Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ ee844e20

History | View | Annotate | Download (87.2 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007 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

27
"""
28

    
29
# pylint: disable-msg=E1103
30

    
31
# E1103: %s %r has no %r member (but some types could not be
32
# inferred), because the _TryOSFromDisk returns either (True, os_obj)
33
# or (False, "string") which confuses pylint
34

    
35

    
36
import os
37
import os.path
38
import shutil
39
import time
40
import stat
41
import errno
42
import re
43
import subprocess
44
import random
45
import logging
46
import tempfile
47
import zlib
48
import base64
49

    
50
from ganeti import errors
51
from ganeti import utils
52
from ganeti import ssh
53
from ganeti import hypervisor
54
from ganeti import constants
55
from ganeti import bdev
56
from ganeti import objects
57
from ganeti import ssconf
58

    
59

    
60
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
61

    
62

    
63
class RPCFail(Exception):
64
  """Class denoting RPC failure.
65

66
  Its argument is the error message.
67

68
  """
69

    
70

    
71
def _Fail(msg, *args, **kwargs):
72
  """Log an error and the raise an RPCFail exception.
73

74
  This exception is then handled specially in the ganeti daemon and
75
  turned into a 'failed' return type. As such, this function is a
76
  useful shortcut for logging the error and returning it to the master
77
  daemon.
78

79
  @type msg: string
80
  @param msg: the text of the exception
81
  @raise RPCFail
82

83
  """
84
  if args:
85
    msg = msg % args
86
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
87
    if "exc" in kwargs and kwargs["exc"]:
88
      logging.exception(msg)
89
    else:
90
      logging.error(msg)
91
  raise RPCFail(msg)
92

    
93

    
94
def _GetConfig():
95
  """Simple wrapper to return a SimpleStore.
96

97
  @rtype: L{ssconf.SimpleStore}
98
  @return: a SimpleStore instance
99

100
  """
101
  return ssconf.SimpleStore()
102

    
103

    
104
def _GetSshRunner(cluster_name):
105
  """Simple wrapper to return an SshRunner.
106

107
  @type cluster_name: str
108
  @param cluster_name: the cluster name, which is needed
109
      by the SshRunner constructor
110
  @rtype: L{ssh.SshRunner}
111
  @return: an SshRunner instance
112

113
  """
114
  return ssh.SshRunner(cluster_name)
115

    
116

    
117
def _Decompress(data):
118
  """Unpacks data compressed by the RPC client.
119

120
  @type data: list or tuple
121
  @param data: Data sent by RPC client
122
  @rtype: str
123
  @return: Decompressed data
124

125
  """
126
  assert isinstance(data, (list, tuple))
127
  assert len(data) == 2
128
  (encoding, content) = data
129
  if encoding == constants.RPC_ENCODING_NONE:
130
    return content
131
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
132
    return zlib.decompress(base64.b64decode(content))
133
  else:
134
    raise AssertionError("Unknown data encoding")
135

    
136

    
137
def _CleanDirectory(path, exclude=None):
138
  """Removes all regular files in a directory.
139

140
  @type path: str
141
  @param path: the directory to clean
142
  @type exclude: list
143
  @param exclude: list of files to be excluded, defaults
144
      to the empty list
145

146
  """
147
  if not os.path.isdir(path):
148
    return
149
  if exclude is None:
150
    exclude = []
151
  else:
152
    # Normalize excluded paths
153
    exclude = [os.path.normpath(i) for i in exclude]
154

    
155
  for rel_name in utils.ListVisibleFiles(path):
156
    full_name = os.path.normpath(os.path.join(path, rel_name))
157
    if full_name in exclude:
158
      continue
159
    if os.path.isfile(full_name) and not os.path.islink(full_name):
160
      utils.RemoveFile(full_name)
161

    
162

    
163
def _BuildUploadFileList():
164
  """Build the list of allowed upload files.
165

166
  This is abstracted so that it's built only once at module import time.
167

168
  """
169
  allowed_files = set([
170
    constants.CLUSTER_CONF_FILE,
171
    constants.ETC_HOSTS,
172
    constants.SSH_KNOWN_HOSTS_FILE,
173
    constants.VNC_PASSWORD_FILE,
174
    constants.RAPI_CERT_FILE,
175
    constants.RAPI_USERS_FILE,
176
    constants.HMAC_CLUSTER_KEY,
177
    ])
178

    
179
  for hv_name in constants.HYPER_TYPES:
180
    hv_class = hypervisor.GetHypervisorClass(hv_name)
181
    allowed_files.update(hv_class.GetAncillaryFiles())
182

    
183
  return frozenset(allowed_files)
184

    
185

    
186
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
187

    
188

    
189
def JobQueuePurge():
190
  """Removes job queue files and archived jobs.
191

192
  @rtype: tuple
193
  @return: True, None
194

195
  """
196
  _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
197
  _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
198

    
199

    
200
def GetMasterInfo():
201
  """Returns master information.
202

203
  This is an utility function to compute master information, either
204
  for consumption here or from the node daemon.
205

206
  @rtype: tuple
207
  @return: master_netdev, master_ip, master_name
208
  @raise RPCFail: in case of errors
209

210
  """
211
  try:
212
    cfg = _GetConfig()
213
    master_netdev = cfg.GetMasterNetdev()
214
    master_ip = cfg.GetMasterIP()
215
    master_node = cfg.GetMasterNode()
216
  except errors.ConfigurationError, err:
217
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
218
  return (master_netdev, master_ip, master_node)
219

    
220

    
221
def StartMaster(start_daemons, no_voting):
222
  """Activate local node as master node.
223

224
  The function will always try activate the IP address of the master
225
  (unless someone else has it). It will also start the master daemons,
226
  based on the start_daemons parameter.
227

228
  @type start_daemons: boolean
229
  @param start_daemons: whether to also start the master
230
      daemons (ganeti-masterd and ganeti-rapi)
231
  @type no_voting: boolean
232
  @param no_voting: whether to start ganeti-masterd without a node vote
233
      (if start_daemons is True), but still non-interactively
234
  @rtype: None
235

236
  """
237
  # GetMasterInfo will raise an exception if not able to return data
238
  master_netdev, master_ip, _ = GetMasterInfo()
239

    
240
  err_msgs = []
241
  if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
242
    if utils.OwnIpAddress(master_ip):
243
      # we already have the ip:
244
      logging.debug("Master IP already configured, doing nothing")
245
    else:
246
      msg = "Someone else has the master ip, not activating"
247
      logging.error(msg)
248
      err_msgs.append(msg)
249
  else:
250
    result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
251
                           "dev", master_netdev, "label",
252
                           "%s:0" % master_netdev])
253
    if result.failed:
254
      msg = "Can't activate master IP: %s" % result.output
255
      logging.error(msg)
256
      err_msgs.append(msg)
257

    
258
    result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
259
                           "-s", master_ip, master_ip])
260
    # we'll ignore the exit code of arping
261

    
262
  # and now start the master and rapi daemons
263
  if start_daemons:
264
    if no_voting:
265
      masterd_args = "--no-voting --yes-do-it"
266
    else:
267
      masterd_args = ""
268

    
269
    env = {
270
      "EXTRA_MASTERD_ARGS": masterd_args,
271
      }
272

    
273
    result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
274
    if result.failed:
275
      msg = "Can't start Ganeti master: %s" % result.output
276
      logging.error(msg)
277
      err_msgs.append(msg)
278

    
279
  if err_msgs:
280
    _Fail("; ".join(err_msgs))
281

    
282

    
283
def StopMaster(stop_daemons):
284
  """Deactivate this node as master.
285

286
  The function will always try to deactivate the IP address of the
287
  master. It will also stop the master daemons depending on the
288
  stop_daemons parameter.
289

290
  @type stop_daemons: boolean
291
  @param stop_daemons: whether to also stop the master daemons
292
      (ganeti-masterd and ganeti-rapi)
293
  @rtype: None
294

295
  """
296
  # TODO: log and report back to the caller the error failures; we
297
  # need to decide in which case we fail the RPC for this
298

    
299
  # GetMasterInfo will raise an exception if not able to return data
300
  master_netdev, master_ip, _ = GetMasterInfo()
301

    
302
  result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
303
                         "dev", master_netdev])
304
  if result.failed:
305
    logging.error("Can't remove the master IP, error: %s", result.output)
306
    # but otherwise ignore the failure
307

    
308
  if stop_daemons:
309
    result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
310
    if result.failed:
311
      logging.error("Could not stop Ganeti master, command %s had exitcode %s"
312
                    " and error %s",
313
                    result.cmd, result.exit_code, result.output)
314

    
315

    
316
def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
317
  """Joins this node to the cluster.
318

319
  This does the following:
320
      - updates the hostkeys of the machine (rsa and dsa)
321
      - adds the ssh private key to the user
322
      - adds the ssh public key to the users' authorized_keys file
323

324
  @type dsa: str
325
  @param dsa: the DSA private key to write
326
  @type dsapub: str
327
  @param dsapub: the DSA public key to write
328
  @type rsa: str
329
  @param rsa: the RSA private key to write
330
  @type rsapub: str
331
  @param rsapub: the RSA public key to write
332
  @type sshkey: str
333
  @param sshkey: the SSH private key to write
334
  @type sshpub: str
335
  @param sshpub: the SSH public key to write
336
  @rtype: boolean
337
  @return: the success of the operation
338

339
  """
340
  sshd_keys =  [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
341
                (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
342
                (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
343
                (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
344
  for name, content, mode in sshd_keys:
345
    utils.WriteFile(name, data=content, mode=mode)
346

    
347
  try:
348
    priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
349
                                                    mkdir=True)
350
  except errors.OpExecError, err:
351
    _Fail("Error while processing user ssh files: %s", err, exc=True)
352

    
353
  for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
354
    utils.WriteFile(name, data=content, mode=0600)
355

    
356
  utils.AddAuthorizedKey(auth_keys, sshpub)
357

    
358
  result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"])
359
  if result.failed:
360
    _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)",
361
          result.cmd, result.exit_code, result.output)
362

    
363

    
364
def LeaveCluster(modify_ssh_setup):
365
  """Cleans up and remove the current node.
366

367
  This function cleans up and prepares the current node to be removed
368
  from the cluster.
369

370
  If processing is successful, then it raises an
371
  L{errors.QuitGanetiException} which is used as a special case to
372
  shutdown the node daemon.
373

374
  @param modify_ssh_setup: boolean
375

376
  """
377
  _CleanDirectory(constants.DATA_DIR)
378
  JobQueuePurge()
379

    
380
  if modify_ssh_setup:
381
    try:
382
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
383

    
384
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
385

    
386
      utils.RemoveFile(priv_key)
387
      utils.RemoveFile(pub_key)
388
    except errors.OpExecError:
389
      logging.exception("Error while processing ssh files")
390

    
391
  try:
392
    utils.RemoveFile(constants.HMAC_CLUSTER_KEY)
393
    utils.RemoveFile(constants.RAPI_CERT_FILE)
394
    utils.RemoveFile(constants.SSL_CERT_FILE)
395
  except: # pylint: disable-msg=W0702
396
    logging.exception("Error while removing cluster secrets")
397

    
398
  result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
399
  if result.failed:
400
    logging.error("Command %s failed with exitcode %s and error %s",
401
                  result.cmd, result.exit_code, result.output)
402

    
403
  # Raise a custom exception (handled in ganeti-noded)
404
  raise errors.QuitGanetiException(True, 'Shutdown scheduled')
405

    
406

    
407
def GetNodeInfo(vgname, hypervisor_type):
408
  """Gives back a hash with different information about the node.
409

410
  @type vgname: C{string}
411
  @param vgname: the name of the volume group to ask for disk space information
412
  @type hypervisor_type: C{str}
413
  @param hypervisor_type: the name of the hypervisor to ask for
414
      memory information
415
  @rtype: C{dict}
416
  @return: dictionary with the following keys:
417
      - vg_size is the size of the configured volume group in MiB
418
      - vg_free is the free size of the volume group in MiB
419
      - memory_dom0 is the memory allocated for domain0 in MiB
420
      - memory_free is the currently available (free) ram in MiB
421
      - memory_total is the total number of ram in MiB
422

423
  """
424
  outputarray = {}
425
  vginfo = _GetVGInfo(vgname)
426
  outputarray['vg_size'] = vginfo['vg_size']
427
  outputarray['vg_free'] = vginfo['vg_free']
428

    
429
  hyper = hypervisor.GetHypervisor(hypervisor_type)
430
  hyp_info = hyper.GetNodeInfo()
431
  if hyp_info is not None:
432
    outputarray.update(hyp_info)
433

    
434
  outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
435

    
436
  return outputarray
437

    
438

    
439
def VerifyNode(what, cluster_name):
440
  """Verify the status of the local node.
441

442
  Based on the input L{what} parameter, various checks are done on the
443
  local node.
444

445
  If the I{filelist} key is present, this list of
446
  files is checksummed and the file/checksum pairs are returned.
447

448
  If the I{nodelist} key is present, we check that we have
449
  connectivity via ssh with the target nodes (and check the hostname
450
  report).
451

452
  If the I{node-net-test} key is present, we check that we have
453
  connectivity to the given nodes via both primary IP and, if
454
  applicable, secondary IPs.
455

456
  @type what: C{dict}
457
  @param what: a dictionary of things to check:
458
      - filelist: list of files for which to compute checksums
459
      - nodelist: list of nodes we should check ssh communication with
460
      - node-net-test: list of nodes we should check node daemon port
461
        connectivity with
462
      - hypervisor: list with hypervisors to run the verify for
463
  @rtype: dict
464
  @return: a dictionary with the same keys as the input dict, and
465
      values representing the result of the checks
466

467
  """
468
  result = {}
469

    
470
  if constants.NV_HYPERVISOR in what:
471
    result[constants.NV_HYPERVISOR] = tmp = {}
472
    for hv_name in what[constants.NV_HYPERVISOR]:
473
      tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
474

    
475
  if constants.NV_FILELIST in what:
476
    result[constants.NV_FILELIST] = utils.FingerprintFiles(
477
      what[constants.NV_FILELIST])
478

    
479
  if constants.NV_NODELIST in what:
480
    result[constants.NV_NODELIST] = tmp = {}
481
    random.shuffle(what[constants.NV_NODELIST])
482
    for node in what[constants.NV_NODELIST]:
483
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
484
      if not success:
485
        tmp[node] = message
486

    
487
  if constants.NV_NODENETTEST in what:
488
    result[constants.NV_NODENETTEST] = tmp = {}
489
    my_name = utils.HostInfo().name
490
    my_pip = my_sip = None
491
    for name, pip, sip in what[constants.NV_NODENETTEST]:
492
      if name == my_name:
493
        my_pip = pip
494
        my_sip = sip
495
        break
496
    if not my_pip:
497
      tmp[my_name] = ("Can't find my own primary/secondary IP"
498
                      " in the node list")
499
    else:
500
      port = utils.GetDaemonPort(constants.NODED)
501
      for name, pip, sip in what[constants.NV_NODENETTEST]:
502
        fail = []
503
        if not utils.TcpPing(pip, port, source=my_pip):
504
          fail.append("primary")
505
        if sip != pip:
506
          if not utils.TcpPing(sip, port, source=my_sip):
507
            fail.append("secondary")
508
        if fail:
509
          tmp[name] = ("failure using the %s interface(s)" %
510
                       " and ".join(fail))
511

    
512
  if constants.NV_LVLIST in what:
513
    result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
514

    
515
  if constants.NV_INSTANCELIST in what:
516
    result[constants.NV_INSTANCELIST] = GetInstanceList(
517
      what[constants.NV_INSTANCELIST])
518

    
519
  if constants.NV_VGLIST in what:
520
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
521

    
522
  if constants.NV_PVLIST in what:
523
    result[constants.NV_PVLIST] = \
524
      bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
525
                                   filter_allocatable=False)
526

    
527
  if constants.NV_VERSION in what:
528
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
529
                                    constants.RELEASE_VERSION)
530

    
531
  if constants.NV_HVINFO in what:
532
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
533
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
534

    
535
  if constants.NV_DRBDLIST in what:
536
    try:
537
      used_minors = bdev.DRBD8.GetUsedDevs().keys()
538
    except errors.BlockDeviceError, err:
539
      logging.warning("Can't get used minors list", exc_info=True)
540
      used_minors = str(err)
541
    result[constants.NV_DRBDLIST] = used_minors
542

    
543
  if constants.NV_NODESETUP in what:
544
    result[constants.NV_NODESETUP] = tmpr = []
545
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
546
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
547
                  " under /sys, missing required directories /sys/block"
548
                  " and /sys/class/net")
549
    if (not os.path.isdir("/proc/sys") or
550
        not os.path.isfile("/proc/sysrq-trigger")):
551
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
552
                  " under /proc, missing required directory /proc/sys and"
553
                  " the file /proc/sysrq-trigger")
554

    
555
  if constants.NV_TIME in what:
556
    result[constants.NV_TIME] = utils.SplitTime(time.time())
557

    
558
  return result
559

    
560

    
561
def GetVolumeList(vg_name):
562
  """Compute list of logical volumes and their size.
563

564
  @type vg_name: str
565
  @param vg_name: the volume group whose LVs we should list
566
  @rtype: dict
567
  @return:
568
      dictionary of all partions (key) with value being a tuple of
569
      their size (in MiB), inactive and online status::
570

571
        {'test1': ('20.06', True, True)}
572

573
      in case of errors, a string is returned with the error
574
      details.
575

576
  """
577
  lvs = {}
578
  sep = '|'
579
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
580
                         "--separator=%s" % sep,
581
                         "-olv_name,lv_size,lv_attr", vg_name])
582
  if result.failed:
583
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
584

    
585
  valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
586
  for line in result.stdout.splitlines():
587
    line = line.strip()
588
    match = valid_line_re.match(line)
589
    if not match:
590
      logging.error("Invalid line returned from lvs output: '%s'", line)
591
      continue
592
    name, size, attr = match.groups()
593
    inactive = attr[4] == '-'
594
    online = attr[5] == 'o'
595
    virtual = attr[0] == 'v'
596
    if virtual:
597
      # we don't want to report such volumes as existing, since they
598
      # don't really hold data
599
      continue
600
    lvs[name] = (size, inactive, online)
601

    
602
  return lvs
603

    
604

    
605
def ListVolumeGroups():
606
  """List the volume groups and their size.
607

608
  @rtype: dict
609
  @return: dictionary with keys volume name and values the
610
      size of the volume
611

612
  """
613
  return utils.ListVolumeGroups()
614

    
615

    
616
def NodeVolumes():
617
  """List all volumes on this node.
618

619
  @rtype: list
620
  @return:
621
    A list of dictionaries, each having four keys:
622
      - name: the logical volume name,
623
      - size: the size of the logical volume
624
      - dev: the physical device on which the LV lives
625
      - vg: the volume group to which it belongs
626

627
    In case of errors, we return an empty list and log the
628
    error.
629

630
    Note that since a logical volume can live on multiple physical
631
    volumes, the resulting list might include a logical volume
632
    multiple times.
633

634
  """
635
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
636
                         "--separator=|",
637
                         "--options=lv_name,lv_size,devices,vg_name"])
638
  if result.failed:
639
    _Fail("Failed to list logical volumes, lvs output: %s",
640
          result.output)
641

    
642
  def parse_dev(dev):
643
    if '(' in dev:
644
      return dev.split('(')[0]
645
    else:
646
      return dev
647

    
648
  def map_line(line):
649
    return {
650
      'name': line[0].strip(),
651
      'size': line[1].strip(),
652
      'dev': parse_dev(line[2].strip()),
653
      'vg': line[3].strip(),
654
    }
655

    
656
  return [map_line(line.split('|')) for line in result.stdout.splitlines()
657
          if line.count('|') >= 3]
658

    
659

    
660
def BridgesExist(bridges_list):
661
  """Check if a list of bridges exist on the current node.
662

663
  @rtype: boolean
664
  @return: C{True} if all of them exist, C{False} otherwise
665

666
  """
667
  missing = []
668
  for bridge in bridges_list:
669
    if not utils.BridgeExists(bridge):
670
      missing.append(bridge)
671

    
672
  if missing:
673
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
674

    
675

    
676
def GetInstanceList(hypervisor_list):
677
  """Provides a list of instances.
678

679
  @type hypervisor_list: list
680
  @param hypervisor_list: the list of hypervisors to query information
681

682
  @rtype: list
683
  @return: a list of all running instances on the current node
684
    - instance1.example.com
685
    - instance2.example.com
686

687
  """
688
  results = []
689
  for hname in hypervisor_list:
690
    try:
691
      names = hypervisor.GetHypervisor(hname).ListInstances()
692
      results.extend(names)
693
    except errors.HypervisorError, err:
694
      _Fail("Error enumerating instances (hypervisor %s): %s",
695
            hname, err, exc=True)
696

    
697
  return results
698

    
699

    
700
def GetInstanceInfo(instance, hname):
701
  """Gives back the information about an instance as a dictionary.
702

703
  @type instance: string
704
  @param instance: the instance name
705
  @type hname: string
706
  @param hname: the hypervisor type of the instance
707

708
  @rtype: dict
709
  @return: dictionary with the following keys:
710
      - memory: memory size of instance (int)
711
      - state: xen state of instance (string)
712
      - time: cpu time of instance (float)
713

714
  """
715
  output = {}
716

    
717
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
718
  if iinfo is not None:
719
    output['memory'] = iinfo[2]
720
    output['state'] = iinfo[4]
721
    output['time'] = iinfo[5]
722

    
723
  return output
724

    
725

    
726
def GetInstanceMigratable(instance):
727
  """Gives whether an instance can be migrated.
728

729
  @type instance: L{objects.Instance}
730
  @param instance: object representing the instance to be checked.
731

732
  @rtype: tuple
733
  @return: tuple of (result, description) where:
734
      - result: whether the instance can be migrated or not
735
      - description: a description of the issue, if relevant
736

737
  """
738
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
739
  iname = instance.name
740
  if iname not in hyper.ListInstances():
741
    _Fail("Instance %s is not running", iname)
742

    
743
  for idx in range(len(instance.disks)):
744
    link_name = _GetBlockDevSymlinkPath(iname, idx)
745
    if not os.path.islink(link_name):
746
      _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
747

    
748

    
749
def GetAllInstancesInfo(hypervisor_list):
750
  """Gather data about all instances.
751

752
  This is the equivalent of L{GetInstanceInfo}, except that it
753
  computes data for all instances at once, thus being faster if one
754
  needs data about more than one instance.
755

756
  @type hypervisor_list: list
757
  @param hypervisor_list: list of hypervisors to query for instance data
758

759
  @rtype: dict
760
  @return: dictionary of instance: data, with data having the following keys:
761
      - memory: memory size of instance (int)
762
      - state: xen state of instance (string)
763
      - time: cpu time of instance (float)
764
      - vcpus: the number of vcpus
765

766
  """
767
  output = {}
768

    
769
  for hname in hypervisor_list:
770
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
771
    if iinfo:
772
      for name, _, memory, vcpus, state, times in iinfo:
773
        value = {
774
          'memory': memory,
775
          'vcpus': vcpus,
776
          'state': state,
777
          'time': times,
778
          }
779
        if name in output:
780
          # we only check static parameters, like memory and vcpus,
781
          # and not state and time which can change between the
782
          # invocations of the different hypervisors
783
          for key in 'memory', 'vcpus':
784
            if value[key] != output[name][key]:
785
              _Fail("Instance %s is running twice"
786
                    " with different parameters", name)
787
        output[name] = value
788

    
789
  return output
790

    
791

    
792
def InstanceOsAdd(instance, reinstall, debug):
793
  """Add an OS to an instance.
794

795
  @type instance: L{objects.Instance}
796
  @param instance: Instance whose OS is to be installed
797
  @type reinstall: boolean
798
  @param reinstall: whether this is an instance reinstall
799
  @type debug: integer
800
  @param debug: debug level, passed to the OS scripts
801
  @rtype: None
802

803
  """
804
  inst_os = OSFromDisk(instance.os)
805

    
806
  create_env = OSEnvironment(instance, inst_os, debug)
807
  if reinstall:
808
    create_env['INSTANCE_REINSTALL'] = "1"
809

    
810
  logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
811
                                     instance.name, int(time.time()))
812

    
813
  result = utils.RunCmd([inst_os.create_script], env=create_env,
814
                        cwd=inst_os.path, output=logfile,)
815
  if result.failed:
816
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
817
                  " output: %s", result.cmd, result.fail_reason, logfile,
818
                  result.output)
819
    lines = [utils.SafeEncode(val)
820
             for val in utils.TailFile(logfile, lines=20)]
821
    _Fail("OS create script failed (%s), last lines in the"
822
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
823

    
824

    
825
def RunRenameInstance(instance, old_name, debug):
826
  """Run the OS rename script for an instance.
827

828
  @type instance: L{objects.Instance}
829
  @param instance: Instance whose OS is to be installed
830
  @type old_name: string
831
  @param old_name: previous instance name
832
  @type debug: integer
833
  @param debug: debug level, passed to the OS scripts
834
  @rtype: boolean
835
  @return: the success of the operation
836

837
  """
838
  inst_os = OSFromDisk(instance.os)
839

    
840
  rename_env = OSEnvironment(instance, inst_os, debug)
841
  rename_env['OLD_INSTANCE_NAME'] = old_name
842

    
843
  logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
844
                                           old_name,
845
                                           instance.name, int(time.time()))
846

    
847
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
848
                        cwd=inst_os.path, output=logfile)
849

    
850
  if result.failed:
851
    logging.error("os create command '%s' returned error: %s output: %s",
852
                  result.cmd, result.fail_reason, result.output)
853
    lines = [utils.SafeEncode(val)
854
             for val in utils.TailFile(logfile, lines=20)]
855
    _Fail("OS rename script failed (%s), last lines in the"
856
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
857

    
858

    
859
def _GetVGInfo(vg_name):
860
  """Get information about the volume group.
861

862
  @type vg_name: str
863
  @param vg_name: the volume group which we query
864
  @rtype: dict
865
  @return:
866
    A dictionary with the following keys:
867
      - C{vg_size} is the total size of the volume group in MiB
868
      - C{vg_free} is the free size of the volume group in MiB
869
      - C{pv_count} are the number of physical disks in that VG
870

871
    If an error occurs during gathering of data, we return the same dict
872
    with keys all set to None.
873

874
  """
875
  retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
876

    
877
  retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
878
                         "--nosuffix", "--units=m", "--separator=:", vg_name])
879

    
880
  if retval.failed:
881
    logging.error("volume group %s not present", vg_name)
882
    return retdic
883
  valarr = retval.stdout.strip().rstrip(':').split(':')
884
  if len(valarr) == 3:
885
    try:
886
      retdic = {
887
        "vg_size": int(round(float(valarr[0]), 0)),
888
        "vg_free": int(round(float(valarr[1]), 0)),
889
        "pv_count": int(valarr[2]),
890
        }
891
    except (TypeError, ValueError), err:
892
      logging.exception("Fail to parse vgs output: %s", err)
893
  else:
894
    logging.error("vgs output has the wrong number of fields (expected"
895
                  " three): %s", str(valarr))
896
  return retdic
897

    
898

    
899
def _GetBlockDevSymlinkPath(instance_name, idx):
900
  return os.path.join(constants.DISK_LINKS_DIR,
901
                      "%s:%d" % (instance_name, idx))
902

    
903

    
904
def _SymlinkBlockDev(instance_name, device_path, idx):
905
  """Set up symlinks to a instance's block device.
906

907
  This is an auxiliary function run when an instance is start (on the primary
908
  node) or when an instance is migrated (on the target node).
909

910

911
  @param instance_name: the name of the target instance
912
  @param device_path: path of the physical block device, on the node
913
  @param idx: the disk index
914
  @return: absolute path to the disk's symlink
915

916
  """
917
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
918
  try:
919
    os.symlink(device_path, link_name)
920
  except OSError, err:
921
    if err.errno == errno.EEXIST:
922
      if (not os.path.islink(link_name) or
923
          os.readlink(link_name) != device_path):
924
        os.remove(link_name)
925
        os.symlink(device_path, link_name)
926
    else:
927
      raise
928

    
929
  return link_name
930

    
931

    
932
def _RemoveBlockDevLinks(instance_name, disks):
933
  """Remove the block device symlinks belonging to the given instance.
934

935
  """
936
  for idx, _ in enumerate(disks):
937
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
938
    if os.path.islink(link_name):
939
      try:
940
        os.remove(link_name)
941
      except OSError:
942
        logging.exception("Can't remove symlink '%s'", link_name)
943

    
944

    
945
def _GatherAndLinkBlockDevs(instance):
946
  """Set up an instance's block device(s).
947

948
  This is run on the primary node at instance startup. The block
949
  devices must be already assembled.
950

951
  @type instance: L{objects.Instance}
952
  @param instance: the instance whose disks we shoul assemble
953
  @rtype: list
954
  @return: list of (disk_object, device_path)
955

956
  """
957
  block_devices = []
958
  for idx, disk in enumerate(instance.disks):
959
    device = _RecursiveFindBD(disk)
960
    if device is None:
961
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
962
                                    str(disk))
963
    device.Open()
964
    try:
965
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
966
    except OSError, e:
967
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
968
                                    e.strerror)
969

    
970
    block_devices.append((disk, link_name))
971

    
972
  return block_devices
973

    
974

    
975
def StartInstance(instance):
976
  """Start an instance.
977

978
  @type instance: L{objects.Instance}
979
  @param instance: the instance object
980
  @rtype: None
981

982
  """
983
  running_instances = GetInstanceList([instance.hypervisor])
984

    
985
  if instance.name in running_instances:
986
    logging.info("Instance %s already running, not starting", instance.name)
987
    return
988

    
989
  try:
990
    block_devices = _GatherAndLinkBlockDevs(instance)
991
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
992
    hyper.StartInstance(instance, block_devices)
993
  except errors.BlockDeviceError, err:
994
    _Fail("Block device error: %s", err, exc=True)
995
  except errors.HypervisorError, err:
996
    _RemoveBlockDevLinks(instance.name, instance.disks)
997
    _Fail("Hypervisor error: %s", err, exc=True)
998

    
999

    
1000
def InstanceShutdown(instance, timeout):
1001
  """Shut an instance down.
1002

1003
  @note: this functions uses polling with a hardcoded timeout.
1004

1005
  @type instance: L{objects.Instance}
1006
  @param instance: the instance object
1007
  @type timeout: integer
1008
  @param timeout: maximum timeout for soft shutdown
1009
  @rtype: None
1010

1011
  """
1012
  hv_name = instance.hypervisor
1013
  hyper = hypervisor.GetHypervisor(hv_name)
1014
  iname = instance.name
1015

    
1016
  if instance.name not in hyper.ListInstances():
1017
    logging.info("Instance %s not running, doing nothing", iname)
1018
    return
1019

    
1020
  class _TryShutdown:
1021
    def __init__(self):
1022
      self.tried_once = False
1023

    
1024
    def __call__(self):
1025
      if iname not in hyper.ListInstances():
1026
        return
1027

    
1028
      try:
1029
        hyper.StopInstance(instance, retry=self.tried_once)
1030
      except errors.HypervisorError, err:
1031
        if iname not in hyper.ListInstances():
1032
          # if the instance is no longer existing, consider this a
1033
          # success and go to cleanup
1034
          return
1035

    
1036
        _Fail("Failed to stop instance %s: %s", iname, err)
1037

    
1038
      self.tried_once = True
1039

    
1040
      raise utils.RetryAgain()
1041

    
1042
  try:
1043
    utils.Retry(_TryShutdown(), 5, timeout)
1044
  except utils.RetryTimeout:
1045
    # the shutdown did not succeed
1046
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1047

    
1048
    try:
1049
      hyper.StopInstance(instance, force=True)
1050
    except errors.HypervisorError, err:
1051
      if iname in hyper.ListInstances():
1052
        # only raise an error if the instance still exists, otherwise
1053
        # the error could simply be "instance ... unknown"!
1054
        _Fail("Failed to force stop instance %s: %s", iname, err)
1055

    
1056
    time.sleep(1)
1057

    
1058
    if iname in hyper.ListInstances():
1059
      _Fail("Could not shutdown instance %s even by destroy", iname)
1060

    
1061
  _RemoveBlockDevLinks(iname, instance.disks)
1062

    
1063

    
1064
def InstanceReboot(instance, reboot_type, shutdown_timeout):
1065
  """Reboot an instance.
1066

1067
  @type instance: L{objects.Instance}
1068
  @param instance: the instance object to reboot
1069
  @type reboot_type: str
1070
  @param reboot_type: the type of reboot, one the following
1071
    constants:
1072
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1073
        instance OS, do not recreate the VM
1074
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1075
        restart the VM (at the hypervisor level)
1076
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1077
        not accepted here, since that mode is handled differently, in
1078
        cmdlib, and translates into full stop and start of the
1079
        instance (instead of a call_instance_reboot RPC)
1080
  @type shutdown_timeout: integer
1081
  @param shutdown_timeout: maximum timeout for soft shutdown
1082
  @rtype: None
1083

1084
  """
1085
  running_instances = GetInstanceList([instance.hypervisor])
1086

    
1087
  if instance.name not in running_instances:
1088
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1089

    
1090
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1091
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1092
    try:
1093
      hyper.RebootInstance(instance)
1094
    except errors.HypervisorError, err:
1095
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1096
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1097
    try:
1098
      InstanceShutdown(instance, shutdown_timeout)
1099
      return StartInstance(instance)
1100
    except errors.HypervisorError, err:
1101
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1102
  else:
1103
    _Fail("Invalid reboot_type received: %s", reboot_type)
1104

    
1105

    
1106
def MigrationInfo(instance):
1107
  """Gather information about an instance to be migrated.
1108

1109
  @type instance: L{objects.Instance}
1110
  @param instance: the instance definition
1111

1112
  """
1113
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1114
  try:
1115
    info = hyper.MigrationInfo(instance)
1116
  except errors.HypervisorError, err:
1117
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1118
  return info
1119

    
1120

    
1121
def AcceptInstance(instance, info, target):
1122
  """Prepare the node to accept an instance.
1123

1124
  @type instance: L{objects.Instance}
1125
  @param instance: the instance definition
1126
  @type info: string/data (opaque)
1127
  @param info: migration information, from the source node
1128
  @type target: string
1129
  @param target: target host (usually ip), on this node
1130

1131
  """
1132
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1133
  try:
1134
    hyper.AcceptInstance(instance, info, target)
1135
  except errors.HypervisorError, err:
1136
    _Fail("Failed to accept instance: %s", err, exc=True)
1137

    
1138

    
1139
def FinalizeMigration(instance, info, success):
1140
  """Finalize any preparation to accept an instance.
1141

1142
  @type instance: L{objects.Instance}
1143
  @param instance: the instance definition
1144
  @type info: string/data (opaque)
1145
  @param info: migration information, from the source node
1146
  @type success: boolean
1147
  @param success: whether the migration was a success or a failure
1148

1149
  """
1150
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1151
  try:
1152
    hyper.FinalizeMigration(instance, info, success)
1153
  except errors.HypervisorError, err:
1154
    _Fail("Failed to finalize migration: %s", err, exc=True)
1155

    
1156

    
1157
def MigrateInstance(instance, target, live):
1158
  """Migrates an instance to another node.
1159

1160
  @type instance: L{objects.Instance}
1161
  @param instance: the instance definition
1162
  @type target: string
1163
  @param target: the target node name
1164
  @type live: boolean
1165
  @param live: whether the migration should be done live or not (the
1166
      interpretation of this parameter is left to the hypervisor)
1167
  @rtype: tuple
1168
  @return: a tuple of (success, msg) where:
1169
      - succes is a boolean denoting the success/failure of the operation
1170
      - msg is a string with details in case of failure
1171

1172
  """
1173
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1174

    
1175
  try:
1176
    hyper.MigrateInstance(instance, target, live)
1177
  except errors.HypervisorError, err:
1178
    _Fail("Failed to migrate instance: %s", err, exc=True)
1179

    
1180

    
1181
def BlockdevCreate(disk, size, owner, on_primary, info):
1182
  """Creates a block device for an instance.
1183

1184
  @type disk: L{objects.Disk}
1185
  @param disk: the object describing the disk we should create
1186
  @type size: int
1187
  @param size: the size of the physical underlying device, in MiB
1188
  @type owner: str
1189
  @param owner: the name of the instance for which disk is created,
1190
      used for device cache data
1191
  @type on_primary: boolean
1192
  @param on_primary:  indicates if it is the primary node or not
1193
  @type info: string
1194
  @param info: string that will be sent to the physical device
1195
      creation, used for example to set (LVM) tags on LVs
1196

1197
  @return: the new unique_id of the device (this can sometime be
1198
      computed only after creation), or None. On secondary nodes,
1199
      it's not required to return anything.
1200

1201
  """
1202
  # TODO: remove the obsolete 'size' argument
1203
  # pylint: disable-msg=W0613
1204
  clist = []
1205
  if disk.children:
1206
    for child in disk.children:
1207
      try:
1208
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1209
      except errors.BlockDeviceError, err:
1210
        _Fail("Can't assemble device %s: %s", child, err)
1211
      if on_primary or disk.AssembleOnSecondary():
1212
        # we need the children open in case the device itself has to
1213
        # be assembled
1214
        try:
1215
          # pylint: disable-msg=E1103
1216
          crdev.Open()
1217
        except errors.BlockDeviceError, err:
1218
          _Fail("Can't make child '%s' read-write: %s", child, err)
1219
      clist.append(crdev)
1220

    
1221
  try:
1222
    device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1223
  except errors.BlockDeviceError, err:
1224
    _Fail("Can't create block device: %s", err)
1225

    
1226
  if on_primary or disk.AssembleOnSecondary():
1227
    try:
1228
      device.Assemble()
1229
    except errors.BlockDeviceError, err:
1230
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1231
    device.SetSyncSpeed(constants.SYNC_SPEED)
1232
    if on_primary or disk.OpenOnSecondary():
1233
      try:
1234
        device.Open(force=True)
1235
      except errors.BlockDeviceError, err:
1236
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1237
    DevCacheManager.UpdateCache(device.dev_path, owner,
1238
                                on_primary, disk.iv_name)
1239

    
1240
  device.SetInfo(info)
1241

    
1242
  return device.unique_id
1243

    
1244

    
1245
def BlockdevRemove(disk):
1246
  """Remove a block device.
1247

1248
  @note: This is intended to be called recursively.
1249

1250
  @type disk: L{objects.Disk}
1251
  @param disk: the disk object we should remove
1252
  @rtype: boolean
1253
  @return: the success of the operation
1254

1255
  """
1256
  msgs = []
1257
  try:
1258
    rdev = _RecursiveFindBD(disk)
1259
  except errors.BlockDeviceError, err:
1260
    # probably can't attach
1261
    logging.info("Can't attach to device %s in remove", disk)
1262
    rdev = None
1263
  if rdev is not None:
1264
    r_path = rdev.dev_path
1265
    try:
1266
      rdev.Remove()
1267
    except errors.BlockDeviceError, err:
1268
      msgs.append(str(err))
1269
    if not msgs:
1270
      DevCacheManager.RemoveCache(r_path)
1271

    
1272
  if disk.children:
1273
    for child in disk.children:
1274
      try:
1275
        BlockdevRemove(child)
1276
      except RPCFail, err:
1277
        msgs.append(str(err))
1278

    
1279
  if msgs:
1280
    _Fail("; ".join(msgs))
1281

    
1282

    
1283
def _RecursiveAssembleBD(disk, owner, as_primary):
1284
  """Activate a block device for an instance.
1285

1286
  This is run on the primary and secondary nodes for an instance.
1287

1288
  @note: this function is called recursively.
1289

1290
  @type disk: L{objects.Disk}
1291
  @param disk: the disk we try to assemble
1292
  @type owner: str
1293
  @param owner: the name of the instance which owns the disk
1294
  @type as_primary: boolean
1295
  @param as_primary: if we should make the block device
1296
      read/write
1297

1298
  @return: the assembled device or None (in case no device
1299
      was assembled)
1300
  @raise errors.BlockDeviceError: in case there is an error
1301
      during the activation of the children or the device
1302
      itself
1303

1304
  """
1305
  children = []
1306
  if disk.children:
1307
    mcn = disk.ChildrenNeeded()
1308
    if mcn == -1:
1309
      mcn = 0 # max number of Nones allowed
1310
    else:
1311
      mcn = len(disk.children) - mcn # max number of Nones
1312
    for chld_disk in disk.children:
1313
      try:
1314
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1315
      except errors.BlockDeviceError, err:
1316
        if children.count(None) >= mcn:
1317
          raise
1318
        cdev = None
1319
        logging.error("Error in child activation (but continuing): %s",
1320
                      str(err))
1321
      children.append(cdev)
1322

    
1323
  if as_primary or disk.AssembleOnSecondary():
1324
    r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1325
    r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1326
    result = r_dev
1327
    if as_primary or disk.OpenOnSecondary():
1328
      r_dev.Open()
1329
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1330
                                as_primary, disk.iv_name)
1331

    
1332
  else:
1333
    result = True
1334
  return result
1335

    
1336

    
1337
def BlockdevAssemble(disk, owner, as_primary):
1338
  """Activate a block device for an instance.
1339

1340
  This is a wrapper over _RecursiveAssembleBD.
1341

1342
  @rtype: str or boolean
1343
  @return: a C{/dev/...} path for primary nodes, and
1344
      C{True} for secondary nodes
1345

1346
  """
1347
  try:
1348
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1349
    if isinstance(result, bdev.BlockDev):
1350
      # pylint: disable-msg=E1103
1351
      result = result.dev_path
1352
  except errors.BlockDeviceError, err:
1353
    _Fail("Error while assembling disk: %s", err, exc=True)
1354

    
1355
  return result
1356

    
1357

    
1358
def BlockdevShutdown(disk):
1359
  """Shut down a block device.
1360

1361
  First, if the device is assembled (Attach() is successful), then
1362
  the device is shutdown. Then the children of the device are
1363
  shutdown.
1364

1365
  This function is called recursively. Note that we don't cache the
1366
  children or such, as oppossed to assemble, shutdown of different
1367
  devices doesn't require that the upper device was active.
1368

1369
  @type disk: L{objects.Disk}
1370
  @param disk: the description of the disk we should
1371
      shutdown
1372
  @rtype: None
1373

1374
  """
1375
  msgs = []
1376
  r_dev = _RecursiveFindBD(disk)
1377
  if r_dev is not None:
1378
    r_path = r_dev.dev_path
1379
    try:
1380
      r_dev.Shutdown()
1381
      DevCacheManager.RemoveCache(r_path)
1382
    except errors.BlockDeviceError, err:
1383
      msgs.append(str(err))
1384

    
1385
  if disk.children:
1386
    for child in disk.children:
1387
      try:
1388
        BlockdevShutdown(child)
1389
      except RPCFail, err:
1390
        msgs.append(str(err))
1391

    
1392
  if msgs:
1393
    _Fail("; ".join(msgs))
1394

    
1395

    
1396
def BlockdevAddchildren(parent_cdev, new_cdevs):
1397
  """Extend a mirrored block device.
1398

1399
  @type parent_cdev: L{objects.Disk}
1400
  @param parent_cdev: the disk to which we should add children
1401
  @type new_cdevs: list of L{objects.Disk}
1402
  @param new_cdevs: the list of children which we should add
1403
  @rtype: None
1404

1405
  """
1406
  parent_bdev = _RecursiveFindBD(parent_cdev)
1407
  if parent_bdev is None:
1408
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1409
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1410
  if new_bdevs.count(None) > 0:
1411
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1412
  parent_bdev.AddChildren(new_bdevs)
1413

    
1414

    
1415
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1416
  """Shrink a mirrored block device.
1417

1418
  @type parent_cdev: L{objects.Disk}
1419
  @param parent_cdev: the disk from which we should remove children
1420
  @type new_cdevs: list of L{objects.Disk}
1421
  @param new_cdevs: the list of children which we should remove
1422
  @rtype: None
1423

1424
  """
1425
  parent_bdev = _RecursiveFindBD(parent_cdev)
1426
  if parent_bdev is None:
1427
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1428
  devs = []
1429
  for disk in new_cdevs:
1430
    rpath = disk.StaticDevPath()
1431
    if rpath is None:
1432
      bd = _RecursiveFindBD(disk)
1433
      if bd is None:
1434
        _Fail("Can't find device %s while removing children", disk)
1435
      else:
1436
        devs.append(bd.dev_path)
1437
    else:
1438
      devs.append(rpath)
1439
  parent_bdev.RemoveChildren(devs)
1440

    
1441

    
1442
def BlockdevGetmirrorstatus(disks):
1443
  """Get the mirroring status of a list of devices.
1444

1445
  @type disks: list of L{objects.Disk}
1446
  @param disks: the list of disks which we should query
1447
  @rtype: disk
1448
  @return:
1449
      a list of (mirror_done, estimated_time) tuples, which
1450
      are the result of L{bdev.BlockDev.CombinedSyncStatus}
1451
  @raise errors.BlockDeviceError: if any of the disks cannot be
1452
      found
1453

1454
  """
1455
  stats = []
1456
  for dsk in disks:
1457
    rbd = _RecursiveFindBD(dsk)
1458
    if rbd is None:
1459
      _Fail("Can't find device %s", dsk)
1460

    
1461
    stats.append(rbd.CombinedSyncStatus())
1462

    
1463
  return stats
1464

    
1465

    
1466
def _RecursiveFindBD(disk):
1467
  """Check if a device is activated.
1468

1469
  If so, return information about the real device.
1470

1471
  @type disk: L{objects.Disk}
1472
  @param disk: the disk object we need to find
1473

1474
  @return: None if the device can't be found,
1475
      otherwise the device instance
1476

1477
  """
1478
  children = []
1479
  if disk.children:
1480
    for chdisk in disk.children:
1481
      children.append(_RecursiveFindBD(chdisk))
1482

    
1483
  return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1484

    
1485

    
1486
def BlockdevFind(disk):
1487
  """Check if a device is activated.
1488

1489
  If it is, return information about the real device.
1490

1491
  @type disk: L{objects.Disk}
1492
  @param disk: the disk to find
1493
  @rtype: None or objects.BlockDevStatus
1494
  @return: None if the disk cannot be found, otherwise a the current
1495
           information
1496

1497
  """
1498
  try:
1499
    rbd = _RecursiveFindBD(disk)
1500
  except errors.BlockDeviceError, err:
1501
    _Fail("Failed to find device: %s", err, exc=True)
1502

    
1503
  if rbd is None:
1504
    return None
1505

    
1506
  return rbd.GetSyncStatus()
1507

    
1508

    
1509
def BlockdevGetsize(disks):
1510
  """Computes the size of the given disks.
1511

1512
  If a disk is not found, returns None instead.
1513

1514
  @type disks: list of L{objects.Disk}
1515
  @param disks: the list of disk to compute the size for
1516
  @rtype: list
1517
  @return: list with elements None if the disk cannot be found,
1518
      otherwise the size
1519

1520
  """
1521
  result = []
1522
  for cf in disks:
1523
    try:
1524
      rbd = _RecursiveFindBD(cf)
1525
    except errors.BlockDeviceError:
1526
      result.append(None)
1527
      continue
1528
    if rbd is None:
1529
      result.append(None)
1530
    else:
1531
      result.append(rbd.GetActualSize())
1532
  return result
1533

    
1534

    
1535
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1536
  """Export a block device to a remote node.
1537

1538
  @type disk: L{objects.Disk}
1539
  @param disk: the description of the disk to export
1540
  @type dest_node: str
1541
  @param dest_node: the destination node to export to
1542
  @type dest_path: str
1543
  @param dest_path: the destination path on the target node
1544
  @type cluster_name: str
1545
  @param cluster_name: the cluster name, needed for SSH hostalias
1546
  @rtype: None
1547

1548
  """
1549
  real_disk = _RecursiveFindBD(disk)
1550
  if real_disk is None:
1551
    _Fail("Block device '%s' is not set up", disk)
1552

    
1553
  real_disk.Open()
1554

    
1555
  # the block size on the read dd is 1MiB to match our units
1556
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1557
                               "dd if=%s bs=1048576 count=%s",
1558
                               real_disk.dev_path, str(disk.size))
1559

    
1560
  # we set here a smaller block size as, due to ssh buffering, more
1561
  # than 64-128k will mostly ignored; we use nocreat to fail if the
1562
  # device is not already there or we pass a wrong path; we use
1563
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1564
  # to not buffer too much memory; this means that at best, we flush
1565
  # every 64k, which will not be very fast
1566
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1567
                                " oflag=dsync", dest_path)
1568

    
1569
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1570
                                                   constants.GANETI_RUNAS,
1571
                                                   destcmd)
1572

    
1573
  # all commands have been checked, so we're safe to combine them
1574
  command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1575

    
1576
  result = utils.RunCmd(["bash", "-c", command])
1577

    
1578
  if result.failed:
1579
    _Fail("Disk copy command '%s' returned error: %s"
1580
          " output: %s", command, result.fail_reason, result.output)
1581

    
1582

    
1583
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1584
  """Write a file to the filesystem.
1585

1586
  This allows the master to overwrite(!) a file. It will only perform
1587
  the operation if the file belongs to a list of configuration files.
1588

1589
  @type file_name: str
1590
  @param file_name: the target file name
1591
  @type data: str
1592
  @param data: the new contents of the file
1593
  @type mode: int
1594
  @param mode: the mode to give the file (can be None)
1595
  @type uid: int
1596
  @param uid: the owner of the file (can be -1 for default)
1597
  @type gid: int
1598
  @param gid: the group of the file (can be -1 for default)
1599
  @type atime: float
1600
  @param atime: the atime to set on the file (can be None)
1601
  @type mtime: float
1602
  @param mtime: the mtime to set on the file (can be None)
1603
  @rtype: None
1604

1605
  """
1606
  if not os.path.isabs(file_name):
1607
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1608

    
1609
  if file_name not in _ALLOWED_UPLOAD_FILES:
1610
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1611
          file_name)
1612

    
1613
  raw_data = _Decompress(data)
1614

    
1615
  utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1616
                  atime=atime, mtime=mtime)
1617

    
1618

    
1619
def WriteSsconfFiles(values):
1620
  """Update all ssconf files.
1621

1622
  Wrapper around the SimpleStore.WriteFiles.
1623

1624
  """
1625
  ssconf.SimpleStore().WriteFiles(values)
1626

    
1627

    
1628
def _ErrnoOrStr(err):
1629
  """Format an EnvironmentError exception.
1630

1631
  If the L{err} argument has an errno attribute, it will be looked up
1632
  and converted into a textual C{E...} description. Otherwise the
1633
  string representation of the error will be returned.
1634

1635
  @type err: L{EnvironmentError}
1636
  @param err: the exception to format
1637

1638
  """
1639
  if hasattr(err, 'errno'):
1640
    detail = errno.errorcode[err.errno]
1641
  else:
1642
    detail = str(err)
1643
  return detail
1644

    
1645

    
1646
def _OSOndiskAPIVersion(os_dir):
1647
  """Compute and return the API version of a given OS.
1648

1649
  This function will try to read the API version of the OS residing in
1650
  the 'os_dir' directory.
1651

1652
  @type os_dir: str
1653
  @param os_dir: the directory in which we should look for the OS
1654
  @rtype: tuple
1655
  @return: tuple (status, data) with status denoting the validity and
1656
      data holding either the vaid versions or an error message
1657

1658
  """
1659
  api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
1660

    
1661
  try:
1662
    st = os.stat(api_file)
1663
  except EnvironmentError, err:
1664
    return False, ("Required file '%s' not found under path %s: %s" %
1665
                   (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1666

    
1667
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1668
    return False, ("File '%s' in %s is not a regular file" %
1669
                   (constants.OS_API_FILE, os_dir))
1670

    
1671
  try:
1672
    api_versions = utils.ReadFile(api_file).splitlines()
1673
  except EnvironmentError, err:
1674
    return False, ("Error while reading the API version file at %s: %s" %
1675
                   (api_file, _ErrnoOrStr(err)))
1676

    
1677
  try:
1678
    api_versions = [int(version.strip()) for version in api_versions]
1679
  except (TypeError, ValueError), err:
1680
    return False, ("API version(s) can't be converted to integer: %s" %
1681
                   str(err))
1682

    
1683
  return True, api_versions
1684

    
1685

    
1686
def DiagnoseOS(top_dirs=None):
1687
  """Compute the validity for all OSes.
1688

1689
  @type top_dirs: list
1690
  @param top_dirs: the list of directories in which to
1691
      search (if not given defaults to
1692
      L{constants.OS_SEARCH_PATH})
1693
  @rtype: list of L{objects.OS}
1694
  @return: a list of tuples (name, path, status, diagnose, variants)
1695
      for all (potential) OSes under all search paths, where:
1696
          - name is the (potential) OS name
1697
          - path is the full path to the OS
1698
          - status True/False is the validity of the OS
1699
          - diagnose is the error message for an invalid OS, otherwise empty
1700
          - variants is a list of supported OS variants, if any
1701

1702
  """
1703
  if top_dirs is None:
1704
    top_dirs = constants.OS_SEARCH_PATH
1705

    
1706
  result = []
1707
  for dir_name in top_dirs:
1708
    if os.path.isdir(dir_name):
1709
      try:
1710
        f_names = utils.ListVisibleFiles(dir_name)
1711
      except EnvironmentError, err:
1712
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1713
        break
1714
      for name in f_names:
1715
        os_path = os.path.sep.join([dir_name, name])
1716
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1717
        if status:
1718
          diagnose = ""
1719
          variants = os_inst.supported_variants
1720
        else:
1721
          diagnose = os_inst
1722
          variants = []
1723
        result.append((name, os_path, status, diagnose, variants))
1724

    
1725
  return result
1726

    
1727

    
1728
def _TryOSFromDisk(name, base_dir=None):
1729
  """Create an OS instance from disk.
1730

1731
  This function will return an OS instance if the given name is a
1732
  valid OS name.
1733

1734
  @type base_dir: string
1735
  @keyword base_dir: Base directory containing OS installations.
1736
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
1737
  @rtype: tuple
1738
  @return: success and either the OS instance if we find a valid one,
1739
      or error message
1740

1741
  """
1742
  if base_dir is None:
1743
    os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1744
  else:
1745
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1746

    
1747
  if os_dir is None:
1748
    return False, "Directory for OS %s not found in search path" % name
1749

    
1750
  status, api_versions = _OSOndiskAPIVersion(os_dir)
1751
  if not status:
1752
    # push the error up
1753
    return status, api_versions
1754

    
1755
  if not constants.OS_API_VERSIONS.intersection(api_versions):
1756
    return False, ("API version mismatch for path '%s': found %s, want %s." %
1757
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
1758

    
1759
  # OS Files dictionary, we will populate it with the absolute path names
1760
  os_files = dict.fromkeys(constants.OS_SCRIPTS)
1761

    
1762
  if max(api_versions) >= constants.OS_API_V15:
1763
    os_files[constants.OS_VARIANTS_FILE] = ''
1764

    
1765
  for filename in os_files:
1766
    os_files[filename] = os.path.sep.join([os_dir, filename])
1767

    
1768
    try:
1769
      st = os.stat(os_files[filename])
1770
    except EnvironmentError, err:
1771
      return False, ("File '%s' under path '%s' is missing (%s)" %
1772
                     (filename, os_dir, _ErrnoOrStr(err)))
1773

    
1774
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1775
      return False, ("File '%s' under path '%s' is not a regular file" %
1776
                     (filename, os_dir))
1777

    
1778
    if filename in constants.OS_SCRIPTS:
1779
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1780
        return False, ("File '%s' under path '%s' is not executable" %
1781
                       (filename, os_dir))
1782

    
1783
  variants = None
1784
  if constants.OS_VARIANTS_FILE in os_files:
1785
    variants_file = os_files[constants.OS_VARIANTS_FILE]
1786
    try:
1787
      variants = utils.ReadFile(variants_file).splitlines()
1788
    except EnvironmentError, err:
1789
      return False, ("Error while reading the OS variants file at %s: %s" %
1790
                     (variants_file, _ErrnoOrStr(err)))
1791
    if not variants:
1792
      return False, ("No supported os variant found")
1793

    
1794
  os_obj = objects.OS(name=name, path=os_dir,
1795
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
1796
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
1797
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
1798
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
1799
                      supported_variants=variants,
1800
                      api_versions=api_versions)
1801
  return True, os_obj
1802

    
1803

    
1804
def OSFromDisk(name, base_dir=None):
1805
  """Create an OS instance from disk.
1806

1807
  This function will return an OS instance if the given name is a
1808
  valid OS name. Otherwise, it will raise an appropriate
1809
  L{RPCFail} exception, detailing why this is not a valid OS.
1810

1811
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1812
  an exception but returns true/false status data.
1813

1814
  @type base_dir: string
1815
  @keyword base_dir: Base directory containing OS installations.
1816
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
1817
  @rtype: L{objects.OS}
1818
  @return: the OS instance if we find a valid one
1819
  @raise RPCFail: if we don't find a valid OS
1820

1821
  """
1822
  name_only = name.split("+", 1)[0]
1823
  status, payload = _TryOSFromDisk(name_only, base_dir)
1824

    
1825
  if not status:
1826
    _Fail(payload)
1827

    
1828
  return payload
1829

    
1830

    
1831
def OSEnvironment(instance, inst_os, debug=0):
1832
  """Calculate the environment for an os script.
1833

1834
  @type instance: L{objects.Instance}
1835
  @param instance: target instance for the os script run
1836
  @type inst_os: L{objects.OS}
1837
  @param inst_os: operating system for which the environment is being built
1838
  @type debug: integer
1839
  @param debug: debug level (0 or 1, for OS Api 10)
1840
  @rtype: dict
1841
  @return: dict of environment variables
1842
  @raise errors.BlockDeviceError: if the block device
1843
      cannot be found
1844

1845
  """
1846
  result = {}
1847
  api_version = \
1848
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1849
  result['OS_API_VERSION'] = '%d' % api_version
1850
  result['INSTANCE_NAME'] = instance.name
1851
  result['INSTANCE_OS'] = instance.os
1852
  result['HYPERVISOR'] = instance.hypervisor
1853
  result['DISK_COUNT'] = '%d' % len(instance.disks)
1854
  result['NIC_COUNT'] = '%d' % len(instance.nics)
1855
  result['DEBUG_LEVEL'] = '%d' % debug
1856
  if api_version >= constants.OS_API_V15:
1857
    try:
1858
      variant = instance.os.split('+', 1)[1]
1859
    except IndexError:
1860
      variant = inst_os.supported_variants[0]
1861
    result['OS_VARIANT'] = variant
1862
  for idx, disk in enumerate(instance.disks):
1863
    real_disk = _RecursiveFindBD(disk)
1864
    if real_disk is None:
1865
      raise errors.BlockDeviceError("Block device '%s' is not set up" %
1866
                                    str(disk))
1867
    real_disk.Open()
1868
    result['DISK_%d_PATH' % idx] = real_disk.dev_path
1869
    result['DISK_%d_ACCESS' % idx] = disk.mode
1870
    if constants.HV_DISK_TYPE in instance.hvparams:
1871
      result['DISK_%d_FRONTEND_TYPE' % idx] = \
1872
        instance.hvparams[constants.HV_DISK_TYPE]
1873
    if disk.dev_type in constants.LDS_BLOCK:
1874
      result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1875
    elif disk.dev_type == constants.LD_FILE:
1876
      result['DISK_%d_BACKEND_TYPE' % idx] = \
1877
        'file:%s' % disk.physical_id[0]
1878
  for idx, nic in enumerate(instance.nics):
1879
    result['NIC_%d_MAC' % idx] = nic.mac
1880
    if nic.ip:
1881
      result['NIC_%d_IP' % idx] = nic.ip
1882
    result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1883
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1884
      result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1885
    if nic.nicparams[constants.NIC_LINK]:
1886
      result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1887
    if constants.HV_NIC_TYPE in instance.hvparams:
1888
      result['NIC_%d_FRONTEND_TYPE' % idx] = \
1889
        instance.hvparams[constants.HV_NIC_TYPE]
1890

    
1891
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1892
    for key, value in source.items():
1893
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1894

    
1895
  return result
1896

    
1897
def BlockdevGrow(disk, amount):
1898
  """Grow a stack of block devices.
1899

1900
  This function is called recursively, with the childrens being the
1901
  first ones to resize.
1902

1903
  @type disk: L{objects.Disk}
1904
  @param disk: the disk to be grown
1905
  @rtype: (status, result)
1906
  @return: a tuple with the status of the operation
1907
      (True/False), and the errors message if status
1908
      is False
1909

1910
  """
1911
  r_dev = _RecursiveFindBD(disk)
1912
  if r_dev is None:
1913
    _Fail("Cannot find block device %s", disk)
1914

    
1915
  try:
1916
    r_dev.Grow(amount)
1917
  except errors.BlockDeviceError, err:
1918
    _Fail("Failed to grow block device: %s", err, exc=True)
1919

    
1920

    
1921
def BlockdevSnapshot(disk):
1922
  """Create a snapshot copy of a block device.
1923

1924
  This function is called recursively, and the snapshot is actually created
1925
  just for the leaf lvm backend device.
1926

1927
  @type disk: L{objects.Disk}
1928
  @param disk: the disk to be snapshotted
1929
  @rtype: string
1930
  @return: snapshot disk path
1931

1932
  """
1933
  if disk.dev_type == constants.LD_DRBD8:
1934
    if not disk.children:
1935
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
1936
            disk.unique_id)
1937
    return BlockdevSnapshot(disk.children[0])
1938
  elif disk.dev_type == constants.LD_LV:
1939
    r_dev = _RecursiveFindBD(disk)
1940
    if r_dev is not None:
1941
      # FIXME: choose a saner value for the snapshot size
1942
      # let's stay on the safe side and ask for the full size, for now
1943
      return r_dev.Snapshot(disk.size)
1944
    else:
1945
      _Fail("Cannot find block device %s", disk)
1946
  else:
1947
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1948
          disk.unique_id, disk.dev_type)
1949

    
1950

    
1951
def ExportSnapshot(disk, dest_node, instance, cluster_name, idx, debug):
1952
  """Export a block device snapshot to a remote node.
1953

1954
  @type disk: L{objects.Disk}
1955
  @param disk: the description of the disk to export
1956
  @type dest_node: str
1957
  @param dest_node: the destination node to export to
1958
  @type instance: L{objects.Instance}
1959
  @param instance: the instance object to whom the disk belongs
1960
  @type cluster_name: str
1961
  @param cluster_name: the cluster name, needed for SSH hostalias
1962
  @type idx: int
1963
  @param idx: the index of the disk in the instance's disk list,
1964
      used to export to the OS scripts environment
1965
  @type debug: integer
1966
  @param debug: debug level, passed to the OS scripts
1967
  @rtype: None
1968

1969
  """
1970
  inst_os = OSFromDisk(instance.os)
1971
  export_env = OSEnvironment(instance, inst_os, debug)
1972

    
1973
  export_script = inst_os.export_script
1974

    
1975
  logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1976
                                     instance.name, int(time.time()))
1977
  if not os.path.exists(constants.LOG_OS_DIR):
1978
    os.mkdir(constants.LOG_OS_DIR, 0750)
1979
  real_disk = _RecursiveFindBD(disk)
1980
  if real_disk is None:
1981
    _Fail("Block device '%s' is not set up", disk)
1982

    
1983
  real_disk.Open()
1984

    
1985
  export_env['EXPORT_DEVICE'] = real_disk.dev_path
1986
  export_env['EXPORT_INDEX'] = str(idx)
1987

    
1988
  destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1989
  destfile = disk.physical_id[1]
1990

    
1991
  # the target command is built out of three individual commands,
1992
  # which are joined by pipes; we check each individual command for
1993
  # valid parameters
1994
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
1995
                               inst_os.path, export_script, logfile)
1996

    
1997
  comprcmd = "gzip"
1998

    
1999
  destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
2000
                                destdir, destdir, destfile)
2001
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2002
                                                   constants.GANETI_RUNAS,
2003
                                                   destcmd)
2004

    
2005
  # all commands have been checked, so we're safe to combine them
2006
  command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
2007

    
2008
  result = utils.RunCmd(["bash", "-c", command], env=export_env)
2009

    
2010
  if result.failed:
2011
    _Fail("OS snapshot export command '%s' returned error: %s"
2012
          " output: %s", command, result.fail_reason, result.output)
2013

    
2014

    
2015
def FinalizeExport(instance, snap_disks):
2016
  """Write out the export configuration information.
2017

2018
  @type instance: L{objects.Instance}
2019
  @param instance: the instance which we export, used for
2020
      saving configuration
2021
  @type snap_disks: list of L{objects.Disk}
2022
  @param snap_disks: list of snapshot block devices, which
2023
      will be used to get the actual name of the dump file
2024

2025
  @rtype: None
2026

2027
  """
2028
  destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
2029
  finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
2030

    
2031
  config = objects.SerializableConfigParser()
2032

    
2033
  config.add_section(constants.INISECT_EXP)
2034
  config.set(constants.INISECT_EXP, 'version', '0')
2035
  config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2036
  config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2037
  config.set(constants.INISECT_EXP, 'os', instance.os)
2038
  config.set(constants.INISECT_EXP, 'compression', 'gzip')
2039

    
2040
  config.add_section(constants.INISECT_INS)
2041
  config.set(constants.INISECT_INS, 'name', instance.name)
2042
  config.set(constants.INISECT_INS, 'memory', '%d' %
2043
             instance.beparams[constants.BE_MEMORY])
2044
  config.set(constants.INISECT_INS, 'vcpus', '%d' %
2045
             instance.beparams[constants.BE_VCPUS])
2046
  config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2047

    
2048
  nic_total = 0
2049
  for nic_count, nic in enumerate(instance.nics):
2050
    nic_total += 1
2051
    config.set(constants.INISECT_INS, 'nic%d_mac' %
2052
               nic_count, '%s' % nic.mac)
2053
    config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2054
    config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
2055
               '%s' % nic.bridge)
2056
  # TODO: redundant: on load can read nics until it doesn't exist
2057
  config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2058

    
2059
  disk_total = 0
2060
  for disk_count, disk in enumerate(snap_disks):
2061
    if disk:
2062
      disk_total += 1
2063
      config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2064
                 ('%s' % disk.iv_name))
2065
      config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2066
                 ('%s' % disk.physical_id[1]))
2067
      config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2068
                 ('%d' % disk.size))
2069

    
2070
  config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2071

    
2072
  utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
2073
                  data=config.Dumps())
2074
  shutil.rmtree(finaldestdir, True)
2075
  shutil.move(destdir, finaldestdir)
2076

    
2077

    
2078
def ExportInfo(dest):
2079
  """Get export configuration information.
2080

2081
  @type dest: str
2082
  @param dest: directory containing the export
2083

2084
  @rtype: L{objects.SerializableConfigParser}
2085
  @return: a serializable config file containing the
2086
      export info
2087

2088
  """
2089
  cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
2090

    
2091
  config = objects.SerializableConfigParser()
2092
  config.read(cff)
2093

    
2094
  if (not config.has_section(constants.INISECT_EXP) or
2095
      not config.has_section(constants.INISECT_INS)):
2096
    _Fail("Export info file doesn't have the required fields")
2097

    
2098
  return config.Dumps()
2099

    
2100

    
2101
def ImportOSIntoInstance(instance, src_node, src_images, cluster_name, debug):
2102
  """Import an os image into an instance.
2103

2104
  @type instance: L{objects.Instance}
2105
  @param instance: instance to import the disks into
2106
  @type src_node: string
2107
  @param src_node: source node for the disk images
2108
  @type src_images: list of string
2109
  @param src_images: absolute paths of the disk images
2110
  @type debug: integer
2111
  @param debug: debug level, passed to the OS scripts
2112
  @rtype: list of boolean
2113
  @return: each boolean represent the success of importing the n-th disk
2114

2115
  """
2116
  inst_os = OSFromDisk(instance.os)
2117
  import_env = OSEnvironment(instance, inst_os, debug)
2118
  import_script = inst_os.import_script
2119

    
2120
  logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2121
                                        instance.name, int(time.time()))
2122
  if not os.path.exists(constants.LOG_OS_DIR):
2123
    os.mkdir(constants.LOG_OS_DIR, 0750)
2124

    
2125
  comprcmd = "gunzip"
2126
  impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2127
                               import_script, logfile)
2128

    
2129
  final_result = []
2130
  for idx, image in enumerate(src_images):
2131
    if image:
2132
      destcmd = utils.BuildShellCmd('cat %s', image)
2133
      remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2134
                                                       constants.GANETI_RUNAS,
2135
                                                       destcmd)
2136
      command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2137
      import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2138
      import_env['IMPORT_INDEX'] = str(idx)
2139
      result = utils.RunCmd(command, env=import_env)
2140
      if result.failed:
2141
        logging.error("Disk import command '%s' returned error: %s"
2142
                      " output: %s", command, result.fail_reason,
2143
                      result.output)
2144
        final_result.append("error importing disk %d: %s, %s" %
2145
                            (idx, result.fail_reason, result.output[-100]))
2146

    
2147
  if final_result:
2148
    _Fail("; ".join(final_result), log=False)
2149

    
2150

    
2151
def ListExports():
2152
  """Return a list of exports currently available on this machine.
2153

2154
  @rtype: list
2155
  @return: list of the exports
2156

2157
  """
2158
  if os.path.isdir(constants.EXPORT_DIR):
2159
    return utils.ListVisibleFiles(constants.EXPORT_DIR)
2160
  else:
2161
    _Fail("No exports directory")
2162

    
2163

    
2164
def RemoveExport(export):
2165
  """Remove an existing export from the node.
2166

2167
  @type export: str
2168
  @param export: the name of the export to remove
2169
  @rtype: None
2170

2171
  """
2172
  target = os.path.join(constants.EXPORT_DIR, export)
2173

    
2174
  try:
2175
    shutil.rmtree(target)
2176
  except EnvironmentError, err:
2177
    _Fail("Error while removing the export: %s", err, exc=True)
2178

    
2179

    
2180
def BlockdevRename(devlist):
2181
  """Rename a list of block devices.
2182

2183
  @type devlist: list of tuples
2184
  @param devlist: list of tuples of the form  (disk,
2185
      new_logical_id, new_physical_id); disk is an
2186
      L{objects.Disk} object describing the current disk,
2187
      and new logical_id/physical_id is the name we
2188
      rename it to
2189
  @rtype: boolean
2190
  @return: True if all renames succeeded, False otherwise
2191

2192
  """
2193
  msgs = []
2194
  result = True
2195
  for disk, unique_id in devlist:
2196
    dev = _RecursiveFindBD(disk)
2197
    if dev is None:
2198
      msgs.append("Can't find device %s in rename" % str(disk))
2199
      result = False
2200
      continue
2201
    try:
2202
      old_rpath = dev.dev_path
2203
      dev.Rename(unique_id)
2204
      new_rpath = dev.dev_path
2205
      if old_rpath != new_rpath:
2206
        DevCacheManager.RemoveCache(old_rpath)
2207
        # FIXME: we should add the new cache information here, like:
2208
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2209
        # but we don't have the owner here - maybe parse from existing
2210
        # cache? for now, we only lose lvm data when we rename, which
2211
        # is less critical than DRBD or MD
2212
    except errors.BlockDeviceError, err:
2213
      msgs.append("Can't rename device '%s' to '%s': %s" %
2214
                  (dev, unique_id, err))
2215
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2216
      result = False
2217
  if not result:
2218
    _Fail("; ".join(msgs))
2219

    
2220

    
2221
def _TransformFileStorageDir(file_storage_dir):
2222
  """Checks whether given file_storage_dir is valid.
2223

2224
  Checks wheter the given file_storage_dir is within the cluster-wide
2225
  default file_storage_dir stored in SimpleStore. Only paths under that
2226
  directory are allowed.
2227

2228
  @type file_storage_dir: str
2229
  @param file_storage_dir: the path to check
2230

2231
  @return: the normalized path if valid, None otherwise
2232

2233
  """
2234
  cfg = _GetConfig()
2235
  file_storage_dir = os.path.normpath(file_storage_dir)
2236
  base_file_storage_dir = cfg.GetFileStorageDir()
2237
  if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2238
      base_file_storage_dir):
2239
    _Fail("File storage directory '%s' is not under base file"
2240
          " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2241
  return file_storage_dir
2242

    
2243

    
2244
def CreateFileStorageDir(file_storage_dir):
2245
  """Create file storage directory.
2246

2247
  @type file_storage_dir: str
2248
  @param file_storage_dir: directory to create
2249

2250
  @rtype: tuple
2251
  @return: tuple with first element a boolean indicating wheter dir
2252
      creation was successful or not
2253

2254
  """
2255
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2256
  if os.path.exists(file_storage_dir):
2257
    if not os.path.isdir(file_storage_dir):
2258
      _Fail("Specified storage dir '%s' is not a directory",
2259
            file_storage_dir)
2260
  else:
2261
    try:
2262
      os.makedirs(file_storage_dir, 0750)
2263
    except OSError, err:
2264
      _Fail("Cannot create file storage directory '%s': %s",
2265
            file_storage_dir, err, exc=True)
2266

    
2267

    
2268
def RemoveFileStorageDir(file_storage_dir):
2269
  """Remove file storage directory.
2270

2271
  Remove it only if it's empty. If not log an error and return.
2272

2273
  @type file_storage_dir: str
2274
  @param file_storage_dir: the directory we should cleanup
2275
  @rtype: tuple (success,)
2276
  @return: tuple of one element, C{success}, denoting
2277
      whether the operation was successful
2278

2279
  """
2280
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2281
  if os.path.exists(file_storage_dir):
2282
    if not os.path.isdir(file_storage_dir):
2283
      _Fail("Specified Storage directory '%s' is not a directory",
2284
            file_storage_dir)
2285
    # deletes dir only if empty, otherwise we want to fail the rpc call
2286
    try:
2287
      os.rmdir(file_storage_dir)
2288
    except OSError, err:
2289
      _Fail("Cannot remove file storage directory '%s': %s",
2290
            file_storage_dir, err)
2291

    
2292

    
2293
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2294
  """Rename the file storage directory.
2295

2296
  @type old_file_storage_dir: str
2297
  @param old_file_storage_dir: the current path
2298
  @type new_file_storage_dir: str
2299
  @param new_file_storage_dir: the name we should rename to
2300
  @rtype: tuple (success,)
2301
  @return: tuple of one element, C{success}, denoting
2302
      whether the operation was successful
2303

2304
  """
2305
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2306
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2307
  if not os.path.exists(new_file_storage_dir):
2308
    if os.path.isdir(old_file_storage_dir):
2309
      try:
2310
        os.rename(old_file_storage_dir, new_file_storage_dir)
2311
      except OSError, err:
2312
        _Fail("Cannot rename '%s' to '%s': %s",
2313
              old_file_storage_dir, new_file_storage_dir, err)
2314
    else:
2315
      _Fail("Specified storage dir '%s' is not a directory",
2316
            old_file_storage_dir)
2317
  else:
2318
    if os.path.exists(old_file_storage_dir):
2319
      _Fail("Cannot rename '%s' to '%s': both locations exist",
2320
            old_file_storage_dir, new_file_storage_dir)
2321

    
2322

    
2323
def _EnsureJobQueueFile(file_name):
2324
  """Checks whether the given filename is in the queue directory.
2325

2326
  @type file_name: str
2327
  @param file_name: the file name we should check
2328
  @rtype: None
2329
  @raises RPCFail: if the file is not valid
2330

2331
  """
2332
  queue_dir = os.path.normpath(constants.QUEUE_DIR)
2333
  result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2334

    
2335
  if not result:
2336
    _Fail("Passed job queue file '%s' does not belong to"
2337
          " the queue directory '%s'", file_name, queue_dir)
2338

    
2339

    
2340
def JobQueueUpdate(file_name, content):
2341
  """Updates a file in the queue directory.
2342

2343
  This is just a wrapper over L{utils.WriteFile}, with proper
2344
  checking.
2345

2346
  @type file_name: str
2347
  @param file_name: the job file name
2348
  @type content: str
2349
  @param content: the new job contents
2350
  @rtype: boolean
2351
  @return: the success of the operation
2352

2353
  """
2354
  _EnsureJobQueueFile(file_name)
2355

    
2356
  # Write and replace the file atomically
2357
  utils.WriteFile(file_name, data=_Decompress(content))
2358

    
2359

    
2360
def JobQueueRename(old, new):
2361
  """Renames a job queue file.
2362

2363
  This is just a wrapper over os.rename with proper checking.
2364

2365
  @type old: str
2366
  @param old: the old (actual) file name
2367
  @type new: str
2368
  @param new: the desired file name
2369
  @rtype: tuple
2370
  @return: the success of the operation and payload
2371

2372
  """
2373
  _EnsureJobQueueFile(old)
2374
  _EnsureJobQueueFile(new)
2375

    
2376
  utils.RenameFile(old, new, mkdir=True)
2377

    
2378

    
2379
def JobQueueSetDrainFlag(drain_flag):
2380
  """Set the drain flag for the queue.
2381

2382
  This will set or unset the queue drain flag.
2383

2384
  @type drain_flag: boolean
2385
  @param drain_flag: if True, will set the drain flag, otherwise reset it.
2386
  @rtype: truple
2387
  @return: always True, None
2388
  @warning: the function always returns True
2389

2390
  """
2391
  if drain_flag:
2392
    utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2393
  else:
2394
    utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2395

    
2396

    
2397
def BlockdevClose(instance_name, disks):
2398
  """Closes the given block devices.
2399

2400
  This means they will be switched to secondary mode (in case of
2401
  DRBD).
2402

2403
  @param instance_name: if the argument is not empty, the symlinks
2404
      of this instance will be removed
2405
  @type disks: list of L{objects.Disk}
2406
  @param disks: the list of disks to be closed
2407
  @rtype: tuple (success, message)
2408
  @return: a tuple of success and message, where success
2409
      indicates the succes of the operation, and message
2410
      which will contain the error details in case we
2411
      failed
2412

2413
  """
2414
  bdevs = []
2415
  for cf in disks:
2416
    rd = _RecursiveFindBD(cf)
2417
    if rd is None:
2418
      _Fail("Can't find device %s", cf)
2419
    bdevs.append(rd)
2420

    
2421
  msg = []
2422
  for rd in bdevs:
2423
    try:
2424
      rd.Close()
2425
    except errors.BlockDeviceError, err:
2426
      msg.append(str(err))
2427
  if msg:
2428
    _Fail("Can't make devices secondary: %s", ",".join(msg))
2429
  else:
2430
    if instance_name:
2431
      _RemoveBlockDevLinks(instance_name, disks)
2432

    
2433

    
2434
def ValidateHVParams(hvname, hvparams):
2435
  """Validates the given hypervisor parameters.
2436

2437
  @type hvname: string
2438
  @param hvname: the hypervisor name
2439
  @type hvparams: dict
2440
  @param hvparams: the hypervisor parameters to be validated
2441
  @rtype: None
2442

2443
  """
2444
  try:
2445
    hv_type = hypervisor.GetHypervisor(hvname)
2446
    hv_type.ValidateParameters(hvparams)
2447
  except errors.HypervisorError, err:
2448
    _Fail(str(err), log=False)
2449

    
2450

    
2451
def DemoteFromMC():
2452
  """Demotes the current node from master candidate role.
2453

2454
  """
2455
  # try to ensure we're not the master by mistake
2456
  master, myself = ssconf.GetMasterAndMyself()
2457
  if master == myself:
2458
    _Fail("ssconf status shows I'm the master node, will not demote")
2459

    
2460
  result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2461
  if not result.failed:
2462
    _Fail("The master daemon is running, will not demote")
2463

    
2464
  try:
2465
    if os.path.isfile(constants.CLUSTER_CONF_FILE):
2466
      utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2467
  except EnvironmentError, err:
2468
    if err.errno != errno.ENOENT:
2469
      _Fail("Error while backing up cluster file: %s", err, exc=True)
2470

    
2471
  utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2472

    
2473

    
2474
def _FindDisks(nodes_ip, disks):
2475
  """Sets the physical ID on disks and returns the block devices.
2476

2477
  """
2478
  # set the correct physical ID
2479
  my_name = utils.HostInfo().name
2480
  for cf in disks:
2481
    cf.SetPhysicalID(my_name, nodes_ip)
2482

    
2483
  bdevs = []
2484

    
2485
  for cf in disks:
2486
    rd = _RecursiveFindBD(cf)
2487
    if rd is None:
2488
      _Fail("Can't find device %s", cf)
2489
    bdevs.append(rd)
2490
  return bdevs
2491

    
2492

    
2493
def DrbdDisconnectNet(nodes_ip, disks):
2494
  """Disconnects the network on a list of drbd devices.
2495

2496
  """
2497
  bdevs = _FindDisks(nodes_ip, disks)
2498

    
2499
  # disconnect disks
2500
  for rd in bdevs:
2501
    try:
2502
      rd.DisconnectNet()
2503
    except errors.BlockDeviceError, err:
2504
      _Fail("Can't change network configuration to standalone mode: %s",
2505
            err, exc=True)
2506

    
2507

    
2508
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2509
  """Attaches the network on a list of drbd devices.
2510

2511
  """
2512
  bdevs = _FindDisks(nodes_ip, disks)
2513

    
2514
  if multimaster:
2515
    for idx, rd in enumerate(bdevs):
2516
      try:
2517
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2518
      except EnvironmentError, err:
2519
        _Fail("Can't create symlink: %s", err)
2520
  # reconnect disks, switch to new master configuration and if
2521
  # needed primary mode
2522
  for rd in bdevs:
2523
    try:
2524
      rd.AttachNet(multimaster)
2525
    except errors.BlockDeviceError, err:
2526
      _Fail("Can't change network configuration: %s", err)
2527

    
2528
  # wait until the disks are connected; we need to retry the re-attach
2529
  # if the device becomes standalone, as this might happen if the one
2530
  # node disconnects and reconnects in a different mode before the
2531
  # other node reconnects; in this case, one or both of the nodes will
2532
  # decide it has wrong configuration and switch to standalone
2533

    
2534
  def _Attach():
2535
    all_connected = True
2536

    
2537
    for rd in bdevs:
2538
      stats = rd.GetProcStatus()
2539

    
2540
      all_connected = (all_connected and
2541
                       (stats.is_connected or stats.is_in_resync))
2542

    
2543
      if stats.is_standalone:
2544
        # peer had different config info and this node became
2545
        # standalone, even though this should not happen with the
2546
        # new staged way of changing disk configs
2547
        try:
2548
          rd.AttachNet(multimaster)
2549
        except errors.BlockDeviceError, err:
2550
          _Fail("Can't change network configuration: %s", err)
2551

    
2552
    if not all_connected:
2553
      raise utils.RetryAgain()
2554

    
2555
  try:
2556
    # Start with a delay of 100 miliseconds and go up to 5 seconds
2557
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2558
  except utils.RetryTimeout:
2559
    _Fail("Timeout in disk reconnecting")
2560

    
2561
  if multimaster:
2562
    # change to primary mode
2563
    for rd in bdevs:
2564
      try:
2565
        rd.Open()
2566
      except errors.BlockDeviceError, err:
2567
        _Fail("Can't change to primary mode: %s", err)
2568

    
2569

    
2570
def DrbdWaitSync(nodes_ip, disks):
2571
  """Wait until DRBDs have synchronized.
2572

2573
  """
2574
  def _helper(rd):
2575
    stats = rd.GetProcStatus()
2576
    if not (stats.is_connected or stats.is_in_resync):
2577
      raise utils.RetryAgain()
2578
    return stats
2579

    
2580
  bdevs = _FindDisks(nodes_ip, disks)
2581

    
2582
  min_resync = 100
2583
  alldone = True
2584
  for rd in bdevs:
2585
    try:
2586
      # poll each second for 15 seconds
2587
      stats = utils.Retry(_helper, 1, 15, args=[rd])
2588
    except utils.RetryTimeout:
2589
      stats = rd.GetProcStatus()
2590
      # last check
2591
      if not (stats.is_connected or stats.is_in_resync):
2592
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2593
    alldone = alldone and (not stats.is_in_resync)
2594
    if stats.sync_percent is not None:
2595
      min_resync = min(min_resync, stats.sync_percent)
2596

    
2597
  return (alldone, min_resync)
2598

    
2599

    
2600
def PowercycleNode(hypervisor_type):
2601
  """Hard-powercycle the node.
2602

2603
  Because we need to return first, and schedule the powercycle in the
2604
  background, we won't be able to report failures nicely.
2605

2606
  """
2607
  hyper = hypervisor.GetHypervisor(hypervisor_type)
2608
  try:
2609
    pid = os.fork()
2610
  except OSError:
2611
    # if we can't fork, we'll pretend that we're in the child process
2612
    pid = 0
2613
  if pid > 0:
2614
    return "Reboot scheduled in 5 seconds"
2615
  time.sleep(5)
2616
  hyper.PowercycleNode()
2617

    
2618

    
2619
class HooksRunner(object):
2620
  """Hook runner.
2621

2622
  This class is instantiated on the node side (ganeti-noded) and not
2623
  on the master side.
2624

2625
  """
2626
  def __init__(self, hooks_base_dir=None):
2627
    """Constructor for hooks runner.
2628

2629
    @type hooks_base_dir: str or None
2630
    @param hooks_base_dir: if not None, this overrides the
2631
        L{constants.HOOKS_BASE_DIR} (useful for unittests)
2632

2633
    """
2634
    if hooks_base_dir is None:
2635
      hooks_base_dir = constants.HOOKS_BASE_DIR
2636
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
2637
    # constant
2638
    self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
2639

    
2640
  @staticmethod
2641
  def ExecHook(script, env):
2642
    """Exec one hook script.
2643

2644
    @type script: str
2645
    @param script: the full path to the script
2646
    @type env: dict
2647
    @param env: the environment with which to exec the script
2648
    @rtype: tuple (success, message)
2649
    @return: a tuple of success and message, where success
2650
        indicates the succes of the operation, and message
2651
        which will contain the error details in case we
2652
        failed
2653

2654
    """
2655
    # exec the process using subprocess and log the output
2656
    fdstdin = None
2657
    try:
2658
      fdstdin = open("/dev/null", "r")
2659
      child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2660
                               stderr=subprocess.STDOUT, close_fds=True,
2661
                               shell=False, cwd="/", env=env)
2662
      output = ""
2663
      try:
2664
        output = child.stdout.read(4096)
2665
        child.stdout.close()
2666
      except EnvironmentError, err:
2667
        output += "Hook script error: %s" % str(err)
2668

    
2669
      while True:
2670
        try:
2671
          result = child.wait()
2672
          break
2673
        except EnvironmentError, err:
2674
          if err.errno == errno.EINTR:
2675
            continue
2676
          raise
2677
    finally:
2678
      # try not to leak fds
2679
      for fd in (fdstdin, ):
2680
        if fd is not None:
2681
          try:
2682
            fd.close()
2683
          except EnvironmentError, err:
2684
            # just log the error
2685
            #logging.exception("Error while closing fd %s", fd)
2686
            pass
2687

    
2688
    return result == 0, utils.SafeEncode(output.strip())
2689

    
2690
  def RunHooks(self, hpath, phase, env):
2691
    """Run the scripts in the hooks directory.
2692

2693
    @type hpath: str
2694
    @param hpath: the path to the hooks directory which
2695
        holds the scripts
2696
    @type phase: str
2697
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
2698
        L{constants.HOOKS_PHASE_POST}
2699
    @type env: dict
2700
    @param env: dictionary with the environment for the hook
2701
    @rtype: list
2702
    @return: list of 3-element tuples:
2703
      - script path
2704
      - script result, either L{constants.HKR_SUCCESS} or
2705
        L{constants.HKR_FAIL}
2706
      - output of the script
2707

2708
    @raise errors.ProgrammerError: for invalid input
2709
        parameters
2710

2711
    """
2712
    if phase == constants.HOOKS_PHASE_PRE:
2713
      suffix = "pre"
2714
    elif phase == constants.HOOKS_PHASE_POST:
2715
      suffix = "post"
2716
    else:
2717
      _Fail("Unknown hooks phase '%s'", phase)
2718

    
2719
    rr = []
2720

    
2721
    subdir = "%s-%s.d" % (hpath, suffix)
2722
    dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2723
    try:
2724
      dir_contents = utils.ListVisibleFiles(dir_name)
2725
    except OSError:
2726
      # FIXME: must log output in case of failures
2727
      return rr
2728

    
2729
    # we use the standard python sort order,
2730
    # so 00name is the recommended naming scheme
2731
    dir_contents.sort()
2732
    for relname in dir_contents:
2733
      fname = os.path.join(dir_name, relname)
2734
      if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2735
              constants.EXT_PLUGIN_MASK.match(relname) is not None):
2736
        rrval = constants.HKR_SKIP
2737
        output = ""
2738
      else:
2739
        result, output = self.ExecHook(fname, env)
2740
        if not result:
2741
          rrval = constants.HKR_FAIL
2742
        else:
2743
          rrval = constants.HKR_SUCCESS
2744
      rr.append(("%s/%s" % (subdir, relname), rrval, output))
2745

    
2746
    return rr
2747

    
2748

    
2749
class IAllocatorRunner(object):
2750
  """IAllocator runner.
2751

2752
  This class is instantiated on the node side (ganeti-noded) and not on
2753
  the master side.
2754

2755
  """
2756
  @staticmethod
2757
  def Run(name, idata):
2758
    """Run an iallocator script.
2759

2760
    @type name: str
2761
    @param name: the iallocator script name
2762
    @type idata: str
2763
    @param idata: the allocator input data
2764

2765
    @rtype: tuple
2766
    @return: two element tuple of:
2767
       - status
2768
       - either error message or stdout of allocator (for success)
2769

2770
    """
2771
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2772
                                  os.path.isfile)
2773
    if alloc_script is None:
2774
      _Fail("iallocator module '%s' not found in the search path", name)
2775

    
2776
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2777
    try:
2778
      os.write(fd, idata)
2779
      os.close(fd)
2780
      result = utils.RunCmd([alloc_script, fin_name])
2781
      if result.failed:
2782
        _Fail("iallocator module '%s' failed: %s, output '%s'",
2783
              name, result.fail_reason, result.output)
2784
    finally:
2785
      os.unlink(fin_name)
2786

    
2787
    return result.stdout
2788

    
2789

    
2790
class DevCacheManager(object):
2791
  """Simple class for managing a cache of block device information.
2792

2793
  """
2794
  _DEV_PREFIX = "/dev/"
2795
  _ROOT_DIR = constants.BDEV_CACHE_DIR
2796

    
2797
  @classmethod
2798
  def _ConvertPath(cls, dev_path):
2799
    """Converts a /dev/name path to the cache file name.
2800

2801
    This replaces slashes with underscores and strips the /dev
2802
    prefix. It then returns the full path to the cache file.
2803

2804
    @type dev_path: str
2805
    @param dev_path: the C{/dev/} path name
2806
    @rtype: str
2807
    @return: the converted path name
2808

2809
    """
2810
    if dev_path.startswith(cls._DEV_PREFIX):
2811
      dev_path = dev_path[len(cls._DEV_PREFIX):]
2812
    dev_path = dev_path.replace("/", "_")
2813
    fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2814
    return fpath
2815

    
2816
  @classmethod
2817
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2818
    """Updates the cache information for a given device.
2819

2820
    @type dev_path: str
2821
    @param dev_path: the pathname of the device
2822
    @type owner: str
2823
    @param owner: the owner (instance name) of the device
2824
    @type on_primary: bool
2825
    @param on_primary: whether this is the primary
2826
        node nor not
2827
    @type iv_name: str
2828
    @param iv_name: the instance-visible name of the
2829
        device, as in objects.Disk.iv_name
2830

2831
    @rtype: None
2832

2833
    """
2834
    if dev_path is None:
2835
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
2836
      return
2837
    fpath = cls._ConvertPath(dev_path)
2838
    if on_primary:
2839
      state = "primary"
2840
    else:
2841
      state = "secondary"
2842
    if iv_name is None:
2843
      iv_name = "not_visible"
2844
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2845
    try:
2846
      utils.WriteFile(fpath, data=fdata)
2847
    except EnvironmentError, err:
2848
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2849

    
2850
  @classmethod
2851
  def RemoveCache(cls, dev_path):
2852
    """Remove data for a dev_path.
2853

2854
    This is just a wrapper over L{utils.RemoveFile} with a converted
2855
    path name and logging.
2856

2857
    @type dev_path: str
2858
    @param dev_path: the pathname of the device
2859

2860
    @rtype: None
2861

2862
    """
2863
    if dev_path is None:
2864
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
2865
      return
2866
    fpath = cls._ConvertPath(dev_path)
2867
    try:
2868
      utils.RemoveFile(fpath)
2869
    except EnvironmentError, err:
2870
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)