv_size,devices,vg_name"])
784
  if result.failed:
785
    _Fail("Failed to list logical volumes, lvs output: %s",
786
          result.output)
787

    
788
  def parse_dev(dev):
789
    return dev.split("(")[0]
790

    
791
  def handle_dev(dev):
792
    return [parse_dev(x) for x in dev.split(",")]
793

    
794
  def map_line(line):
795
    line = [v.strip() for v in line]
796
    return [{"name": line[0], "size": line[1],
797
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
798

    
799
  all_devs = []
800
  for line in result.stdout.splitlines():
801
    if line.count("|") >= 3:
802
      all_devs.extend(map_line(line.split("|")))
803
    else:
804
      logging.warning("Strange line in the output from lvs: '%s'", line)
805
  return all_devs
806

    
807

    
808
def BridgesExist(bridges_list):
809
  """Check if a list of bridges exist on the current node.
810

811
  @rtype: boolean
812
  @return: C{True} if all of them exist, C{False} otherwise
813

814
  """
815
  missing = []
816
  for bridge in bridges_list:
817
    if not utils.BridgeExists(bridge):
818
      missing.append(bridge)
819

    
820
  if missing:
821
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
822

    
823

    
824
def GetInstanceList(hypervisor_list):
825
  """Provides a list of instances.
826

827
  @type hypervisor_list: list
828
  @param hypervisor_list: the list of hypervisors to query information
829

830
  @rtype: list
831
  @return: a list of all running instances on the current node
832
    - instance1.example.com
833
    - instance2.example.com
834

835
  """
836
  results = []
837
  for hname in hypervisor_list:
838
    try:
839
      names = hypervisor.GetHypervisor(hname).ListInstances()
840
      results.extend(names)
841
    except errors.HypervisorError, err:
842
      _Fail("Error enumerating instances (hypervisor %s): %s",
843
            hname, err, exc=True)
844

    
845
  return results
846

    
847

    
848
def GetInstanceInfo(instance, hname):
849
  """Gives back the information about an instance as a dictionary.
850

851
  @type instance: string
852
  @param instance: the instance name
853
  @type hname: string
854
  @param hname: the hypervisor type of the instance
855

856
  @rtype: dict
857
  @return: dictionary with the following keys:
858
      - memory: memory size of instance (int)
859
      - state: xen state of instance (string)
860
      - time: cpu time of instance (float)
861

862
  """
863
  output = {}
864

    
865
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
866
  if iinfo is not None:
867
    output["memory"] = iinfo[2]
868
    output["state"] = iinfo[4]
869
    output["time"] = iinfo[5]
870

    
871
  return output
872

    
873

    
874
def GetInstanceMigratable(instance):
875
  """Gives whether an instance can be migrated.
876

877
  @type instance: L{objects.Instance}
878
  @param instance: object representing the instance to be checked.
879

880
  @rtype: tuple
881
  @return: tuple of (result, description) where:
882
      - result: whether the instance can be migrated or not
883
      - description: a description of the issue, if relevant
884

885
  """
886
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
887
  iname = instance.name
888
  if iname not in hyper.ListInstances():
889
    _Fail("Instance %s is not running", iname)
890

    
891
  for idx in range(len(instance.disks)):
892
    link_name = _GetBlockDevSymlinkPath(iname, idx)
893
    if not os.path.islink(link_name):
894
      logging.warning("Instance %s is missing symlink %s for disk %d",
895
                      iname, link_name, idx)
896

    
897

    
898
def GetAllInstancesInfo(hypervisor_list):
899
  """Gather data about all instances.
900

901
  This is the equivalent of L{GetInstanceInfo}, except that it
902
  computes data for all instances at once, thus being faster if one
903
  needs data about more than one instance.
904

905
  @type hypervisor_list: list
906
  @param hypervisor_list: list of hypervisors to query for instance data
907

908
  @rtype: dict
909
  @return: dictionary of instance: data, with data having the following keys:
910
      - memory: memory size of instance (int)
911
      - state: xen state of instance (string)
912
      - time: cpu time of instance (float)
913
      - vcpus: the number of vcpus
914

915
  """
916
  output = {}
917

    
918
  for hname in hypervisor_list:
919
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
920
    if iinfo:
921
      for name, _, memory, vcpus, state, times in iinfo:
922
        value = {
923
          "memory": memory,
924
          "vcpus": vcpus,
925
          "state": state,
926
          "time": times,
927
          }
928
        if name in output:
929
          # we only check static parameters, like memory and vcpus,
930
          # and not state and time which can change between the
931
          # invocations of the different hypervisors
932
          for key in "memory", "vcpus":
933
            if value[key] != output[name][key]:
934
              _Fail("Instance %s is running twice"
935
                    " with different parameters", name)
936
        output[name] = value
937

    
938
  return output
939

    
940

    
941
def _InstanceLogName(kind, os_name, instance, component):
942
  """Compute the OS log filename for a given instance and operation.
943

944
  The instance name and os name are passed in as strings since not all
945
  operations have these as part of an instance object.
946

947
  @type kind: string
948
  @param kind: the operation type (e.g. add, import, etc.)
949
  @type os_name: string
950
  @param os_name: the os name
951
  @type instance: string
952
  @param instance: the name of the instance being imported/added/etc.
953
  @type component: string or None
954
  @param component: the name of the component of the instance being
955
      transferred
956

957
  """
958
  # TODO: Use tempfile.mkstemp to create unique filename
959
  if component:
960
    assert "/" not in component
961
    c_msg = "-%s" % component
962
  else:
963
    c_msg = ""
964
  base = ("%s-%s-%s%s-%s.log" %
965
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
966
  return utils.PathJoin(constants.LOG_OS_DIR, base)
967

    
968

    
969
def InstanceOsAdd(instance, reinstall, debug):
970
  """Add an OS to an instance.
971

972
  @type instance: L{objects.Instance}
973
  @param instance: Instance whose OS is to be installed
974
  @type reinstall: boolean
975
  @param reinstall: whether this is an instance reinstall
976
  @type debug: integer
977
  @param debug: debug level, passed to the OS scripts
978
  @rtype: None
979

980
  """
981
  inst_os = OSFromDisk(instance.os)
982

    
983
  create_env = OSEnvironment(instance, inst_os, debug)
984
  if reinstall:
985
    create_env["INSTANCE_REINSTALL"] = "1"
986

    
987
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
988

    
989
  result = utils.RunCmd([inst_os.create_script], env=create_env,
990
                        cwd=inst_os.path, output=logfile, reset_env=True)
991
  if result.failed:
992
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
993
                  " output: %s", result.cmd, result.fail_reason, logfile,
994
                  result.output)
995
    lines = [utils.SafeEncode(val)
996
             for val in utils.TailFile(logfile, lines=20)]
997
    _Fail("OS create script failed (%s), last lines in the"
998
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
999

    
1000

    
1001
def RunRenameInstance(instance, old_name, debug):
1002
  """Run the OS rename script for an instance.
1003

1004
  @type instance: L{objects.Instance}
1005
  @param instance: Instance whose OS is to be installed
1006
  @type old_name: string
1007
  @param old_name: previous instance name
1008
  @type debug: integer
1009
  @param debug: debug level, passed to the OS scripts
1010
  @rtype: boolean
1011
  @return: the success of the operation
1012

1013
  """
1014
  inst_os = OSFromDisk(instance.os)
1015

    
1016
  rename_env = OSEnvironment(instance, inst_os, debug)
1017
  rename_env["OLD_INSTANCE_NAME"] = old_name
1018

    
1019
  logfile = _InstanceLogName("rename", instance.os,
1020
                             "%s-%s" % (old_name, instance.name), None)
1021

    
1022
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1023
                        cwd=inst_os.path, output=logfile, reset_env=True)
1024

    
1025
  if result.failed:
1026
    logging.error("os create command '%s' returned error: %s output: %s",
1027
                  result.cmd, result.fail_reason, result.output)
1028
    lines = [utils.SafeEncode(val)
1029
             for val in utils.TailFile(logfile, lines=20)]
1030
    _Fail("OS rename script failed (%s), last lines in the"
1031
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1032

    
1033

    
1034
def _GetBlockDevSymlinkPath(instance_name, idx):
1035
  return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1036
                        (instance_name, constants.DISK_SEPARATOR, idx))
1037

    
1038

    
1039
def _SymlinkBlockDev(instance_name, device_path, idx):
1040
  """Set up symlinks to a instance's block device.
1041

1042
  This is an auxiliary function run when an instance is start (on the primary
1043
  node) or when an instance is migrated (on the target node).
1044

1045

1046
  @param instance_name: the name of the target instance
1047
  @param device_path: path of the physical block device, on the node
1048
  @param idx: the disk index
1049
  @return: absolute path to the disk's symlink
1050

1051
  """
1052
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1053
  try:
1054
    os.symlink(device_path, link_name)
1055
  except OSError, err:
1056
    if err.errno == errno.EEXIST:
1057
      if (not os.path.islink(link_name) or
1058
          os.readlink(link_name) != device_path):
1059
        os.remove(link_name)
1060
        os.symlink(device_path, link_name)
1061
    else:
1062
      raise
1063

    
1064
  return link_name
1065

    
1066

    
1067
def _RemoveBlockDevLinks(instance_name, disks):
1068
  """Remove the block device symlinks belonging to the given instance.
1069

1070
  """
1071
  for idx, _ in enumerate(disks):
1072
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1073
    if os.path.islink(link_name):
1074
      try:
1075
        os.remove(link_name)
1076
      except OSError:
1077
        logging.exception("Can't remove symlink '%s'", link_name)
1078

    
1079

    
1080
def _GatherAndLinkBlockDevs(instance):
1081
  """Set up an instance's block device(s).
1082

1083
  This is run on the primary node at instance startup. The block
1084
  devices must be already assembled.
1085

1086
  @type instance: L{objects.Instance}
1087
  @param instance: the instance whose disks we shoul assemble
1088
  @rtype: list
1089
  @return: list of (disk_object, device_path)
1090

1091
  """
1092
  block_devices = []
1093
  for idx, disk in enumerate(instance.disks):
1094
    device = _RecursiveFindBD(disk)
1095
    if device is None:
1096
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1097
                                    str(disk))
1098
    device.Open()
1099
    try:
1100
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1101
    except OSError, e:
1102
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1103
                                    e.strerror)
1104

    
1105
    block_devices.append((disk, link_name))
1106

    
1107
  return block_devices
1108

    
1109

    
1110
def StartInstance(instance, startup_paused):
1111
  """Start an instance.
1112

1113
  @type instance: L{objects.Instance}
1114
  @param instance: the instance object
1115
  @type startup_paused: bool
1116
  @param instance: pause instance at startup?
1117
  @rtype: None
1118

1119
  """
1120
  running_instances = GetInstanceList([instance.hypervisor])
1121

    
1122
  if instance.name in running_instances:
1123
    logging.info("Instance %s already running, not starting", instance.name)
1124
    return
1125

    
1126
  try:
1127
    block_devices = _GatherAndLinkBlockDevs(instance)
1128
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1129
    hyper.StartInstance(instance, block_devices, startup_paused)
1130
  except errors.BlockDeviceError, err:
1131
    _Fail("Block device error: %s", err, exc=True)
1132
  except errors.HypervisorError, err:
1133
    _RemoveBlockDevLinks(instance.name, instance.disks)
1134
    _Fail("Hypervisor error: %s", err, exc=True)
1135

    
1136

    
1137
def InstanceShutdown(instance, timeout):
1138
  """Shut an instance down.
1139

1140
  @note: this functions uses polling with a hardcoded timeout.
1141

1142
  @type instance: L{objects.Instance}
1143
  @param instance: the instance object
1144
  @type timeout: integer
1145
  @param timeout: maximum timeout for soft shutdown
1146
  @rtype: None
1147

1148
  """
1149
  hv_name = instance.hypervisor
1150
  hyper = hypervisor.GetHypervisor(hv_name)
1151
  iname = instance.name
1152

    
1153
  if instance.name not in hyper.ListInstances():
1154
    logging.info("Instance %s not running, doing nothing", iname)
1155
    return
1156

    
1157
  class _TryShutdown:
1158
    def __init__(self):
1159
      self.tried_once = False
1160

    
1161
    def __call__(self):
1162
      if iname not in hyper.ListInstances():
1163
        return
1164

    
1165
      try:
1166
        hyper.StopInstance(instance, retry=self.tried_once)
1167
      except errors.HypervisorError, err:
1168
        if iname not in hyper.ListInstances():
1169
          # if the instance is no longer existing, consider this a
1170
          # success and go to cleanup
1171
          return
1172

    
1173
        _Fail("Failed to stop instance %s: %s", iname, err)
1174

    
1175
      self.tried_once = True
1176

    
1177
      raise utils.RetryAgain()
1178

    
1179
  try:
1180
    utils.Retry(_TryShutdown(), 5, timeout)
1181
  except utils.RetryTimeout:
1182
    # the shutdown did not succeed
1183
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1184

    
1185
    try:
1186
      hyper.StopInstance(instance, force=True)
1187
    except errors.HypervisorError, err:
1188
      if iname in hyper.ListInstances():
1189
        # only raise an error if the instance still exists, otherwise
1190
        # the error could simply be "instance ... unknown"!
1191
        _Fail("Failed to force stop instance %s: %s", iname, err)
1192

    
1193
    time.sleep(1)
1194

    
1195
    if iname in hyper.ListInstances():
1196
      _Fail("Could not shutdown instance %s even by destroy", iname)
1197

    
1198
  try:
1199
    hyper.CleanupInstance(instance.name)
1200
  except errors.HypervisorError, err:
1201
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1202

    
1203
  _RemoveBlockDevLinks(iname, instance.disks)
1204

    
1205

    
1206
def InstanceReboot(instance, reboot_type, shutdown_timeout):
1207
  """Reboot an instance.
1208

1209
  @type instance: L{objects.Instance}
1210
  @param instance: the instance object to reboot
1211
  @type reboot_type: str
1212
  @param reboot_type: the type of reboot, one the following
1213
    constants:
1214
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1215
        instance OS, do not recreate the VM
1216
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1217
        restart the VM (at the hypervisor level)
1218
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1219
        not accepted here, since that mode is handled differently, in
1220
        cmdlib, and translates into full stop and start of the
1221
        instance (instead of a call_instance_reboot RPC)
1222
  @type shutdown_timeout: integer
1223
  @param shutdown_timeout: maximum timeout for soft shutdown
1224
  @rtype: None
1225

1226
  """
1227
  running_instances = GetInstanceList([instance.hypervisor])
1228

    
1229
  if instance.name not in running_instances:
1230
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1231

    
1232
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1233
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1234
    try:
1235
      hyper.RebootInstance(instance)
1236
    except errors.HypervisorError, err:
1237
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1238
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1239
    try:
1240
      InstanceShutdown(instance, shutdown_timeout)
1241
      return StartInstance(instance, False)
1242
    except errors.HypervisorError, err:
1243
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1244
  else:
1245
    _Fail("Invalid reboot_type received: %s", reboot_type)
1246

    
1247

    
1248
def MigrationInfo(instance):
1249
  """Gather information about an instance to be migrated.
1250

1251
  @type instance: L{objects.Instance}
1252
  @param instance: the instance definition
1253

1254
  """
1255
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1256
  try:
1257
    info = hyper.MigrationInfo(instance)
1258
  except errors.HypervisorError, err:
1259
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1260
  return info
1261

    
1262

    
1263
def AcceptInstance(instance, info, target):
1264
  """Prepare the node to accept an instance.
1265

1266
  @type instance: L{objects.Instance}
1267
  @param instance: the instance definition
1268
  @type info: string/data (opaque)
1269
  @param info: migration information, from the source node
1270
  @type target: string
1271
  @param target: target host (usually ip), on this node
1272

1273
  """
1274
  # TODO: why is this required only for DTS_EXT_MIRROR?
1275
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1276
    # Create the symlinks, as the disks are not active
1277
    # in any way
1278
    try:
1279
      _GatherAndLinkBlockDevs(instance)
1280
    except errors.BlockDeviceError, err:
1281
      _Fail("Block device error: %s", err, exc=True)
1282

    
1283
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1284
  try:
1285
    hyper.AcceptInstance(instance, info, target)
1286
  except errors.HypervisorError, err:
1287
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1288
      _RemoveBlockDevLinks(instance.name, instance.disks)
1289
    _Fail("Failed to accept instance: %s", err, exc=True)
1290

    
1291

    
1292
def FinalizeMigration(instance, info, success):
1293
  """Finalize any preparation to accept an instance.
1294

1295
  @type instance: L{objects.Instance}
1296
  @param instance: the instance definition
1297
  @type info: string/data (opaque)
1298
  @param info: migration information, from the source node
1299
  @type success: boolean
1300
  @param success: whether the migration was a success or a failure
1301

1302
  """
1303
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1304
  try:
1305
    hyper.FinalizeMigration(instance, info, success)
1306
  except errors.HypervisorError, err:
1307
    _Fail("Failed to finalize migration: %s", err, exc=True)
1308

    
1309

    
1310
def MigrateInstance(instance, target, live):
1311
  """Migrates an instance to another node.
1312

1313
  @type instance: L{objects.Instance}
1314
  @param instance: the instance definition
1315
  @type target: string
1316
  @param target: the target node name
1317
  @type live: boolean
1318
  @param live: whether the migration should be done live or not (the
1319
      interpretation of this parameter is left to the hypervisor)
1320
  @rtype: tuple
1321
  @return: a tuple of (success, msg) where:
1322
      - succes is a boolean denoting the success/failure of the operation
1323
      - msg is a string with details in case of failure
1324

1325
  """
1326
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1327

    
1328
  try:
1329
    hyper.MigrateInstance(instance, target, live)
1330
  except errors.HypervisorError, err:
1331
    _Fail("Failed to migrate instance: %s", err, exc=True)
1332

    
1333

    
1334
def BlockdevCreate(disk, size, owner, on_primary, info):
1335
  """Creates a block device for an instance.
1336

1337
  @type disk: L{objects.Disk}
1338
  @param disk: the object describing the disk we should create
1339
  @type size: int
1340
  @param size: the size of the physical underlying device, in MiB
1341
  @type owner: str
1342
  @param owner: the name of the instance for which disk is created,
1343
      used for device cache data
1344
  @type on_primary: boolean
1345
  @param on_primary:  indicates if it is the primary node or not
1346
  @type info: string
1347
  @param info: string that will be sent to the physical device
1348
      creation, used for example to set (LVM) tags on LVs
1349

1350
  @return: the new unique_id of the device (this can sometime be
1351
      computed only after creation), or None. On secondary nodes,
1352
      it's not required to return anything.
1353

1354
  """
1355
  # TODO: remove the obsolete "size" argument
1356
  # pylint: disable=W0613
1357
  clist = []
1358
  if disk.children:
1359
    for child in disk.children:
1360
      try:
1361
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1362
      except errors.BlockDeviceError, err:
1363
        _Fail("Can't assemble device %s: %s", child, err)
1364
      if on_primary or disk.AssembleOnSecondary():
1365
        # we need the children open in case the device itself has to
1366
        # be assembled
1367
        try:
1368
          # pylint: disable=E1103
1369
          crdev.Open()
1370
        except errors.BlockDeviceError, err:
1371
          _Fail("Can't make child '%s' read-write: %s", child, err)
1372
      clist.append(crdev)
1373

    
1374
  try:
1375
    device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1376
  except errors.BlockDeviceError, err:
1377
    _Fail("Can't create block device: %s", err)
1378

    
1379
  if on_primary or disk.AssembleOnSecondary():
1380
    try:
1381
      device.Assemble()
1382
    except errors.BlockDeviceError, err:
1383
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1384
    device.SetSyncSpeed(constants.SYNC_SPEED)
1385
    if on_primary or disk.OpenOnSecondary():
1386
      try:
1387
        device.Open(force=True)
1388
      except errors.BlockDeviceError, err:
1389
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1390
    DevCacheManager.UpdateCache(device.dev_path, owner,
1391
                                on_primary, disk.iv_name)
1392

    
1393
  device.SetInfo(info)
1394

    
1395
  return device.unique_id
1396

    
1397

    
1398
def _WipeDevice(path, offset, size):
1399
  """This function actually wipes the device.
1400

1401
  @param path: The path to the device to wipe
1402
  @param offset: The offset in MiB in the file
1403
  @param size: The size in MiB to write
1404

1405
  """
1406
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1407
         "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1408
         "count=%d" % size]
1409
  result = utils.RunCmd(cmd)
1410

    
1411
  if result.failed:
1412
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1413
          result.fail_reason, result.output)
1414

    
1415

    
1416
def BlockdevWipe(disk, offset, size):
1417
  """Wipes a block device.
1418

1419
  @type disk: L{objects.Disk}
1420
  @param disk: the disk object we want to wipe
1421
  @type offset: int
1422
  @param offset: The offset in MiB in the file
1423
  @type size: int
1424
  @param size: The size in MiB to write
1425

1426
  """
1427
  try:
1428
    rdev = _RecursiveFindBD(disk)
1429
  except errors.BlockDeviceError:
1430
    rdev = None
1431

    
1432
  if not rdev:
1433
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1434

    
1435
  # Do cross verify some of the parameters
1436
  if offset > rdev.size:
1437
    _Fail("Offset is bigger than device size")
1438
  if (offset + size) > rdev.size:
1439
    _Fail("The provided offset and size to wipe is bigger than device size")
1440

    
1441
  _WipeDevice(rdev.dev_path, offset, size)
1442

    
1443

    
1444
def BlockdevPauseResumeSync(disks, pause):
1445
  """Pause or resume the sync of the block device.
1446

1447
  @type disks: list of L{objects.Disk}
1448
  @param disks: the disks object we want to pause/resume
1449
  @type pause: bool
1450
  @param pause: Wheater to pause or resume
1451

1452
  """
1453
  success = []
1454
  for disk in disks:
1455
    try:
1456
      rdev = _RecursiveFindBD(disk)
1457
    except errors.BlockDeviceError:
1458
      rdev = None
1459

    
1460
    if not rdev:
1461
      success.append((False, ("Cannot change sync for device %s:"
1462
                              " device not found" % disk.iv_name)))
1463
      continue
1464

    
1465
    result = rdev.PauseResumeSync(pause)
1466

    
1467
    if result:
1468
      success.append((result, None))
1469
    else:
1470
      if pause:
1471
        msg = "Pause"
1472
      else:
1473
        msg = "Resume"
1474
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1475

    
1476
  return success
1477

    
1478

    
1479
def BlockdevRemove(disk):
1480
  """Remove a block device.
1481

1482
  @note: This is intended to be called recursively.
1483

1484
  @type disk: L{objects.Disk}
1485
  @param disk: the disk object we should remove
1486
  @rtype: boolean
1487
  @return: the success of the operation
1488

1489
  """
1490
  msgs = []
1491
  try:
1492
    rdev = _RecursiveFindBD(disk)
1493
  except errors.BlockDeviceError, err:
1494
    # probably can't attach
1495
    logging.info("Can't attach to device %s in remove", disk)
1496
    rdev = None
1497
  if rdev is not None:
1498
    r_path = rdev.dev_path
1499
    try:
1500
      rdev.Remove()
1501
    except errors.BlockDeviceError, err:
1502
      msgs.append(str(err))
1503
    if not msgs:
1504
      DevCacheManager.RemoveCache(r_path)
1505

    
1506
  if disk.children:
1507
    for child in disk.children:
1508
      try:
1509
        BlockdevRemove(child)
1510
      except RPCFail, err:
1511
        msgs.append(str(err))
1512

    
1513
  if msgs:
1514
    _Fail("; ".join(msgs))
1515

    
1516

    
1517
def _RecursiveAssembleBD(disk, owner, as_primary):
1518
  """Activate a block device for an instance.
1519

1520
  This is run on the primary and secondary nodes for an instance.
1521

1522
  @note: this function is called recursively.
1523

1524
  @type disk: L{objects.Disk}
1525
  @param disk: the disk we try to assemble
1526
  @type owner: str
1527
  @param owner: the name of the instance which owns the disk
1528
  @type as_primary: boolean
1529
  @param as_primary: if we should make the block device
1530
      read/write
1531

1532
  @return: the assembled device or None (in case no device
1533
      was assembled)
1534
  @raise errors.BlockDeviceError: in case there is an error
1535
      during the activation of the children or the device
1536
      itself
1537

1538
  """
1539
  children = []
1540
  if disk.children:
1541
    mcn = disk.ChildrenNeeded()
1542
    if mcn == -1:
1543
      mcn = 0 # max number of Nones allowed
1544
    else:
1545
      mcn = len(disk.children) - mcn # max number of Nones
1546
    for chld_disk in disk.children:
1547
      try:
1548
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1549
      except errors.BlockDeviceError, err:
1550
        if children.count(None) >= mcn:
1551
          raise
1552
        cdev = None
1553
        logging.error("Error in child activation (but continuing): %s",
1554
                      str(err))
1555
      children.append(cdev)
1556

    
1557
  if as_primary or disk.AssembleOnSecondary():
1558
    r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1559
    r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1560
    result = r_dev
1561
    if as_primary or disk.OpenOnSecondary():
1562
      r_dev.Open()
1563
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1564
                                as_primary, disk.iv_name)
1565

    
1566
  else:
1567
    result = True
1568
  return result
1569

    
1570

    
1571
def BlockdevAssemble(disk, owner, as_primary, idx):
1572
  """Activate a block device for an instance.
1573

1574
  This is a wrapper over _RecursiveAssembleBD.
1575

1576
  @rtype: str or boolean
1577
  @return: a C{/dev/...} path for primary nodes, and
1578
      C{True} for secondary nodes
1579

1580
  """
1581
  try:
1582
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1583
    if isinstance(result, bdev.BlockDev):
1584
      # pylint: disable=E1103
1585
      result = result.dev_path
1586
      if as_primary:
1587
        _SymlinkBlockDev(owner, result, idx)
1588
  except errors.BlockDeviceError, err:
1589
    _Fail("Error while assembling disk: %s", err, exc=True)
1590
  except OSError, err:
1591
    _Fail("Error while symlinking disk: %s", err, exc=True)
1592

    
1593
  return result
1594

    
1595

    
1596
def BlockdevShutdown(disk):
1597
  """Shut down a block device.
1598

1599
  First, if the device is assembled (Attach() is successful), then
1600
  the device is shutdown. Then the children of the device are
1601
  shutdown.
1602

1603
  This function is called recursively. Note that we don't cache the
1604
  children or such, as oppossed to assemble, shutdown of different
1605
  devices doesn't require that the upper device was active.
1606

1607
  @type disk: L{objects.Disk}
1608
  @param disk: the description of the disk we should
1609
      shutdown
1610
  @rtype: None
1611

1612
  """
1613
  msgs = []
1614
  r_dev = _RecursiveFindBD(disk)
1615
  if r_dev is not None:
1616
    r_path = r_dev.dev_path
1617
    try:
1618
      r_dev.Shutdown()
1619
      DevCacheManager.RemoveCache(r_path)
1620
    except errors.BlockDeviceError, err:
1621
      msgs.append(str(err))
1622

    
1623
  if disk.children:
1624
    for child in disk.children:
1625
      try:
1626
        BlockdevShutdown(child)
1627
      except RPCFail, err:
1628
        msgs.append(str(err))
1629

    
1630
  if msgs:
1631
    _Fail("; ".join(msgs))
1632

    
1633

    
1634
def BlockdevAddchildren(parent_cdev, new_cdevs):
1635
  """Extend a mirrored block device.
1636

1637
  @type parent_cdev: L{objects.Disk}
1638
  @param parent_cdev: the disk to which we should add children
1639
  @type new_cdevs: list of L{objects.Disk}
1640
  @param new_cdevs: the list of children which we should add
1641
  @rtype: None
1642

1643
  """
1644
  parent_bdev = _RecursiveFindBD(parent_cdev)
1645
  if parent_bdev is None:
1646
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1647
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1648
  if new_bdevs.count(None) > 0:
1649
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1650
  parent_bdev.AddChildren(new_bdevs)
1651

    
1652

    
1653
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1654
  """Shrink a mirrored block device.
1655

1656
  @type parent_cdev: L{objects.Disk}
1657
  @param parent_cdev: the disk from which we should remove children
1658
  @type new_cdevs: list of L{objects.Disk}
1659
  @param new_cdevs: the list of children which we should remove
1660
  @rtype: None
1661

1662
  """
1663
  parent_bdev = _RecursiveFindBD(parent_cdev)
1664
  if parent_bdev is None:
1665
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1666
  devs = []
1667
  for disk in new_cdevs:
1668
    rpath = disk.StaticDevPath()
1669
    if rpath is None:
1670
      bd = _RecursiveFindBD(disk)
1671
      if bd is None:
1672
        _Fail("Can't find device %s while removing children", disk)
1673
      else:
1674
        devs.append(bd.dev_path)
1675
    else:
1676
      if not utils.IsNormAbsPath(rpath):
1677
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1678
      devs.append(rpath)
1679
  parent_bdev.RemoveChildren(devs)
1680

    
1681

    
1682
def BlockdevGetmirrorstatus(disks):
1683
  """Get the mirroring status of a list of devices.
1684

1685
  @type disks: list of L{objects.Disk}
1686
  @param disks: the list of disks which we should query
1687
  @rtype: disk
1688
  @return: List of L{objects.BlockDevStatus}, one for each disk
1689
  @raise errors.BlockDeviceError: if any of the disks cannot be
1690
      found
1691

1692
  """
1693
  stats = []
1694
  for dsk in disks:
1695
    rbd = _RecursiveFindBD(dsk)
1696
    if rbd is None:
1697
      _Fail("Can't find device %s", dsk)
1698

    
1699
    stats.append(rbd.CombinedSyncStatus())
1700

    
1701
  return stats
1702

    
1703

    
1704
def BlockdevGetmirrorstatusMulti(disks):
1705
  """Get the mirroring status of a list of devices.
1706

1707
  @type disks: list of L{objects.Disk}
1708
  @param disks: the list of disks which we should query
1709
  @rtype: disk
1710
  @return: List of tuples, (bool, status), one for each disk; bool denotes
1711
    success/failure, status is L{objects.BlockDevStatus} on success, string
1712
    otherwise
1713

1714
  """
1715
  result = []
1716
  for disk in disks:
1717
    try:
1718
      rbd = _RecursiveFindBD(disk)
1719
      if rbd is None:
1720
        result.append((False, "Can't find device %s" % disk))
1721
        continue
1722

    
1723
      status = rbd.CombinedSyncStatus()
1724
    except errors.BlockDeviceError, err:
1725
      logging.exception("Error while getting disk status")
1726
      result.append((False, str(err)))
1727
    else:
1728
      result.append((True, status))
1729

    
1730
  assert len(disks) == len(result)
1731

    
1732
  return result
1733

    
1734

    
1735
def _RecursiveFindBD(disk):
1736
  """Check if a device is activated.
1737

1738
  If so, return information about the real device.
1739

1740
  @type disk: L{objects.Disk}
1741
  @param disk: the disk object we need to find
1742

1743
  @return: None if the device can't be found,
1744
      otherwise the device instance
1745

1746
  """
1747
  children = []
1748
  if disk.children:
1749
    for chdisk in disk.children:
1750
      children.append(_RecursiveFindBD(chdisk))
1751

    
1752
  return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1753

    
1754

    
1755
def _OpenRealBD(disk):
1756
  """Opens the underlying block device of a disk.
1757

1758
  @type disk: L{objects.Disk}
1759
  @param disk: the disk object we want to open
1760

1761
  """
1762
  real_disk = _RecursiveFindBD(disk)
1763
  if real_disk is None:
1764
    _Fail("Block device '%s' is not set up", disk)
1765

    
1766
  real_disk.Open()
1767

    
1768
  return real_disk
1769

    
1770

    
1771
def BlockdevFind(disk):
1772
  """Check if a device is activated.
1773

1774
  If it is, return information about the real device.
1775

1776
  @type disk: L{objects.Disk}
1777
  @param disk: the disk to find
1778
  @rtype: None or objects.BlockDevStatus
1779
  @return: None if the disk cannot be found, otherwise a the current
1780
           information
1781

1782
  """
1783
  try:
1784
    rbd = _RecursiveFindBD(disk)
1785
  except errors.BlockDeviceError, err:
1786
    _Fail("Failed to find device: %s", err, exc=True)
1787

    
1788
  if rbd is None:
1789
    return None
1790

    
1791
  return rbd.GetSyncStatus()
1792

    
1793

    
1794
def BlockdevGetsize(disks):
1795
  """Computes the size of the given disks.
1796

1797
  If a disk is not found, returns None instead.
1798

1799
  @type disks: list of L{objects.Disk}
1800
  @param disks: the list of disk to compute the size for
1801
  @rtype: list
1802
  @return: list with elements None if the disk cannot be found,
1803
      otherwise the size
1804

1805
  """
1806
  result = []
1807
  for cf in disks:
1808
    try:
1809
      rbd = _RecursiveFindBD(cf)
1810
    except errors.BlockDeviceError:
1811
      result.append(None)
1812
      continue
1813
    if rbd is None:
1814
      result.append(None)
1815
    else:
1816
      result.append(rbd.GetActualSize())
1817
  return result
1818

    
1819

    
1820
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1821
  """Export a block device to a remote node.
1822

1823
  @type disk: L{objects.Disk}
1824
  @param disk: the description of the disk to export
1825
  @type dest_node: str
1826
  @param dest_node: the destination node to export to
1827
  @type dest_path: str
1828
  @param dest_path: the destination path on the target node
1829
  @type cluster_name: str
1830
  @param cluster_name: the cluster name, needed for SSH hostalias
1831
  @rtype: None
1832

1833
  """
1834
  real_disk = _OpenRealBD(disk)
1835

    
1836
  # the block size on the read dd is 1MiB to match our units
1837
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1838
                               "dd if=%s bs=1048576 count=%s",
1839
                               real_disk.dev_path, str(disk.size))
1840

    
1841
  # we set here a smaller block size as, due to ssh buffering, more
1842
  # than 64-128k will mostly ignored; we use nocreat to fail if the
1843
  # device is not already there or we pass a wrong path; we use
1844
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1845
  # to not buffer too much memory; this means that at best, we flush
1846
  # every 64k, which will not be very fast
1847
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1848
                                " oflag=dsync", dest_path)
1849

    
1850
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1851
                                                   constants.GANETI_RUNAS,
1852
                                                   destcmd)
1853

    
1854
  # all commands have been checked, so we're safe to combine them
1855
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1856

    
1857
  result = utils.RunCmd(["bash", "-c", command])
1858

    
1859
  if result.failed:
1860
    _Fail("Disk copy command '%s' returned error: %s"
1861
          " output: %s", command, result.fail_reason, result.output)
1862

    
1863

    
1864
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1865
  """Write a file to the filesystem.
1866

1867
  This allows the master to overwrite(!) a file. It will only perform
1868
  the operation if the file belongs to a list of configuration files.
1869

1870
  @type file_name: str
1871
  @param file_name: the target file name
1872
  @type data: str
1873
  @param data: the new contents of the file
1874
  @type mode: int
1875
  @param mode: the mode to give the file (can be None)
1876
  @type uid: string
1877
  @param uid: the owner of the file
1878
  @type gid: string
1879
  @param gid: the group of the file
1880
  @type atime: float
1881
  @param atime: the atime to set on the file (can be None)
1882
  @type mtime: float
1883
  @param mtime: the mtime to set on the file (can be None)
1884
  @rtype: None
1885

1886
  """
1887
  if not os.path.isabs(file_name):
1888
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1889

    
1890
  if file_name not in _ALLOWED_UPLOAD_FILES:
1891
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1892
          file_name)
1893

    
1894
  raw_data = _Decompress(data)
1895

    
1896
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
1897
    _Fail("Invalid username/groupname type")
1898

    
1899
  getents = runtime.GetEnts()
1900
  uid = getents.LookupUser(uid)
1901
  gid = getents.LookupGroup(gid)
1902

    
1903
  utils.SafeWriteFile(file_name, None,
1904
                      data=raw_data, mode=mode, uid=uid, gid=gid,
1905
                      atime=atime, mtime=mtime)
1906

    
1907

    
1908
def RunOob(oob_program, command, node, timeout):
1909
  """Executes oob_program with given command on given node.
1910

1911
  @param oob_program: The path to the executable oob_program
1912
  @param command: The command to invoke on oob_program
1913
  @param node: The node given as an argument to the program
1914
  @param timeout: Timeout after which we kill the oob program
1915

1916
  @return: stdout
1917
  @raise RPCFail: If execution fails for some reason
1918

1919
  """
1920
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
1921

    
1922
  if result.failed:
1923
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
1924
          result.fail_reason, result.output)
1925

    
1926
  return result.stdout
1927

    
1928

    
1929
def WriteSsconfFiles(values):
1930
  """Update all ssconf files.
1931

1932
  Wrapper around the SimpleStore.WriteFiles.
1933

1934
  """
1935
  ssconf.SimpleStore().WriteFiles(values)
1936

    
1937

    
1938
def _ErrnoOrStr(err):
1939
  """Format an EnvironmentError exception.
1940

1941
  If the L{err} argument has an errno attribute, it will be looked up
1942
  and converted into a textual C{E...} description. Otherwise the
1943
  string representation of the error will be returned.
1944

1945
  @type err: L{EnvironmentError}
1946
  @param err: the exception to format
1947

1948
  """
1949
  if hasattr(err, "errno"):
1950
    detail = errno.errorcode[err.errno]
1951
  else:
1952
    detail = str(err)
1953
  return detail
1954

    
1955

    
1956
def _OSOndiskAPIVersion(os_dir):
1957
  """Compute and return the API version of a given OS.
1958

1959
  This function will try to read the API version of the OS residing in
1960
  the 'os_dir' directory.
1961

1962
  @type os_dir: str
1963
  @param os_dir: the directory in which we should look for the OS
1964
  @rtype: tuple
1965
  @return: tuple (status, data) with status denoting the validity and
1966
      data holding either the vaid versions or an error message
1967

1968
  """
1969
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1970

    
1971
  try:
1972
    st = os.stat(api_file)
1973
  except EnvironmentError, err:
1974
    return False, ("Required file '%s' not found under path %s: %s" %
1975
                   (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1976

    
1977
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1978
    return False, ("File '%s' in %s is not a regular file" %
1979
                   (constants.OS_API_FILE, os_dir))
1980

    
1981
  try:
1982
    api_versions = utils.ReadFile(api_file).splitlines()
1983
  except EnvironmentError, err:
1984
    return False, ("Error while reading the API version file at %s: %s" %
1985
                   (api_file, _ErrnoOrStr(err)))
1986

    
1987
  try:
1988
    api_versions = [int(version.strip()) for version in api_versions]
1989
  except (TypeError, ValueError), err:
1990
    return False, ("API version(s) can't be converted to integer: %s" %
1991
                   str(err))
1992

    
1993
  return True, api_versions
1994

    
1995

    
1996
def DiagnoseOS(top_dirs=None):
1997
  """Compute the validity for all OSes.
1998

1999
  @type top_dirs: list
2000
  @param top_dirs: the list of directories in which to
2001
      search (if not given defaults to
2002
      L{constants.OS_SEARCH_PATH})
2003
  @rtype: list of L{objects.OS}
2004
  @return: a list of tuples (name, path, status, diagnose, variants,
2005
      parameters, api_version) for all (potential) OSes under all
2006
      search paths, where:
2007
          - name is the (potential) OS name
2008
          - path is the full path to the OS
2009
          - status True/False is the validity of the OS
2010
          - diagnose is the error message for an invalid OS, otherwise empty
2011
          - variants is a list of supported OS variants, if any
2012
          - parameters is a list of (name, help) parameters, if any
2013
          - api_version is a list of support OS API versions
2014

2015
  """
2016
  if top_dirs is None:
2017
    top_dirs = constants.OS_SEARCH_PATH
2018

    
2019
  result = []
2020
  for dir_name in top_dirs:
2021
    if os.path.isdir(dir_name):
2022
      try:
2023
        f_names = utils.ListVisibleFiles(dir_name)
2024
      except EnvironmentError, err:
2025
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2026
        break
2027
      for name in f_names:
2028
        os_path = utils.PathJoin(dir_name, name)
2029
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2030
        if status:
2031
          diagnose = ""
2032
          variants = os_inst.supported_variants
2033
          parameters = os_inst.supported_parameters
2034
          api_versions = os_inst.api_versions
2035
        else:
2036
          diagnose = os_inst
2037
          variants = parameters = api_versions = []
2038
        result.append((name, os_path, status, diagnose, variants,
2039
                       parameters, api_versions))
2040

    
2041
  return result
2042

    
2043

    
2044
def _TryOSFromDisk(name, base_dir=None):
2045
  """Create an OS instance from disk.
2046

2047
  This function will return an OS instance if the given name is a
2048
  valid OS name.
2049

2050
  @type base_dir: string
2051
  @keyword base_dir: Base directory containing OS installations.
2052
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2053
  @rtype: tuple
2054
  @return: success and either the OS instance if we find a valid one,
2055
      or error message
2056

2057
  """
2058
  if base_dir is None:
2059
    os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2060
  else:
2061
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2062

    
2063
  if os_dir is None:
2064
    return False, "Directory for OS %s not found in search path" % name
2065

    
2066
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2067
  if not status:
2068
    # push the error up
2069
    return status, api_versions
2070

    
2071
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2072
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2073
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2074

    
2075
  # OS Files dictionary, we will populate it with the absolute path
2076
  # names; if the value is True, then it is a required file, otherwise
2077
  # an optional one
2078
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2079

    
2080
  if max(api_versions) >= constants.OS_API_V15:
2081
    os_files[constants.OS_VARIANTS_FILE] = False
2082

    
2083
  if max(api_versions) >= constants.OS_API_V20:
2084
    os_files[constants.OS_PARAMETERS_FILE] = True
2085
  else:
2086
    del os_files[constants.OS_SCRIPT_VERIFY]
2087

    
2088
  for (filename, required) in os_files.items():
2089
    os_files[filename] = utils.PathJoin(os_dir, filename)
2090

    
2091
    try:
2092
      st = os.stat(os_files[filename])
2093
    except EnvironmentError, err:
2094
      if err.errno == errno.ENOENT and not required:
2095
        del os_files[filename]
2096
        continue
2097
      return False, ("File '%s' under path '%s' is missing (%s)" %
2098
                     (filename, os_dir, _ErrnoOrStr(err)))
2099

    
2100
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2101
      return False, ("File '%s' under path '%s' is not a regular file" %
2102
                     (filename, os_dir))
2103

    
2104
    if filename in constants.OS_SCRIPTS:
2105
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2106
        return False, ("File '%s' under path '%s' is not executable" %
2107
                       (filename, os_dir))
2108

    
2109
  variants = []
2110
  if constants.OS_VARIANTS_FILE in os_files:
2111
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2112
    try:
2113
      variants = utils.ReadFile(variants_file).splitlines()
2114
    except EnvironmentError, err:
2115
      # we accept missing files, but not other errors
2116
      if err.errno != errno.ENOENT:
2117
        return False, ("Error while reading the OS variants file at %s: %s" %
2118
                       (variants_file, _ErrnoOrStr(err)))
2119

    
2120
  parameters = []
2121
  if constants.OS_PARAMETERS_FILE in os_files:
2122
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2123
    try:
2124
      parameters = utils.ReadFile(parameters_file).splitlines()
2125
    except EnvironmentError, err:
2126
      return False, ("Error while reading the OS parameters file at %s: %s" %
2127
                     (parameters_file, _ErrnoOrStr(err)))
2128
    parameters = [v.split(None, 1) for v in parameters]
2129

    
2130
  os_obj = objects.OS(name=name, path=os_dir,
2131
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2132
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2133
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2134
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2135
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2136
                                                 None),
2137
                      supported_variants=variants,
2138
                      supported_parameters=parameters,
2139
                      api_versions=api_versions)
2140
  return True, os_obj
2141

    
2142

    
2143
def OSFromDisk(name, base_dir=None):
2144
  """Create an OS instance from disk.
2145

2146
  This function will return an OS instance if the given name is a
2147
  valid OS name. Otherwise, it will raise an appropriate
2148
  L{RPCFail} exception, detailing why this is not a valid OS.
2149

2150
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2151
  an exception but returns true/false status data.
2152

2153
  @type base_dir: string
2154
  @keyword base_dir: Base directory containing OS installations.
2155
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2156
  @rtype: L{objects.OS}
2157
  @return: the OS instance if we find a valid one
2158
  @raise RPCFail: if we don't find a valid OS
2159

2160
  """
2161
  name_only = objects.OS.GetName(name)
2162
  status, payload = _TryOSFromDisk(name_only, base_dir)
2163

    
2164
  if not status:
2165
    _Fail(payload)
2166

    
2167
  return payload
2168

    
2169

    
2170
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2171
  """Calculate the basic environment for an os script.
2172

2173
  @type os_name: str
2174
  @param os_name: full operating system name (including variant)
2175
  @type inst_os: L{objects.OS}
2176
  @param inst_os: operating system for which the environment is being built
2177
  @type os_params: dict
2178
  @param os_params: the OS parameters
2179
  @type debug: integer
2180
  @param debug: debug level (0 or 1, for OS Api 10)
2181
  @rtype: dict
2182
  @return: dict of environment variables
2183
  @raise errors.BlockDeviceError: if the block device
2184
      cannot be found
2185

2186
  """
2187
  result = {}
2188
  api_version = \
2189
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2190
  result["OS_API_VERSION"] = "%d" % api_version
2191
  result["OS_NAME"] = inst_os.name
2192
  result["DEBUG_LEVEL"] = "%d" % debug
2193

    
2194
  # OS variants
2195
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2196
    variant = objects.OS.GetVariant(os_name)
2197
    if not variant:
2198
      variant = inst_os.supported_variants[0]
2199
  else:
2200
    variant = ""
2201
  result["OS_VARIANT"] = variant
2202

    
2203
  # OS params
2204
  for pname, pvalue in os_params.items():
2205
    result["OSP_%s" % pname.upper()] = pvalue
2206

    
2207
  return result
2208

    
2209

    
2210
def OSEnvironment(instance, inst_os, debug=0):
2211
  """Calculate the environment for an os script.
2212

2213
  @type instance: L{objects.Instance}
2214
  @param instance: target instance for the os script run
2215
  @type inst_os: L{objects.OS}
2216
  @param inst_os: operating system for which the environment is being built
2217
  @type debug: integer
2218
  @param debug: debug level (0 or 1, for OS Api 10)
2219
  @rtype: dict
2220
  @return: dict of environment variables
2221
  @raise errors.BlockDeviceError: if the block device
2222
      cannot be found
2223

2224
  """
2225
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2226

    
2227
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2228
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2229

    
2230
  result["HYPERVISOR"] = instance.hypervisor
2231
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2232
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2233
  result["INSTANCE_SECONDARY_NODES"] = \
2234
      ("%s" % " ".join(instance.secondary_nodes))
2235

    
2236
  # Disks
2237
  for idx, disk in enumerate(instance.disks):
2238
    real_disk = _OpenRealBD(disk)
2239
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2240
    result["DISK_%d_ACCESS" % idx] = disk.mode
2241
    if constants.HV_DISK_TYPE in instance.hvparams:
2242
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2243
        instance.hvparams[constants.HV_DISK_TYPE]
2244
    if disk.dev_type in constants.LDS_BLOCK:
2245
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2246
    elif disk.dev_type == constants.LD_FILE:
2247
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2248
        "file:%s" % disk.physical_id[0]
2249

    
2250
  # NICs
2251
  for idx, nic in enumerate(instance.nics):
2252
    result["NIC_%d_MAC" % idx] = nic.mac
2253
    if nic.ip:
2254
      result["NIC_%d_IP" % idx] = nic.ip
2255
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2256
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2257
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2258
    if nic.nicparams[constants.NIC_LINK]:
2259
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2260
    if constants.HV_NIC_TYPE in instance.hvparams:
2261
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2262
        instance.hvparams[constants.HV_NIC_TYPE]
2263

    
2264
  # HV/BE params
2265
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2266
    for key, value in source.items():
2267
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2268

    
2269
  return result
2270

    
2271

    
2272
def BlockdevGrow(disk, amount, dryrun):
2273
  """Grow a stack of block devices.
2274

2275
  This function is called recursively, with the childrens being the
2276
  first ones to resize.
2277

2278
  @type disk: L{objects.Disk}
2279
  @param disk: the disk to be grown
2280
  @type amount: integer
2281
  @param amount: the amount (in mebibytes) to grow with
2282
  @type dryrun: boolean
2283
  @param dryrun: whether to execute the operation in simulation mode
2284
      only, without actually increasing the size
2285
  @rtype: (status, result)
2286
  @return: a tuple with the status of the operation (True/False), and
2287
      the errors message if status is False
2288

2289
  """
2290
  r_dev = _RecursiveFindBD(disk)
2291
  if r_dev is None:
2292
    _Fail("Cannot find block device %s", disk)
2293

    
2294
  try:
2295
    r_dev.Grow(amount, dryrun)
2296
  except errors.BlockDeviceError, err:
2297
    _Fail("Failed to grow block device: %s", err, exc=True)
2298

    
2299

    
2300
def BlockdevSnapshot(disk):
2301
  """Create a snapshot copy of a block device.
2302

2303
  This function is called recursively, and the snapshot is actually created
2304
  just for the leaf lvm backend device.
2305

2306
  @type disk: L{objects.Disk}
2307
  @param disk: the disk to be snapshotted
2308
  @rtype: string
2309
  @return: snapshot disk ID as (vg, lv)
2310

2311
  """
2312
  if disk.dev_type == constants.LD_DRBD8:
2313
    if not disk.children:
2314
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2315
            disk.unique_id)
2316
    return BlockdevSnapshot(disk.children[0])
2317
  elif disk.dev_type == constants.LD_LV:
2318
    r_dev = _RecursiveFindBD(disk)
2319
    if r_dev is not None:
2320
      # FIXME: choose a saner value for the snapshot size
2321
      # let's stay on the safe side and ask for the full size, for now
2322
      return r_dev.Snapshot(disk.size)
2323
    else:
2324
      _Fail("Cannot find block device %s", disk)
2325
  else:
2326
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2327
          disk.unique_id, disk.dev_type)
2328

    
2329

    
2330
def FinalizeExport(instance, snap_disks):
2331
  """Write out the export configuration information.
2332

2333
  @type instance: L{objects.Instance}
2334
  @param instance: the instance which we export, used for
2335
      saving configuration
2336
  @type snap_disks: list of L{objects.Disk}
2337
  @param snap_disks: list of snapshot block devices, which
2338
      will be used to get the actual name of the dump file
2339

2340
  @rtype: None
2341

2342
  """
2343
  destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2344
  finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2345

    
2346
  config = objects.SerializableConfigParser()
2347

    
2348
  config.add_section(constants.INISECT_EXP)
2349
  config.set(constants.INISECT_EXP, "version", "0")
2350
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2351
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
2352
  config.set(constants.INISECT_EXP, "os", instance.os)
2353
  config.set(constants.INISECT_EXP, "compression", "none")
2354

    
2355
  config.add_section(constants.INISECT_INS)
2356
  config.set(constants.INISECT_INS, "name", instance.name)
2357
  config.set(constants.INISECT_INS, "memory", "%d" %
2358
             instance.beparams[constants.BE_MEMORY])
2359
  config.set(constants.INISECT_INS, "vcpus", "%d" %
2360
             instance.beparams[constants.BE_VCPUS])
2361
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2362
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2363
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2364

    
2365
  nic_total = 0
2366
  for nic_count, nic in enumerate(instance.nics):
2367
    nic_total += 1
2368
    config.set(constants.INISECT_INS, "nic%d_mac" %
2369
               nic_count, "%s" % nic.mac)
2370
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2371
    for param in constants.NICS_PARAMETER_TYPES:
2372
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2373
                 "%s" % nic.nicparams.get(param, None))
2374
  # TODO: redundant: on load can read nics until it doesn't exist
2375
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2376

    
2377
  disk_total = 0
2378
  for disk_count, disk in enumerate(snap_disks):
2379
    if disk:
2380
      disk_total += 1
2381
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2382
                 ("%s" % disk.iv_name))
2383
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2384
                 ("%s" % disk.physical_id[1]))
2385
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2386
                 ("%d" % disk.size))
2387

    
2388
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2389

    
2390
  # New-style hypervisor/backend parameters
2391

    
2392
  config.add_section(constants.INISECT_HYP)
2393
  for name, value in instance.hvparams.items():
2394
    if name not in constants.HVC_GLOBALS:
2395
      config.set(constants.INISECT_HYP, name, str(value))
2396

    
2397
  config.add_section(constants.INISECT_BEP)
2398
  for name, value in instance.beparams.items():
2399
    config.set(constants.INISECT_BEP, name, str(value))
2400

    
2401
  config.add_section(constants.INISECT_OSP)
2402
  for name, value in instance.osparams.items():
2403
    config.set(constants.INISECT_OSP, name, str(value))
2404

    
2405
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2406
                  data=config.Dumps())
2407
  shutil.rmtree(finaldestdir, ignore_errors=True)
2408
  shutil.move(destdir, finaldestdir)
2409

    
2410

    
2411
def ExportInfo(dest):
2412
  """Get export configuration information.
2413

2414
  @type dest: str
2415
  @param dest: directory containing the export
2416

2417
  @rtype: L{objects.SerializableConfigParser}
2418
  @return: a serializable config file containing the
2419
      export info
2420

2421
  """
2422
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2423

    
2424
  config = objects.SerializableConfigParser()
2425
  config.read(cff)
2426

    
2427
  if (not config.has_section(constants.INISECT_EXP) or
2428
      not config.has_section(constants.INISECT_INS)):
2429
    _Fail("Export info file doesn't have the required fields")
2430

    
2431
  return config.Dumps()
2432

    
2433

    
2434
def ListExports():
2435
  """Return a list of exports currently available on this machine.
2436

2437
  @rtype: list
2438
  @return: list of the exports
2439

2440
  """
2441
  if os.path.isdir(constants.EXPORT_DIR):
2442
    return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2443
  else:
2444
    _Fail("No exports directory")
2445

    
2446

    
2447
def RemoveExport(export):
2448
  """Remove an existing export from the node.
2449

2450
  @type export: str
2451
  @param export: the name of the export to remove
2452
  @rtype: None
2453

2454
  """
2455
  target = utils.PathJoin(constants.EXPORT_DIR, export)
2456

    
2457
  try:
2458
    shutil.rmtree(target)
2459
  except EnvironmentError, err:
2460
    _Fail("Error while removing the export: %s", err, exc=True)
2461

    
2462

    
2463
def BlockdevRename(devlist):
2464
  """Rename a list of block devices.
2465

2466
  @type devlist: list of tuples
2467
  @param devlist: list of tuples of the form  (disk,
2468
      new_logical_id, new_physical_id); disk is an
2469
      L{objects.Disk} object describing the current disk,
2470
      and new logical_id/physical_id is the name we
2471
      rename it to
2472
  @rtype: boolean
2473
  @return: True if all renames succeeded, False otherwise
2474

2475
  """
2476
  msgs = []
2477
  result = True
2478
  for disk, unique_id in devlist:
2479
    dev = _RecursiveFindBD(disk)
2480
    if dev is None:
2481
      msgs.append("Can't find device %s in rename" % str(disk))
2482
      result = False
2483
      continue
2484
    try:
2485
      old_rpath = dev.dev_path
2486
      dev.Rename(unique_id)
2487
      new_rpath = dev.dev_path
2488
      if old_rpath != new_rpath:
2489
        DevCacheManager.RemoveCache(old_rpath)
2490
        # FIXME: we should add the new cache information here, like:
2491
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2492
        # but we don't have the owner here - maybe parse from existing
2493
        # cache? for now, we only lose lvm data when we rename, which
2494
        # is less critical than DRBD or MD
2495
    except errors.BlockDeviceError, err:
2496
      msgs.append("Can't rename device '%s' to '%s': %s" %
2497
                  (dev, unique_id, err))
2498
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2499
      result = False
2500
  if not result:
2501
    _Fail("; ".join(msgs))
2502

    
2503

    
2504
def _TransformFileStorageDir(fs_dir):
2505
  """Checks whether given file_storage_dir is valid.
2506

2507
  Checks wheter the given fs_dir is within the cluster-wide default
2508
  file_storage_dir or the shared_file_storage_dir, which are stored in
2509
  SimpleStore. Only paths under those directories are allowed.
2510

2511
  @type fs_dir: str
2512
  @param fs_dir: the path to check
2513

2514
  @return: the normalized path if valid, None otherwise
2515

2516
  """
2517
  if not constants.ENABLE_FILE_STORAGE:
2518
    _Fail("File storage disabled at configure time")
2519
  cfg = _GetConfig()
2520
  fs_dir = os.path.normpath(fs_dir)
2521
  base_fstore = cfg.GetFileStorageDir()
2522
  base_shared = cfg.GetSharedFileStorageDir()
2523
  if ((os.path.commonprefix([fs_dir, base_fstore]) != base_fstore) and
2524
      (os.path.commonprefix([fs_dir, base_shared]) != base_shared)):
2525
    _Fail("File storage directory '%s' is not under base file"
2526
          " storage directory '%s' or shared storage directory '%s'",
2527
          fs_dir, base_fstore, base_shared)
2528
  return fs_dir
2529

    
2530

    
2531
def CreateFileStorageDir(file_storage_dir):
2532
  """Create file storage directory.
2533

2534
  @type file_storage_dir: str
2535
  @param file_storage_dir: directory to create
2536

2537
  @rtype: tuple
2538
  @return: tuple with first element a boolean indicating wheter dir
2539
      creation was successful or not
2540

2541
  """
2542
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2543
  if os.path.exists(file_storage_dir):
2544
    if not os.path.isdir(file_storage_dir):
2545
      _Fail("Specified storage dir '%s' is not a directory",
2546
            file_storage_dir)
2547
  else:
2548
    try:
2549
      os.makedirs(file_storage_dir, 0750)
2550
    except OSError, err:
2551
      _Fail("Cannot create file storage directory '%s': %s",
2552
            file_storage_dir, err, exc=True)
2553

    
2554

    
2555
def RemoveFileStorageDir(file_storage_dir):
2556
  """Remove file storage directory.
2557

2558
  Remove it only if it's empty. If not log an error and return.
2559

2560
  @type file_storage_dir: str
2561
  @param file_storage_dir: the directory we should cleanup
2562
  @rtype: tuple (success,)
2563
  @return: tuple of one element, C{success}, denoting
2564
      whether the operation was successful
2565

2566
  """
2567
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2568
  if os.path.exists(file_storage_dir):
2569
    if not os.path.isdir(file_storage_dir):
2570
      _Fail("Specified Storage directory '%s' is not a directory",
2571
            file_storage_dir)
2572
    # deletes dir only if empty, otherwise we want to fail the rpc call
2573
    try:
2574
      os.rmdir(file_storage_dir)
2575
    except OSError, err:
2576
      _Fail("Cannot remove file storage directory '%s': %s",
2577
            file_storage_dir, err)
2578

    
2579

    
2580
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2581
  """Rename the file storage directory.
2582

2583
  @type old_file_storage_dir: str
2584
  @param old_file_storage_dir: the current path
2585
  @type new_file_storage_dir: str
2586
  @param new_file_storage_dir: the name we should rename to
2587
  @rtype: tuple (success,)
2588
  @return: tuple of one element, C{success}, denoting
2589
      whether the operation was successful
2590

2591
  """
2592
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2593
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2594
  if not os.path.exists(new_file_storage_dir):
2595
    if os.path.isdir(old_file_storage_dir):
2596
      try:
2597
        os.rename(old_file_storage_dir, new_file_storage_dir)
2598
      except OSError, err:
2599
        _Fail("Cannot rename '%s' to '%s': %s",
2600
              old_file_storage_dir, new_file_storage_dir, err)
2601
    else:
2602
      _Fail("Specified storage dir '%s' is not a directory",
2603
            old_file_storage_dir)
2604
  else:
2605
    if os.path.exists(old_file_storage_dir):
2606
      _Fail("Cannot rename '%s' to '%s': both locations exist",
2607
            old_file_storage_dir, new_file_storage_dir)
2608

    
2609

    
2610
def _EnsureJobQueueFile(file_name):
2611
  """Checks whether the given filename is in the queue directory.
2612

2613
  @type file_name: str
2614
  @param file_name: the file name we should check
2615
  @rtype: None
2616
  @raises RPCFail: if the file is not valid
2617

2618
  """
2619
  queue_dir = os.path.normpath(constants.QUEUE_DIR)
2620
  result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2621

    
2622
  if not result:
2623
    _Fail("Passed job queue file '%s' does not belong to"
2624
          " the queue directory '%s'", file_name, queue_dir)
2625

    
2626

    
2627
def JobQueueUpdate(file_name, content):
2628
  """Updates a file in the queue directory.
2629

2630
  This is just a wrapper over L{utils.io.WriteFile}, with proper
2631
  checking.
2632

2633
  @type file_name: str
2634
  @param file_name: the job file name
2635
  @type content: str
2636
  @param content: the new job contents
2637
  @rtype: boolean
2638
  @return: the success of the operation
2639

2640
  """
2641
  _EnsureJobQueueFile(file_name)
2642
  getents = runtime.GetEnts()
2643

    
2644
  # Write and replace the file atomically
2645
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2646
                  gid=getents.masterd_gid)
2647

    
2648

    
2649
def JobQueueRename(old, new):
2650
  """Renames a job queue file.
2651

2652
  This is just a wrapper over os.rename with proper checking.
2653

2654
  @type old: str
2655
  @param old: the old (actual) file name
2656
  @type new: str
2657
  @param new: the desired file name
2658
  @rtype: tuple
2659
  @return: the success of the operation and payload
2660

2661
  """
2662
  _EnsureJobQueueFile(old)
2663
  _EnsureJobQueueFile(new)
2664

    
2665
  utils.RenameFile(old, new, mkdir=True)
2666

    
2667

    
2668
def BlockdevClose(instance_name, disks):
2669
  """Closes the given block devices.
2670

2671
  This means they will be switched to secondary mode (in case of
2672
  DRBD).
2673

2674
  @param instance_name: if the argument is not empty, the symlinks
2675
      of this instance will be removed
2676
  @type disks: list of L{objects.Disk}
2677
  @param disks: the list of disks to be closed
2678
  @rtype: tuple (success, message)
2679
  @return: a tuple of success and message, where success
2680
      indicates the succes of the operation, and message
2681
      which will contain the error details in case we
2682
      failed
2683

2684
  """
2685
  bdevs = []
2686
  for cf in disks:
2687
    rd = _RecursiveFindBD(cf)
2688
    if rd is None:
2689
      _Fail("Can't find device %s", cf)
2690
    bdevs.append(rd)
2691

    
2692
  msg = []
2693
  for rd in bdevs:
2694
    try:
2695
      rd.Close()
2696
    except errors.BlockDeviceError, err:
2697
      msg.append(str(err))
2698
  if msg:
2699
    _Fail("Can't make devices secondary: %s", ",".join(msg))
2700
  else:
2701
    if instance_name:
2702
      _RemoveBlockDevLinks(instance_name, disks)
2703

    
2704

    
2705
def ValidateHVParams(hvname, hvparams):
2706
  """Validates the given hypervisor parameters.
2707

2708
  @type hvname: string
2709
  @param hvname: the hypervisor name
2710
  @type hvparams: dict
2711
  @param hvparams: the hypervisor parameters to be validated
2712
  @rtype: None
2713

2714
  """
2715
  try:
2716
    hv_type = hypervisor.GetHypervisor(hvname)
2717
    hv_type.ValidateParameters(hvparams)
2718
  except errors.HypervisorError, err:
2719
    _Fail(str(err), log=False)
2720

    
2721

    
2722
def _CheckOSPList(os_obj, parameters):
2723
  """Check whether a list of parameters is supported by the OS.
2724

2725
  @type os_obj: L{objects.OS}
2726
  @param os_obj: OS object to check
2727
  @type parameters: list
2728
  @param parameters: the list of parameters to check
2729

2730
  """
2731
  supported = [v[0] for v in os_obj.supported_parameters]
2732
  delta = frozenset(parameters).difference(supported)
2733
  if delta:
2734
    _Fail("The following parameters are not supported"
2735
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2736

    
2737

    
2738
def ValidateOS(required, osname, checks, osparams):
2739
  """Validate the given OS' parameters.
2740

2741
  @type required: boolean
2742
  @param required: whether absence of the OS should translate into
2743
      failure or not
2744
  @type osname: string
2745
  @param osname: the OS to be validated
2746
  @type checks: list
2747
  @param checks: list of the checks to run (currently only 'parameters')
2748
  @type osparams: dict
2749
  @param osparams: dictionary with OS parameters
2750
  @rtype: boolean
2751
  @return: True if the validation passed, or False if the OS was not
2752
      found and L{required} was false
2753

2754
  """
2755
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2756
    _Fail("Unknown checks required for OS %s: %s", osname,
2757
          set(checks).difference(constants.OS_VALIDATE_CALLS))
2758

    
2759
  name_only = objects.OS.GetName(osname)
2760
  status, tbv = _TryOSFromDisk(name_only, None)
2761

    
2762
  if not status:
2763
    if required:
2764
      _Fail(tbv)
2765
    else:
2766
      return False
2767

    
2768
  if max(tbv.api_versions) < constants.OS_API_V20:
2769
    return True
2770

    
2771
  if constants.OS_VALIDATE_PARAMETERS in checks:
2772
    _CheckOSPList(tbv, osparams.keys())
2773

    
2774
  validate_env = OSCoreEnv(osname, tbv, osparams)
2775
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2776
                        cwd=tbv.path, reset_env=True)
2777
  if result.failed:
2778
    logging.error("os validate command '%s' returned error: %s output: %s",
2779
                  result.cmd, result.fail_reason, result.output)
2780
    _Fail("OS validation script failed (%s), output: %s",
2781
          result.fail_reason, result.output, log=False)
2782

    
2783
  return True
2784

    
2785

    
2786
def DemoteFromMC():
2787
  """Demotes the current node from master candidate role.
2788

2789
  """
2790
  # try to ensure we're not the master by mistake
2791
  master, myself = ssconf.GetMasterAndMyself()
2792
  if master == myself:
2793
    _Fail("ssconf status shows I'm the master node, will not demote")
2794

    
2795
  result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2796
  if not result.failed:
2797
    _Fail("The master daemon is running, will not demote")
2798

    
2799
  try:
2800
    if os.path.isfile(constants.CLUSTER_CONF_FILE):
2801
      utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2802
  except EnvironmentError, err:
2803
    if err.errno != errno.ENOENT:
2804
      _Fail("Error while backing up cluster file: %s", err, exc=True)
2805

    
2806
  utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2807

    
2808

    
2809
def _GetX509Filenames(cryptodir, name):
2810
  """Returns the full paths for the private key and certificate.
2811

2812
  """
2813
  return (utils.PathJoin(cryptodir, name),
2814
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2815
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2816

    
2817

    
2818
def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2819
  """Creates a new X509 certificate for SSL/TLS.
2820

2821
  @type validity: int
2822
  @param validity: Validity in seconds
2823
  @rtype: tuple; (string, string)
2824
  @return: Certificate name and public part
2825

2826
  """
2827
  (key_pem, cert_pem) = \
2828
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2829
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
2830

    
2831
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
2832
                              prefix="x509-%s-" % utils.TimestampForFilename())
2833
  try:
2834
    name = os.path.basename(cert_dir)
2835
    assert len(name) > 5
2836

    
2837
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2838

    
2839
    utils.WriteFile(key_file, mode=0400, data=key_pem)
2840
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2841

    
2842
    # Never return private key as it shouldn't leave the node
2843
    return (name, cert_pem)
2844
  except Exception:
2845
    shutil.rmtree(cert_dir, ignore_errors=True)
2846
    raise
2847

    
2848

    
2849
def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2850
  """Removes a X509 certificate.
2851

2852
  @type name: string
2853
  @param name: Certificate name
2854

2855
  """
2856
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2857

    
2858
  utils.RemoveFile(key_file)
2859
  utils.RemoveFile(cert_file)
2860

    
2861
  try:
2862
    os.rmdir(cert_dir)
2863
  except EnvironmentError, err:
2864
    _Fail("Cannot remove certificate directory '%s': %s",
2865
          cert_dir, err)
2866

    
2867

    
2868
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2869
  """Returns the command for the requested input/output.
2870

2871
  @type instance: L{objects.Instance}
2872
  @param instance: The instance object
2873
  @param mode: Import/export mode
2874
  @param ieio: Input/output type
2875
  @param ieargs: Input/output arguments
2876

2877
  """
2878
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2879

    
2880
  env = None
2881
  prefix = None
2882
  suffix = None
2883
  exp_size = None
2884

    
2885
  if ieio == constants.IEIO_FILE:
2886
    (filename, ) = ieargs
2887

    
2888
    if not utils.IsNormAbsPath(filename):
2889
      _Fail("Path '%s' is not normalized or absolute", filename)
2890

    
2891
    directory = os.path.normpath(os.path.dirname(filename))
2892

    
2893
    if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2894
        constants.EXPORT_DIR):
2895
      _Fail("File '%s' is not under exports directory '%s'",
2896
            filename, constants.EXPORT_DIR)
2897

    
2898
    # Create directory
2899
    utils.Makedirs(directory, mode=0750)
2900

    
2901
    quoted_filename = utils.ShellQuote(filename)
2902

    
2903
    if mode == constants.IEM_IMPORT:
2904
      suffix = "> %s" % quoted_filename
2905
    elif mode == constants.IEM_EXPORT:
2906
      suffix = "< %s" % quoted_filename
2907

    
2908
      # Retrieve file size
2909
      try:
2910
        st = os.stat(filename)
2911
      except EnvironmentError, err:
2912
        logging.error("Can't stat(2) %s: %s", filename, err)
2913
      else:
2914
        exp_size = utils.BytesToMebibyte(st.st_size)
2915

    
2916
  elif ieio == constants.IEIO_RAW_DISK:
2917
    (disk, ) = ieargs
2918

    
2919
    real_disk = _OpenRealBD(disk)
2920

    
2921
    if mode == constants.IEM_IMPORT:
2922
      # we set here a smaller block size as, due to transport buffering, more
2923
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
2924
      # is not already there or we pass a wrong path; we use notrunc to no
2925
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2926
      # much memory; this means that at best, we flush every 64k, which will
2927
      # not be very fast
2928
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2929
                                    " bs=%s oflag=dsync"),
2930
                                    real_disk.dev_path,
2931
                                    str(64 * 1024))
2932

    
2933
    elif mode == constants.IEM_EXPORT:
2934
      # the block size on the read dd is 1MiB to match our units
2935
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2936
                                   real_disk.dev_path,
2937
                                   str(1024 * 1024), # 1 MB
2938
                                   str(disk.size))
2939
      exp_size = disk.size
2940

    
2941
  elif ieio == constants.IEIO_SCRIPT:
2942
    (disk, disk_index, ) = ieargs
2943

    
2944
    assert isinstance(disk_index, (int, long))
2945

    
2946
    real_disk = _OpenRealBD(disk)
2947

    
2948
    inst_os = OSFromDisk(instance.os)
2949
    env = OSEnvironment(instance, inst_os)
2950

    
2951
    if mode == constants.IEM_IMPORT:
2952
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2953
      env["IMPORT_INDEX"] = str(disk_index)
2954
      script = inst_os.import_script
2955

    
2956
    elif mode == constants.IEM_EXPORT:
2957
      env["EXPORT_DEVICE"] = real_disk.dev_path
2958
      env["EXPORT_INDEX"] = str(disk_index)
2959
      script = inst_os.export_script
2960

    
2961
    # TODO: Pass special environment only to script
2962
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2963

    
2964
    if mode == constants.IEM_IMPORT:
2965
      suffix = "| %s" % script_cmd
2966

    
2967
    elif mode == constants.IEM_EXPORT:
2968
      prefix = "%s |" % script_cmd
2969

    
2970
    # Let script predict size
2971
    exp_size = constants.IE_CUSTOM_SIZE
2972

    
2973
  else:
2974
    _Fail("Invalid %s I/O mode %r", mode, ieio)
2975

    
2976
  return (env, prefix, suffix, exp_size)
2977

    
2978

    
2979
def _CreateImportExportStatusDir(prefix):
2980
  """Creates status directory for import/export.
2981

2982
  """
2983
  return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2984
                          prefix=("%s-%s-" %
2985
                                  (prefix, utils.TimestampForFilename())))
2986

    
2987

    
2988
def StartImportExportDaemon(mode, opts, host, port, instance, component,
2989
                            ieio, ieioargs):
2990
  """Starts an import or export daemon.
2991

2992
  @param mode: Import/output mode
2993
  @type opts: L{objects.ImportExportOptions}
2994
  @param opts: Daemon options
2995
  @type host: string
2996
  @param host: Remote host for export (None for import)
2997
  @type port: int
2998
  @param port: Remote port for export (None for import)
2999
  @type instance: L{objects.Instance}
3000
  @param instance: Instance object
3001
  @type component: string
3002
  @param component: which part of the instance is transferred now,
3003
      e.g. 'disk/0'
3004
  @param ieio: Input/output type
3005
  @param ieioargs: Input/output arguments
3006

3007
  """
3008
  if mode == constants.IEM_IMPORT:
3009
    prefix = "import"
3010

    
3011
    if not (host is None and port is None):
3012
      _Fail("Can not specify host or port on import")
3013

    
3014
  elif mode == constants.IEM_EXPORT:
3015
    prefix = "export"
3016

    
3017
    if host is None or port is None:
3018
      _Fail("Host and port must be specified for an export")
3019

    
3020
  else:
3021
    _Fail("Invalid mode %r", mode)
3022

    
3023
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3024
    _Fail("Cluster certificate can only be used for both key and CA")
3025

    
3026
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3027
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3028

    
3029
  if opts.key_name is None:
3030
    # Use server.pem
3031
    key_path = constants.NODED_CERT_FILE
3032
    cert_path = constants.NODED_CERT_FILE
3033
    assert opts.ca_pem is None
3034
  else:
3035
    (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
3036
                                                 opts.key_name)
3037
    assert opts.ca_pem is not None
3038

    
3039
  for i in [key_path, cert_path]:
3040
    if not os.path.exists(i):
3041
      _Fail("File '%s' does not exist" % i)
3042

    
3043
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3044
  try:
3045
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3046
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3047
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3048

    
3049
    if opts.ca_pem is None:
3050
      # Use server.pem
3051
      ca = utils.ReadFile(constants.NODED_CERT_FILE)
3052
    else:
3053
      ca = opts.ca_pem
3054

    
3055
    # Write CA file
3056
    utils.WriteFile(ca_file, data=ca, mode=0400)
3057

    
3058
    cmd = [
3059
      constants.IMPORT_EXPORT_DAEMON,
3060
      status_file, mode,
3061
      "--key=%s" % key_path,
3062
      "--cert=%s" % cert_path,
3063
      "--ca=%s" % ca_file,
3064
      ]
3065

    
3066
    if host:
3067
      cmd.append("--host=%s" % host)
3068

    
3069
    if port:
3070
      cmd.append("--port=%s" % port)
3071

    
3072
    if opts.ipv6:
3073
      cmd.append("--ipv6")
3074
    else:
3075
      cmd.append("--ipv4")
3076

    
3077
    if opts.compress:
3078
      cmd.append("--compress=%s" % opts.compress)
3079

    
3080
    if opts.magic:
3081
      cmd.append("--magic=%s" % opts.magic)
3082

    
3083
    if exp_size is not None:
3084
      cmd.append("--expected-size=%s" % exp_size)
3085

    
3086
    if cmd_prefix:
3087
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3088

    
3089
    if cmd_suffix:
3090
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3091

    
3092
    if mode == constants.IEM_EXPORT:
3093
      # Retry connection a few times when connecting to remote peer
3094
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3095
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3096
    elif opts.connect_timeout is not None:
3097
      assert mode == constants.IEM_IMPORT
3098
      # Overall timeout for establishing connection while listening
3099
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3100

    
3101
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3102

    
3103
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3104
    # support for receiving a file descriptor for output
3105
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3106
                      output=logfile)
3107

    
3108
    # The import/export name is simply the status directory name
3109
    return os.path.basename(status_dir)
3110

    
3111
  except Exception:
3112
    shutil.rmtree(status_dir, ignore_errors=True)
3113
    raise
3114

    
3115

    
3116
def GetImportExportStatus(names):
3117
  """Returns import/export daemon status.
3118

3119
  @type names: sequence
3120
  @param names: List of names
3121
  @rtype: List of dicts
3122
  @return: Returns a list of the state of each named import/export or None if a
3123
           status couldn't be read
3124

3125
  """
3126
  result = []
3127

    
3128
  for name in names:
3129
    status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3130
                                 _IES_STATUS_FILE)
3131

    
3132
    try:
3133
      data = utils.ReadFile(status_file)
3134
    except EnvironmentError, err:
3135
      if err.errno != errno.ENOENT:
3136
        raise
3137
      data = None
3138

    
3139
    if not data:
3140
      result.append(None)
3141
      continue
3142

    
3143
    result.append(serializer.LoadJson(data))
3144

    
3145
  return result
3146

    
3147

    
3148
def AbortImportExport(name):
3149
  """Sends SIGTERM to a running import/export daemon.
3150

3151
  """
3152
  logging.info("Abort import/export %s", name)
3153

    
3154
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3155
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3156

    
3157
  if pid:
3158
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3159
                 name, pid)
3160
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3161

    
3162

    
3163
def CleanupImportExport(name):
3164
  """Cleanup after an import or export.
3165

3166
  If the import/export daemon is still running it's killed. Afterwards the
3167
  whole status directory is removed.
3168

3169
  """
3170
  logging.info("Finalizing import/export %s", name)
3171

    
3172
  status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3173

    
3174
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3175

    
3176
  if pid:
3177
    logging.info("Import/export %s is still running with PID %s",
3178
                 name, pid)
3179
    utils.KillProcess(pid, waitpid=False)
3180

    
3181
  shutil.rmtree(status_dir, ignore_errors=True)
3182

    
3183

    
3184
def _FindDisks(nodes_ip, disks):
3185
  """Sets the physical ID on disks and returns the block devices.
3186

3187
  """
3188
  # set the correct physical ID
3189
  my_name = netutils.Hostname.GetSysName()
3190
  for cf in disks:
3191
    cf.SetPhysicalID(my_name, nodes_ip)
3192

    
3193
  bdevs = []
3194

    
3195
  for cf in disks:
3196
    rd = _RecursiveFindBD(cf)
3197
    if rd is None:
3198
      _Fail("Can't find device %s", cf)
3199
    bdevs.append(rd)
3200
  return bdevs
3201

    
3202

    
3203
def DrbdDisconnectNet(nodes_ip, disks):
3204
  """Disconnects the network on a list of drbd devices.
3205

3206
  """
3207
  bdevs = _FindDisks(nodes_ip, disks)
3208

    
3209
  # disconnect disks
3210
  for rd in bdevs:
3211
    try:
3212
      rd.DisconnectNet()
3213
    except errors.BlockDeviceError, err:
3214
      _Fail("Can't change network configuration to standalone mode: %s",
3215
            err, exc=True)
3216

    
3217

    
3218
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3219
  """Attaches the network on a list of drbd devices.
3220

3221
  """
3222
  bdevs = _FindDisks(nodes_ip, disks)
3223

    
3224
  if multimaster:
3225
    for idx, rd in enumerate(bdevs):
3226
      try:
3227
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3228
      except EnvironmentError, err:
3229
        _Fail("Can't create symlink: %s", err)
3230
  # reconnect disks, switch to new master configuration and if
3231
  # needed primary mode
3232
  for rd in bdevs:
3233
    try:
3234
      rd.AttachNet(multimaster)
3235
    except errors.BlockDeviceError, err:
3236
      _Fail("Can't change network configuration: %s", err)
3237

    
3238
  # wait until the disks are connected; we need to retry the re-attach
3239
  # if the device becomes standalone, as this might happen if the one
3240
  # node disconnects and reconnects in a different mode before the
3241
  # other node reconnects; in this case, one or both of the nodes will
3242
  # decide it has wrong configuration and switch to standalone
3243

    
3244
  def _Attach():
3245
    all_connected = True
3246

    
3247
    for rd in bdevs:
3248
      stats = rd.GetProcStatus()
3249

    
3250
      all_connected = (all_connected and
3251
                       (stats.is_connected or stats.is_in_resync))
3252

    
3253
      if stats.is_standalone:
3254
        # peer had different config info and this node became
3255
        # standalone, even though this should not happen with the
3256
        # new staged way of changing disk configs
3257
        try:
3258
          rd.AttachNet(multimaster)
3259
        except errors.BlockDeviceError, err:
3260
          _Fail("Can't change network configuration: %s", err)
3261

    
3262
    if not all_connected:
3263
      raise utils.RetryAgain()
3264

    
3265
  try:
3266
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3267
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3268
  except utils.RetryTimeout:
3269
    _Fail("Timeout in disk reconnecting")
3270

    
3271
  if multimaster:
3272
    # change to primary mode
3273
    for rd in bdevs:
3274
      try:
3275
        rd.Open()
3276
      except errors.BlockDeviceError, err:
3277
        _Fail("Can't change to primary mode: %s", err)
3278

    
3279

    
3280
def DrbdWaitSync(nodes_ip, disks):
3281
  """Wait until DRBDs have synchronized.
3282

3283
  """
3284
  def _helper(rd):
3285
    stats = rd.GetProcStatus()
3286
    if not (stats.is_connected or stats.is_in_resync):
3287
      raise utils.RetryAgain()
3288
    return stats
3289

    
3290
  bdevs = _FindDisks(nodes_ip, disks)
3291

    
3292
  min_resync = 100
3293
  alldone = True
3294
  for rd in bdevs:
3295
    try:
3296
      # poll each second for 15 seconds
3297
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3298
    except utils.RetryTimeout:
3299
      stats = rd.GetProcStatus()
3300
      # last check
3301
      if not (stats.is_connected or stats.is_in_resync):
3302
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3303
    alldone = alldone and (not stats.is_in_resync)
3304
    if stats.sync_percent is not None:
3305
      min_resync = min(min_resync, stats.sync_percent)
3306

    
3307
  return (alldone, min_resync)
3308

    
3309

    
3310
def GetDrbdUsermodeHelper():
3311
  """Returns DRBD usermode helper currently configured.
3312

3313
  """
3314
  try:
3315
    return bdev.BaseDRBD.GetUsermodeHelper()
3316
  except errors.BlockDeviceError, err:
3317
    _Fail(str(err))
3318

    
3319

    
3320
def PowercycleNode(hypervisor_type):
3321
  """Hard-powercycle the node.
3322

3323
  Because we need to return first, and schedule the powercycle in the
3324
  background, we won't be able to report failures nicely.
3325

3326
  """
3327
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3328
  try:
3329
    pid = os.fork()
3330
  except OSError:
3331
    # if we can't fork, we'll pretend that we're in the child process
3332
    pid = 0
3333
  if pid > 0:
3334
    return "Reboot scheduled in 5 seconds"
3335
  # ensure the child is running on ram
3336
  try:
3337
    utils.Mlockall()
3338
  except Exception: # pylint: disable=W0703
3339
    pass
3340
  time.sleep(5)
3341
  hyper.PowercycleNode()
3342

    
3343

    
3344
class HooksRunner(object):
3345
  """Hook runner.
3346

3347
  This class is instantiated on the node side (ganeti-noded) and not
3348
  on the master side.
3349

3350
  """
3351
  def __init__(self, hooks_base_dir=None):
3352
    """Constructor for hooks runner.
3353

3354
    @type hooks_base_dir: str or None
3355
    @param hooks_base_dir: if not None, this overrides the
3356
        L{constants.HOOKS_BASE_DIR} (useful for unittests)
3357

3358
    """
3359
    if hooks_base_dir is None:
3360
      hooks_base_dir = constants.HOOKS_BASE_DIR
3361
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
3362
    # constant
3363
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
3364

    
3365
  def RunHooks(self, hpath, phase, env):
3366
    """Run the scripts in the hooks directory.
3367

3368
    @type hpath: str
3369
    @param hpath: the path to the hooks directory which
3370
        holds the scripts
3371
    @type phase: str
3372
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
3373
        L{constants.HOOKS_PHASE_POST}
3374
    @type env: dict
3375
    @param env: dictionary with the environment for the hook
3376
    @rtype: list
3377
    @return: list of 3-element tuples:
3378
      - script path
3379
      - script result, either L{constants.HKR_SUCCESS} or
3380
        L{constants.HKR_FAIL}
3381
      - output of the script
3382

3383
    @raise errors.ProgrammerError: for invalid input
3384
        parameters
3385

3386
    """
3387
    if phase == constants.HOOKS_PHASE_PRE:
3388
      suffix = "pre"
3389
    elif phase == constants.HOOKS_PHASE_POST:
3390
      suffix = "post"
3391
    else:
3392
      _Fail("Unknown hooks phase '%s'", phase)
3393

    
3394
    subdir = "%s-%s.d" % (hpath, suffix)
3395
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3396

    
3397
    results = []
3398

    
3399
    if not os.path.isdir(dir_name):
3400
      # for non-existing/non-dirs, we simply exit instead of logging a
3401
      # warning at every operation
3402
      return results
3403

    
3404
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3405

    
3406
    for (relname, relstatus, runresult)  in runparts_results:
3407
      if relstatus == constants.RUNPARTS_SKIP:
3408
        rrval = constants.HKR_SKIP
3409
        output = ""
3410
      elif relstatus == constants.RUNPARTS_ERR:
3411
        rrval = constants.HKR_FAIL
3412
        output = "Hook script execution error: %s" % runresult
3413
      elif relstatus == constants.RUNPARTS_RUN:
3414
        if runresult.failed:
3415
          rrval = constants.HKR_FAIL
3416
        else:
3417
          rrval = constants.HKR_SUCCESS
3418
        output = utils.SafeEncode(runresult.output.strip())
3419
      results.append(("%s/%s" % (subdir, relname), rrval, output))
3420

    
3421
    return results
3422

    
3423

    
3424
class IAllocatorRunner(object):
3425
  """IAllocator runner.
3426

3427
  This class is instantiated on the node side (ganeti-noded) and not on
3428
  the master side.
3429

3430
  """
3431
  @staticmethod
3432
  def Run(name, idata):
3433
    """Run an iallocator script.
3434

3435
    @type name: str
3436
    @param name: the iallocator script name
3437
    @type idata: str
3438
    @param idata: the allocator input data
3439

3440
    @rtype: tuple
3441
    @return: two element tuple of:
3442
       - status
3443
       - either error message or stdout of allocator (for success)
3444

3445
    """
3446
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3447
                                  os.path.isfile)
3448
    if alloc_script is None:
3449
      _Fail("iallocator module '%s' not found in the search path", name)
3450

    
3451
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3452
    try:
3453
      os.write(fd, idata)
3454
      os.close(fd)
3455
      result = utils.RunCmd([alloc_script, fin_name])
3456
      if result.failed:
3457
        _Fail("iallocator module '%s' failed: %s, output '%s'",
3458
              name, result.fail_reason, result.output)
3459
    finally:
3460
      os.unlink(fin_name)
3461

    
3462
    return result.stdout
3463

    
3464

    
3465
class DevCacheManager(object):
3466
  """Simple class for managing a cache of block device information.
3467

3468
  """
3469
  _DEV_PREFIX = "/dev/"
3470
  _ROOT_DIR = constants.BDEV_CACHE_DIR
3471

    
3472
  @classmethod
3473
  def _ConvertPath(cls, dev_path):
3474
    """Converts a /dev/name path to the cache file name.
3475

3476
    This replaces slashes with underscores and strips the /dev
3477
    prefix. It then returns the full path to the cache file.
3478

3479
    @type dev_path: str
3480
    @param dev_path: the C{/dev/} path name
3481
    @rtype: str
3482
    @return: the converted path name
3483

3484
    """
3485
    if dev_path.startswith(cls._DEV_PREFIX):
3486
      dev_path = dev_path[len(cls._DEV_PREFIX):]
3487
    dev_path = dev_path.replace("/", "_")
3488
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3489
    return fpath
3490

    
3491
  @classmethod
3492
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3493
    """Updates the cache information for a given device.
3494

3495
    @type dev_path: str
3496
    @param dev_path: the pathname of the device
3497
    @type owner: str
3498
    @param owner: the owner (instance name) of the device
3499
    @type on_primary: bool
3500
    @param on_primary: whether this is the primary
3501
        node nor not
3502
    @type iv_name: str
3503
    @param iv_name: the instance-visible name of the
3504
        device, as in objects.Disk.iv_name
3505

3506
    @rtype: None
3507

3508
    """
3509
    if dev_path is None:
3510
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
3511
      return
3512
    fpath = cls._ConvertPath(dev_path)
3513
    if on_primary:
3514
      state = "primary"
3515
    else:
3516
      state = "secondary"
3517
    if iv_name is None:
3518
      iv_name = "not_visible"
3519
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3520
    try:
3521
      utils.WriteFile(fpath, data=fdata)
3522
    except EnvironmentError, err:
3523
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3524

    
3525
  @classmethod
3526
  def RemoveCache(cls, dev_path):
3527
    """Remove data for a dev_path.
3528

3529
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
3530
    path name and logging.
3531

3532
    @type dev_path: str
3533
    @param dev_path: the pathname of the device
3534

3535
    @rtype: None
3536

3537
    """
3538
    if dev_path is None:
3539
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
3540
      return
3541
    fpath = cls._ConvertPath(dev_path)
3542
    try:
3543
      utils.RemoveFile(fpath)
3544
    except EnvironmentError, err:
3545
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)