Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 860bf930

History | View | Annotate | Download (127.5 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc.
5
#
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
10
#
11
# This program is distributed in the hope that it will be useful, but
12
# WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
# General Public License for more details.
15
#
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19
# 02110-1301, USA.
20

    
21

    
22
"""Functions used by the node daemon
23

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

29
"""
30

    
31
# pylint: disable=E1103
32

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

    
37

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

    
52
from ganeti import errors
53
from ganeti import utils
54
from ganeti import ssh
55
from ganeti import hypervisor
56
from ganeti import constants
57
from ganeti import bdev
58
from ganeti import objects
59
from ganeti import ssconf
60
from ganeti import serializer
61
from ganeti import netutils
62
from ganeti import runtime
63
from ganeti import compat
64
from ganeti import pathutils
65
from ganeti import vcluster
66
from ganeti import ht
67
from ganeti import hooksmaster
68

    
69

    
70
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
71
_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
72
  pathutils.DATA_DIR,
73
  pathutils.JOB_QUEUE_ARCHIVE_DIR,
74
  pathutils.QUEUE_DIR,
75
  pathutils.CRYPTO_KEYS_DIR,
76
  ])
77
_MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
78
_X509_KEY_FILE = "key"
79
_X509_CERT_FILE = "cert"
80
_IES_STATUS_FILE = "status"
81
_IES_PID_FILE = "pid"
82
_IES_CA_FILE = "ca"
83

    
84
#: Valid LVS output line regex
85
_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
86

    
87
# Actions for the master setup script
88
_MASTER_START = "start"
89
_MASTER_STOP = "stop"
90

    
91
#: Maximum file permissions for restricted command directory and executables
92
_RCMD_MAX_MODE = (stat.S_IRWXU |
93
                  stat.S_IRGRP | stat.S_IXGRP |
94
                  stat.S_IROTH | stat.S_IXOTH)
95

    
96
#: Delay before returning an error for restricted commands
97
_RCMD_INVALID_DELAY = 10
98

    
99
#: How long to wait to acquire lock for restricted commands (shorter than
100
#: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
101
#: command requests arrive
102
_RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
103

    
104

    
105
class RPCFail(Exception):
106
  """Class denoting RPC failure.
107

108
  Its argument is the error message.
109

110
  """
111

    
112

    
113
def _GetInstReasonFilename(instance_name):
114
  """Path of the file containing the reason of the instance status change.
115

116
  @type instance_name: string
117
  @param instance_name: The name of the instance
118
  @rtype: string
119
  @return: The path of the file
120

121
  """
122
  return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
123

    
124

    
125
def _StoreInstReasonTrail(instance_name, trail):
126
  """Serialize a reason trail related to an instance change of state to file.
127

128
  The exact location of the file depends on the name of the instance and on
129
  the configuration of the Ganeti cluster defined at deploy time.
130

131
  @type instance_name: string
132
  @param instance_name: The name of the instance
133
  @rtype: None
134

135
  """
136
  json = serializer.DumpJson(trail)
137
  filename = _GetInstReasonFilename(instance_name)
138
  utils.WriteFile(filename, data=json)
139

    
140

    
141
def _Fail(msg, *args, **kwargs):
142
  """Log an error and the raise an RPCFail exception.
143

144
  This exception is then handled specially in the ganeti daemon and
145
  turned into a 'failed' return type. As such, this function is a
146
  useful shortcut for logging the error and returning it to the master
147
  daemon.
148

149
  @type msg: string
150
  @param msg: the text of the exception
151
  @raise RPCFail
152

153
  """
154
  if args:
155
    msg = msg % args
156
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
157
    if "exc" in kwargs and kwargs["exc"]:
158
      logging.exception(msg)
159
    else:
160
      logging.error(msg)
161
  raise RPCFail(msg)
162

    
163

    
164
def _GetConfig():
165
  """Simple wrapper to return a SimpleStore.
166

167
  @rtype: L{ssconf.SimpleStore}
168
  @return: a SimpleStore instance
169

170
  """
171
  return ssconf.SimpleStore()
172

    
173

    
174
def _GetSshRunner(cluster_name):
175
  """Simple wrapper to return an SshRunner.
176

177
  @type cluster_name: str
178
  @param cluster_name: the cluster name, which is needed
179
      by the SshRunner constructor
180
  @rtype: L{ssh.SshRunner}
181
  @return: an SshRunner instance
182

183
  """
184
  return ssh.SshRunner(cluster_name)
185

    
186

    
187
def _Decompress(data):
188
  """Unpacks data compressed by the RPC client.
189

190
  @type data: list or tuple
191
  @param data: Data sent by RPC client
192
  @rtype: str
193
  @return: Decompressed data
194

195
  """
196
  assert isinstance(data, (list, tuple))
197
  assert len(data) == 2
198
  (encoding, content) = data
199
  if encoding == constants.RPC_ENCODING_NONE:
200
    return content
201
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
202
    return zlib.decompress(base64.b64decode(content))
203
  else:
204
    raise AssertionError("Unknown data encoding")
205

    
206

    
207
def _CleanDirectory(path, exclude=None):
208
  """Removes all regular files in a directory.
209

210
  @type path: str
211
  @param path: the directory to clean
212
  @type exclude: list
213
  @param exclude: list of files to be excluded, defaults
214
      to the empty list
215

216
  """
217
  if path not in _ALLOWED_CLEAN_DIRS:
218
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
219
          path)
220

    
221
  if not os.path.isdir(path):
222
    return
223
  if exclude is None:
224
    exclude = []
225
  else:
226
    # Normalize excluded paths
227
    exclude = [os.path.normpath(i) for i in exclude]
228

    
229
  for rel_name in utils.ListVisibleFiles(path):
230
    full_name = utils.PathJoin(path, rel_name)
231
    if full_name in exclude:
232
      continue
233
    if os.path.isfile(full_name) and not os.path.islink(full_name):
234
      utils.RemoveFile(full_name)
235

    
236

    
237
def _BuildUploadFileList():
238
  """Build the list of allowed upload files.
239

240
  This is abstracted so that it's built only once at module import time.
241

242
  """
243
  allowed_files = set([
244
    pathutils.CLUSTER_CONF_FILE,
245
    pathutils.ETC_HOSTS,
246
    pathutils.SSH_KNOWN_HOSTS_FILE,
247
    pathutils.VNC_PASSWORD_FILE,
248
    pathutils.RAPI_CERT_FILE,
249
    pathutils.SPICE_CERT_FILE,
250
    pathutils.SPICE_CACERT_FILE,
251
    pathutils.RAPI_USERS_FILE,
252
    pathutils.CONFD_HMAC_KEY,
253
    pathutils.CLUSTER_DOMAIN_SECRET_FILE,
254
    ])
255

    
256
  for hv_name in constants.HYPER_TYPES:
257
    hv_class = hypervisor.GetHypervisorClass(hv_name)
258
    allowed_files.update(hv_class.GetAncillaryFiles()[0])
259

    
260
  assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
261
    "Allowed file storage paths should never be uploaded via RPC"
262

    
263
  return frozenset(allowed_files)
264

    
265

    
266
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
267

    
268

    
269
def JobQueuePurge():
270
  """Removes job queue files and archived jobs.
271

272
  @rtype: tuple
273
  @return: True, None
274

275
  """
276
  _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
277
  _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
278

    
279

    
280
def GetMasterInfo():
281
  """Returns master information.
282

283
  This is an utility function to compute master information, either
284
  for consumption here or from the node daemon.
285

286
  @rtype: tuple
287
  @return: master_netdev, master_ip, master_name, primary_ip_family,
288
    master_netmask
289
  @raise RPCFail: in case of errors
290

291
  """
292
  try:
293
    cfg = _GetConfig()
294
    master_netdev = cfg.GetMasterNetdev()
295
    master_ip = cfg.GetMasterIP()
296
    master_netmask = cfg.GetMasterNetmask()
297
    master_node = cfg.GetMasterNode()
298
    primary_ip_family = cfg.GetPrimaryIPFamily()
299
  except errors.ConfigurationError, err:
300
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
301
  return (master_netdev, master_ip, master_node, primary_ip_family,
302
          master_netmask)
303

    
304

    
305
def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
306
  """Decorator that runs hooks before and after the decorated function.
307

308
  @type hook_opcode: string
309
  @param hook_opcode: opcode of the hook
310
  @type hooks_path: string
311
  @param hooks_path: path of the hooks
312
  @type env_builder_fn: function
313
  @param env_builder_fn: function that returns a dictionary containing the
314
    environment variables for the hooks. Will get all the parameters of the
315
    decorated function.
316
  @raise RPCFail: in case of pre-hook failure
317

318
  """
319
  def decorator(fn):
320
    def wrapper(*args, **kwargs):
321
      _, myself = ssconf.GetMasterAndMyself()
322
      nodes = ([myself], [myself])  # these hooks run locally
323

    
324
      env_fn = compat.partial(env_builder_fn, *args, **kwargs)
325

    
326
      cfg = _GetConfig()
327
      hr = HooksRunner()
328
      hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
329
                                   hr.RunLocalHooks, None, env_fn,
330
                                   logging.warning, cfg.GetClusterName(),
331
                                   cfg.GetMasterNode())
332
      hm.RunPhase(constants.HOOKS_PHASE_PRE)
333
      result = fn(*args, **kwargs)
334
      hm.RunPhase(constants.HOOKS_PHASE_POST)
335

    
336
      return result
337
    return wrapper
338
  return decorator
339

    
340

    
341
def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
342
  """Builds environment variables for master IP hooks.
343

344
  @type master_params: L{objects.MasterNetworkParameters}
345
  @param master_params: network parameters of the master
346
  @type use_external_mip_script: boolean
347
  @param use_external_mip_script: whether to use an external master IP
348
    address setup script (unused, but necessary per the implementation of the
349
    _RunLocalHooks decorator)
350

351
  """
352
  # pylint: disable=W0613
353
  ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
354
  env = {
355
    "MASTER_NETDEV": master_params.netdev,
356
    "MASTER_IP": master_params.ip,
357
    "MASTER_NETMASK": str(master_params.netmask),
358
    "CLUSTER_IP_VERSION": str(ver),
359
  }
360

    
361
  return env
362

    
363

    
364
def _RunMasterSetupScript(master_params, action, use_external_mip_script):
365
  """Execute the master IP address setup script.
366

367
  @type master_params: L{objects.MasterNetworkParameters}
368
  @param master_params: network parameters of the master
369
  @type action: string
370
  @param action: action to pass to the script. Must be one of
371
    L{backend._MASTER_START} or L{backend._MASTER_STOP}
372
  @type use_external_mip_script: boolean
373
  @param use_external_mip_script: whether to use an external master IP
374
    address setup script
375
  @raise backend.RPCFail: if there are errors during the execution of the
376
    script
377

378
  """
379
  env = _BuildMasterIpEnv(master_params)
380

    
381
  if use_external_mip_script:
382
    setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
383
  else:
384
    setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
385

    
386
  result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
387

    
388
  if result.failed:
389
    _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
390
          (action, result.exit_code, result.output), log=True)
391

    
392

    
393
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
394
               _BuildMasterIpEnv)
395
def ActivateMasterIp(master_params, use_external_mip_script):
396
  """Activate the IP address of the master daemon.
397

398
  @type master_params: L{objects.MasterNetworkParameters}
399
  @param master_params: network parameters of the master
400
  @type use_external_mip_script: boolean
401
  @param use_external_mip_script: whether to use an external master IP
402
    address setup script
403
  @raise RPCFail: in case of errors during the IP startup
404

405
  """
406
  _RunMasterSetupScript(master_params, _MASTER_START,
407
                        use_external_mip_script)
408

    
409

    
410
def StartMasterDaemons(no_voting):
411
  """Activate local node as master node.
412

413
  The function will start the master daemons (ganeti-masterd and ganeti-rapi).
414

415
  @type no_voting: boolean
416
  @param no_voting: whether to start ganeti-masterd without a node vote
417
      but still non-interactively
418
  @rtype: None
419

420
  """
421

    
422
  if no_voting:
423
    masterd_args = "--no-voting --yes-do-it"
424
  else:
425
    masterd_args = ""
426

    
427
  env = {
428
    "EXTRA_MASTERD_ARGS": masterd_args,
429
    }
430

    
431
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
432
  if result.failed:
433
    msg = "Can't start Ganeti master: %s" % result.output
434
    logging.error(msg)
435
    _Fail(msg)
436

    
437

    
438
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
439
               _BuildMasterIpEnv)
440
def DeactivateMasterIp(master_params, use_external_mip_script):
441
  """Deactivate the master IP on this node.
442

443
  @type master_params: L{objects.MasterNetworkParameters}
444
  @param master_params: network parameters of the master
445
  @type use_external_mip_script: boolean
446
  @param use_external_mip_script: whether to use an external master IP
447
    address setup script
448
  @raise RPCFail: in case of errors during the IP turndown
449

450
  """
451
  _RunMasterSetupScript(master_params, _MASTER_STOP,
452
                        use_external_mip_script)
453

    
454

    
455
def StopMasterDaemons():
456
  """Stop the master daemons on this node.
457

458
  Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
459

460
  @rtype: None
461

462
  """
463
  # TODO: log and report back to the caller the error failures; we
464
  # need to decide in which case we fail the RPC for this
465

    
466
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
467
  if result.failed:
468
    logging.error("Could not stop Ganeti master, command %s had exitcode %s"
469
                  " and error %s",
470
                  result.cmd, result.exit_code, result.output)
471

    
472

    
473
def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
474
  """Change the netmask of the master IP.
475

476
  @param old_netmask: the old value of the netmask
477
  @param netmask: the new value of the netmask
478
  @param master_ip: the master IP
479
  @param master_netdev: the master network device
480

481
  """
482
  if old_netmask == netmask:
483
    return
484

    
485
  if not netutils.IPAddress.Own(master_ip):
486
    _Fail("The master IP address is not up, not attempting to change its"
487
          " netmask")
488

    
489
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
490
                         "%s/%s" % (master_ip, netmask),
491
                         "dev", master_netdev, "label",
492
                         "%s:0" % master_netdev])
493
  if result.failed:
494
    _Fail("Could not set the new netmask on the master IP address")
495

    
496
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
497
                         "%s/%s" % (master_ip, old_netmask),
498
                         "dev", master_netdev, "label",
499
                         "%s:0" % master_netdev])
500
  if result.failed:
501
    _Fail("Could not bring down the master IP address with the old netmask")
502

    
503

    
504
def EtcHostsModify(mode, host, ip):
505
  """Modify a host entry in /etc/hosts.
506

507
  @param mode: The mode to operate. Either add or remove entry
508
  @param host: The host to operate on
509
  @param ip: The ip associated with the entry
510

511
  """
512
  if mode == constants.ETC_HOSTS_ADD:
513
    if not ip:
514
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
515
              " present")
516
    utils.AddHostToEtcHosts(host, ip)
517
  elif mode == constants.ETC_HOSTS_REMOVE:
518
    if ip:
519
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
520
              " parameter is present")
521
    utils.RemoveHostFromEtcHosts(host)
522
  else:
523
    RPCFail("Mode not supported")
524

    
525

    
526
def LeaveCluster(modify_ssh_setup):
527
  """Cleans up and remove the current node.
528

529
  This function cleans up and prepares the current node to be removed
530
  from the cluster.
531

532
  If processing is successful, then it raises an
533
  L{errors.QuitGanetiException} which is used as a special case to
534
  shutdown the node daemon.
535

536
  @param modify_ssh_setup: boolean
537

538
  """
539
  _CleanDirectory(pathutils.DATA_DIR)
540
  _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
541
  JobQueuePurge()
542

    
543
  if modify_ssh_setup:
544
    try:
545
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
546

    
547
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
548

    
549
      utils.RemoveFile(priv_key)
550
      utils.RemoveFile(pub_key)
551
    except errors.OpExecError:
552
      logging.exception("Error while processing ssh files")
553

    
554
  try:
555
    utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
556
    utils.RemoveFile(pathutils.RAPI_CERT_FILE)
557
    utils.RemoveFile(pathutils.SPICE_CERT_FILE)
558
    utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
559
    utils.RemoveFile(pathutils.NODED_CERT_FILE)
560
  except: # pylint: disable=W0702
561
    logging.exception("Error while removing cluster secrets")
562

    
563
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
564
  if result.failed:
565
    logging.error("Command %s failed with exitcode %s and error %s",
566
                  result.cmd, result.exit_code, result.output)
567

    
568
  # Raise a custom exception (handled in ganeti-noded)
569
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
570

    
571

    
572
def _GetVgInfo(name, excl_stor):
573
  """Retrieves information about a LVM volume group.
574

575
  """
576
  # TODO: GetVGInfo supports returning information for multiple VGs at once
577
  vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
578
  if vginfo:
579
    vg_free = int(round(vginfo[0][0], 0))
580
    vg_size = int(round(vginfo[0][1], 0))
581
  else:
582
    vg_free = None
583
    vg_size = None
584

    
585
  return {
586
    "name": name,
587
    "vg_free": vg_free,
588
    "vg_size": vg_size,
589
    }
590

    
591

    
592
def _GetHvInfo(name):
593
  """Retrieves node information from a hypervisor.
594

595
  The information returned depends on the hypervisor. Common items:
596

597
    - vg_size is the size of the configured volume group in MiB
598
    - vg_free is the free size of the volume group in MiB
599
    - memory_dom0 is the memory allocated for domain0 in MiB
600
    - memory_free is the currently available (free) ram in MiB
601
    - memory_total is the total number of ram in MiB
602
    - hv_version: the hypervisor version, if available
603

604
  """
605
  return hypervisor.GetHypervisor(name).GetNodeInfo()
606

    
607

    
608
def _GetNamedNodeInfo(names, fn):
609
  """Calls C{fn} for all names in C{names} and returns a dictionary.
610

611
  @rtype: None or dict
612

613
  """
614
  if names is None:
615
    return None
616
  else:
617
    return map(fn, names)
618

    
619

    
620
def GetNodeInfo(vg_names, hv_names, excl_stor):
621
  """Gives back a hash with different information about the node.
622

623
  @type vg_names: list of string
624
  @param vg_names: Names of the volume groups to ask for disk space information
625
  @type hv_names: list of string
626
  @param hv_names: Names of the hypervisors to ask for node information
627
  @type excl_stor: boolean
628
  @param excl_stor: Whether exclusive_storage is active
629
  @rtype: tuple; (string, None/dict, None/dict)
630
  @return: Tuple containing boot ID, volume group information and hypervisor
631
    information
632

633
  """
634
  bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
635
  vg_info = _GetNamedNodeInfo(vg_names, (lambda vg: _GetVgInfo(vg, excl_stor)))
636
  hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo)
637

    
638
  return (bootid, vg_info, hv_info)
639

    
640

    
641
def _CheckExclusivePvs(pvi_list):
642
  """Check that PVs are not shared among LVs
643

644
  @type pvi_list: list of L{objects.LvmPvInfo} objects
645
  @param pvi_list: information about the PVs
646

647
  @rtype: list of tuples (string, list of strings)
648
  @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
649

650
  """
651
  res = []
652
  for pvi in pvi_list:
653
    if len(pvi.lv_list) > 1:
654
      res.append((pvi.name, pvi.lv_list))
655
  return res
656

    
657

    
658
def VerifyNode(what, cluster_name):
659
  """Verify the status of the local node.
660

661
  Based on the input L{what} parameter, various checks are done on the
662
  local node.
663

664
  If the I{filelist} key is present, this list of
665
  files is checksummed and the file/checksum pairs are returned.
666

667
  If the I{nodelist} key is present, we check that we have
668
  connectivity via ssh with the target nodes (and check the hostname
669
  report).
670

671
  If the I{node-net-test} key is present, we check that we have
672
  connectivity to the given nodes via both primary IP and, if
673
  applicable, secondary IPs.
674

675
  @type what: C{dict}
676
  @param what: a dictionary of things to check:
677
      - filelist: list of files for which to compute checksums
678
      - nodelist: list of nodes we should check ssh communication with
679
      - node-net-test: list of nodes we should check node daemon port
680
        connectivity with
681
      - hypervisor: list with hypervisors to run the verify for
682
  @rtype: dict
683
  @return: a dictionary with the same keys as the input dict, and
684
      values representing the result of the checks
685

686
  """
687
  result = {}
688
  my_name = netutils.Hostname.GetSysName()
689
  port = netutils.GetDaemonPort(constants.NODED)
690
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
691

    
692
  if constants.NV_HYPERVISOR in what and vm_capable:
693
    result[constants.NV_HYPERVISOR] = tmp = {}
694
    for hv_name in what[constants.NV_HYPERVISOR]:
695
      try:
696
        val = hypervisor.GetHypervisor(hv_name).Verify()
697
      except errors.HypervisorError, err:
698
        val = "Error while checking hypervisor: %s" % str(err)
699
      tmp[hv_name] = val
700

    
701
  if constants.NV_HVPARAMS in what and vm_capable:
702
    result[constants.NV_HVPARAMS] = tmp = []
703
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
704
      try:
705
        logging.info("Validating hv %s, %s", hv_name, hvparms)
706
        hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
707
      except errors.HypervisorError, err:
708
        tmp.append((source, hv_name, str(err)))
709

    
710
  if constants.NV_FILELIST in what:
711
    fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
712
                                              what[constants.NV_FILELIST]))
713
    result[constants.NV_FILELIST] = \
714
      dict((vcluster.MakeVirtualPath(key), value)
715
           for (key, value) in fingerprints.items())
716

    
717
  if constants.NV_NODELIST in what:
718
    (nodes, bynode) = what[constants.NV_NODELIST]
719

    
720
    # Add nodes from other groups (different for each node)
721
    try:
722
      nodes.extend(bynode[my_name])
723
    except KeyError:
724
      pass
725

    
726
    # Use a random order
727
    random.shuffle(nodes)
728

    
729
    # Try to contact all nodes
730
    val = {}
731
    for node in nodes:
732
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
733
      if not success:
734
        val[node] = message
735

    
736
    result[constants.NV_NODELIST] = val
737

    
738
  if constants.NV_NODENETTEST in what:
739
    result[constants.NV_NODENETTEST] = tmp = {}
740
    my_pip = my_sip = None
741
    for name, pip, sip in what[constants.NV_NODENETTEST]:
742
      if name == my_name:
743
        my_pip = pip
744
        my_sip = sip
745
        break
746
    if not my_pip:
747
      tmp[my_name] = ("Can't find my own primary/secondary IP"
748
                      " in the node list")
749
    else:
750
      for name, pip, sip in what[constants.NV_NODENETTEST]:
751
        fail = []
752
        if not netutils.TcpPing(pip, port, source=my_pip):
753
          fail.append("primary")
754
        if sip != pip:
755
          if not netutils.TcpPing(sip, port, source=my_sip):
756
            fail.append("secondary")
757
        if fail:
758
          tmp[name] = ("failure using the %s interface(s)" %
759
                       " and ".join(fail))
760

    
761
  if constants.NV_MASTERIP in what:
762
    # FIXME: add checks on incoming data structures (here and in the
763
    # rest of the function)
764
    master_name, master_ip = what[constants.NV_MASTERIP]
765
    if master_name == my_name:
766
      source = constants.IP4_ADDRESS_LOCALHOST
767
    else:
768
      source = None
769
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
770
                                                     source=source)
771

    
772
  if constants.NV_USERSCRIPTS in what:
773
    result[constants.NV_USERSCRIPTS] = \
774
      [script for script in what[constants.NV_USERSCRIPTS]
775
       if not utils.IsExecutable(script)]
776

    
777
  if constants.NV_OOB_PATHS in what:
778
    result[constants.NV_OOB_PATHS] = tmp = []
779
    for path in what[constants.NV_OOB_PATHS]:
780
      try:
781
        st = os.stat(path)
782
      except OSError, err:
783
        tmp.append("error stating out of band helper: %s" % err)
784
      else:
785
        if stat.S_ISREG(st.st_mode):
786
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
787
            tmp.append(None)
788
          else:
789
            tmp.append("out of band helper %s is not executable" % path)
790
        else:
791
          tmp.append("out of band helper %s is not a file" % path)
792

    
793
  if constants.NV_LVLIST in what and vm_capable:
794
    try:
795
      val = GetVolumeList(utils.ListVolumeGroups().keys())
796
    except RPCFail, err:
797
      val = str(err)
798
    result[constants.NV_LVLIST] = val
799

    
800
  if constants.NV_INSTANCELIST in what and vm_capable:
801
    # GetInstanceList can fail
802
    try:
803
      val = GetInstanceList(what[constants.NV_INSTANCELIST])
804
    except RPCFail, err:
805
      val = str(err)
806
    result[constants.NV_INSTANCELIST] = val
807

    
808
  if constants.NV_VGLIST in what and vm_capable:
809
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
810

    
811
  if constants.NV_PVLIST in what and vm_capable:
812
    check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
813
    val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
814
                                       filter_allocatable=False,
815
                                       include_lvs=check_exclusive_pvs)
816
    if check_exclusive_pvs:
817
      result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
818
      for pvi in val:
819
        # Avoid sending useless data on the wire
820
        pvi.lv_list = []
821
    result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
822

    
823
  if constants.NV_VERSION in what:
824
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
825
                                    constants.RELEASE_VERSION)
826

    
827
  if constants.NV_HVINFO in what and vm_capable:
828
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
829
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
830

    
831
  if constants.NV_DRBDLIST in what and vm_capable:
832
    try:
833
      used_minors = bdev.DRBD8.GetUsedDevs().keys()
834
    except errors.BlockDeviceError, err:
835
      logging.warning("Can't get used minors list", exc_info=True)
836
      used_minors = str(err)
837
    result[constants.NV_DRBDLIST] = used_minors
838

    
839
  if constants.NV_DRBDHELPER in what and vm_capable:
840
    status = True
841
    try:
842
      payload = bdev.BaseDRBD.GetUsermodeHelper()
843
    except errors.BlockDeviceError, err:
844
      logging.error("Can't get DRBD usermode helper: %s", str(err))
845
      status = False
846
      payload = str(err)
847
    result[constants.NV_DRBDHELPER] = (status, payload)
848

    
849
  if constants.NV_NODESETUP in what:
850
    result[constants.NV_NODESETUP] = tmpr = []
851
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
852
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
853
                  " under /sys, missing required directories /sys/block"
854
                  " and /sys/class/net")
855
    if (not os.path.isdir("/proc/sys") or
856
        not os.path.isfile("/proc/sysrq-trigger")):
857
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
858
                  " under /proc, missing required directory /proc/sys and"
859
                  " the file /proc/sysrq-trigger")
860

    
861
  if constants.NV_TIME in what:
862
    result[constants.NV_TIME] = utils.SplitTime(time.time())
863

    
864
  if constants.NV_OSLIST in what and vm_capable:
865
    result[constants.NV_OSLIST] = DiagnoseOS()
866

    
867
  if constants.NV_BRIDGES in what and vm_capable:
868
    result[constants.NV_BRIDGES] = [bridge
869
                                    for bridge in what[constants.NV_BRIDGES]
870
                                    if not utils.BridgeExists(bridge)]
871

    
872
  if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
873
    result[constants.NV_FILE_STORAGE_PATHS] = \
874
      bdev.ComputeWrongFileStoragePaths()
875

    
876
  return result
877

    
878

    
879
def GetBlockDevSizes(devices):
880
  """Return the size of the given block devices
881

882
  @type devices: list
883
  @param devices: list of block device nodes to query
884
  @rtype: dict
885
  @return:
886
    dictionary of all block devices under /dev (key). The value is their
887
    size in MiB.
888

889
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
890

891
  """
892
  DEV_PREFIX = "/dev/"
893
  blockdevs = {}
894

    
895
  for devpath in devices:
896
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
897
      continue
898

    
899
    try:
900
      st = os.stat(devpath)
901
    except EnvironmentError, err:
902
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
903
      continue
904

    
905
    if stat.S_ISBLK(st.st_mode):
906
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
907
      if result.failed:
908
        # We don't want to fail, just do not list this device as available
909
        logging.warning("Cannot get size for block device %s", devpath)
910
        continue
911

    
912
      size = int(result.stdout) / (1024 * 1024)
913
      blockdevs[devpath] = size
914
  return blockdevs
915

    
916

    
917
def GetVolumeList(vg_names):
918
  """Compute list of logical volumes and their size.
919

920
  @type vg_names: list
921
  @param vg_names: the volume groups whose LVs we should list, or
922
      empty for all volume groups
923
  @rtype: dict
924
  @return:
925
      dictionary of all partions (key) with value being a tuple of
926
      their size (in MiB), inactive and online status::
927

928
        {'xenvg/test1': ('20.06', True, True)}
929

930
      in case of errors, a string is returned with the error
931
      details.
932

933
  """
934
  lvs = {}
935
  sep = "|"
936
  if not vg_names:
937
    vg_names = []
938
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
939
                         "--separator=%s" % sep,
940
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
941
  if result.failed:
942
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
943

    
944
  for line in result.stdout.splitlines():
945
    line = line.strip()
946
    match = _LVSLINE_REGEX.match(line)
947
    if not match:
948
      logging.error("Invalid line returned from lvs output: '%s'", line)
949
      continue
950
    vg_name, name, size, attr = match.groups()
951
    inactive = attr[4] == "-"
952
    online = attr[5] == "o"
953
    virtual = attr[0] == "v"
954
    if virtual:
955
      # we don't want to report such volumes as existing, since they
956
      # don't really hold data
957
      continue
958
    lvs[vg_name + "/" + name] = (size, inactive, online)
959

    
960
  return lvs
961

    
962

    
963
def ListVolumeGroups():
964
  """List the volume groups and their size.
965

966
  @rtype: dict
967
  @return: dictionary with keys volume name and values the
968
      size of the volume
969

970
  """
971
  return utils.ListVolumeGroups()
972

    
973

    
974
def NodeVolumes():
975
  """List all volumes on this node.
976

977
  @rtype: list
978
  @return:
979
    A list of dictionaries, each having four keys:
980
      - name: the logical volume name,
981
      - size: the size of the logical volume
982
      - dev: the physical device on which the LV lives
983
      - vg: the volume group to which it belongs
984

985
    In case of errors, we return an empty list and log the
986
    error.
987

988
    Note that since a logical volume can live on multiple physical
989
    volumes, the resulting list might include a logical volume
990
    multiple times.
991

992
  """
993
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
994
                         "--separator=|",
995
                         "--options=lv_name,lv_size,devices,vg_name"])
996
  if result.failed:
997
    _Fail("Failed to list logical volumes, lvs output: %s",
998
          result.output)
999

    
1000
  def parse_dev(dev):
1001
    return dev.split("(")[0]
1002

    
1003
  def handle_dev(dev):
1004
    return [parse_dev(x) for x in dev.split(",")]
1005

    
1006
  def map_line(line):
1007
    line = [v.strip() for v in line]
1008
    return [{"name": line[0], "size": line[1],
1009
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1010

    
1011
  all_devs = []
1012
  for line in result.stdout.splitlines():
1013
    if line.count("|") >= 3:
1014
      all_devs.extend(map_line(line.split("|")))
1015
    else:
1016
      logging.warning("Strange line in the output from lvs: '%s'", line)
1017
  return all_devs
1018

    
1019

    
1020
def BridgesExist(bridges_list):
1021
  """Check if a list of bridges exist on the current node.
1022

1023
  @rtype: boolean
1024
  @return: C{True} if all of them exist, C{False} otherwise
1025

1026
  """
1027
  missing = []
1028
  for bridge in bridges_list:
1029
    if not utils.BridgeExists(bridge):
1030
      missing.append(bridge)
1031

    
1032
  if missing:
1033
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
1034

    
1035

    
1036
def GetInstanceList(hypervisor_list):
1037
  """Provides a list of instances.
1038

1039
  @type hypervisor_list: list
1040
  @param hypervisor_list: the list of hypervisors to query information
1041

1042
  @rtype: list
1043
  @return: a list of all running instances on the current node
1044
    - instance1.example.com
1045
    - instance2.example.com
1046

1047
  """
1048
  results = []
1049
  for hname in hypervisor_list:
1050
    try:
1051
      names = hypervisor.GetHypervisor(hname).ListInstances()
1052
      results.extend(names)
1053
    except errors.HypervisorError, err:
1054
      _Fail("Error enumerating instances (hypervisor %s): %s",
1055
            hname, err, exc=True)
1056

    
1057
  return results
1058

    
1059

    
1060
def GetInstanceInfo(instance, hname):
1061
  """Gives back the information about an instance as a dictionary.
1062

1063
  @type instance: string
1064
  @param instance: the instance name
1065
  @type hname: string
1066
  @param hname: the hypervisor type of the instance
1067

1068
  @rtype: dict
1069
  @return: dictionary with the following keys:
1070
      - memory: memory size of instance (int)
1071
      - state: xen state of instance (string)
1072
      - time: cpu time of instance (float)
1073
      - vcpus: the number of vcpus (int)
1074

1075
  """
1076
  output = {}
1077

    
1078
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
1079
  if iinfo is not None:
1080
    output["memory"] = iinfo[2]
1081
    output["vcpus"] = iinfo[3]
1082
    output["state"] = iinfo[4]
1083
    output["time"] = iinfo[5]
1084

    
1085
  return output
1086

    
1087

    
1088
def GetInstanceMigratable(instance):
1089
  """Gives whether an instance can be migrated.
1090

1091
  @type instance: L{objects.Instance}
1092
  @param instance: object representing the instance to be checked.
1093

1094
  @rtype: tuple
1095
  @return: tuple of (result, description) where:
1096
      - result: whether the instance can be migrated or not
1097
      - description: a description of the issue, if relevant
1098

1099
  """
1100
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1101
  iname = instance.name
1102
  if iname not in hyper.ListInstances():
1103
    _Fail("Instance %s is not running", iname)
1104

    
1105
  for idx in range(len(instance.disks)):
1106
    link_name = _GetBlockDevSymlinkPath(iname, idx)
1107
    if not os.path.islink(link_name):
1108
      logging.warning("Instance %s is missing symlink %s for disk %d",
1109
                      iname, link_name, idx)
1110

    
1111

    
1112
def GetAllInstancesInfo(hypervisor_list):
1113
  """Gather data about all instances.
1114

1115
  This is the equivalent of L{GetInstanceInfo}, except that it
1116
  computes data for all instances at once, thus being faster if one
1117
  needs data about more than one instance.
1118

1119
  @type hypervisor_list: list
1120
  @param hypervisor_list: list of hypervisors to query for instance data
1121

1122
  @rtype: dict
1123
  @return: dictionary of instance: data, with data having the following keys:
1124
      - memory: memory size of instance (int)
1125
      - state: xen state of instance (string)
1126
      - time: cpu time of instance (float)
1127
      - vcpus: the number of vcpus
1128

1129
  """
1130
  output = {}
1131

    
1132
  for hname in hypervisor_list:
1133
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
1134
    if iinfo:
1135
      for name, _, memory, vcpus, state, times in iinfo:
1136
        value = {
1137
          "memory": memory,
1138
          "vcpus": vcpus,
1139
          "state": state,
1140
          "time": times,
1141
          }
1142
        if name in output:
1143
          # we only check static parameters, like memory and vcpus,
1144
          # and not state and time which can change between the
1145
          # invocations of the different hypervisors
1146
          for key in "memory", "vcpus":
1147
            if value[key] != output[name][key]:
1148
              _Fail("Instance %s is running twice"
1149
                    " with different parameters", name)
1150
        output[name] = value
1151

    
1152
  return output
1153

    
1154

    
1155
def _InstanceLogName(kind, os_name, instance, component):
1156
  """Compute the OS log filename for a given instance and operation.
1157

1158
  The instance name and os name are passed in as strings since not all
1159
  operations have these as part of an instance object.
1160

1161
  @type kind: string
1162
  @param kind: the operation type (e.g. add, import, etc.)
1163
  @type os_name: string
1164
  @param os_name: the os name
1165
  @type instance: string
1166
  @param instance: the name of the instance being imported/added/etc.
1167
  @type component: string or None
1168
  @param component: the name of the component of the instance being
1169
      transferred
1170

1171
  """
1172
  # TODO: Use tempfile.mkstemp to create unique filename
1173
  if component:
1174
    assert "/" not in component
1175
    c_msg = "-%s" % component
1176
  else:
1177
    c_msg = ""
1178
  base = ("%s-%s-%s%s-%s.log" %
1179
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1180
  return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1181

    
1182

    
1183
def InstanceOsAdd(instance, reinstall, debug):
1184
  """Add an OS to an instance.
1185

1186
  @type instance: L{objects.Instance}
1187
  @param instance: Instance whose OS is to be installed
1188
  @type reinstall: boolean
1189
  @param reinstall: whether this is an instance reinstall
1190
  @type debug: integer
1191
  @param debug: debug level, passed to the OS scripts
1192
  @rtype: None
1193

1194
  """
1195
  inst_os = OSFromDisk(instance.os)
1196

    
1197
  create_env = OSEnvironment(instance, inst_os, debug)
1198
  if reinstall:
1199
    create_env["INSTANCE_REINSTALL"] = "1"
1200

    
1201
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
1202

    
1203
  result = utils.RunCmd([inst_os.create_script], env=create_env,
1204
                        cwd=inst_os.path, output=logfile, reset_env=True)
1205
  if result.failed:
1206
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
1207
                  " output: %s", result.cmd, result.fail_reason, logfile,
1208
                  result.output)
1209
    lines = [utils.SafeEncode(val)
1210
             for val in utils.TailFile(logfile, lines=20)]
1211
    _Fail("OS create script failed (%s), last lines in the"
1212
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1213

    
1214

    
1215
def RunRenameInstance(instance, old_name, debug):
1216
  """Run the OS rename script for an instance.
1217

1218
  @type instance: L{objects.Instance}
1219
  @param instance: Instance whose OS is to be installed
1220
  @type old_name: string
1221
  @param old_name: previous instance name
1222
  @type debug: integer
1223
  @param debug: debug level, passed to the OS scripts
1224
  @rtype: boolean
1225
  @return: the success of the operation
1226

1227
  """
1228
  inst_os = OSFromDisk(instance.os)
1229

    
1230
  rename_env = OSEnvironment(instance, inst_os, debug)
1231
  rename_env["OLD_INSTANCE_NAME"] = old_name
1232

    
1233
  logfile = _InstanceLogName("rename", instance.os,
1234
                             "%s-%s" % (old_name, instance.name), None)
1235

    
1236
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1237
                        cwd=inst_os.path, output=logfile, reset_env=True)
1238

    
1239
  if result.failed:
1240
    logging.error("os create command '%s' returned error: %s output: %s",
1241
                  result.cmd, result.fail_reason, result.output)
1242
    lines = [utils.SafeEncode(val)
1243
             for val in utils.TailFile(logfile, lines=20)]
1244
    _Fail("OS rename script failed (%s), last lines in the"
1245
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1246

    
1247

    
1248
def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1249
  """Returns symlink path for block device.
1250

1251
  """
1252
  if _dir is None:
1253
    _dir = pathutils.DISK_LINKS_DIR
1254

    
1255
  return utils.PathJoin(_dir,
1256
                        ("%s%s%s" %
1257
                         (instance_name, constants.DISK_SEPARATOR, idx)))
1258

    
1259

    
1260
def _SymlinkBlockDev(instance_name, device_path, idx):
1261
  """Set up symlinks to a instance's block device.
1262

1263
  This is an auxiliary function run when an instance is start (on the primary
1264
  node) or when an instance is migrated (on the target node).
1265

1266

1267
  @param instance_name: the name of the target instance
1268
  @param device_path: path of the physical block device, on the node
1269
  @param idx: the disk index
1270
  @return: absolute path to the disk's symlink
1271

1272
  """
1273
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1274
  try:
1275
    os.symlink(device_path, link_name)
1276
  except OSError, err:
1277
    if err.errno == errno.EEXIST:
1278
      if (not os.path.islink(link_name) or
1279
          os.readlink(link_name) != device_path):
1280
        os.remove(link_name)
1281
        os.symlink(device_path, link_name)
1282
    else:
1283
      raise
1284

    
1285
  return link_name
1286

    
1287

    
1288
def _RemoveBlockDevLinks(instance_name, disks):
1289
  """Remove the block device symlinks belonging to the given instance.
1290

1291
  """
1292
  for idx, _ in enumerate(disks):
1293
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1294
    if os.path.islink(link_name):
1295
      try:
1296
        os.remove(link_name)
1297
      except OSError:
1298
        logging.exception("Can't remove symlink '%s'", link_name)
1299

    
1300

    
1301
def _GatherAndLinkBlockDevs(instance):
1302
  """Set up an instance's block device(s).
1303

1304
  This is run on the primary node at instance startup. The block
1305
  devices must be already assembled.
1306

1307
  @type instance: L{objects.Instance}
1308
  @param instance: the instance whose disks we shoul assemble
1309
  @rtype: list
1310
  @return: list of (disk_object, device_path)
1311

1312
  """
1313
  block_devices = []
1314
  for idx, disk in enumerate(instance.disks):
1315
    device = _RecursiveFindBD(disk)
1316
    if device is None:
1317
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1318
                                    str(disk))
1319
    device.Open()
1320
    try:
1321
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1322
    except OSError, e:
1323
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1324
                                    e.strerror)
1325

    
1326
    block_devices.append((disk, link_name))
1327

    
1328
  return block_devices
1329

    
1330

    
1331
def StartInstance(instance, startup_paused, reason, store_reason=True):
1332
  """Start an instance.
1333

1334
  @type instance: L{objects.Instance}
1335
  @param instance: the instance object
1336
  @type startup_paused: bool
1337
  @param instance: pause instance at startup?
1338
  @type reason: list of reasons
1339
  @param reason: the reason trail for this startup
1340
  @type store_reason: boolean
1341
  @param store_reason: whether to store the shutdown reason trail on file
1342
  @rtype: None
1343

1344
  """
1345
  running_instances = GetInstanceList([instance.hypervisor])
1346

    
1347
  if instance.name in running_instances:
1348
    logging.info("Instance %s already running, not starting", instance.name)
1349
    return
1350

    
1351
  try:
1352
    block_devices = _GatherAndLinkBlockDevs(instance)
1353
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1354
    hyper.StartInstance(instance, block_devices, startup_paused)
1355
    if store_reason:
1356
      _StoreInstReasonTrail(instance.name, reason)
1357
  except errors.BlockDeviceError, err:
1358
    _Fail("Block device error: %s", err, exc=True)
1359
  except errors.HypervisorError, err:
1360
    _RemoveBlockDevLinks(instance.name, instance.disks)
1361
    _Fail("Hypervisor error: %s", err, exc=True)
1362

    
1363

    
1364
def InstanceShutdown(instance, timeout, reason, store_reason=True):
1365
  """Shut an instance down.
1366

1367
  @note: this functions uses polling with a hardcoded timeout.
1368

1369
  @type instance: L{objects.Instance}
1370
  @param instance: the instance object
1371
  @type timeout: integer
1372
  @param timeout: maximum timeout for soft shutdown
1373
  @type reason: list of reasons
1374
  @param reason: the reason trail for this shutdown
1375
  @type store_reason: boolean
1376
  @param store_reason: whether to store the shutdown reason trail on file
1377
  @rtype: None
1378

1379
  """
1380
  hv_name = instance.hypervisor
1381
  hyper = hypervisor.GetHypervisor(hv_name)
1382
  iname = instance.name
1383

    
1384
  if instance.name not in hyper.ListInstances():
1385
    logging.info("Instance %s not running, doing nothing", iname)
1386
    return
1387

    
1388
  class _TryShutdown:
1389
    def __init__(self):
1390
      self.tried_once = False
1391

    
1392
    def __call__(self):
1393
      if iname not in hyper.ListInstances():
1394
        return
1395

    
1396
      try:
1397
        hyper.StopInstance(instance, retry=self.tried_once)
1398
        if store_reason:
1399
          _StoreInstReasonTrail(instance.name, reason)
1400
      except errors.HypervisorError, err:
1401
        if iname not in hyper.ListInstances():
1402
          # if the instance is no longer existing, consider this a
1403
          # success and go to cleanup
1404
          return
1405

    
1406
        _Fail("Failed to stop instance %s: %s", iname, err)
1407

    
1408
      self.tried_once = True
1409

    
1410
      raise utils.RetryAgain()
1411

    
1412
  try:
1413
    utils.Retry(_TryShutdown(), 5, timeout)
1414
  except utils.RetryTimeout:
1415
    # the shutdown did not succeed
1416
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1417

    
1418
    try:
1419
      hyper.StopInstance(instance, force=True)
1420
    except errors.HypervisorError, err:
1421
      if iname in hyper.ListInstances():
1422
        # only raise an error if the instance still exists, otherwise
1423
        # the error could simply be "instance ... unknown"!
1424
        _Fail("Failed to force stop instance %s: %s", iname, err)
1425

    
1426
    time.sleep(1)
1427

    
1428
    if iname in hyper.ListInstances():
1429
      _Fail("Could not shutdown instance %s even by destroy", iname)
1430

    
1431
  try:
1432
    hyper.CleanupInstance(instance.name)
1433
  except errors.HypervisorError, err:
1434
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1435

    
1436
  _RemoveBlockDevLinks(iname, instance.disks)
1437

    
1438

    
1439
def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1440
  """Reboot an instance.
1441

1442
  @type instance: L{objects.Instance}
1443
  @param instance: the instance object to reboot
1444
  @type reboot_type: str
1445
  @param reboot_type: the type of reboot, one the following
1446
    constants:
1447
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1448
        instance OS, do not recreate the VM
1449
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1450
        restart the VM (at the hypervisor level)
1451
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1452
        not accepted here, since that mode is handled differently, in
1453
        cmdlib, and translates into full stop and start of the
1454
        instance (instead of a call_instance_reboot RPC)
1455
  @type shutdown_timeout: integer
1456
  @param shutdown_timeout: maximum timeout for soft shutdown
1457
  @type reason: list of reasons
1458
  @param reason: the reason trail for this reboot
1459
  @rtype: None
1460

1461
  """
1462
  running_instances = GetInstanceList([instance.hypervisor])
1463

    
1464
  if instance.name not in running_instances:
1465
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1466

    
1467
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1468
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1469
    try:
1470
      hyper.RebootInstance(instance)
1471
    except errors.HypervisorError, err:
1472
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1473
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1474
    try:
1475
      InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1476
      result = StartInstance(instance, False, reason, store_reason=False)
1477
      _StoreInstReasonTrail(instance.name, reason)
1478
      return result
1479
    except errors.HypervisorError, err:
1480
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1481
  else:
1482
    _Fail("Invalid reboot_type received: %s", reboot_type)
1483

    
1484

    
1485
def InstanceBalloonMemory(instance, memory):
1486
  """Resize an instance's memory.
1487

1488
  @type instance: L{objects.Instance}
1489
  @param instance: the instance object
1490
  @type memory: int
1491
  @param memory: new memory amount in MB
1492
  @rtype: None
1493

1494
  """
1495
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1496
  running = hyper.ListInstances()
1497
  if instance.name not in running:
1498
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1499
    return
1500
  try:
1501
    hyper.BalloonInstanceMemory(instance, memory)
1502
  except errors.HypervisorError, err:
1503
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1504

    
1505

    
1506
def MigrationInfo(instance):
1507
  """Gather information about an instance to be migrated.
1508

1509
  @type instance: L{objects.Instance}
1510
  @param instance: the instance definition
1511

1512
  """
1513
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1514
  try:
1515
    info = hyper.MigrationInfo(instance)
1516
  except errors.HypervisorError, err:
1517
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1518
  return info
1519

    
1520

    
1521
def AcceptInstance(instance, info, target):
1522
  """Prepare the node to accept an instance.
1523

1524
  @type instance: L{objects.Instance}
1525
  @param instance: the instance definition
1526
  @type info: string/data (opaque)
1527
  @param info: migration information, from the source node
1528
  @type target: string
1529
  @param target: target host (usually ip), on this node
1530

1531
  """
1532
  # TODO: why is this required only for DTS_EXT_MIRROR?
1533
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1534
    # Create the symlinks, as the disks are not active
1535
    # in any way
1536
    try:
1537
      _GatherAndLinkBlockDevs(instance)
1538
    except errors.BlockDeviceError, err:
1539
      _Fail("Block device error: %s", err, exc=True)
1540

    
1541
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1542
  try:
1543
    hyper.AcceptInstance(instance, info, target)
1544
  except errors.HypervisorError, err:
1545
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1546
      _RemoveBlockDevLinks(instance.name, instance.disks)
1547
    _Fail("Failed to accept instance: %s", err, exc=True)
1548

    
1549

    
1550
def FinalizeMigrationDst(instance, info, success):
1551
  """Finalize any preparation to accept an instance.
1552

1553
  @type instance: L{objects.Instance}
1554
  @param instance: the instance definition
1555
  @type info: string/data (opaque)
1556
  @param info: migration information, from the source node
1557
  @type success: boolean
1558
  @param success: whether the migration was a success or a failure
1559

1560
  """
1561
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1562
  try:
1563
    hyper.FinalizeMigrationDst(instance, info, success)
1564
  except errors.HypervisorError, err:
1565
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1566

    
1567

    
1568
def MigrateInstance(instance, target, live):
1569
  """Migrates an instance to another node.
1570

1571
  @type instance: L{objects.Instance}
1572
  @param instance: the instance definition
1573
  @type target: string
1574
  @param target: the target node name
1575
  @type live: boolean
1576
  @param live: whether the migration should be done live or not (the
1577
      interpretation of this parameter is left to the hypervisor)
1578
  @raise RPCFail: if migration fails for some reason
1579

1580
  """
1581
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1582

    
1583
  try:
1584
    hyper.MigrateInstance(instance, target, live)
1585
  except errors.HypervisorError, err:
1586
    _Fail("Failed to migrate instance: %s", err, exc=True)
1587

    
1588

    
1589
def FinalizeMigrationSource(instance, success, live):
1590
  """Finalize the instance migration on the source node.
1591

1592
  @type instance: L{objects.Instance}
1593
  @param instance: the instance definition of the migrated instance
1594
  @type success: bool
1595
  @param success: whether the migration succeeded or not
1596
  @type live: bool
1597
  @param live: whether the user requested a live migration or not
1598
  @raise RPCFail: If the execution fails for some reason
1599

1600
  """
1601
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1602

    
1603
  try:
1604
    hyper.FinalizeMigrationSource(instance, success, live)
1605
  except Exception, err:  # pylint: disable=W0703
1606
    _Fail("Failed to finalize the migration on the source node: %s", err,
1607
          exc=True)
1608

    
1609

    
1610
def GetMigrationStatus(instance):
1611
  """Get the migration status
1612

1613
  @type instance: L{objects.Instance}
1614
  @param instance: the instance that is being migrated
1615
  @rtype: L{objects.MigrationStatus}
1616
  @return: the status of the current migration (one of
1617
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1618
           progress info that can be retrieved from the hypervisor
1619
  @raise RPCFail: If the migration status cannot be retrieved
1620

1621
  """
1622
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1623
  try:
1624
    return hyper.GetMigrationStatus(instance)
1625
  except Exception, err:  # pylint: disable=W0703
1626
    _Fail("Failed to get migration status: %s", err, exc=True)
1627

    
1628

    
1629
def HotplugDevice(instance, action, dev_type, device, extra, seq):
1630
  """Hotplug a device
1631

1632
  Hotplug is currently supported only for KVM Hypervisor.
1633
  @type instance: L{objects.Instance}
1634
  @param instance: the instance to which we hotplug a device
1635
  @type action: string
1636
  @param action: the hotplug action to perform
1637
  @type dev_type: string
1638
  @param dev_type: the device type to hotplug
1639
  @type device: either L{objects.NIC} or L{objects.Disk}
1640
  @param device: the device object to hotplug
1641
  @type extra: string
1642
  @param extra: extra info used by hotplug code (e.g. disk link)
1643
  @type seq: int
1644
  @param seq: the index of the device from master perspective
1645
  @raise RPCFail: in case instance does not have KVM hypervisor
1646

1647
  """
1648
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1649
  try:
1650
    hyper.HotplugSupported(instance, action, dev_type)
1651
  except (errors.HotplugError, errors.HypervisorError), err:
1652
    _Fail("Hotplug is not supported: %s", err)
1653

    
1654
  if action == constants.HOTPLUG_ACTION_ADD:
1655
    fn = hyper.HotAddDevice
1656
  elif action == constants.HOTPLUG_ACTION_REMOVE:
1657
    fn = hyper.HotDelDevice
1658
  elif action == constants.HOTPLUG_ACTION_MODIFY:
1659
    fn = hyper.HotModDevice
1660
  else:
1661
    assert action in constants.HOTPLUG_ALL_ACTIONS
1662

    
1663
  return fn(instance, dev_type, device, extra, seq)
1664

    
1665

    
1666
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1667
  """Creates a block device for an instance.
1668

1669
  @type disk: L{objects.Disk}
1670
  @param disk: the object describing the disk we should create
1671
  @type size: int
1672
  @param size: the size of the physical underlying device, in MiB
1673
  @type owner: str
1674
  @param owner: the name of the instance for which disk is created,
1675
      used for device cache data
1676
  @type on_primary: boolean
1677
  @param on_primary:  indicates if it is the primary node or not
1678
  @type info: string
1679
  @param info: string that will be sent to the physical device
1680
      creation, used for example to set (LVM) tags on LVs
1681
  @type excl_stor: boolean
1682
  @param excl_stor: Whether exclusive_storage is active
1683

1684
  @return: the new unique_id of the device (this can sometime be
1685
      computed only after creation), or None. On secondary nodes,
1686
      it's not required to return anything.
1687

1688
  """
1689
  # TODO: remove the obsolete "size" argument
1690
  # pylint: disable=W0613
1691
  clist = []
1692
  if disk.children:
1693
    for child in disk.children:
1694
      try:
1695
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1696
      except errors.BlockDeviceError, err:
1697
        _Fail("Can't assemble device %s: %s", child, err)
1698
      if on_primary or disk.AssembleOnSecondary():
1699
        # we need the children open in case the device itself has to
1700
        # be assembled
1701
        try:
1702
          # pylint: disable=E1103
1703
          crdev.Open()
1704
        except errors.BlockDeviceError, err:
1705
          _Fail("Can't make child '%s' read-write: %s", child, err)
1706
      clist.append(crdev)
1707

    
1708
  try:
1709
    device = bdev.Create(disk, clist, excl_stor)
1710
  except errors.BlockDeviceError, err:
1711
    _Fail("Can't create block device: %s", err)
1712

    
1713
  if on_primary or disk.AssembleOnSecondary():
1714
    try:
1715
      device.Assemble()
1716
    except errors.BlockDeviceError, err:
1717
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1718
    if on_primary or disk.OpenOnSecondary():
1719
      try:
1720
        device.Open(force=True)
1721
      except errors.BlockDeviceError, err:
1722
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1723
    DevCacheManager.UpdateCache(device.dev_path, owner,
1724
                                on_primary, disk.iv_name)
1725

    
1726
  device.SetInfo(info)
1727

    
1728
  return device.unique_id
1729

    
1730

    
1731
def _WipeDevice(path, offset, size):
1732
  """This function actually wipes the device.
1733

1734
  @param path: The path to the device to wipe
1735
  @param offset: The offset in MiB in the file
1736
  @param size: The size in MiB to write
1737

1738
  """
1739
  # Internal sizes are always in Mebibytes; if the following "dd" command
1740
  # should use a different block size the offset and size given to this
1741
  # function must be adjusted accordingly before being passed to "dd".
1742
  block_size = 1024 * 1024
1743

    
1744
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1745
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1746
         "count=%d" % size]
1747
  result = utils.RunCmd(cmd)
1748

    
1749
  if result.failed:
1750
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1751
          result.fail_reason, result.output)
1752

    
1753

    
1754
def BlockdevWipe(disk, offset, size):
1755
  """Wipes a block device.
1756

1757
  @type disk: L{objects.Disk}
1758
  @param disk: the disk object we want to wipe
1759
  @type offset: int
1760
  @param offset: The offset in MiB in the file
1761
  @type size: int
1762
  @param size: The size in MiB to write
1763

1764
  """
1765
  try:
1766
    rdev = _RecursiveFindBD(disk)
1767
  except errors.BlockDeviceError:
1768
    rdev = None
1769

    
1770
  if not rdev:
1771
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1772

    
1773
  # Do cross verify some of the parameters
1774
  if offset < 0:
1775
    _Fail("Negative offset")
1776
  if size < 0:
1777
    _Fail("Negative size")
1778
  if offset > rdev.size:
1779
    _Fail("Offset is bigger than device size")
1780
  if (offset + size) > rdev.size:
1781
    _Fail("The provided offset and size to wipe is bigger than device size")
1782

    
1783
  _WipeDevice(rdev.dev_path, offset, size)
1784

    
1785

    
1786
def BlockdevPauseResumeSync(disks, pause):
1787
  """Pause or resume the sync of the block device.
1788

1789
  @type disks: list of L{objects.Disk}
1790
  @param disks: the disks object we want to pause/resume
1791
  @type pause: bool
1792
  @param pause: Wheater to pause or resume
1793

1794
  """
1795
  success = []
1796
  for disk in disks:
1797
    try:
1798
      rdev = _RecursiveFindBD(disk)
1799
    except errors.BlockDeviceError:
1800
      rdev = None
1801

    
1802
    if not rdev:
1803
      success.append((False, ("Cannot change sync for device %s:"
1804
                              " device not found" % disk.iv_name)))
1805
      continue
1806

    
1807
    result = rdev.PauseResumeSync(pause)
1808

    
1809
    if result:
1810
      success.append((result, None))
1811
    else:
1812
      if pause:
1813
        msg = "Pause"
1814
      else:
1815
        msg = "Resume"
1816
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1817

    
1818
  return success
1819

    
1820

    
1821
def BlockdevRemove(disk):
1822
  """Remove a block device.
1823

1824
  @note: This is intended to be called recursively.
1825

1826
  @type disk: L{objects.Disk}
1827
  @param disk: the disk object we should remove
1828
  @rtype: boolean
1829
  @return: the success of the operation
1830

1831
  """
1832
  msgs = []
1833
  try:
1834
    rdev = _RecursiveFindBD(disk)
1835
  except errors.BlockDeviceError, err:
1836
    # probably can't attach
1837
    logging.info("Can't attach to device %s in remove", disk)
1838
    rdev = None
1839
  if rdev is not None:
1840
    r_path = rdev.dev_path
1841
    try:
1842
      rdev.Remove()
1843
    except errors.BlockDeviceError, err:
1844
      msgs.append(str(err))
1845
    if not msgs:
1846
      DevCacheManager.RemoveCache(r_path)
1847

    
1848
  if disk.children:
1849
    for child in disk.children:
1850
      try:
1851
        BlockdevRemove(child)
1852
      except RPCFail, err:
1853
        msgs.append(str(err))
1854

    
1855
  if msgs:
1856
    _Fail("; ".join(msgs))
1857

    
1858

    
1859
def _RecursiveAssembleBD(disk, owner, as_primary):
1860
  """Activate a block device for an instance.
1861

1862
  This is run on the primary and secondary nodes for an instance.
1863

1864
  @note: this function is called recursively.
1865

1866
  @type disk: L{objects.Disk}
1867
  @param disk: the disk we try to assemble
1868
  @type owner: str
1869
  @param owner: the name of the instance which owns the disk
1870
  @type as_primary: boolean
1871
  @param as_primary: if we should make the block device
1872
      read/write
1873

1874
  @return: the assembled device or None (in case no device
1875
      was assembled)
1876
  @raise errors.BlockDeviceError: in case there is an error
1877
      during the activation of the children or the device
1878
      itself
1879

1880
  """
1881
  children = []
1882
  if disk.children:
1883
    mcn = disk.ChildrenNeeded()
1884
    if mcn == -1:
1885
      mcn = 0 # max number of Nones allowed
1886
    else:
1887
      mcn = len(disk.children) - mcn # max number of Nones
1888
    for chld_disk in disk.children:
1889
      try:
1890
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1891
      except errors.BlockDeviceError, err:
1892
        if children.count(None) >= mcn:
1893
          raise
1894
        cdev = None
1895
        logging.error("Error in child activation (but continuing): %s",
1896
                      str(err))
1897
      children.append(cdev)
1898

    
1899
  if as_primary or disk.AssembleOnSecondary():
1900
    r_dev = bdev.Assemble(disk, children)
1901
    result = r_dev
1902
    if as_primary or disk.OpenOnSecondary():
1903
      r_dev.Open()
1904
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1905
                                as_primary, disk.iv_name)
1906

    
1907
  else:
1908
    result = True
1909
  return result
1910

    
1911

    
1912
def BlockdevAssemble(disk, owner, as_primary, idx):
1913
  """Activate a block device for an instance.
1914

1915
  This is a wrapper over _RecursiveAssembleBD.
1916

1917
  @rtype: str or boolean
1918
  @return: a C{/dev/...} path for primary nodes, and
1919
      C{True} for secondary nodes
1920

1921
  """
1922
  try:
1923
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1924
    if isinstance(result, bdev.BlockDev):
1925
      # pylint: disable=E1103
1926
      result = result.dev_path
1927
      if as_primary:
1928
        _SymlinkBlockDev(owner, result, idx)
1929
  except errors.BlockDeviceError, err:
1930
    _Fail("Error while assembling disk: %s", err, exc=True)
1931
  except OSError, err:
1932
    _Fail("Error while symlinking disk: %s", err, exc=True)
1933

    
1934
  return result
1935

    
1936

    
1937
def BlockdevShutdown(disk):
1938
  """Shut down a block device.
1939

1940
  First, if the device is assembled (Attach() is successful), then
1941
  the device is shutdown. Then the children of the device are
1942
  shutdown.
1943

1944
  This function is called recursively. Note that we don't cache the
1945
  children or such, as oppossed to assemble, shutdown of different
1946
  devices doesn't require that the upper device was active.
1947

1948
  @type disk: L{objects.Disk}
1949
  @param disk: the description of the disk we should
1950
      shutdown
1951
  @rtype: None
1952

1953
  """
1954
  msgs = []
1955
  r_dev = _RecursiveFindBD(disk)
1956
  if r_dev is not None:
1957
    r_path = r_dev.dev_path
1958
    try:
1959
      r_dev.Shutdown()
1960
      DevCacheManager.RemoveCache(r_path)
1961
    except errors.BlockDeviceError, err:
1962
      msgs.append(str(err))
1963

    
1964
  if disk.children:
1965
    for child in disk.children:
1966
      try:
1967
        BlockdevShutdown(child)
1968
      except RPCFail, err:
1969
        msgs.append(str(err))
1970

    
1971
  if msgs:
1972
    _Fail("; ".join(msgs))
1973

    
1974

    
1975
def BlockdevAddchildren(parent_cdev, new_cdevs):
1976
  """Extend a mirrored block device.
1977

1978
  @type parent_cdev: L{objects.Disk}
1979
  @param parent_cdev: the disk to which we should add children
1980
  @type new_cdevs: list of L{objects.Disk}
1981
  @param new_cdevs: the list of children which we should add
1982
  @rtype: None
1983

1984
  """
1985
  parent_bdev = _RecursiveFindBD(parent_cdev)
1986
  if parent_bdev is None:
1987
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1988
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1989
  if new_bdevs.count(None) > 0:
1990
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1991
  parent_bdev.AddChildren(new_bdevs)
1992

    
1993

    
1994
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1995
  """Shrink a mirrored block device.
1996

1997
  @type parent_cdev: L{objects.Disk}
1998
  @param parent_cdev: the disk from which we should remove children
1999
  @type new_cdevs: list of L{objects.Disk}
2000
  @param new_cdevs: the list of children which we should remove
2001
  @rtype: None
2002

2003
  """
2004
  parent_bdev = _RecursiveFindBD(parent_cdev)
2005
  if parent_bdev is None:
2006
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2007
  devs = []
2008
  for disk in new_cdevs:
2009
    rpath = disk.StaticDevPath()
2010
    if rpath is None:
2011
      bd = _RecursiveFindBD(disk)
2012
      if bd is None:
2013
        _Fail("Can't find device %s while removing children", disk)
2014
      else:
2015
        devs.append(bd.dev_path)
2016
    else:
2017
      if not utils.IsNormAbsPath(rpath):
2018
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2019
      devs.append(rpath)
2020
  parent_bdev.RemoveChildren(devs)
2021

    
2022

    
2023
def BlockdevGetmirrorstatus(disks):
2024
  """Get the mirroring status of a list of devices.
2025

2026
  @type disks: list of L{objects.Disk}
2027
  @param disks: the list of disks which we should query
2028
  @rtype: disk
2029
  @return: List of L{objects.BlockDevStatus}, one for each disk
2030
  @raise errors.BlockDeviceError: if any of the disks cannot be
2031
      found
2032

2033
  """
2034
  stats = []
2035
  for dsk in disks:
2036
    rbd = _RecursiveFindBD(dsk)
2037
    if rbd is None:
2038
      _Fail("Can't find device %s", dsk)
2039

    
2040
    stats.append(rbd.CombinedSyncStatus())
2041

    
2042
  return stats
2043

    
2044

    
2045
def BlockdevGetmirrorstatusMulti(disks):
2046
  """Get the mirroring status of a list of devices.
2047

2048
  @type disks: list of L{objects.Disk}
2049
  @param disks: the list of disks which we should query
2050
  @rtype: disk
2051
  @return: List of tuples, (bool, status), one for each disk; bool denotes
2052
    success/failure, status is L{objects.BlockDevStatus} on success, string
2053
    otherwise
2054

2055
  """
2056
  result = []
2057
  for disk in disks:
2058
    try:
2059
      rbd = _RecursiveFindBD(disk)
2060
      if rbd is None:
2061
        result.append((False, "Can't find device %s" % disk))
2062
        continue
2063

    
2064
      status = rbd.CombinedSyncStatus()
2065
    except errors.BlockDeviceError, err:
2066
      logging.exception("Error while getting disk status")
2067
      result.append((False, str(err)))
2068
    else:
2069
      result.append((True, status))
2070

    
2071
  assert len(disks) == len(result)
2072

    
2073
  return result
2074

    
2075

    
2076
def _RecursiveFindBD(disk):
2077
  """Check if a device is activated.
2078

2079
  If so, return information about the real device.
2080

2081
  @type disk: L{objects.Disk}
2082
  @param disk: the disk object we need to find
2083

2084
  @return: None if the device can't be found,
2085
      otherwise the device instance
2086

2087
  """
2088
  children = []
2089
  if disk.children:
2090
    for chdisk in disk.children:
2091
      children.append(_RecursiveFindBD(chdisk))
2092

    
2093
  return bdev.FindDevice(disk, children)
2094

    
2095

    
2096
def _OpenRealBD(disk):
2097
  """Opens the underlying block device of a disk.
2098

2099
  @type disk: L{objects.Disk}
2100
  @param disk: the disk object we want to open
2101

2102
  """
2103
  real_disk = _RecursiveFindBD(disk)
2104
  if real_disk is None:
2105
    _Fail("Block device '%s' is not set up", disk)
2106

    
2107
  real_disk.Open()
2108

    
2109
  return real_disk
2110

    
2111

    
2112
def BlockdevFind(disk):
2113
  """Check if a device is activated.
2114

2115
  If it is, return information about the real device.
2116

2117
  @type disk: L{objects.Disk}
2118
  @param disk: the disk to find
2119
  @rtype: None or objects.BlockDevStatus
2120
  @return: None if the disk cannot be found, otherwise a the current
2121
           information
2122

2123
  """
2124
  try:
2125
    rbd = _RecursiveFindBD(disk)
2126
  except errors.BlockDeviceError, err:
2127
    _Fail("Failed to find device: %s", err, exc=True)
2128

    
2129
  if rbd is None:
2130
    return None
2131

    
2132
  return rbd.GetSyncStatus()
2133

    
2134

    
2135
def BlockdevGetsize(disks):
2136
  """Computes the size of the given disks.
2137

2138
  If a disk is not found, returns None instead.
2139

2140
  @type disks: list of L{objects.Disk}
2141
  @param disks: the list of disk to compute the size for
2142
  @rtype: list
2143
  @return: list with elements None if the disk cannot be found,
2144
      otherwise the size
2145

2146
  """
2147
  result = []
2148
  for cf in disks:
2149
    try:
2150
      rbd = _RecursiveFindBD(cf)
2151
    except errors.BlockDeviceError:
2152
      result.append(None)
2153
      continue
2154
    if rbd is None:
2155
      result.append(None)
2156
    else:
2157
      result.append(rbd.GetActualSize())
2158
  return result
2159

    
2160

    
2161
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2162
  """Export a block device to a remote node.
2163

2164
  @type disk: L{objects.Disk}
2165
  @param disk: the description of the disk to export
2166
  @type dest_node: str
2167
  @param dest_node: the destination node to export to
2168
  @type dest_path: str
2169
  @param dest_path: the destination path on the target node
2170
  @type cluster_name: str
2171
  @param cluster_name: the cluster name, needed for SSH hostalias
2172
  @rtype: None
2173

2174
  """
2175
  real_disk = _OpenRealBD(disk)
2176

    
2177
  # the block size on the read dd is 1MiB to match our units
2178
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2179
                               "dd if=%s bs=1048576 count=%s",
2180
                               real_disk.dev_path, str(disk.size))
2181

    
2182
  # we set here a smaller block size as, due to ssh buffering, more
2183
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2184
  # device is not already there or we pass a wrong path; we use
2185
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2186
  # to not buffer too much memory; this means that at best, we flush
2187
  # every 64k, which will not be very fast
2188
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2189
                                " oflag=dsync", dest_path)
2190

    
2191
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2192
                                                   constants.SSH_LOGIN_USER,
2193
                                                   destcmd)
2194

    
2195
  # all commands have been checked, so we're safe to combine them
2196
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2197

    
2198
  result = utils.RunCmd(["bash", "-c", command])
2199

    
2200
  if result.failed:
2201
    _Fail("Disk copy command '%s' returned error: %s"
2202
          " output: %s", command, result.fail_reason, result.output)
2203

    
2204

    
2205
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2206
  """Write a file to the filesystem.
2207

2208
  This allows the master to overwrite(!) a file. It will only perform
2209
  the operation if the file belongs to a list of configuration files.
2210

2211
  @type file_name: str
2212
  @param file_name: the target file name
2213
  @type data: str
2214
  @param data: the new contents of the file
2215
  @type mode: int
2216
  @param mode: the mode to give the file (can be None)
2217
  @type uid: string
2218
  @param uid: the owner of the file
2219
  @type gid: string
2220
  @param gid: the group of the file
2221
  @type atime: float
2222
  @param atime: the atime to set on the file (can be None)
2223
  @type mtime: float
2224
  @param mtime: the mtime to set on the file (can be None)
2225
  @rtype: None
2226

2227
  """
2228
  file_name = vcluster.LocalizeVirtualPath(file_name)
2229

    
2230
  if not os.path.isabs(file_name):
2231
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2232

    
2233
  if file_name not in _ALLOWED_UPLOAD_FILES:
2234
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2235
          file_name)
2236

    
2237
  raw_data = _Decompress(data)
2238

    
2239
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2240
    _Fail("Invalid username/groupname type")
2241

    
2242
  getents = runtime.GetEnts()
2243
  uid = getents.LookupUser(uid)
2244
  gid = getents.LookupGroup(gid)
2245

    
2246
  utils.SafeWriteFile(file_name, None,
2247
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2248
                      atime=atime, mtime=mtime)
2249

    
2250

    
2251
def RunOob(oob_program, command, node, timeout):
2252
  """Executes oob_program with given command on given node.
2253

2254
  @param oob_program: The path to the executable oob_program
2255
  @param command: The command to invoke on oob_program
2256
  @param node: The node given as an argument to the program
2257
  @param timeout: Timeout after which we kill the oob program
2258

2259
  @return: stdout
2260
  @raise RPCFail: If execution fails for some reason
2261

2262
  """
2263
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2264

    
2265
  if result.failed:
2266
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2267
          result.fail_reason, result.output)
2268

    
2269
  return result.stdout
2270

    
2271

    
2272
def _OSOndiskAPIVersion(os_dir):
2273
  """Compute and return the API version of a given OS.
2274

2275
  This function will try to read the API version of the OS residing in
2276
  the 'os_dir' directory.
2277

2278
  @type os_dir: str
2279
  @param os_dir: the directory in which we should look for the OS
2280
  @rtype: tuple
2281
  @return: tuple (status, data) with status denoting the validity and
2282
      data holding either the vaid versions or an error message
2283

2284
  """
2285
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2286

    
2287
  try:
2288
    st = os.stat(api_file)
2289
  except EnvironmentError, err:
2290
    return False, ("Required file '%s' not found under path %s: %s" %
2291
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2292

    
2293
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2294
    return False, ("File '%s' in %s is not a regular file" %
2295
                   (constants.OS_API_FILE, os_dir))
2296

    
2297
  try:
2298
    api_versions = utils.ReadFile(api_file).splitlines()
2299
  except EnvironmentError, err:
2300
    return False, ("Error while reading the API version file at %s: %s" %
2301
                   (api_file, utils.ErrnoOrStr(err)))
2302

    
2303
  try:
2304
    api_versions = [int(version.strip()) for version in api_versions]
2305
  except (TypeError, ValueError), err:
2306
    return False, ("API version(s) can't be converted to integer: %s" %
2307
                   str(err))
2308

    
2309
  return True, api_versions
2310

    
2311

    
2312
def DiagnoseOS(top_dirs=None):
2313
  """Compute the validity for all OSes.
2314

2315
  @type top_dirs: list
2316
  @param top_dirs: the list of directories in which to
2317
      search (if not given defaults to
2318
      L{pathutils.OS_SEARCH_PATH})
2319
  @rtype: list of L{objects.OS}
2320
  @return: a list of tuples (name, path, status, diagnose, variants,
2321
      parameters, api_version) for all (potential) OSes under all
2322
      search paths, where:
2323
          - name is the (potential) OS name
2324
          - path is the full path to the OS
2325
          - status True/False is the validity of the OS
2326
          - diagnose is the error message for an invalid OS, otherwise empty
2327
          - variants is a list of supported OS variants, if any
2328
          - parameters is a list of (name, help) parameters, if any
2329
          - api_version is a list of support OS API versions
2330

2331
  """
2332
  if top_dirs is None:
2333
    top_dirs = pathutils.OS_SEARCH_PATH
2334

    
2335
  result = []
2336
  for dir_name in top_dirs:
2337
    if os.path.isdir(dir_name):
2338
      try:
2339
        f_names = utils.ListVisibleFiles(dir_name)
2340
      except EnvironmentError, err:
2341
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2342
        break
2343
      for name in f_names:
2344
        os_path = utils.PathJoin(dir_name, name)
2345
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2346
        if status:
2347
          diagnose = ""
2348
          variants = os_inst.supported_variants
2349
          parameters = os_inst.supported_parameters
2350
          api_versions = os_inst.api_versions
2351
        else:
2352
          diagnose = os_inst
2353
          variants = parameters = api_versions = []
2354
        result.append((name, os_path, status, diagnose, variants,
2355
                       parameters, api_versions))
2356

    
2357
  return result
2358

    
2359

    
2360
def _TryOSFromDisk(name, base_dir=None):
2361
  """Create an OS instance from disk.
2362

2363
  This function will return an OS instance if the given name is a
2364
  valid OS name.
2365

2366
  @type base_dir: string
2367
  @keyword base_dir: Base directory containing OS installations.
2368
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2369
  @rtype: tuple
2370
  @return: success and either the OS instance if we find a valid one,
2371
      or error message
2372

2373
  """
2374
  if base_dir is None:
2375
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2376
  else:
2377
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2378

    
2379
  if os_dir is None:
2380
    return False, "Directory for OS %s not found in search path" % name
2381

    
2382
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2383
  if not status:
2384
    # push the error up
2385
    return status, api_versions
2386

    
2387
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2388
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2389
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2390

    
2391
  # OS Files dictionary, we will populate it with the absolute path
2392
  # names; if the value is True, then it is a required file, otherwise
2393
  # an optional one
2394
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2395

    
2396
  if max(api_versions) >= constants.OS_API_V15:
2397
    os_files[constants.OS_VARIANTS_FILE] = False
2398

    
2399
  if max(api_versions) >= constants.OS_API_V20:
2400
    os_files[constants.OS_PARAMETERS_FILE] = True
2401
  else:
2402
    del os_files[constants.OS_SCRIPT_VERIFY]
2403

    
2404
  for (filename, required) in os_files.items():
2405
    os_files[filename] = utils.PathJoin(os_dir, filename)
2406

    
2407
    try:
2408
      st = os.stat(os_files[filename])
2409
    except EnvironmentError, err:
2410
      if err.errno == errno.ENOENT and not required:
2411
        del os_files[filename]
2412
        continue
2413
      return False, ("File '%s' under path '%s' is missing (%s)" %
2414
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2415

    
2416
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2417
      return False, ("File '%s' under path '%s' is not a regular file" %
2418
                     (filename, os_dir))
2419

    
2420
    if filename in constants.OS_SCRIPTS:
2421
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2422
        return False, ("File '%s' under path '%s' is not executable" %
2423
                       (filename, os_dir))
2424

    
2425
  variants = []
2426
  if constants.OS_VARIANTS_FILE in os_files:
2427
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2428
    try:
2429
      variants = \
2430
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2431
    except EnvironmentError, err:
2432
      # we accept missing files, but not other errors
2433
      if err.errno != errno.ENOENT:
2434
        return False, ("Error while reading the OS variants file at %s: %s" %
2435
                       (variants_file, utils.ErrnoOrStr(err)))
2436

    
2437
  parameters = []
2438
  if constants.OS_PARAMETERS_FILE in os_files:
2439
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2440
    try:
2441
      parameters = utils.ReadFile(parameters_file).splitlines()
2442
    except EnvironmentError, err:
2443
      return False, ("Error while reading the OS parameters file at %s: %s" %
2444
                     (parameters_file, utils.ErrnoOrStr(err)))
2445
    parameters = [v.split(None, 1) for v in parameters]
2446

    
2447
  os_obj = objects.OS(name=name, path=os_dir,
2448
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2449
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2450
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2451
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2452
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2453
                                                 None),
2454
                      supported_variants=variants,
2455
                      supported_parameters=parameters,
2456
                      api_versions=api_versions)
2457
  return True, os_obj
2458

    
2459

    
2460
def OSFromDisk(name, base_dir=None):
2461
  """Create an OS instance from disk.
2462

2463
  This function will return an OS instance if the given name is a
2464
  valid OS name. Otherwise, it will raise an appropriate
2465
  L{RPCFail} exception, detailing why this is not a valid OS.
2466

2467
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2468
  an exception but returns true/false status data.
2469

2470
  @type base_dir: string
2471
  @keyword base_dir: Base directory containing OS installations.
2472
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2473
  @rtype: L{objects.OS}
2474
  @return: the OS instance if we find a valid one
2475
  @raise RPCFail: if we don't find a valid OS
2476

2477
  """
2478
  name_only = objects.OS.GetName(name)
2479
  status, payload = _TryOSFromDisk(name_only, base_dir)
2480

    
2481
  if not status:
2482
    _Fail(payload)
2483

    
2484
  return payload
2485

    
2486

    
2487
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2488
  """Calculate the basic environment for an os script.
2489

2490
  @type os_name: str
2491
  @param os_name: full operating system name (including variant)
2492
  @type inst_os: L{objects.OS}
2493
  @param inst_os: operating system for which the environment is being built
2494
  @type os_params: dict
2495
  @param os_params: the OS parameters
2496
  @type debug: integer
2497
  @param debug: debug level (0 or 1, for OS Api 10)
2498
  @rtype: dict
2499
  @return: dict of environment variables
2500
  @raise errors.BlockDeviceError: if the block device
2501
      cannot be found
2502

2503
  """
2504
  result = {}
2505
  api_version = \
2506
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2507
  result["OS_API_VERSION"] = "%d" % api_version
2508
  result["OS_NAME"] = inst_os.name
2509
  result["DEBUG_LEVEL"] = "%d" % debug
2510

    
2511
  # OS variants
2512
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2513
    variant = objects.OS.GetVariant(os_name)
2514
    if not variant:
2515
      variant = inst_os.supported_variants[0]
2516
  else:
2517
    variant = ""
2518
  result["OS_VARIANT"] = variant
2519

    
2520
  # OS params
2521
  for pname, pvalue in os_params.items():
2522
    result["OSP_%s" % pname.upper()] = pvalue
2523

    
2524
  # Set a default path otherwise programs called by OS scripts (or
2525
  # even hooks called from OS scripts) might break, and we don't want
2526
  # to have each script require setting a PATH variable
2527
  result["PATH"] = constants.HOOKS_PATH
2528

    
2529
  return result
2530

    
2531

    
2532
def OSEnvironment(instance, inst_os, debug=0):
2533
  """Calculate the environment for an os script.
2534

2535
  @type instance: L{objects.Instance}
2536
  @param instance: target instance for the os script run
2537
  @type inst_os: L{objects.OS}
2538
  @param inst_os: operating system for which the environment is being built
2539
  @type debug: integer
2540
  @param debug: debug level (0 or 1, for OS Api 10)
2541
  @rtype: dict
2542
  @return: dict of environment variables
2543
  @raise errors.BlockDeviceError: if the block device
2544
      cannot be found
2545

2546
  """
2547
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2548

    
2549
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2550
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2551

    
2552
  result["HYPERVISOR"] = instance.hypervisor
2553
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2554
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2555
  result["INSTANCE_SECONDARY_NODES"] = \
2556
      ("%s" % " ".join(instance.secondary_nodes))
2557

    
2558
  # Disks
2559
  for idx, disk in enumerate(instance.disks):
2560
    real_disk = _OpenRealBD(disk)
2561
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2562
    result["DISK_%d_ACCESS" % idx] = disk.mode
2563
    result["DISK_%d_UUID" % idx] = disk.uuid
2564
    if disk.name:
2565
      result["DISK_%d_NAME" % idx] = disk.name
2566
    if constants.HV_DISK_TYPE in instance.hvparams:
2567
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2568
        instance.hvparams[constants.HV_DISK_TYPE]
2569
    if disk.dev_type in constants.LDS_BLOCK:
2570
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2571
    elif disk.dev_type == constants.LD_FILE:
2572
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2573
        "file:%s" % disk.physical_id[0]
2574

    
2575
  # NICs
2576
  for idx, nic in enumerate(instance.nics):
2577
    result["NIC_%d_MAC" % idx] = nic.mac
2578
    result["NIC_%d_UUID" % idx] = nic.uuid
2579
    if nic.name:
2580
      result["NIC_%d_NAME" % idx] = nic.name
2581
    if nic.ip:
2582
      result["NIC_%d_IP" % idx] = nic.ip
2583
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2584
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2585
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2586
    if nic.nicparams[constants.NIC_LINK]:
2587
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2588
    if nic.netinfo:
2589
      nobj = objects.Network.FromDict(nic.netinfo)
2590
      result.update(nobj.HooksDict("NIC_%d_" % idx))
2591
    if constants.HV_NIC_TYPE in instance.hvparams:
2592
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2593
        instance.hvparams[constants.HV_NIC_TYPE]
2594

    
2595
  # HV/BE params
2596
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2597
    for key, value in source.items():
2598
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2599

    
2600
  return result
2601

    
2602

    
2603
def DiagnoseExtStorage(top_dirs=None):
2604
  """Compute the validity for all ExtStorage Providers.
2605

2606
  @type top_dirs: list
2607
  @param top_dirs: the list of directories in which to
2608
      search (if not given defaults to
2609
      L{pathutils.ES_SEARCH_PATH})
2610
  @rtype: list of L{objects.ExtStorage}
2611
  @return: a list of tuples (name, path, status, diagnose, parameters)
2612
      for all (potential) ExtStorage Providers under all
2613
      search paths, where:
2614
          - name is the (potential) ExtStorage Provider
2615
          - path is the full path to the ExtStorage Provider
2616
          - status True/False is the validity of the ExtStorage Provider
2617
          - diagnose is the error message for an invalid ExtStorage Provider,
2618
            otherwise empty
2619
          - parameters is a list of (name, help) parameters, if any
2620

2621
  """
2622
  if top_dirs is None:
2623
    top_dirs = pathutils.ES_SEARCH_PATH
2624

    
2625
  result = []
2626
  for dir_name in top_dirs:
2627
    if os.path.isdir(dir_name):
2628
      try:
2629
        f_names = utils.ListVisibleFiles(dir_name)
2630
      except EnvironmentError, err:
2631
        logging.exception("Can't list the ExtStorage directory %s: %s",
2632
                          dir_name, err)
2633
        break
2634
      for name in f_names:
2635
        es_path = utils.PathJoin(dir_name, name)
2636
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2637
        if status:
2638
          diagnose = ""
2639
          parameters = es_inst.supported_parameters
2640
        else:
2641
          diagnose = es_inst
2642
          parameters = []
2643
        result.append((name, es_path, status, diagnose, parameters))
2644

    
2645
  return result
2646

    
2647

    
2648
def BlockdevGrow(disk, amount, dryrun, backingstore):
2649
  """Grow a stack of block devices.
2650

2651
  This function is called recursively, with the childrens being the
2652
  first ones to resize.
2653

2654
  @type disk: L{objects.Disk}
2655
  @param disk: the disk to be grown
2656
  @type amount: integer
2657
  @param amount: the amount (in mebibytes) to grow with
2658
  @type dryrun: boolean
2659
  @param dryrun: whether to execute the operation in simulation mode
2660
      only, without actually increasing the size
2661
  @param backingstore: whether to execute the operation on backing storage
2662
      only, or on "logical" storage only; e.g. DRBD is logical storage,
2663
      whereas LVM, file, RBD are backing storage
2664
  @rtype: (status, result)
2665
  @return: a tuple with the status of the operation (True/False), and
2666
      the errors message if status is False
2667

2668
  """
2669
  r_dev = _RecursiveFindBD(disk)
2670
  if r_dev is None:
2671
    _Fail("Cannot find block device %s", disk)
2672

    
2673
  try:
2674
    r_dev.Grow(amount, dryrun, backingstore)
2675
  except errors.BlockDeviceError, err:
2676
    _Fail("Failed to grow block device: %s", err, exc=True)
2677

    
2678

    
2679
def BlockdevSnapshot(disk):
2680
  """Create a snapshot copy of a block device.
2681

2682
  This function is called recursively, and the snapshot is actually created
2683
  just for the leaf lvm backend device.
2684

2685
  @type disk: L{objects.Disk}
2686
  @param disk: the disk to be snapshotted
2687
  @rtype: string
2688
  @return: snapshot disk ID as (vg, lv)
2689

2690
  """
2691
  if disk.dev_type == constants.LD_DRBD8:
2692
    if not disk.children:
2693
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2694
            disk.unique_id)
2695
    return BlockdevSnapshot(disk.children[0])
2696
  elif disk.dev_type == constants.LD_LV:
2697
    r_dev = _RecursiveFindBD(disk)
2698
    if r_dev is not None:
2699
      # FIXME: choose a saner value for the snapshot size
2700
      # let's stay on the safe side and ask for the full size, for now
2701
      return r_dev.Snapshot(disk.size)
2702
    else:
2703
      _Fail("Cannot find block device %s", disk)
2704
  else:
2705
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2706
          disk.unique_id, disk.dev_type)
2707

    
2708

    
2709
def BlockdevSetInfo(disk, info):
2710
  """Sets 'metadata' information on block devices.
2711

2712
  This function sets 'info' metadata on block devices. Initial
2713
  information is set at device creation; this function should be used
2714
  for example after renames.
2715

2716
  @type disk: L{objects.Disk}
2717
  @param disk: the disk to be grown
2718
  @type info: string
2719
  @param info: new 'info' metadata
2720
  @rtype: (status, result)
2721
  @return: a tuple with the status of the operation (True/False), and
2722
      the errors message if status is False
2723

2724
  """
2725
  r_dev = _RecursiveFindBD(disk)
2726
  if r_dev is None:
2727
    _Fail("Cannot find block device %s", disk)
2728

    
2729
  try:
2730
    r_dev.SetInfo(info)
2731
  except errors.BlockDeviceError, err:
2732
    _Fail("Failed to set information on block device: %s", err, exc=True)
2733

    
2734

    
2735
def FinalizeExport(instance, snap_disks):
2736
  """Write out the export configuration information.
2737

2738
  @type instance: L{objects.Instance}
2739
  @param instance: the instance which we export, used for
2740
      saving configuration
2741
  @type snap_disks: list of L{objects.Disk}
2742
  @param snap_disks: list of snapshot block devices, which
2743
      will be used to get the actual name of the dump file
2744

2745
  @rtype: None
2746

2747
  """
2748
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2749
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2750

    
2751
  config = objects.SerializableConfigParser()
2752

    
2753
  config.add_section(constants.INISECT_EXP)
2754
  config.set(constants.INISECT_EXP, "version", "0")
2755
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2756
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
2757
  config.set(constants.INISECT_EXP, "os", instance.os)
2758
  config.set(constants.INISECT_EXP, "compression", "none")
2759

    
2760
  config.add_section(constants.INISECT_INS)
2761
  config.set(constants.INISECT_INS, "name", instance.name)
2762
  config.set(constants.INISECT_INS, "maxmem", "%d" %
2763
             instance.beparams[constants.BE_MAXMEM])
2764
  config.set(constants.INISECT_INS, "minmem", "%d" %
2765
             instance.beparams[constants.BE_MINMEM])
2766
  # "memory" is deprecated, but useful for exporting to old ganeti versions
2767
  config.set(constants.INISECT_INS, "memory", "%d" %
2768
             instance.beparams[constants.BE_MAXMEM])
2769
  config.set(constants.INISECT_INS, "vcpus", "%d" %
2770
             instance.beparams[constants.BE_VCPUS])
2771
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2772
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2773
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2774

    
2775
  nic_total = 0
2776
  for nic_count, nic in enumerate(instance.nics):
2777
    nic_total += 1
2778
    config.set(constants.INISECT_INS, "nic%d_mac" %
2779
               nic_count, "%s" % nic.mac)
2780
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2781
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2782
               "%s" % nic.network)
2783
    for param in constants.NICS_PARAMETER_TYPES:
2784
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2785
                 "%s" % nic.nicparams.get(param, None))
2786
  # TODO: redundant: on load can read nics until it doesn't exist
2787
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2788

    
2789
  disk_total = 0
2790
  for disk_count, disk in enumerate(snap_disks):
2791
    if disk:
2792
      disk_total += 1
2793
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2794
                 ("%s" % disk.iv_name))
2795
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2796
                 ("%s" % disk.physical_id[1]))
2797
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2798
                 ("%d" % disk.size))
2799

    
2800
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2801

    
2802
  # New-style hypervisor/backend parameters
2803

    
2804
  config.add_section(constants.INISECT_HYP)
2805
  for name, value in instance.hvparams.items():
2806
    if name not in constants.HVC_GLOBALS:
2807
      config.set(constants.INISECT_HYP, name, str(value))
2808

    
2809
  config.add_section(constants.INISECT_BEP)
2810
  for name, value in instance.beparams.items():
2811
    config.set(constants.INISECT_BEP, name, str(value))
2812

    
2813
  config.add_section(constants.INISECT_OSP)
2814
  for name, value in instance.osparams.items():
2815
    config.set(constants.INISECT_OSP, name, str(value))
2816

    
2817
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2818
                  data=config.Dumps())
2819
  shutil.rmtree(finaldestdir, ignore_errors=True)
2820
  shutil.move(destdir, finaldestdir)
2821

    
2822

    
2823
def ExportInfo(dest):
2824
  """Get export configuration information.
2825

2826
  @type dest: str
2827
  @param dest: directory containing the export
2828

2829
  @rtype: L{objects.SerializableConfigParser}
2830
  @return: a serializable config file containing the
2831
      export info
2832

2833
  """
2834
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2835

    
2836
  config = objects.SerializableConfigParser()
2837
  config.read(cff)
2838

    
2839
  if (not config.has_section(constants.INISECT_EXP) or
2840
      not config.has_section(constants.INISECT_INS)):
2841
    _Fail("Export info file doesn't have the required fields")
2842

    
2843
  return config.Dumps()
2844

    
2845

    
2846
def ListExports():
2847
  """Return a list of exports currently available on this machine.
2848

2849
  @rtype: list
2850
  @return: list of the exports
2851

2852
  """
2853
  if os.path.isdir(pathutils.EXPORT_DIR):
2854
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
2855
  else:
2856
    _Fail("No exports directory")
2857

    
2858

    
2859
def RemoveExport(export):
2860
  """Remove an existing export from the node.
2861

2862
  @type export: str
2863
  @param export: the name of the export to remove
2864
  @rtype: None
2865

2866
  """
2867
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
2868

    
2869
  try:
2870
    shutil.rmtree(target)
2871
  except EnvironmentError, err:
2872
    _Fail("Error while removing the export: %s", err, exc=True)
2873

    
2874

    
2875
def BlockdevRename(devlist):
2876
  """Rename a list of block devices.
2877

2878
  @type devlist: list of tuples
2879
  @param devlist: list of tuples of the form  (disk,
2880
      new_logical_id, new_physical_id); disk is an
2881
      L{objects.Disk} object describing the current disk,
2882
      and new logical_id/physical_id is the name we
2883
      rename it to
2884
  @rtype: boolean
2885
  @return: True if all renames succeeded, False otherwise
2886

2887
  """
2888
  msgs = []
2889
  result = True
2890
  for disk, unique_id in devlist:
2891
    dev = _RecursiveFindBD(disk)
2892
    if dev is None:
2893
      msgs.append("Can't find device %s in rename" % str(disk))
2894
      result = False
2895
      continue
2896
    try:
2897
      old_rpath = dev.dev_path
2898
      dev.Rename(unique_id)
2899
      new_rpath = dev.dev_path
2900
      if old_rpath != new_rpath:
2901
        DevCacheManager.RemoveCache(old_rpath)
2902
        # FIXME: we should add the new cache information here, like:
2903
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2904
        # but we don't have the owner here - maybe parse from existing
2905
        # cache? for now, we only lose lvm data when we rename, which
2906
        # is less critical than DRBD or MD
2907
    except errors.BlockDeviceError, err:
2908
      msgs.append("Can't rename device '%s' to '%s': %s" %
2909
                  (dev, unique_id, err))
2910
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2911
      result = False
2912
  if not result:
2913
    _Fail("; ".join(msgs))
2914

    
2915

    
2916
def _TransformFileStorageDir(fs_dir):
2917
  """Checks whether given file_storage_dir is valid.
2918

2919
  Checks wheter the given fs_dir is within the cluster-wide default
2920
  file_storage_dir or the shared_file_storage_dir, which are stored in
2921
  SimpleStore. Only paths under those directories are allowed.
2922

2923
  @type fs_dir: str
2924
  @param fs_dir: the path to check
2925

2926
  @return: the normalized path if valid, None otherwise
2927

2928
  """
2929
  if not (constants.ENABLE_FILE_STORAGE or
2930
          constants.ENABLE_SHARED_FILE_STORAGE):
2931
    _Fail("File storage disabled at configure time")
2932

    
2933
  bdev.CheckFileStoragePath(fs_dir)
2934

    
2935
  return os.path.normpath(fs_dir)
2936

    
2937

    
2938
def CreateFileStorageDir(file_storage_dir):
2939
  """Create file storage directory.
2940

2941
  @type file_storage_dir: str
2942
  @param file_storage_dir: directory to create
2943

2944
  @rtype: tuple
2945
  @return: tuple with first element a boolean indicating wheter dir
2946
      creation was successful or not
2947

2948
  """
2949
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2950
  if os.path.exists(file_storage_dir):
2951
    if not os.path.isdir(file_storage_dir):
2952
      _Fail("Specified storage dir '%s' is not a directory",
2953
            file_storage_dir)
2954
  else:
2955
    try:
2956
      os.makedirs(file_storage_dir, 0750)
2957
    except OSError, err:
2958
      _Fail("Cannot create file storage directory '%s': %s",
2959
            file_storage_dir, err, exc=True)
2960

    
2961

    
2962
def RemoveFileStorageDir(file_storage_dir):
2963
  """Remove file storage directory.
2964

2965
  Remove it only if it's empty. If not log an error and return.
2966

2967
  @type file_storage_dir: str
2968
  @param file_storage_dir: the directory we should cleanup
2969
  @rtype: tuple (success,)
2970
  @return: tuple of one element, C{success}, denoting
2971
      whether the operation was successful
2972

2973
  """
2974
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2975
  if os.path.exists(file_storage_dir):
2976
    if not os.path.isdir(file_storage_dir):
2977
      _Fail("Specified Storage directory '%s' is not a directory",
2978
            file_storage_dir)
2979
    # deletes dir only if empty, otherwise we want to fail the rpc call
2980
    try:
2981
      os.rmdir(file_storage_dir)
2982
    except OSError, err:
2983
      _Fail("Cannot remove file storage directory '%s': %s",
2984
            file_storage_dir, err)
2985

    
2986

    
2987
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2988
  """Rename the file storage directory.
2989

2990
  @type old_file_storage_dir: str
2991
  @param old_file_storage_dir: the current path
2992
  @type new_file_storage_dir: str
2993
  @param new_file_storage_dir: the name we should rename to
2994
  @rtype: tuple (success,)
2995
  @return: tuple of one element, C{success}, denoting
2996
      whether the operation was successful
2997

2998
  """
2999
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3000
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3001
  if not os.path.exists(new_file_storage_dir):
3002
    if os.path.isdir(old_file_storage_dir):
3003
      try:
3004
        os.rename(old_file_storage_dir, new_file_storage_dir)
3005
      except OSError, err:
3006
        _Fail("Cannot rename '%s' to '%s': %s",
3007
              old_file_storage_dir, new_file_storage_dir, err)
3008
    else:
3009
      _Fail("Specified storage dir '%s' is not a directory",
3010
            old_file_storage_dir)
3011
  else:
3012
    if os.path.exists(old_file_storage_dir):
3013
      _Fail("Cannot rename '%s' to '%s': both locations exist",
3014
            old_file_storage_dir, new_file_storage_dir)
3015

    
3016

    
3017
def _EnsureJobQueueFile(file_name):
3018
  """Checks whether the given filename is in the queue directory.
3019

3020
  @type file_name: str
3021
  @param file_name: the file name we should check
3022
  @rtype: None
3023
  @raises RPCFail: if the file is not valid
3024

3025
  """
3026
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3027
    _Fail("Passed job queue file '%s' does not belong to"
3028
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3029

    
3030

    
3031
def JobQueueUpdate(file_name, content):
3032
  """Updates a file in the queue directory.
3033

3034
  This is just a wrapper over L{utils.io.WriteFile}, with proper
3035
  checking.
3036

3037
  @type file_name: str
3038
  @param file_name: the job file name
3039
  @type content: str
3040
  @param content: the new job contents
3041
  @rtype: boolean
3042
  @return: the success of the operation
3043

3044
  """
3045
  file_name = vcluster.LocalizeVirtualPath(file_name)
3046

    
3047
  _EnsureJobQueueFile(file_name)
3048
  getents = runtime.GetEnts()
3049

    
3050
  # Write and replace the file atomically
3051
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3052
                  gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3053

    
3054

    
3055
def JobQueueRename(old, new):
3056
  """Renames a job queue file.
3057

3058
  This is just a wrapper over os.rename with proper checking.
3059

3060
  @type old: str
3061
  @param old: the old (actual) file name
3062
  @type new: str
3063
  @param new: the desired file name
3064
  @rtype: tuple
3065
  @return: the success of the operation and payload
3066

3067
  """
3068
  old = vcluster.LocalizeVirtualPath(old)
3069
  new = vcluster.LocalizeVirtualPath(new)
3070

    
3071
  _EnsureJobQueueFile(old)
3072
  _EnsureJobQueueFile(new)
3073

    
3074
  getents = runtime.GetEnts()
3075

    
3076
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3077
                   dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3078

    
3079

    
3080
def BlockdevClose(instance_name, disks):
3081
  """Closes the given block devices.
3082

3083
  This means they will be switched to secondary mode (in case of
3084
  DRBD).
3085

3086
  @param instance_name: if the argument is not empty, the symlinks
3087
      of this instance will be removed
3088
  @type disks: list of L{objects.Disk}
3089
  @param disks: the list of disks to be closed
3090
  @rtype: tuple (success, message)
3091
  @return: a tuple of success and message, where success
3092
      indicates the succes of the operation, and message
3093
      which will contain the error details in case we
3094
      failed
3095

3096
  """
3097
  bdevs = []
3098
  for cf in disks:
3099
    rd = _RecursiveFindBD(cf)
3100
    if rd is None:
3101
      _Fail("Can't find device %s", cf)
3102
    bdevs.append(rd)
3103

    
3104
  msg = []
3105
  for rd in bdevs:
3106
    try:
3107
      rd.Close()
3108
    except errors.BlockDeviceError, err:
3109
      msg.append(str(err))
3110
  if msg:
3111
    _Fail("Can't make devices secondary: %s", ",".join(msg))
3112
  else:
3113
    if instance_name:
3114
      _RemoveBlockDevLinks(instance_name, disks)
3115

    
3116

    
3117
def ValidateHVParams(hvname, hvparams):
3118
  """Validates the given hypervisor parameters.
3119

3120
  @type hvname: string
3121
  @param hvname: the hypervisor name
3122
  @type hvparams: dict
3123
  @param hvparams: the hypervisor parameters to be validated
3124
  @rtype: None
3125

3126
  """
3127
  try:
3128
    hv_type = hypervisor.GetHypervisor(hvname)
3129
    hv_type.ValidateParameters(hvparams)
3130
  except errors.HypervisorError, err:
3131
    _Fail(str(err), log=False)
3132

    
3133

    
3134
def _CheckOSPList(os_obj, parameters):
3135
  """Check whether a list of parameters is supported by the OS.
3136

3137
  @type os_obj: L{objects.OS}
3138
  @param os_obj: OS object to check
3139
  @type parameters: list
3140
  @param parameters: the list of parameters to check
3141

3142
  """
3143
  supported = [v[0] for v in os_obj.supported_parameters]
3144
  delta = frozenset(parameters).difference(supported)
3145
  if delta:
3146
    _Fail("The following parameters are not supported"
3147
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3148

    
3149

    
3150
def ValidateOS(required, osname, checks, osparams):
3151
  """Validate the given OS' parameters.
3152

3153
  @type required: boolean
3154
  @param required: whether absence of the OS should translate into
3155
      failure or not
3156
  @type osname: string
3157
  @param osname: the OS to be validated
3158
  @type checks: list
3159
  @param checks: list of the checks to run (currently only 'parameters')
3160
  @type osparams: dict
3161
  @param osparams: dictionary with OS parameters
3162
  @rtype: boolean
3163
  @return: True if the validation passed, or False if the OS was not
3164
      found and L{required} was false
3165

3166
  """
3167
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3168
    _Fail("Unknown checks required for OS %s: %s", osname,
3169
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3170

    
3171
  name_only = objects.OS.GetName(osname)
3172
  status, tbv = _TryOSFromDisk(name_only, None)
3173

    
3174
  if not status:
3175
    if required:
3176
      _Fail(tbv)
3177
    else:
3178
      return False
3179

    
3180
  if max(tbv.api_versions) < constants.OS_API_V20:
3181
    return True
3182

    
3183
  if constants.OS_VALIDATE_PARAMETERS in checks:
3184
    _CheckOSPList(tbv, osparams.keys())
3185

    
3186
  validate_env = OSCoreEnv(osname, tbv, osparams)
3187
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3188
                        cwd=tbv.path, reset_env=True)
3189
  if result.failed:
3190
    logging.error("os validate command '%s' returned error: %s output: %s",
3191
                  result.cmd, result.fail_reason, result.output)
3192
    _Fail("OS validation script failed (%s), output: %s",
3193
          result.fail_reason, result.output, log=False)
3194

    
3195
  return True
3196

    
3197

    
3198
def DemoteFromMC():
3199
  """Demotes the current node from master candidate role.
3200

3201
  """
3202
  # try to ensure we're not the master by mistake
3203
  master, myself = ssconf.GetMasterAndMyself()
3204
  if master == myself:
3205
    _Fail("ssconf status shows I'm the master node, will not demote")
3206

    
3207
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3208
  if not result.failed:
3209
    _Fail("The master daemon is running, will not demote")
3210

    
3211
  try:
3212
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3213
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3214
  except EnvironmentError, err:
3215
    if err.errno != errno.ENOENT:
3216
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3217

    
3218
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3219

    
3220

    
3221
def _GetX509Filenames(cryptodir, name):
3222
  """Returns the full paths for the private key and certificate.
3223

3224
  """
3225
  return (utils.PathJoin(cryptodir, name),
3226
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3227
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3228

    
3229

    
3230
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3231
  """Creates a new X509 certificate for SSL/TLS.
3232

3233
  @type validity: int
3234
  @param validity: Validity in seconds
3235
  @rtype: tuple; (string, string)
3236
  @return: Certificate name and public part
3237

3238
  """
3239
  (key_pem, cert_pem) = \
3240
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3241
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3242

    
3243
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3244
                              prefix="x509-%s-" % utils.TimestampForFilename())
3245
  try:
3246
    name = os.path.basename(cert_dir)
3247
    assert len(name) > 5
3248

    
3249
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3250

    
3251
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3252
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3253

    
3254
    # Never return private key as it shouldn't leave the node
3255
    return (name, cert_pem)
3256
  except Exception:
3257
    shutil.rmtree(cert_dir, ignore_errors=True)
3258
    raise
3259

    
3260

    
3261
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3262
  """Removes a X509 certificate.
3263

3264
  @type name: string
3265
  @param name: Certificate name
3266

3267
  """
3268
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3269

    
3270
  utils.RemoveFile(key_file)
3271
  utils.RemoveFile(cert_file)
3272

    
3273
  try:
3274
    os.rmdir(cert_dir)
3275
  except EnvironmentError, err:
3276
    _Fail("Cannot remove certificate directory '%s': %s",
3277
          cert_dir, err)
3278

    
3279

    
3280
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3281
  """Returns the command for the requested input/output.
3282

3283
  @type instance: L{objects.Instance}
3284
  @param instance: The instance object
3285
  @param mode: Import/export mode
3286
  @param ieio: Input/output type
3287
  @param ieargs: Input/output arguments
3288

3289
  """
3290
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3291

    
3292
  env = None
3293
  prefix = None
3294
  suffix = None
3295
  exp_size = None
3296

    
3297
  if ieio == constants.IEIO_FILE:
3298
    (filename, ) = ieargs
3299

    
3300
    if not utils.IsNormAbsPath(filename):
3301
      _Fail("Path '%s' is not normalized or absolute", filename)
3302

    
3303
    real_filename = os.path.realpath(filename)
3304
    directory = os.path.dirname(real_filename)
3305

    
3306
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3307
      _Fail("File '%s' is not under exports directory '%s': %s",
3308
            filename, pathutils.EXPORT_DIR, real_filename)
3309

    
3310
    # Create directory
3311
    utils.Makedirs(directory, mode=0750)
3312

    
3313
    quoted_filename = utils.ShellQuote(filename)
3314

    
3315
    if mode == constants.IEM_IMPORT:
3316
      suffix = "> %s" % quoted_filename
3317
    elif mode == constants.IEM_EXPORT:
3318
      suffix = "< %s" % quoted_filename
3319

    
3320
      # Retrieve file size
3321
      try:
3322
        st = os.stat(filename)
3323
      except EnvironmentError, err:
3324
        logging.error("Can't stat(2) %s: %s", filename, err)
3325
      else:
3326
        exp_size = utils.BytesToMebibyte(st.st_size)
3327

    
3328
  elif ieio == constants.IEIO_RAW_DISK:
3329
    (disk, ) = ieargs
3330

    
3331
    real_disk = _OpenRealBD(disk)
3332

    
3333
    if mode == constants.IEM_IMPORT:
3334
      # we set here a smaller block size as, due to transport buffering, more
3335
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3336
      # is not already there or we pass a wrong path; we use notrunc to no
3337
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3338
      # much memory; this means that at best, we flush every 64k, which will
3339
      # not be very fast
3340
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3341
                                    " bs=%s oflag=dsync"),
3342
                                    real_disk.dev_path,
3343
                                    str(64 * 1024))
3344

    
3345
    elif mode == constants.IEM_EXPORT:
3346
      # the block size on the read dd is 1MiB to match our units
3347
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3348
                                   real_disk.dev_path,
3349
                                   str(1024 * 1024), # 1 MB
3350
                                   str(disk.size))
3351
      exp_size = disk.size
3352

    
3353
  elif ieio == constants.IEIO_SCRIPT:
3354
    (disk, disk_index, ) = ieargs
3355

    
3356
    assert isinstance(disk_index, (int, long))
3357

    
3358
    real_disk = _OpenRealBD(disk)
3359

    
3360
    inst_os = OSFromDisk(instance.os)
3361
    env = OSEnvironment(instance, inst_os)
3362

    
3363
    if mode == constants.IEM_IMPORT:
3364
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3365
      env["IMPORT_INDEX"] = str(disk_index)
3366
      script = inst_os.import_script
3367

    
3368
    elif mode == constants.IEM_EXPORT:
3369
      env["EXPORT_DEVICE"] = real_disk.dev_path
3370
      env["EXPORT_INDEX"] = str(disk_index)
3371
      script = inst_os.export_script
3372

    
3373
    # TODO: Pass special environment only to script
3374
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3375

    
3376
    if mode == constants.IEM_IMPORT:
3377
      suffix = "| %s" % script_cmd
3378

    
3379
    elif mode == constants.IEM_EXPORT:
3380
      prefix = "%s |" % script_cmd
3381

    
3382
    # Let script predict size
3383
    exp_size = constants.IE_CUSTOM_SIZE
3384

    
3385
  else:
3386
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3387

    
3388
  return (env, prefix, suffix, exp_size)
3389

    
3390

    
3391
def _CreateImportExportStatusDir(prefix):
3392
  """Creates status directory for import/export.
3393

3394
  """
3395
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3396
                          prefix=("%s-%s-" %
3397
                                  (prefix, utils.TimestampForFilename())))
3398

    
3399

    
3400
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3401
                            ieio, ieioargs):
3402
  """Starts an import or export daemon.
3403

3404
  @param mode: Import/output mode
3405
  @type opts: L{objects.ImportExportOptions}
3406
  @param opts: Daemon options
3407
  @type host: string
3408
  @param host: Remote host for export (None for import)
3409
  @type port: int
3410
  @param port: Remote port for export (None for import)
3411
  @type instance: L{objects.Instance}
3412
  @param instance: Instance object
3413
  @type component: string
3414
  @param component: which part of the instance is transferred now,
3415
      e.g. 'disk/0'
3416
  @param ieio: Input/output type
3417
  @param ieioargs: Input/output arguments
3418

3419
  """
3420
  if mode == constants.IEM_IMPORT:
3421
    prefix = "import"
3422

    
3423
    if not (host is None and port is None):
3424
      _Fail("Can not specify host or port on import")
3425

    
3426
  elif mode == constants.IEM_EXPORT:
3427
    prefix = "export"
3428

    
3429
    if host is None or port is None:
3430
      _Fail("Host and port must be specified for an export")
3431

    
3432
  else:
3433
    _Fail("Invalid mode %r", mode)
3434

    
3435
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3436
    _Fail("Cluster certificate can only be used for both key and CA")
3437

    
3438
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3439
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3440

    
3441
  if opts.key_name is None:
3442
    # Use server.pem
3443
    key_path = pathutils.NODED_CERT_FILE
3444
    cert_path = pathutils.NODED_CERT_FILE
3445
    assert opts.ca_pem is None
3446
  else:
3447
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3448
                                                 opts.key_name)
3449
    assert opts.ca_pem is not None
3450

    
3451
  for i in [key_path, cert_path]:
3452
    if not os.path.exists(i):
3453
      _Fail("File '%s' does not exist" % i)
3454

    
3455
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3456
  try:
3457
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3458
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3459
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3460

    
3461
    if opts.ca_pem is None:
3462
      # Use server.pem
3463
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3464
    else:
3465
      ca = opts.ca_pem
3466

    
3467
    # Write CA file
3468
    utils.WriteFile(ca_file, data=ca, mode=0400)
3469

    
3470
    cmd = [
3471
      pathutils.IMPORT_EXPORT_DAEMON,
3472
      status_file, mode,
3473
      "--key=%s" % key_path,
3474
      "--cert=%s" % cert_path,
3475
      "--ca=%s" % ca_file,
3476
      ]
3477

    
3478
    if host:
3479
      cmd.append("--host=%s" % host)
3480

    
3481
    if port:
3482
      cmd.append("--port=%s" % port)
3483

    
3484
    if opts.ipv6:
3485
      cmd.append("--ipv6")
3486
    else:
3487
      cmd.append("--ipv4")
3488

    
3489
    if opts.compress:
3490
      cmd.append("--compress=%s" % opts.compress)
3491

    
3492
    if opts.magic:
3493
      cmd.append("--magic=%s" % opts.magic)
3494

    
3495
    if exp_size is not None:
3496
      cmd.append("--expected-size=%s" % exp_size)
3497

    
3498
    if cmd_prefix:
3499
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3500

    
3501
    if cmd_suffix:
3502
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3503

    
3504
    if mode == constants.IEM_EXPORT:
3505
      # Retry connection a few times when connecting to remote peer
3506
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3507
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3508
    elif opts.connect_timeout is not None:
3509
      assert mode == constants.IEM_IMPORT
3510
      # Overall timeout for establishing connection while listening
3511
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3512

    
3513
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3514

    
3515
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3516
    # support for receiving a file descriptor for output
3517
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3518
                      output=logfile)
3519

    
3520
    # The import/export name is simply the status directory name
3521
    return os.path.basename(status_dir)
3522

    
3523
  except Exception:
3524
    shutil.rmtree(status_dir, ignore_errors=True)
3525
    raise
3526

    
3527

    
3528
def GetImportExportStatus(names):
3529
  """Returns import/export daemon status.
3530

3531
  @type names: sequence
3532
  @param names: List of names
3533
  @rtype: List of dicts
3534
  @return: Returns a list of the state of each named import/export or None if a
3535
           status couldn't be read
3536

3537
  """
3538
  result = []
3539

    
3540
  for name in names:
3541
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3542
                                 _IES_STATUS_FILE)
3543

    
3544
    try:
3545
      data = utils.ReadFile(status_file)
3546
    except EnvironmentError, err:
3547
      if err.errno != errno.ENOENT:
3548
        raise
3549
      data = None
3550

    
3551
    if not data:
3552
      result.append(None)
3553
      continue
3554

    
3555
    result.append(serializer.LoadJson(data))
3556

    
3557
  return result
3558

    
3559

    
3560
def AbortImportExport(name):
3561
  """Sends SIGTERM to a running import/export daemon.
3562

3563
  """
3564
  logging.info("Abort import/export %s", name)
3565

    
3566
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3567
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3568

    
3569
  if pid:
3570
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3571
                 name, pid)
3572
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3573

    
3574

    
3575
def CleanupImportExport(name):
3576
  """Cleanup after an import or export.
3577

3578
  If the import/export daemon is still running it's killed. Afterwards the
3579
  whole status directory is removed.
3580

3581
  """
3582
  logging.info("Finalizing import/export %s", name)
3583

    
3584
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3585

    
3586
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3587

    
3588
  if pid:
3589
    logging.info("Import/export %s is still running with PID %s",
3590
                 name, pid)
3591
    utils.KillProcess(pid, waitpid=False)
3592

    
3593
  shutil.rmtree(status_dir, ignore_errors=True)
3594

    
3595

    
3596
def _FindDisks(nodes_ip, disks):
3597
  """Sets the physical ID on disks and returns the block devices.
3598

3599
  """
3600
  # set the correct physical ID
3601
  my_name = netutils.Hostname.GetSysName()
3602
  for cf in disks:
3603
    cf.SetPhysicalID(my_name, nodes_ip)
3604

    
3605
  bdevs = []
3606

    
3607
  for cf in disks:
3608
    rd = _RecursiveFindBD(cf)
3609
    if rd is None:
3610
      _Fail("Can't find device %s", cf)
3611
    bdevs.append(rd)
3612
  return bdevs
3613

    
3614

    
3615
def DrbdDisconnectNet(nodes_ip, disks):
3616
  """Disconnects the network on a list of drbd devices.
3617

3618
  """
3619
  bdevs = _FindDisks(nodes_ip, disks)
3620

    
3621
  # disconnect disks
3622
  for rd in bdevs:
3623
    try:
3624
      rd.DisconnectNet()
3625
    except errors.BlockDeviceError, err:
3626
      _Fail("Can't change network configuration to standalone mode: %s",
3627
            err, exc=True)
3628

    
3629

    
3630
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3631
  """Attaches the network on a list of drbd devices.
3632

3633
  """
3634
  bdevs = _FindDisks(nodes_ip, disks)
3635

    
3636
  if multimaster:
3637
    for idx, rd in enumerate(bdevs):
3638
      try:
3639
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3640
      except EnvironmentError, err:
3641
        _Fail("Can't create symlink: %s", err)
3642
  # reconnect disks, switch to new master configuration and if
3643
  # needed primary mode
3644
  for rd in bdevs:
3645
    try:
3646
      rd.AttachNet(multimaster)
3647
    except errors.BlockDeviceError, err:
3648
      _Fail("Can't change network configuration: %s", err)
3649

    
3650
  # wait until the disks are connected; we need to retry the re-attach
3651
  # if the device becomes standalone, as this might happen if the one
3652
  # node disconnects and reconnects in a different mode before the
3653
  # other node reconnects; in this case, one or both of the nodes will
3654
  # decide it has wrong configuration and switch to standalone
3655

    
3656
  def _Attach():
3657
    all_connected = True
3658

    
3659
    for rd in bdevs:
3660
      stats = rd.GetProcStatus()
3661

    
3662
      if multimaster:
3663
        # In the multimaster case we have to wait explicitly until
3664
        # the resource is Connected and UpToDate/UpToDate, because
3665
        # we promote *both nodes* to primary directly afterwards.
3666
        # Being in resync is not enough, since there is a race during which we
3667
        # may promote a node with an Outdated disk to primary, effectively
3668
        # tearing down the connection.
3669
        all_connected = (all_connected and
3670
                         stats.is_connected and
3671
                         stats.is_disk_uptodate and
3672
                         stats.peer_disk_uptodate)
3673
      else:
3674
        all_connected = (all_connected and
3675
                         (stats.is_connected or stats.is_in_resync))
3676

    
3677
      if stats.is_standalone:
3678
        # peer had different config info and this node became
3679
        # standalone, even though this should not happen with the
3680
        # new staged way of changing disk configs
3681
        try:
3682
          rd.AttachNet(multimaster)
3683
        except errors.BlockDeviceError, err:
3684
          _Fail("Can't change network configuration: %s", err)
3685

    
3686
    if not all_connected:
3687
      raise utils.RetryAgain()
3688

    
3689
  try:
3690
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3691
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3692
  except utils.RetryTimeout:
3693
    _Fail("Timeout in disk reconnecting")
3694

    
3695
  if multimaster:
3696
    # change to primary mode
3697
    for rd in bdevs:
3698
      try:
3699
        rd.Open()
3700
      except errors.BlockDeviceError, err:
3701
        _Fail("Can't change to primary mode: %s", err)
3702

    
3703

    
3704
def DrbdWaitSync(nodes_ip, disks):
3705
  """Wait until DRBDs have synchronized.
3706

3707
  """
3708
  def _helper(rd):
3709
    stats = rd.GetProcStatus()
3710
    if not (stats.is_connected or stats.is_in_resync):
3711
      raise utils.RetryAgain()
3712
    return stats
3713

    
3714
  bdevs = _FindDisks(nodes_ip, disks)
3715

    
3716
  min_resync = 100
3717
  alldone = True
3718
  for rd in bdevs:
3719
    try:
3720
      # poll each second for 15 seconds
3721
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3722
    except utils.RetryTimeout:
3723
      stats = rd.GetProcStatus()
3724
      # last check
3725
      if not (stats.is_connected or stats.is_in_resync):
3726
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3727
    alldone = alldone and (not stats.is_in_resync)
3728
    if stats.sync_percent is not None:
3729
      min_resync = min(min_resync, stats.sync_percent)
3730

    
3731
  return (alldone, min_resync)
3732

    
3733

    
3734
def GetDrbdUsermodeHelper():
3735
  """Returns DRBD usermode helper currently configured.
3736

3737
  """
3738
  try:
3739
    return bdev.BaseDRBD.GetUsermodeHelper()
3740
  except errors.BlockDeviceError, err:
3741
    _Fail(str(err))
3742

    
3743

    
3744
def PowercycleNode(hypervisor_type):
3745
  """Hard-powercycle the node.
3746

3747
  Because we need to return first, and schedule the powercycle in the
3748
  background, we won't be able to report failures nicely.
3749

3750
  """
3751
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3752
  try:
3753
    pid = os.fork()
3754
  except OSError:
3755
    # if we can't fork, we'll pretend that we're in the child process
3756
    pid = 0
3757
  if pid > 0:
3758
    return "Reboot scheduled in 5 seconds"
3759
  # ensure the child is running on ram
3760
  try:
3761
    utils.Mlockall()
3762
  except Exception: # pylint: disable=W0703
3763
    pass
3764
  time.sleep(5)
3765
  hyper.PowercycleNode()
3766

    
3767

    
3768
def _VerifyRestrictedCmdName(cmd):
3769
  """Verifies a restricted command name.
3770

3771
  @type cmd: string
3772
  @param cmd: Command name
3773
  @rtype: tuple; (boolean, string or None)
3774
  @return: The tuple's first element is the status; if C{False}, the second
3775
    element is an error message string, otherwise it's C{None}
3776

3777
  """
3778
  if not cmd.strip():
3779
    return (False, "Missing command name")
3780

    
3781
  if os.path.basename(cmd) != cmd:
3782
    return (False, "Invalid command name")
3783

    
3784
  if not constants.EXT_PLUGIN_MASK.match(cmd):
3785
    return (False, "Command name contains forbidden characters")
3786

    
3787
  return (True, None)
3788

    
3789

    
3790
def _CommonRestrictedCmdCheck(path, owner):
3791
  """Common checks for restricted command file system directories and files.
3792

3793
  @type path: string
3794
  @param path: Path to check
3795
  @param owner: C{None} or tuple containing UID and GID
3796
  @rtype: tuple; (boolean, string or C{os.stat} result)
3797
  @return: The tuple's first element is the status; if C{False}, the second
3798
    element is an error message string, otherwise it's the result of C{os.stat}
3799

3800
  """
3801
  if owner is None:
3802
    # Default to root as owner
3803
    owner = (0, 0)
3804

    
3805
  try:
3806
    st = os.stat(path)
3807
  except EnvironmentError, err:
3808
    return (False, "Can't stat(2) '%s': %s" % (path, err))
3809

    
3810
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
3811
    return (False, "Permissions on '%s' are too permissive" % path)
3812

    
3813
  if (st.st_uid, st.st_gid) != owner:
3814
    (owner_uid, owner_gid) = owner
3815
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
3816

    
3817
  return (True, st)
3818

    
3819

    
3820
def _VerifyRestrictedCmdDirectory(path, _owner=None):
3821
  """Verifies restricted command directory.
3822

3823
  @type path: string
3824
  @param path: Path to check
3825
  @rtype: tuple; (boolean, string or None)
3826
  @return: The tuple's first element is the status; if C{False}, the second
3827
    element is an error message string, otherwise it's C{None}
3828

3829
  """
3830
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
3831

    
3832
  if not status:
3833
    return (False, value)
3834

    
3835
  if not stat.S_ISDIR(value.st_mode):
3836
    return (False, "Path '%s' is not a directory" % path)
3837

    
3838
  return (True, None)
3839

    
3840

    
3841
def _VerifyRestrictedCmd(path, cmd, _owner=None):
3842
  """Verifies a whole restricted command and returns its executable filename.
3843

3844
  @type path: string
3845
  @param path: Directory containing restricted commands
3846
  @type cmd: string
3847
  @param cmd: Command name
3848
  @rtype: tuple; (boolean, string)
3849
  @return: The tuple's first element is the status; if C{False}, the second
3850
    element is an error message string, otherwise the second element is the
3851
    absolute path to the executable
3852

3853
  """
3854
  executable = utils.PathJoin(path, cmd)
3855

    
3856
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
3857

    
3858
  if not status:
3859
    return (False, msg)
3860

    
3861
  if not utils.IsExecutable(executable):
3862
    return (False, "access(2) thinks '%s' can't be executed" % executable)
3863

    
3864
  return (True, executable)
3865

    
3866

    
3867
def _PrepareRestrictedCmd(path, cmd,
3868
                          _verify_dir=_VerifyRestrictedCmdDirectory,
3869
                          _verify_name=_VerifyRestrictedCmdName,
3870
                          _verify_cmd=_VerifyRestrictedCmd):
3871
  """Performs a number of tests on a restricted command.
3872

3873
  @type path: string
3874
  @param path: Directory containing restricted commands
3875
  @type cmd: string
3876
  @param cmd: Command name
3877
  @return: Same as L{_VerifyRestrictedCmd}
3878

3879
  """
3880
  # Verify the directory first
3881
  (status, msg) = _verify_dir(path)
3882
  if status:
3883
    # Check command if everything was alright
3884
    (status, msg) = _verify_name(cmd)
3885

    
3886
  if not status:
3887
    return (False, msg)
3888

    
3889
  # Check actual executable
3890
  return _verify_cmd(path, cmd)
3891

    
3892

    
3893
def RunRestrictedCmd(cmd,
3894
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
3895
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
3896
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
3897
                     _sleep_fn=time.sleep,
3898
                     _prepare_fn=_PrepareRestrictedCmd,
3899
                     _runcmd_fn=utils.RunCmd,
3900
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
3901
  """Executes a restricted command after performing strict tests.
3902

3903
  @type cmd: string
3904
  @param cmd: Command name
3905
  @rtype: string
3906
  @return: Command output
3907
  @raise RPCFail: In case of an error
3908

3909
  """
3910
  logging.info("Preparing to run restricted command '%s'", cmd)
3911

    
3912
  if not _enabled:
3913
    _Fail("Restricted commands disabled at configure time")
3914

    
3915
  lock = None
3916
  try:
3917
    cmdresult = None
3918
    try:
3919
      lock = utils.FileLock.Open(_lock_file)
3920
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
3921

    
3922
      (status, value) = _prepare_fn(_path, cmd)
3923

    
3924
      if status:
3925
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
3926
                               postfork_fn=lambda _: lock.Unlock())
3927
      else:
3928
        logging.error(value)
3929
    except Exception: # pylint: disable=W0703
3930
      # Keep original error in log
3931
      logging.exception("Caught exception")
3932

    
3933
    if cmdresult is None:
3934
      logging.info("Sleeping for %0.1f seconds before returning",
3935
                   _RCMD_INVALID_DELAY)
3936
      _sleep_fn(_RCMD_INVALID_DELAY)
3937

    
3938
      # Do not include original error message in returned error
3939
      _Fail("Executing command '%s' failed" % cmd)
3940
    elif cmdresult.failed or cmdresult.fail_reason:
3941
      _Fail("Restricted command '%s' failed: %s; output: %s",
3942
            cmd, cmdresult.fail_reason, cmdresult.output)
3943
    else:
3944
      return cmdresult.output
3945
  finally:
3946
    if lock is not None:
3947
      # Release lock at last
3948
      lock.Close()
3949
      lock = None
3950

    
3951

    
3952
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
3953
  """Creates or removes the watcher pause file.
3954

3955
  @type until: None or number
3956
  @param until: Unix timestamp saying until when the watcher shouldn't run
3957

3958
  """
3959
  if until is None:
3960
    logging.info("Received request to no longer pause watcher")
3961
    utils.RemoveFile(_filename)
3962
  else:
3963
    logging.info("Received request to pause watcher until %s", until)
3964

    
3965
    if not ht.TNumber(until):
3966
      _Fail("Duration must be numeric")
3967

    
3968
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
3969

    
3970

    
3971
class HooksRunner(object):
3972
  """Hook runner.
3973

3974
  This class is instantiated on the node side (ganeti-noded) and not
3975
  on the master side.
3976

3977
  """
3978
  def __init__(self, hooks_base_dir=None):
3979
    """Constructor for hooks runner.
3980

3981
    @type hooks_base_dir: str or None
3982
    @param hooks_base_dir: if not None, this overrides the
3983
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
3984

3985
    """
3986
    if hooks_base_dir is None:
3987
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
3988
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
3989
    # constant
3990
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
3991

    
3992
  def RunLocalHooks(self, node_list, hpath, phase, env):
3993
    """Check that the hooks will be run only locally and then run them.
3994

3995
    """
3996
    assert len(node_list) == 1
3997
    node = node_list[0]
3998
    _, myself = ssconf.GetMasterAndMyself()
3999
    assert node == myself
4000

    
4001
    results = self.RunHooks(hpath, phase, env)
4002

    
4003
    # Return values in the form expected by HooksMaster
4004
    return {node: (None, False, results)}
4005

    
4006
  def RunHooks(self, hpath, phase, env):
4007
    """Run the scripts in the hooks directory.
4008

4009
    @type hpath: str
4010
    @param hpath: the path to the hooks directory which
4011
        holds the scripts
4012
    @type phase: str
4013
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
4014
        L{constants.HOOKS_PHASE_POST}
4015
    @type env: dict
4016
    @param env: dictionary with the environment for the hook
4017
    @rtype: list
4018
    @return: list of 3-element tuples:
4019
      - script path
4020
      - script result, either L{constants.HKR_SUCCESS} or
4021
        L{constants.HKR_FAIL}
4022
      - output of the script
4023

4024
    @raise errors.ProgrammerError: for invalid input
4025
        parameters
4026

4027
    """
4028
    if phase == constants.HOOKS_PHASE_PRE:
4029
      suffix = "pre"
4030
    elif phase == constants.HOOKS_PHASE_POST:
4031
      suffix = "post"
4032
    else:
4033
      _Fail("Unknown hooks phase '%s'", phase)
4034

    
4035
    subdir = "%s-%s.d" % (hpath, suffix)
4036
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4037

    
4038
    results = []
4039

    
4040
    if not os.path.isdir(dir_name):
4041
      # for non-existing/non-dirs, we simply exit instead of logging a
4042
      # warning at every operation
4043
      return results
4044

    
4045
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4046

    
4047
    for (relname, relstatus, runresult) in runparts_results:
4048
      if relstatus == constants.RUNPARTS_SKIP:
4049
        rrval = constants.HKR_SKIP
4050
        output = ""
4051
      elif relstatus == constants.RUNPARTS_ERR:
4052
        rrval = constants.HKR_FAIL
4053
        output = "Hook script execution error: %s" % runresult
4054
      elif relstatus == constants.RUNPARTS_RUN:
4055
        if runresult.failed:
4056
          rrval = constants.HKR_FAIL
4057
        else:
4058
          rrval = constants.HKR_SUCCESS
4059
        output = utils.SafeEncode(runresult.output.strip())
4060
      results.append(("%s/%s" % (subdir, relname), rrval, output))
4061

    
4062
    return results
4063

    
4064

    
4065
class IAllocatorRunner(object):
4066
  """IAllocator runner.
4067

4068
  This class is instantiated on the node side (ganeti-noded) and not on
4069
  the master side.
4070

4071
  """
4072
  @staticmethod
4073
  def Run(name, idata):
4074
    """Run an iallocator script.
4075

4076
    @type name: str
4077
    @param name: the iallocator script name
4078
    @type idata: str
4079
    @param idata: the allocator input data
4080

4081
    @rtype: tuple
4082
    @return: two element tuple of:
4083
       - status
4084
       - either error message or stdout of allocator (for success)
4085

4086
    """
4087
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4088
                                  os.path.isfile)
4089
    if alloc_script is None:
4090
      _Fail("iallocator module '%s' not found in the search path", name)
4091

    
4092
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4093
    try:
4094
      os.write(fd, idata)
4095
      os.close(fd)
4096
      result = utils.RunCmd([alloc_script, fin_name])
4097
      if result.failed:
4098
        _Fail("iallocator module '%s' failed: %s, output '%s'",
4099
              name, result.fail_reason, result.output)
4100
    finally:
4101
      os.unlink(fin_name)
4102

    
4103
    return result.stdout
4104

    
4105

    
4106
class DevCacheManager(object):
4107
  """Simple class for managing a cache of block device information.
4108

4109
  """
4110
  _DEV_PREFIX = "/dev/"
4111
  _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4112

    
4113
  @classmethod
4114
  def _ConvertPath(cls, dev_path):
4115
    """Converts a /dev/name path to the cache file name.
4116

4117
    This replaces slashes with underscores and strips the /dev
4118
    prefix. It then returns the full path to the cache file.
4119

4120
    @type dev_path: str
4121
    @param dev_path: the C{/dev/} path name
4122
    @rtype: str
4123
    @return: the converted path name
4124

4125
    """
4126
    if dev_path.startswith(cls._DEV_PREFIX):
4127
      dev_path = dev_path[len(cls._DEV_PREFIX):]
4128
    dev_path = dev_path.replace("/", "_")
4129
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4130
    return fpath
4131

    
4132
  @classmethod
4133
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4134
    """Updates the cache information for a given device.
4135

4136
    @type dev_path: str
4137
    @param dev_path: the pathname of the device
4138
    @type owner: str
4139
    @param owner: the owner (instance name) of the device
4140
    @type on_primary: bool
4141
    @param on_primary: whether this is the primary
4142
        node nor not
4143
    @type iv_name: str
4144
    @param iv_name: the instance-visible name of the
4145
        device, as in objects.Disk.iv_name
4146

4147
    @rtype: None
4148

4149
    """
4150
    if dev_path is None:
4151
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
4152
      return
4153
    fpath = cls._ConvertPath(dev_path)
4154
    if on_primary:
4155
      state = "primary"
4156
    else:
4157
      state = "secondary"
4158
    if iv_name is None:
4159
      iv_name = "not_visible"
4160
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4161
    try:
4162
      utils.WriteFile(fpath, data=fdata)
4163
    except EnvironmentError, err:
4164
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4165

    
4166
  @classmethod
4167
  def RemoveCache(cls, dev_path):
4168
    """Remove data for a dev_path.
4169

4170
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
4171
    path name and logging.
4172

4173
    @type dev_path: str
4174
    @param dev_path: the pathname of the device
4175

4176
    @rtype: None
4177

4178
    """
4179
    if dev_path is None:
4180
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
4181
      return
4182
    fpath = cls._ConvertPath(dev_path)
4183
    try:
4184
      utils.RemoveFile(fpath)
4185
    except EnvironmentError, err:
4186
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)