Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 8fa42c7c

History | View | Annotate | Download (43.5 kB)

1
#
2
#
3

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

    
21

    
22
"""Functions used by the node daemon"""
23

    
24

    
25
import os
26
import os.path
27
import shutil
28
import time
29
import tempfile
30
import stat
31
import errno
32
import re
33
import subprocess
34

    
35
from ganeti import logger
36
from ganeti import errors
37
from ganeti import utils
38
from ganeti import ssh
39
from ganeti import hypervisor
40
from ganeti import constants
41
from ganeti import bdev
42
from ganeti import objects
43
from ganeti import ssconf
44

    
45

    
46
def StartMaster():
47
  """Activate local node as master node.
48

49
  There are two needed steps for this:
50
    - run the master script
51
    - register the cron script
52

53
  """
54
  result = utils.RunCmd([constants.MASTER_SCRIPT, "-d", "start"])
55

    
56
  if result.failed:
57
    logger.Error("could not activate cluster interface with command %s,"
58
                 " error: '%s'" % (result.cmd, result.output))
59
    return False
60

    
61
  return True
62

    
63

    
64
def StopMaster():
65
  """Deactivate this node as master.
66

67
  This does two things:
68
    - run the master stop script
69
    - remove link to master cron script.
70

71
  """
72
  result = utils.RunCmd([constants.MASTER_SCRIPT, "-d", "stop"])
73

    
74
  if result.failed:
75
    logger.Error("could not deactivate cluster interface with command %s,"
76
                 " error: '%s'" % (result.cmd, result.output))
77
    return False
78

    
79
  return True
80

    
81

    
82
def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
83
  """Joins this node to the cluster.
84

85
  This does the following:
86
      - updates the hostkeys of the machine (rsa and dsa)
87
      - adds the ssh private key to the user
88
      - adds the ssh public key to the users' authorized_keys file
89

90
  """
91
  sshd_keys =  [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
92
                (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
93
                (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
94
                (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
95
  for name, content, mode in sshd_keys:
96
    utils.WriteFile(name, data=content, mode=mode)
97

    
98
  try:
99
    priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
100
                                                    mkdir=True)
101
  except errors.OpExecError, err:
102
    logger.Error("Error while processing user ssh files: %s" % err)
103
    return False
104

    
105
  for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
106
    utils.WriteFile(name, data=content, mode=0600)
107

    
108
  utils.AddAuthorizedKey(auth_keys, sshpub)
109

    
110
  utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
111

    
112
  return True
113

    
114

    
115
def LeaveCluster():
116
  """Cleans up the current node and prepares it to be removed from the cluster.
117

118
  """
119
  if os.path.isdir(constants.DATA_DIR):
120
    for rel_name in utils.ListVisibleFiles(constants.DATA_DIR):
121
      full_name = os.path.join(constants.DATA_DIR, rel_name)
122
      if os.path.isfile(full_name) and not os.path.islink(full_name):
123
        utils.RemoveFile(full_name)
124

    
125

    
126
  try:
127
    priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
128
  except errors.OpExecError, err:
129
    logger.Error("Error while processing ssh files: %s" % err)
130
    return
131

    
132
  f = open(pub_key, 'r')
133
  try:
134
    utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
135
  finally:
136
    f.close()
137

    
138
  utils.RemoveFile(priv_key)
139
  utils.RemoveFile(pub_key)
140

    
141

    
142
def GetNodeInfo(vgname):
143
  """Gives back a hash with different informations about the node.
144

145
  Returns:
146
    { 'vg_size' : xxx,  'vg_free' : xxx, 'memory_domain0': xxx,
147
      'memory_free' : xxx, 'memory_total' : xxx }
148
    where
149
    vg_size is the size of the configured volume group in MiB
150
    vg_free is the free size of the volume group in MiB
151
    memory_dom0 is the memory allocated for domain0 in MiB
152
    memory_free is the currently available (free) ram in MiB
153
    memory_total is the total number of ram in MiB
154

155
  """
156
  outputarray = {}
157
  vginfo = _GetVGInfo(vgname)
158
  outputarray['vg_size'] = vginfo['vg_size']
159
  outputarray['vg_free'] = vginfo['vg_free']
160

    
161
  hyper = hypervisor.GetHypervisor()
162
  hyp_info = hyper.GetNodeInfo()
163
  if hyp_info is not None:
164
    outputarray.update(hyp_info)
165

    
166
  f = open("/proc/sys/kernel/random/boot_id", 'r')
167
  try:
168
    outputarray["bootid"] = f.read(128).rstrip("\n")
169
  finally:
170
    f.close()
171

    
172
  return outputarray
173

    
174

    
175
def VerifyNode(what):
176
  """Verify the status of the local node.
177

178
  Args:
179
    what - a dictionary of things to check:
180
      'filelist' : list of files for which to compute checksums
181
      'nodelist' : list of nodes we should check communication with
182
      'hypervisor': run the hypervisor-specific verify
183

184
  Requested files on local node are checksummed and the result returned.
185

186
  The nodelist is traversed, with the following checks being made
187
  for each node:
188
  - known_hosts key correct
189
  - correct resolving of node name (target node returns its own hostname
190
    by ssh-execution of 'hostname', result compared against name in list.
191

192
  """
193
  result = {}
194

    
195
  if 'hypervisor' in what:
196
    result['hypervisor'] = hypervisor.GetHypervisor().Verify()
197

    
198
  if 'filelist' in what:
199
    result['filelist'] = utils.FingerprintFiles(what['filelist'])
200

    
201
  if 'nodelist' in what:
202
    result['nodelist'] = {}
203
    for node in what['nodelist']:
204
      success, message = ssh.VerifyNodeHostname(node)
205
      if not success:
206
        result['nodelist'][node] = message
207
  return result
208

    
209

    
210
def GetVolumeList(vg_name):
211
  """Compute list of logical volumes and their size.
212

213
  Returns:
214
    dictionary of all partions (key) with their size:
215
    test1: 20.06MiB
216

217
  """
218
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m",
219
                         "-oname,size", vg_name])
220
  if result.failed:
221
    logger.Error("Failed to list logical volumes, lvs output: %s" %
222
                 result.output)
223
    return {}
224

    
225
  lvlist = [line.split() for line in result.output.splitlines()]
226
  return dict(lvlist)
227

    
228

    
229
def ListVolumeGroups():
230
  """List the volume groups and their size.
231

232
  Returns:
233
    Dictionary with keys volume name and values the size of the volume
234

235
  """
236
  return utils.ListVolumeGroups()
237

    
238

    
239
def NodeVolumes():
240
  """List all volumes on this node.
241

242
  """
243
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
244
                         "--separator=|",
245
                         "--options=lv_name,lv_size,devices,vg_name"])
246
  if result.failed:
247
    logger.Error("Failed to list logical volumes, lvs output: %s" %
248
                 result.output)
249
    return {}
250

    
251
  def parse_dev(dev):
252
    if '(' in dev:
253
      return dev.split('(')[0]
254
    else:
255
      return dev
256

    
257
  def map_line(line):
258
    return {
259
      'name': line[0].strip(),
260
      'size': line[1].strip(),
261
      'dev': parse_dev(line[2].strip()),
262
      'vg': line[3].strip(),
263
    }
264

    
265
  return [map_line(line.split('|')) for line in result.output.splitlines()]
266

    
267

    
268
def BridgesExist(bridges_list):
269
  """Check if a list of bridges exist on the current node.
270

271
  Returns:
272
    True if all of them exist, false otherwise
273

274
  """
275
  for bridge in bridges_list:
276
    if not utils.BridgeExists(bridge):
277
      return False
278

    
279
  return True
280

    
281

    
282
def GetInstanceList():
283
  """Provides a list of instances.
284

285
  Returns:
286
    A list of all running instances on the current node
287
    - instance1.example.com
288
    - instance2.example.com
289

290
  """
291
  try:
292
    names = hypervisor.GetHypervisor().ListInstances()
293
  except errors.HypervisorError, err:
294
    logger.Error("error enumerating instances: %s" % str(err))
295
    raise
296

    
297
  return names
298

    
299

    
300
def GetInstanceInfo(instance):
301
  """Gives back the informations about an instance as a dictionary.
302

303
  Args:
304
    instance: name of the instance (ex. instance1.example.com)
305

306
  Returns:
307
    { 'memory' : 511, 'state' : '-b---', 'time' : 3188.8, }
308
    where
309
    memory: memory size of instance (int)
310
    state: xen state of instance (string)
311
    time: cpu time of instance (float)
312

313
  """
314
  output = {}
315

    
316
  iinfo = hypervisor.GetHypervisor().GetInstanceInfo(instance)
317
  if iinfo is not None:
318
    output['memory'] = iinfo[2]
319
    output['state'] = iinfo[4]
320
    output['time'] = iinfo[5]
321

    
322
  return output
323

    
324

    
325
def GetAllInstancesInfo():
326
  """Gather data about all instances.
327

328
  This is the equivalent of `GetInstanceInfo()`, except that it
329
  computes data for all instances at once, thus being faster if one
330
  needs data about more than one instance.
331

332
  Returns: a dictionary of dictionaries, keys being the instance name,
333
    and with values:
334
    { 'memory' : 511, 'state' : '-b---', 'time' : 3188.8, }
335
    where
336
    memory: memory size of instance (int)
337
    state: xen state of instance (string)
338
    time: cpu time of instance (float)
339
    vcpus: the number of cpus
340

341
  """
342
  output = {}
343

    
344
  iinfo = hypervisor.GetHypervisor().GetAllInstancesInfo()
345
  if iinfo:
346
    for name, inst_id, memory, vcpus, state, times in iinfo:
347
      output[name] = {
348
        'memory': memory,
349
        'vcpus': vcpus,
350
        'state': state,
351
        'time': times,
352
        }
353

    
354
  return output
355

    
356

    
357
def AddOSToInstance(instance, os_disk, swap_disk):
358
  """Add an OS to an instance.
359

360
  Args:
361
    instance: the instance object
362
    os_disk: the instance-visible name of the os device
363
    swap_disk: the instance-visible name of the swap device
364

365
  """
366
  inst_os = OSFromDisk(instance.os)
367

    
368
  create_script = inst_os.create_script
369

    
370
  os_device = instance.FindDisk(os_disk)
371
  if os_device is None:
372
    logger.Error("Can't find this device-visible name '%s'" % os_disk)
373
    return False
374

    
375
  swap_device = instance.FindDisk(swap_disk)
376
  if swap_device is None:
377
    logger.Error("Can't find this device-visible name '%s'" % swap_disk)
378
    return False
379

    
380
  real_os_dev = _RecursiveFindBD(os_device)
381
  if real_os_dev is None:
382
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
383
                                  str(os_device))
384
  real_os_dev.Open()
385

    
386
  real_swap_dev = _RecursiveFindBD(swap_device)
387
  if real_swap_dev is None:
388
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
389
                                  str(swap_device))
390
  real_swap_dev.Open()
391

    
392
  logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
393
                                     instance.name, int(time.time()))
394
  if not os.path.exists(constants.LOG_OS_DIR):
395
    os.mkdir(constants.LOG_OS_DIR, 0750)
396

    
397
  command = utils.BuildShellCmd("cd %s && %s -i %s -b %s -s %s &>%s",
398
                                inst_os.path, create_script, instance.name,
399
                                real_os_dev.dev_path, real_swap_dev.dev_path,
400
                                logfile)
401

    
402
  result = utils.RunCmd(command)
403

    
404
  if result.failed:
405
    logger.Error("os create command '%s' returned error: %s"
406
                 " output: %s" %
407
                 (command, result.fail_reason, result.output))
408
    return False
409

    
410
  return True
411

    
412

    
413
def RunRenameInstance(instance, old_name, os_disk, swap_disk):
414
  """Run the OS rename script for an instance.
415

416
  Args:
417
    instance: the instance object
418
    old_name: the old name of the instance
419
    os_disk: the instance-visible name of the os device
420
    swap_disk: the instance-visible name of the swap device
421

422
  """
423
  inst_os = OSFromDisk(instance.os)
424

    
425
  script = inst_os.rename_script
426

    
427
  os_device = instance.FindDisk(os_disk)
428
  if os_device is None:
429
    logger.Error("Can't find this device-visible name '%s'" % os_disk)
430
    return False
431

    
432
  swap_device = instance.FindDisk(swap_disk)
433
  if swap_device is None:
434
    logger.Error("Can't find this device-visible name '%s'" % swap_disk)
435
    return False
436

    
437
  real_os_dev = _RecursiveFindBD(os_device)
438
  if real_os_dev is None:
439
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
440
                                  str(os_device))
441
  real_os_dev.Open()
442

    
443
  real_swap_dev = _RecursiveFindBD(swap_device)
444
  if real_swap_dev is None:
445
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
446
                                  str(swap_device))
447
  real_swap_dev.Open()
448

    
449
  logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
450
                                           old_name,
451
                                           instance.name, int(time.time()))
452
  if not os.path.exists(constants.LOG_OS_DIR):
453
    os.mkdir(constants.LOG_OS_DIR, 0750)
454

    
455
  command = utils.BuildShellCmd("cd %s && %s -o %s -n %s -b %s -s %s &>%s",
456
                                inst_os.path, script, old_name, instance.name,
457
                                real_os_dev.dev_path, real_swap_dev.dev_path,
458
                                logfile)
459

    
460
  result = utils.RunCmd(command)
461

    
462
  if result.failed:
463
    logger.Error("os create command '%s' returned error: %s"
464
                 " output: %s" %
465
                 (command, result.fail_reason, result.output))
466
    return False
467

    
468
  return True
469

    
470

    
471
def _GetVGInfo(vg_name):
472
  """Get informations about the volume group.
473

474
  Args:
475
    vg_name: the volume group
476

477
  Returns:
478
    { 'vg_size' : xxx, 'vg_free' : xxx, 'pv_count' : xxx }
479
    where
480
    vg_size is the total size of the volume group in MiB
481
    vg_free is the free size of the volume group in MiB
482
    pv_count are the number of physical disks in that vg
483

484
  """
485
  retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
486
                         "--nosuffix", "--units=m", "--separator=:", vg_name])
487

    
488
  if retval.failed:
489
    errmsg = "volume group %s not present" % vg_name
490
    logger.Error(errmsg)
491
    raise errors.LVMError(errmsg)
492
  valarr = retval.stdout.strip().split(':')
493
  retdic = {
494
    "vg_size": int(round(float(valarr[0]), 0)),
495
    "vg_free": int(round(float(valarr[1]), 0)),
496
    "pv_count": int(valarr[2]),
497
    }
498
  return retdic
499

    
500

    
501
def _GatherBlockDevs(instance):
502
  """Set up an instance's block device(s).
503

504
  This is run on the primary node at instance startup. The block
505
  devices must be already assembled.
506

507
  """
508
  block_devices = []
509
  for disk in instance.disks:
510
    device = _RecursiveFindBD(disk)
511
    if device is None:
512
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
513
                                    str(disk))
514
    device.Open()
515
    block_devices.append((disk, device))
516
  return block_devices
517

    
518

    
519
def StartInstance(instance, extra_args):
520
  """Start an instance.
521

522
  Args:
523
    instance - name of instance to start.
524

525
  """
526
  running_instances = GetInstanceList()
527

    
528
  if instance.name in running_instances:
529
    return True
530

    
531
  block_devices = _GatherBlockDevs(instance)
532
  hyper = hypervisor.GetHypervisor()
533

    
534
  try:
535
    hyper.StartInstance(instance, block_devices, extra_args)
536
  except errors.HypervisorError, err:
537
    logger.Error("Failed to start instance: %s" % err)
538
    return False
539

    
540
  return True
541

    
542

    
543
def ShutdownInstance(instance):
544
  """Shut an instance down.
545

546
  Args:
547
    instance - name of instance to shutdown.
548

549
  """
550
  running_instances = GetInstanceList()
551

    
552
  if instance.name not in running_instances:
553
    return True
554

    
555
  hyper = hypervisor.GetHypervisor()
556
  try:
557
    hyper.StopInstance(instance)
558
  except errors.HypervisorError, err:
559
    logger.Error("Failed to stop instance: %s" % err)
560
    return False
561

    
562
  # test every 10secs for 2min
563
  shutdown_ok = False
564

    
565
  time.sleep(1)
566
  for dummy in range(11):
567
    if instance.name not in GetInstanceList():
568
      break
569
    time.sleep(10)
570
  else:
571
    # the shutdown did not succeed
572
    logger.Error("shutdown of '%s' unsuccessful, using destroy" % instance)
573

    
574
    try:
575
      hyper.StopInstance(instance, force=True)
576
    except errors.HypervisorError, err:
577
      logger.Error("Failed to stop instance: %s" % err)
578
      return False
579

    
580
    time.sleep(1)
581
    if instance.name in GetInstanceList():
582
      logger.Error("could not shutdown instance '%s' even by destroy")
583
      return False
584

    
585
  return True
586

    
587

    
588
def RebootInstance(instance, reboot_type, extra_args):
589
  """Reboot an instance.
590

591
  Args:
592
    instance    - name of instance to reboot
593
    reboot_type - how to reboot [soft,hard,full]
594

595
  """
596
  running_instances = GetInstanceList()
597

    
598
  if instance.name not in running_instances:
599
    logger.Error("Cannot reboot instance that is not running")
600
    return False
601

    
602
  hyper = hypervisor.GetHypervisor()
603
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
604
    try:
605
      hyper.RebootInstance(instance)
606
    except errors.HypervisorError, err:
607
      logger.Error("Failed to soft reboot instance: %s" % err)
608
      return False
609
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
610
    try:
611
      ShutdownInstance(instance)
612
      StartInstance(instance, extra_args)
613
    except errors.HypervisorError, err:
614
      logger.Error("Failed to hard reboot instance: %s" % err)
615
      return False
616
  else:
617
    raise errors.ParameterError("reboot_type invalid")
618

    
619

    
620
  return True
621

    
622

    
623
def CreateBlockDevice(disk, size, owner, on_primary, info):
624
  """Creates a block device for an instance.
625

626
  Args:
627
   bdev: a ganeti.objects.Disk object
628
   size: the size of the physical underlying devices
629
   do_open: if the device should be `Assemble()`-d and
630
            `Open()`-ed after creation
631

632
  Returns:
633
    the new unique_id of the device (this can sometime be
634
    computed only after creation), or None. On secondary nodes,
635
    it's not required to return anything.
636

637
  """
638
  clist = []
639
  if disk.children:
640
    for child in disk.children:
641
      crdev = _RecursiveAssembleBD(child, owner, on_primary)
642
      if on_primary or disk.AssembleOnSecondary():
643
        # we need the children open in case the device itself has to
644
        # be assembled
645
        crdev.Open()
646
      else:
647
        crdev.Close()
648
      clist.append(crdev)
649
  try:
650
    device = bdev.FindDevice(disk.dev_type, disk.physical_id, clist)
651
    if device is not None:
652
      logger.Info("removing existing device %s" % disk)
653
      device.Remove()
654
  except errors.BlockDeviceError, err:
655
    pass
656

    
657
  device = bdev.Create(disk.dev_type, disk.physical_id,
658
                       clist, size)
659
  if device is None:
660
    raise ValueError("Can't create child device for %s, %s" %
661
                     (disk, size))
662
  if on_primary or disk.AssembleOnSecondary():
663
    device.Assemble()
664
    device.SetSyncSpeed(constants.SYNC_SPEED)
665
    if on_primary or disk.OpenOnSecondary():
666
      device.Open(force=True)
667
    DevCacheManager.UpdateCache(device.dev_path, owner,
668
                                on_primary, disk.iv_name)
669

    
670
  device.SetInfo(info)
671

    
672
  physical_id = device.unique_id
673
  return physical_id
674

    
675

    
676
def RemoveBlockDevice(disk):
677
  """Remove a block device.
678

679
  This is intended to be called recursively.
680

681
  """
682
  try:
683
    # since we are removing the device, allow a partial match
684
    # this allows removal of broken mirrors
685
    rdev = _RecursiveFindBD(disk, allow_partial=True)
686
  except errors.BlockDeviceError, err:
687
    # probably can't attach
688
    logger.Info("Can't attach to device %s in remove" % disk)
689
    rdev = None
690
  if rdev is not None:
691
    r_path = rdev.dev_path
692
    result = rdev.Remove()
693
    if result:
694
      DevCacheManager.RemoveCache(r_path)
695
  else:
696
    result = True
697
  if disk.children:
698
    for child in disk.children:
699
      result = result and RemoveBlockDevice(child)
700
  return result
701

    
702

    
703
def _RecursiveAssembleBD(disk, owner, as_primary):
704
  """Activate a block device for an instance.
705

706
  This is run on the primary and secondary nodes for an instance.
707

708
  This function is called recursively.
709

710
  Args:
711
    disk: a objects.Disk object
712
    as_primary: if we should make the block device read/write
713

714
  Returns:
715
    the assembled device or None (in case no device was assembled)
716

717
  If the assembly is not successful, an exception is raised.
718

719
  """
720
  children = []
721
  if disk.children:
722
    for chld_disk in disk.children:
723
      children.append(_RecursiveAssembleBD(chld_disk, owner, as_primary))
724

    
725
  if as_primary or disk.AssembleOnSecondary():
726
    r_dev = bdev.AttachOrAssemble(disk.dev_type, disk.physical_id, children)
727
    r_dev.SetSyncSpeed(constants.SYNC_SPEED)
728
    result = r_dev
729
    if as_primary or disk.OpenOnSecondary():
730
      r_dev.Open()
731
    else:
732
      r_dev.Close()
733
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
734
                                as_primary, disk.iv_name)
735

    
736
  else:
737
    result = True
738
  return result
739

    
740

    
741
def AssembleBlockDevice(disk, owner, as_primary):
742
  """Activate a block device for an instance.
743

744
  This is a wrapper over _RecursiveAssembleBD.
745

746
  Returns:
747
    a /dev path for primary nodes
748
    True for secondary nodes
749

750
  """
751
  result = _RecursiveAssembleBD(disk, owner, as_primary)
752
  if isinstance(result, bdev.BlockDev):
753
    result = result.dev_path
754
  return result
755

    
756

    
757
def ShutdownBlockDevice(disk):
758
  """Shut down a block device.
759

760
  First, if the device is assembled (can `Attach()`), then the device
761
  is shutdown. Then the children of the device are shutdown.
762

763
  This function is called recursively. Note that we don't cache the
764
  children or such, as oppossed to assemble, shutdown of different
765
  devices doesn't require that the upper device was active.
766

767
  """
768
  r_dev = _RecursiveFindBD(disk)
769
  if r_dev is not None:
770
    r_path = r_dev.dev_path
771
    result = r_dev.Shutdown()
772
    if result:
773
      DevCacheManager.RemoveCache(r_path)
774
  else:
775
    result = True
776
  if disk.children:
777
    for child in disk.children:
778
      result = result and ShutdownBlockDevice(child)
779
  return result
780

    
781

    
782
def MirrorAddChildren(parent_cdev, new_cdevs):
783
  """Extend a mirrored block device.
784

785
  """
786
  parent_bdev = _RecursiveFindBD(parent_cdev, allow_partial=True)
787
  if parent_bdev is None:
788
    logger.Error("Can't find parent device")
789
    return False
790
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
791
  if new_bdevs.count(None) > 0:
792
    logger.Error("Can't find new device(s) to add: %s:%s" %
793
                 (new_bdevs, new_cdevs))
794
    return False
795
  parent_bdev.AddChildren(new_bdevs)
796
  return True
797

    
798

    
799
def MirrorRemoveChildren(parent_cdev, new_cdevs):
800
  """Shrink a mirrored block device.
801

802
  """
803
  parent_bdev = _RecursiveFindBD(parent_cdev)
804
  if parent_bdev is None:
805
    logger.Error("Can't find parent in remove children: %s" % parent_cdev)
806
    return False
807
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
808
  if new_bdevs.count(None) > 0:
809
    logger.Error("Can't find some devices while removing children: %s %s" %
810
                 (new_cdevs, new_bdevs))
811
    return False
812
  parent_bdev.RemoveChildren(new_bdevs)
813
  return True
814

    
815

    
816
def GetMirrorStatus(disks):
817
  """Get the mirroring status of a list of devices.
818

819
  Args:
820
    disks: list of `objects.Disk`
821

822
  Returns:
823
    list of (mirror_done, estimated_time) tuples, which
824
    are the result of bdev.BlockDevice.CombinedSyncStatus()
825

826
  """
827
  stats = []
828
  for dsk in disks:
829
    rbd = _RecursiveFindBD(dsk)
830
    if rbd is None:
831
      raise errors.BlockDeviceError("Can't find device %s" % str(dsk))
832
    stats.append(rbd.CombinedSyncStatus())
833
  return stats
834

    
835

    
836
def _RecursiveFindBD(disk, allow_partial=False):
837
  """Check if a device is activated.
838

839
  If so, return informations about the real device.
840

841
  Args:
842
    disk: the objects.Disk instance
843
    allow_partial: don't abort the find if a child of the
844
                   device can't be found; this is intended to be
845
                   used when repairing mirrors
846

847
  Returns:
848
    None if the device can't be found
849
    otherwise the device instance
850

851
  """
852
  children = []
853
  if disk.children:
854
    for chdisk in disk.children:
855
      children.append(_RecursiveFindBD(chdisk))
856

    
857
  return bdev.FindDevice(disk.dev_type, disk.physical_id, children)
858

    
859

    
860
def FindBlockDevice(disk):
861
  """Check if a device is activated.
862

863
  If so, return informations about the real device.
864

865
  Args:
866
    disk: the objects.Disk instance
867
  Returns:
868
    None if the device can't be found
869
    (device_path, major, minor, sync_percent, estimated_time, is_degraded)
870

871
  """
872
  rbd = _RecursiveFindBD(disk)
873
  if rbd is None:
874
    return rbd
875
  sync_p, est_t, is_degr = rbd.GetSyncStatus()
876
  return rbd.dev_path, rbd.major, rbd.minor, sync_p, est_t, is_degr
877

    
878

    
879
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
880
  """Write a file to the filesystem.
881

882
  This allows the master to overwrite(!) a file. It will only perform
883
  the operation if the file belongs to a list of configuration files.
884

885
  """
886
  if not os.path.isabs(file_name):
887
    logger.Error("Filename passed to UploadFile is not absolute: '%s'" %
888
                 file_name)
889
    return False
890

    
891
  allowed_files = [constants.CLUSTER_CONF_FILE, "/etc/hosts",
892
                   constants.SSH_KNOWN_HOSTS_FILE]
893
  allowed_files.extend(ssconf.SimpleStore().GetFileList())
894
  if file_name not in allowed_files:
895
    logger.Error("Filename passed to UploadFile not in allowed"
896
                 " upload targets: '%s'" % file_name)
897
    return False
898

    
899
  dir_name, small_name = os.path.split(file_name)
900
  fd, new_name = tempfile.mkstemp('.new', small_name, dir_name)
901
  # here we need to make sure we remove the temp file, if any error
902
  # leaves it in place
903
  try:
904
    os.chown(new_name, uid, gid)
905
    os.chmod(new_name, mode)
906
    os.write(fd, data)
907
    os.fsync(fd)
908
    os.utime(new_name, (atime, mtime))
909
    os.rename(new_name, file_name)
910
  finally:
911
    os.close(fd)
912
    utils.RemoveFile(new_name)
913
  return True
914

    
915

    
916
def _ErrnoOrStr(err):
917
  """Format an EnvironmentError exception.
918

919
  If the `err` argument has an errno attribute, it will be looked up
920
  and converted into a textual EXXXX description. Otherwise the string
921
  representation of the error will be returned.
922

923
  """
924
  if hasattr(err, 'errno'):
925
    detail = errno.errorcode[err.errno]
926
  else:
927
    detail = str(err)
928
  return detail
929

    
930

    
931
def _OSSearch(name, search_path=None):
932
  """Search for OSes with the given name in the search_path.
933

934
  Args:
935
    name: The name of the OS to look for
936
    search_path: List of dirs to search (defaults to constants.OS_SEARCH_PATH)
937

938
  Returns:
939
    The base_dir the OS resides in
940

941
  """
942
  if search_path is None:
943
    search_path = constants.OS_SEARCH_PATH
944

    
945
  for dir in search_path:
946
    t_os_dir = os.path.sep.join([dir, name])
947
    if os.path.isdir(t_os_dir):
948
        return dir
949

    
950
  return None
951

    
952

    
953
def _OSOndiskVersion(name, os_dir):
954
  """Compute and return the API version of a given OS.
955

956
  This function will try to read the API version of the os given by
957
  the 'name' parameter and residing in the 'os_dir' directory.
958

959
  Return value will be either an integer denoting the version or None in the
960
  case when this is not a valid OS name.
961

962
  """
963
  api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
964

    
965
  try:
966
    st = os.stat(api_file)
967
  except EnvironmentError, err:
968
    raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file not"
969
                           " found (%s)" % _ErrnoOrStr(err))
970

    
971
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
972
    raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file is not"
973
                           " a regular file")
974

    
975
  try:
976
    f = open(api_file)
977
    try:
978
      api_version = f.read(256)
979
    finally:
980
      f.close()
981
  except EnvironmentError, err:
982
    raise errors.InvalidOS(name, os_dir, "error while reading the"
983
                           " API version (%s)" % _ErrnoOrStr(err))
984

    
985
  api_version = api_version.strip()
986
  try:
987
    api_version = int(api_version)
988
  except (TypeError, ValueError), err:
989
    raise errors.InvalidOS(name, os_dir,
990
                           "API version is not integer (%s)" % str(err))
991

    
992
  return api_version
993

    
994

    
995
def DiagnoseOS(top_dirs=None):
996
  """Compute the validity for all OSes.
997

998
  Returns an OS object for each name in all the given top directories
999
  (if not given defaults to constants.OS_SEARCH_PATH)
1000

1001
  Returns:
1002
    list of OS objects
1003

1004
  """
1005
  if top_dirs is None:
1006
    top_dirs = constants.OS_SEARCH_PATH
1007

    
1008
  result = []
1009
  for dir in top_dirs:
1010
    if os.path.isdir(dir):
1011
      try:
1012
        f_names = utils.ListVisibleFiles(dir)
1013
      except EnvironmentError, err:
1014
        logger.Error("Can't list the OS directory %s: %s" % (dir,str(err)))
1015
        break
1016
      for name in f_names:
1017
        try:
1018
          os_inst = OSFromDisk(name, base_dir=dir)
1019
          result.append(os_inst)
1020
        except errors.InvalidOS, err:
1021
          result.append(objects.OS.FromInvalidOS(err))
1022

    
1023
  return result
1024

    
1025

    
1026
def OSFromDisk(name, base_dir=None):
1027
  """Create an OS instance from disk.
1028

1029
  This function will return an OS instance if the given name is a
1030
  valid OS name. Otherwise, it will raise an appropriate
1031
  `errors.InvalidOS` exception, detailing why this is not a valid
1032
  OS.
1033

1034
  Args:
1035
    os_dir: Directory containing the OS scripts. Defaults to a search
1036
            in all the OS_SEARCH_PATH directories.
1037

1038
  """
1039

    
1040
  if base_dir is None:
1041
    base_dir = _OSSearch(name)
1042

    
1043
  if base_dir is None:
1044
    raise errors.InvalidOS(name, None, "OS dir not found in search path")
1045

    
1046
  os_dir = os.path.sep.join([base_dir, name])
1047
  api_version = _OSOndiskVersion(name, os_dir)
1048

    
1049
  if api_version != constants.OS_API_VERSION:
1050
    raise errors.InvalidOS(name, os_dir, "API version mismatch"
1051
                           " (found %s want %s)"
1052
                           % (api_version, constants.OS_API_VERSION))
1053

    
1054
  # OS Scripts dictionary, we will populate it with the actual script names
1055
  os_scripts = {'create': '', 'export': '', 'import': '', 'rename': ''}
1056

    
1057
  for script in os_scripts:
1058
    os_scripts[script] = os.path.sep.join([os_dir, script])
1059

    
1060
    try:
1061
      st = os.stat(os_scripts[script])
1062
    except EnvironmentError, err:
1063
      raise errors.InvalidOS(name, os_dir, "'%s' script missing (%s)" %
1064
                             (script, _ErrnoOrStr(err)))
1065

    
1066
    if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1067
      raise errors.InvalidOS(name, os_dir, "'%s' script not executable" %
1068
                             script)
1069

    
1070
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1071
      raise errors.InvalidOS(name, os_dir, "'%s' is not a regular file" %
1072
                             script)
1073

    
1074

    
1075
  return objects.OS(name=name, path=os_dir, status=constants.OS_VALID_STATUS,
1076
                    create_script=os_scripts['create'],
1077
                    export_script=os_scripts['export'],
1078
                    import_script=os_scripts['import'],
1079
                    rename_script=os_scripts['rename'],
1080
                    api_version=api_version)
1081

    
1082

    
1083
def SnapshotBlockDevice(disk):
1084
  """Create a snapshot copy of a block device.
1085

1086
  This function is called recursively, and the snapshot is actually created
1087
  just for the leaf lvm backend device.
1088

1089
  Args:
1090
    disk: the disk to be snapshotted
1091

1092
  Returns:
1093
    a config entry for the actual lvm device snapshotted.
1094

1095
  """
1096
  if disk.children:
1097
    if len(disk.children) == 1:
1098
      # only one child, let's recurse on it
1099
      return SnapshotBlockDevice(disk.children[0])
1100
    else:
1101
      # more than one child, choose one that matches
1102
      for child in disk.children:
1103
        if child.size == disk.size:
1104
          # return implies breaking the loop
1105
          return SnapshotBlockDevice(child)
1106
  elif disk.dev_type == constants.LD_LV:
1107
    r_dev = _RecursiveFindBD(disk)
1108
    if r_dev is not None:
1109
      # let's stay on the safe side and ask for the full size, for now
1110
      return r_dev.Snapshot(disk.size)
1111
    else:
1112
      return None
1113
  else:
1114
    raise errors.ProgrammerError("Cannot snapshot non-lvm block device"
1115
                                 "'%s' of type '%s'" %
1116
                                 (disk.unique_id, disk.dev_type))
1117

    
1118

    
1119
def ExportSnapshot(disk, dest_node, instance):
1120
  """Export a block device snapshot to a remote node.
1121

1122
  Args:
1123
    disk: the snapshot block device
1124
    dest_node: the node to send the image to
1125
    instance: instance being exported
1126

1127
  Returns:
1128
    True if successful, False otherwise.
1129

1130
  """
1131
  inst_os = OSFromDisk(instance.os)
1132
  export_script = inst_os.export_script
1133

    
1134
  logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1135
                                     instance.name, int(time.time()))
1136
  if not os.path.exists(constants.LOG_OS_DIR):
1137
    os.mkdir(constants.LOG_OS_DIR, 0750)
1138

    
1139
  real_os_dev = _RecursiveFindBD(disk)
1140
  if real_os_dev is None:
1141
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
1142
                                  str(disk))
1143
  real_os_dev.Open()
1144

    
1145
  destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1146
  destfile = disk.physical_id[1]
1147

    
1148
  # the target command is built out of three individual commands,
1149
  # which are joined by pipes; we check each individual command for
1150
  # valid parameters
1151

    
1152
  expcmd = utils.BuildShellCmd("cd %s; %s -i %s -b %s 2>%s", inst_os.path,
1153
                               export_script, instance.name,
1154
                               real_os_dev.dev_path, logfile)
1155

    
1156
  comprcmd = "gzip"
1157

    
1158
  destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1159
                                destdir, destdir, destfile)
1160
  remotecmd = ssh.BuildSSHCmd(dest_node, constants.GANETI_RUNAS, destcmd)
1161

    
1162

    
1163

    
1164
  # all commands have been checked, so we're safe to combine them
1165
  command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1166

    
1167
  result = utils.RunCmd(command)
1168

    
1169
  if result.failed:
1170
    logger.Error("os snapshot export command '%s' returned error: %s"
1171
                 " output: %s" %
1172
                 (command, result.fail_reason, result.output))
1173
    return False
1174

    
1175
  return True
1176

    
1177

    
1178
def FinalizeExport(instance, snap_disks):
1179
  """Write out the export configuration information.
1180

1181
  Args:
1182
    instance: instance configuration
1183
    snap_disks: snapshot block devices
1184

1185
  Returns:
1186
    False in case of error, True otherwise.
1187

1188
  """
1189
  destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1190
  finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1191

    
1192
  config = objects.SerializableConfigParser()
1193

    
1194
  config.add_section(constants.INISECT_EXP)
1195
  config.set(constants.INISECT_EXP, 'version', '0')
1196
  config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1197
  config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1198
  config.set(constants.INISECT_EXP, 'os', instance.os)
1199
  config.set(constants.INISECT_EXP, 'compression', 'gzip')
1200

    
1201
  config.add_section(constants.INISECT_INS)
1202
  config.set(constants.INISECT_INS, 'name', instance.name)
1203
  config.set(constants.INISECT_INS, 'memory', '%d' % instance.memory)
1204
  config.set(constants.INISECT_INS, 'vcpus', '%d' % instance.vcpus)
1205
  config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1206
  for nic_count, nic in enumerate(instance.nics):
1207
    config.set(constants.INISECT_INS, 'nic%d_mac' %
1208
               nic_count, '%s' % nic.mac)
1209
    config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
1210
  # TODO: redundant: on load can read nics until it doesn't exist
1211
  config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_count)
1212

    
1213
  for disk_count, disk in enumerate(snap_disks):
1214
    config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
1215
               ('%s' % disk.iv_name))
1216
    config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
1217
               ('%s' % disk.physical_id[1]))
1218
    config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
1219
               ('%d' % disk.size))
1220
  config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_count)
1221

    
1222
  cff = os.path.join(destdir, constants.EXPORT_CONF_FILE)
1223
  cfo = open(cff, 'w')
1224
  try:
1225
    config.write(cfo)
1226
  finally:
1227
    cfo.close()
1228

    
1229
  shutil.rmtree(finaldestdir, True)
1230
  shutil.move(destdir, finaldestdir)
1231

    
1232
  return True
1233

    
1234

    
1235
def ExportInfo(dest):
1236
  """Get export configuration information.
1237

1238
  Args:
1239
    dest: directory containing the export
1240

1241
  Returns:
1242
    A serializable config file containing the export info.
1243

1244
  """
1245
  cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
1246

    
1247
  config = objects.SerializableConfigParser()
1248
  config.read(cff)
1249

    
1250
  if (not config.has_section(constants.INISECT_EXP) or
1251
      not config.has_section(constants.INISECT_INS)):
1252
    return None
1253

    
1254
  return config
1255

    
1256

    
1257
def ImportOSIntoInstance(instance, os_disk, swap_disk, src_node, src_image):
1258
  """Import an os image into an instance.
1259

1260
  Args:
1261
    instance: the instance object
1262
    os_disk: the instance-visible name of the os device
1263
    swap_disk: the instance-visible name of the swap device
1264
    src_node: node holding the source image
1265
    src_image: path to the source image on src_node
1266

1267
  Returns:
1268
    False in case of error, True otherwise.
1269

1270
  """
1271
  inst_os = OSFromDisk(instance.os)
1272
  import_script = inst_os.import_script
1273

    
1274
  os_device = instance.FindDisk(os_disk)
1275
  if os_device is None:
1276
    logger.Error("Can't find this device-visible name '%s'" % os_disk)
1277
    return False
1278

    
1279
  swap_device = instance.FindDisk(swap_disk)
1280
  if swap_device is None:
1281
    logger.Error("Can't find this device-visible name '%s'" % swap_disk)
1282
    return False
1283

    
1284
  real_os_dev = _RecursiveFindBD(os_device)
1285
  if real_os_dev is None:
1286
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
1287
                                  str(os_device))
1288
  real_os_dev.Open()
1289

    
1290
  real_swap_dev = _RecursiveFindBD(swap_device)
1291
  if real_swap_dev is None:
1292
    raise errors.BlockDeviceError("Block device '%s' is not set up" %
1293
                                  str(swap_device))
1294
  real_swap_dev.Open()
1295

    
1296
  logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
1297
                                        instance.name, int(time.time()))
1298
  if not os.path.exists(constants.LOG_OS_DIR):
1299
    os.mkdir(constants.LOG_OS_DIR, 0750)
1300

    
1301
  destcmd = utils.BuildShellCmd('cat %s', src_image)
1302
  remotecmd = ssh.BuildSSHCmd(src_node, constants.GANETI_RUNAS, destcmd)
1303

    
1304
  comprcmd = "gunzip"
1305
  impcmd = utils.BuildShellCmd("(cd %s; %s -i %s -b %s -s %s &>%s)",
1306
                               inst_os.path, import_script, instance.name,
1307
                               real_os_dev.dev_path, real_swap_dev.dev_path,
1308
                               logfile)
1309

    
1310
  command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
1311

    
1312
  result = utils.RunCmd(command)
1313

    
1314
  if result.failed:
1315
    logger.Error("os import command '%s' returned error: %s"
1316
                 " output: %s" %
1317
                 (command, result.fail_reason, result.output))
1318
    return False
1319

    
1320
  return True
1321

    
1322

    
1323
def ListExports():
1324
  """Return a list of exports currently available on this machine.
1325

1326
  """
1327
  if os.path.isdir(constants.EXPORT_DIR):
1328
    return utils.ListVisibleFiles(constants.EXPORT_DIR)
1329
  else:
1330
    return []
1331

    
1332

    
1333
def RemoveExport(export):
1334
  """Remove an existing export from the node.
1335

1336
  Args:
1337
    export: the name of the export to remove
1338

1339
  Returns:
1340
    False in case of error, True otherwise.
1341

1342
  """
1343
  target = os.path.join(constants.EXPORT_DIR, export)
1344

    
1345
  shutil.rmtree(target)
1346
  # TODO: catch some of the relevant exceptions and provide a pretty
1347
  # error message if rmtree fails.
1348

    
1349
  return True
1350

    
1351

    
1352
def RenameBlockDevices(devlist):
1353
  """Rename a list of block devices.
1354

1355
  The devlist argument is a list of tuples (disk, new_logical,
1356
  new_physical). The return value will be a combined boolean result
1357
  (True only if all renames succeeded).
1358

1359
  """
1360
  result = True
1361
  for disk, unique_id in devlist:
1362
    dev = _RecursiveFindBD(disk)
1363
    if dev is None:
1364
      result = False
1365
      continue
1366
    try:
1367
      old_rpath = dev.dev_path
1368
      dev.Rename(unique_id)
1369
      new_rpath = dev.dev_path
1370
      if old_rpath != new_rpath:
1371
        DevCacheManager.RemoveCache(old_rpath)
1372
        # FIXME: we should add the new cache information here, like:
1373
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
1374
        # but we don't have the owner here - maybe parse from existing
1375
        # cache? for now, we only lose lvm data when we rename, which
1376
        # is less critical than DRBD or MD
1377
    except errors.BlockDeviceError, err:
1378
      logger.Error("Can't rename device '%s' to '%s': %s" %
1379
                   (dev, unique_id, err))
1380
      result = False
1381
  return result
1382

    
1383

    
1384
class HooksRunner(object):
1385
  """Hook runner.
1386

1387
  This class is instantiated on the node side (ganeti-noded) and not on
1388
  the master side.
1389

1390
  """
1391
  RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
1392

    
1393
  def __init__(self, hooks_base_dir=None):
1394
    """Constructor for hooks runner.
1395

1396
    Args:
1397
      - hooks_base_dir: if not None, this overrides the
1398
        constants.HOOKS_BASE_DIR (useful for unittests)
1399
      - logs_base_dir: if not None, this overrides the
1400
        constants.LOG_HOOKS_DIR (useful for unittests)
1401
      - logging: enable or disable logging of script output
1402

1403
    """
1404
    if hooks_base_dir is None:
1405
      hooks_base_dir = constants.HOOKS_BASE_DIR
1406
    self._BASE_DIR = hooks_base_dir
1407

    
1408
  @staticmethod
1409
  def ExecHook(script, env):
1410
    """Exec one hook script.
1411

1412
    Args:
1413
     - phase: the phase
1414
     - script: the full path to the script
1415
     - env: the environment with which to exec the script
1416

1417
    """
1418
    # exec the process using subprocess and log the output
1419
    fdstdin = None
1420
    try:
1421
      fdstdin = open("/dev/null", "r")
1422
      child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
1423
                               stderr=subprocess.STDOUT, close_fds=True,
1424
                               shell=False, cwd="/",env=env)
1425
      output = ""
1426
      try:
1427
        output = child.stdout.read(4096)
1428
        child.stdout.close()
1429
      except EnvironmentError, err:
1430
        output += "Hook script error: %s" % str(err)
1431

    
1432
      while True:
1433
        try:
1434
          result = child.wait()
1435
          break
1436
        except EnvironmentError, err:
1437
          if err.errno == errno.EINTR:
1438
            continue
1439
          raise
1440
    finally:
1441
      # try not to leak fds
1442
      for fd in (fdstdin, ):
1443
        if fd is not None:
1444
          try:
1445
            fd.close()
1446
          except EnvironmentError, err:
1447
            # just log the error
1448
            #logger.Error("While closing fd %s: %s" % (fd, err))
1449
            pass
1450

    
1451
    return result == 0, output
1452

    
1453
  def RunHooks(self, hpath, phase, env):
1454
    """Run the scripts in the hooks directory.
1455

1456
    This method will not be usually overriden by child opcodes.
1457

1458
    """
1459
    if phase == constants.HOOKS_PHASE_PRE:
1460
      suffix = "pre"
1461
    elif phase == constants.HOOKS_PHASE_POST:
1462
      suffix = "post"
1463
    else:
1464
      raise errors.ProgrammerError("Unknown hooks phase: '%s'" % phase)
1465
    rr = []
1466

    
1467
    subdir = "%s-%s.d" % (hpath, suffix)
1468
    dir_name = "%s/%s" % (self._BASE_DIR, subdir)
1469
    try:
1470
      dir_contents = utils.ListVisibleFiles(dir_name)
1471
    except OSError, err:
1472
      # must log
1473
      return rr
1474

    
1475
    # we use the standard python sort order,
1476
    # so 00name is the recommended naming scheme
1477
    dir_contents.sort()
1478
    for relname in dir_contents:
1479
      fname = os.path.join(dir_name, relname)
1480
      if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
1481
          self.RE_MASK.match(relname) is not None):
1482
        rrval = constants.HKR_SKIP
1483
        output = ""
1484
      else:
1485
        result, output = self.ExecHook(fname, env)
1486
        if not result:
1487
          rrval = constants.HKR_FAIL
1488
        else:
1489
          rrval = constants.HKR_SUCCESS
1490
      rr.append(("%s/%s" % (subdir, relname), rrval, output))
1491

    
1492
    return rr
1493

    
1494

    
1495
class DevCacheManager(object):
1496
  """Simple class for managing a chache of block device information.
1497

1498
  """
1499
  _DEV_PREFIX = "/dev/"
1500
  _ROOT_DIR = constants.BDEV_CACHE_DIR
1501

    
1502
  @classmethod
1503
  def _ConvertPath(cls, dev_path):
1504
    """Converts a /dev/name path to the cache file name.
1505

1506
    This replaces slashes with underscores and strips the /dev
1507
    prefix. It then returns the full path to the cache file
1508

1509
    """
1510
    if dev_path.startswith(cls._DEV_PREFIX):
1511
      dev_path = dev_path[len(cls._DEV_PREFIX):]
1512
    dev_path = dev_path.replace("/", "_")
1513
    fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
1514
    return fpath
1515

    
1516
  @classmethod
1517
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
1518
    """Updates the cache information for a given device.
1519

1520
    """
1521
    fpath = cls._ConvertPath(dev_path)
1522
    if on_primary:
1523
      state = "primary"
1524
    else:
1525
      state = "secondary"
1526
    if iv_name is None:
1527
      iv_name = "not_visible"
1528
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
1529
    try:
1530
      utils.WriteFile(fpath, data=fdata)
1531
    except EnvironmentError, err:
1532
      logger.Error("Can't update bdev cache for %s, error %s" %
1533
                   (dev_path, str(err)))
1534

    
1535
  @classmethod
1536
  def RemoveCache(cls, dev_path):
1537
    """Remove data for a dev_path.
1538

1539
    """
1540
    fpath = cls._ConvertPath(dev_path)
1541
    try:
1542
      utils.RemoveFile(fpath)
1543
    except EnvironmentError, err:
1544
      logger.Error("Can't update bdev cache for %s, error %s" %
1545
                   (dev_path, str(err)))