Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ a0c3fea1

History | View | Annotate | Download (112.1 kB)

1
#!/usr/bin/python
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
"""Module implementing the master-side code."""
23

    
24
# pylint: disable-msg=W0613,W0201
25

    
26
import os
27
import os.path
28
import sha
29
import socket
30
import time
31
import tempfile
32
import re
33
import platform
34

    
35
from ganeti import rpc
36
from ganeti import ssh
37
from ganeti import logger
38
from ganeti import utils
39
from ganeti import errors
40
from ganeti import hypervisor
41
from ganeti import config
42
from ganeti import constants
43
from ganeti import objects
44
from ganeti import opcodes
45
from ganeti import ssconf
46

    
47
class LogicalUnit(object):
48
  """Logical Unit base class.
49

50
  Subclasses must follow these rules:
51
    - implement CheckPrereq which also fills in the opcode instance
52
      with all the fields (even if as None)
53
    - implement Exec
54
    - implement BuildHooksEnv
55
    - redefine HPATH and HTYPE
56
    - optionally redefine their run requirements (REQ_CLUSTER,
57
      REQ_MASTER); note that all commands require root permissions
58

59
  """
60
  HPATH = None
61
  HTYPE = None
62
  _OP_REQP = []
63
  REQ_CLUSTER = True
64
  REQ_MASTER = True
65

    
66
  def __init__(self, processor, op, cfg, sstore):
67
    """Constructor for LogicalUnit.
68

69
    This needs to be overriden in derived classes in order to check op
70
    validity.
71

72
    """
73
    self.processor = processor
74
    self.op = op
75
    self.cfg = cfg
76
    self.sstore = sstore
77
    for attr_name in self._OP_REQP:
78
      attr_val = getattr(op, attr_name, None)
79
      if attr_val is None:
80
        raise errors.OpPrereqError, ("Required parameter '%s' missing" %
81
                                     attr_name)
82
    if self.REQ_CLUSTER:
83
      if not cfg.IsCluster():
84
        raise errors.OpPrereqError, ("Cluster not initialized yet,"
85
                                     " use 'gnt-cluster init' first.")
86
      if self.REQ_MASTER:
87
        master = sstore.GetMasterNode()
88
        if master != socket.gethostname():
89
          raise errors.OpPrereqError, ("Commands must be run on the master"
90
                                       " node %s" % master)
91

    
92
  def CheckPrereq(self):
93
    """Check prerequisites for this LU.
94

95
    This method should check that the prerequisites for the execution
96
    of this LU are fulfilled. It can do internode communication, but
97
    it should be idempotent - no cluster or system changes are
98
    allowed.
99

100
    The method should raise errors.OpPrereqError in case something is
101
    not fulfilled. Its return value is ignored.
102

103
    This method should also update all the parameters of the opcode to
104
    their canonical form; e.g. a short node name must be fully
105
    expanded after this method has successfully completed (so that
106
    hooks, logging, etc. work correctly).
107

108
    """
109
    raise NotImplementedError
110

    
111
  def Exec(self, feedback_fn):
112
    """Execute the LU.
113

114
    This method should implement the actual work. It should raise
115
    errors.OpExecError for failures that are somewhat dealt with in
116
    code, or expected.
117

118
    """
119
    raise NotImplementedError
120

    
121
  def BuildHooksEnv(self):
122
    """Build hooks environment for this LU.
123

124
    This method should return a three-node tuple consisting of: a dict
125
    containing the environment that will be used for running the
126
    specific hook for this LU, a list of node names on which the hook
127
    should run before the execution, and a list of node names on which
128
    the hook should run after the execution.
129

130
    The keys of the dict must not have 'GANETI_' prefixed as this will
131
    be handled in the hooks runner. Also note additional keys will be
132
    added by the hooks runner. If the LU doesn't define any
133
    environment, an empty dict (and not None) should be returned.
134

135
    As for the node lists, the master should not be included in the
136
    them, as it will be added by the hooks runner in case this LU
137
    requires a cluster to run on (otherwise we don't have a node
138
    list). No nodes should be returned as an empty list (and not
139
    None).
140

141
    Note that if the HPATH for a LU class is None, this function will
142
    not be called.
143

144
    """
145
    raise NotImplementedError
146

    
147

    
148
class NoHooksLU(LogicalUnit):
149
  """Simple LU which runs no hooks.
150

151
  This LU is intended as a parent for other LogicalUnits which will
152
  run no hooks, in order to reduce duplicate code.
153

154
  """
155
  HPATH = None
156
  HTYPE = None
157

    
158
  def BuildHooksEnv(self):
159
    """Build hooks env.
160

161
    This is a no-op, since we don't run hooks.
162

163
    """
164
    return
165

    
166

    
167
def _GetWantedNodes(lu, nodes):
168
  """Returns list of checked and expanded nodes.
169

170
  Args:
171
    nodes: List of nodes (strings) or None for all
172

173
  """
174
  if nodes is not None and not isinstance(nodes, list):
175
    raise errors.OpPrereqError, "Invalid argument type 'nodes'"
176

    
177
  if nodes:
178
    wanted_nodes = []
179

    
180
    for name in nodes:
181
      node = lu.cfg.GetNodeInfo(lu.cfg.ExpandNodeName(name))
182
      if node is None:
183
        raise errors.OpPrereqError, ("No such node name '%s'" % name)
184
    wanted_nodes.append(node)
185

    
186
    return wanted_nodes
187
  else:
188
    return [lu.cfg.GetNodeInfo(name) for name in lu.cfg.GetNodeList()]
189

    
190

    
191
def _CheckOutputFields(static, dynamic, selected):
192
  """Checks whether all selected fields are valid.
193

194
  Args:
195
    static: Static fields
196
    dynamic: Dynamic fields
197

198
  """
199
  static_fields = frozenset(static)
200
  dynamic_fields = frozenset(dynamic)
201

    
202
  all_fields = static_fields | dynamic_fields
203

    
204
  if not all_fields.issuperset(selected):
205
    raise errors.OpPrereqError, ("Unknown output fields selected: %s"
206
                                 % ",".join(frozenset(selected).
207
                                            difference(all_fields)))
208

    
209

    
210
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
211
                          memory, vcpus, nics):
212
  """Builds instance related env variables for hooks from single variables.
213

214
  Args:
215
    secondary_nodes: List of secondary nodes as strings
216
  """
217
  env = {
218
    "INSTANCE_NAME": name,
219
    "INSTANCE_PRIMARY": primary_node,
220
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
221
    "INSTANCE_OS_TYPE": os_type,
222
    "INSTANCE_STATUS": status,
223
    "INSTANCE_MEMORY": memory,
224
    "INSTANCE_VCPUS": vcpus,
225
  }
226

    
227
  if nics:
228
    nic_count = len(nics)
229
    for idx, (ip, bridge) in enumerate(nics):
230
      if ip is None:
231
        ip = ""
232
      env["INSTANCE_NIC%d_IP" % idx] = ip
233
      env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
234
  else:
235
    nic_count = 0
236

    
237
  env["INSTANCE_NIC_COUNT"] = nic_count
238

    
239
  return env
240

    
241

    
242
def _BuildInstanceHookEnvByObject(instance, override=None):
243
  """Builds instance related env variables for hooks from an object.
244

245
  Args:
246
    instance: objects.Instance object of instance
247
    override: dict of values to override
248
  """
249
  args = {
250
    'name': instance.name,
251
    'primary_node': instance.primary_node,
252
    'secondary_nodes': instance.secondary_nodes,
253
    'os_type': instance.os,
254
    'status': instance.os,
255
    'memory': instance.memory,
256
    'vcpus': instance.vcpus,
257
    'nics': [(nic.ip, nic.bridge) for nic in instance.nics],
258
  }
259
  if override:
260
    args.update(override)
261
  return _BuildInstanceHookEnv(**args)
262

    
263

    
264
def _UpdateEtcHosts(fullnode, ip):
265
  """Ensure a node has a correct entry in /etc/hosts.
266

267
  Args:
268
    fullnode - Fully qualified domain name of host. (str)
269
    ip       - IPv4 address of host (str)
270

271
  """
272
  node = fullnode.split(".", 1)[0]
273

    
274
  f = open('/etc/hosts', 'r+')
275

    
276
  inthere = False
277

    
278
  save_lines = []
279
  add_lines = []
280
  removed = False
281

    
282
  while True:
283
    rawline = f.readline()
284

    
285
    if not rawline:
286
      # End of file
287
      break
288

    
289
    line = rawline.split('\n')[0]
290

    
291
    # Strip off comments
292
    line = line.split('#')[0]
293

    
294
    if not line:
295
      # Entire line was comment, skip
296
      save_lines.append(rawline)
297
      continue
298

    
299
    fields = line.split()
300

    
301
    haveall = True
302
    havesome = False
303
    for spec in [ ip, fullnode, node ]:
304
      if spec not in fields:
305
        haveall = False
306
      if spec in fields:
307
        havesome = True
308

    
309
    if haveall:
310
      inthere = True
311
      save_lines.append(rawline)
312
      continue
313

    
314
    if havesome and not haveall:
315
      # Line (old, or manual?) which is missing some.  Remove.
316
      removed = True
317
      continue
318

    
319
    save_lines.append(rawline)
320

    
321
  if not inthere:
322
    add_lines.append('%s\t%s %s\n' % (ip, fullnode, node))
323

    
324
  if removed:
325
    if add_lines:
326
      save_lines = save_lines + add_lines
327

    
328
    # We removed a line, write a new file and replace old.
329
    fd, tmpname = tempfile.mkstemp('tmp', 'hosts_', '/etc')
330
    newfile = os.fdopen(fd, 'w')
331
    newfile.write(''.join(save_lines))
332
    newfile.close()
333
    os.rename(tmpname, '/etc/hosts')
334

    
335
  elif add_lines:
336
    # Simply appending a new line will do the trick.
337
    f.seek(0, 2)
338
    for add in add_lines:
339
      f.write(add)
340

    
341
  f.close()
342

    
343

    
344
def _UpdateKnownHosts(fullnode, ip, pubkey):
345
  """Ensure a node has a correct known_hosts entry.
346

347
  Args:
348
    fullnode - Fully qualified domain name of host. (str)
349
    ip       - IPv4 address of host (str)
350
    pubkey   - the public key of the cluster
351

352
  """
353
  if os.path.exists('/etc/ssh/ssh_known_hosts'):
354
    f = open('/etc/ssh/ssh_known_hosts', 'r+')
355
  else:
356
    f = open('/etc/ssh/ssh_known_hosts', 'w+')
357

    
358
  inthere = False
359

    
360
  save_lines = []
361
  add_lines = []
362
  removed = False
363

    
364
  while True:
365
    rawline = f.readline()
366
    logger.Debug('read %s' % (repr(rawline),))
367

    
368
    if not rawline:
369
      # End of file
370
      break
371

    
372
    line = rawline.split('\n')[0]
373

    
374
    parts = line.split(' ')
375
    fields = parts[0].split(',')
376
    key = parts[2]
377

    
378
    haveall = True
379
    havesome = False
380
    for spec in [ ip, fullnode ]:
381
      if spec not in fields:
382
        haveall = False
383
      if spec in fields:
384
        havesome = True
385

    
386
    logger.Debug("key, pubkey = %s." % (repr((key, pubkey)),))
387
    if haveall and key == pubkey:
388
      inthere = True
389
      save_lines.append(rawline)
390
      logger.Debug("Keeping known_hosts '%s'." % (repr(rawline),))
391
      continue
392

    
393
    if havesome and (not haveall or key != pubkey):
394
      removed = True
395
      logger.Debug("Discarding known_hosts '%s'." % (repr(rawline),))
396
      continue
397

    
398
    save_lines.append(rawline)
399

    
400
  if not inthere:
401
    add_lines.append('%s,%s ssh-rsa %s\n' % (fullnode, ip, pubkey))
402
    logger.Debug("Adding known_hosts '%s'." % (repr(add_lines[-1]),))
403

    
404
  if removed:
405
    save_lines = save_lines + add_lines
406

    
407
    # Write a new file and replace old.
408
    fd, tmpname = tempfile.mkstemp('tmp', 'ssh_known_hosts_', '/etc/ssh')
409
    newfile = os.fdopen(fd, 'w')
410
    newfile.write(''.join(save_lines))
411
    newfile.close()
412
    logger.Debug("Wrote new known_hosts.")
413
    os.rename(tmpname, '/etc/ssh/ssh_known_hosts')
414

    
415
  elif add_lines:
416
    # Simply appending a new line will do the trick.
417
    f.seek(0, 2)
418
    for add in add_lines:
419
      f.write(add)
420

    
421
  f.close()
422

    
423

    
424
def _HasValidVG(vglist, vgname):
425
  """Checks if the volume group list is valid.
426

427
  A non-None return value means there's an error, and the return value
428
  is the error message.
429

430
  """
431
  vgsize = vglist.get(vgname, None)
432
  if vgsize is None:
433
    return "volume group '%s' missing" % vgname
434
  elif vgsize < 20480:
435
    return ("volume group '%s' too small (20480MiB required, %dMib found)" %
436
            (vgname, vgsize))
437
  return None
438

    
439

    
440
def _InitSSHSetup(node):
441
  """Setup the SSH configuration for the cluster.
442

443

444
  This generates a dsa keypair for root, adds the pub key to the
445
  permitted hosts and adds the hostkey to its own known hosts.
446

447
  Args:
448
    node: the name of this host as a fqdn
449

450
  """
451
  utils.RemoveFile('/root/.ssh/known_hosts')
452

    
453
  if os.path.exists('/root/.ssh/id_dsa'):
454
    utils.CreateBackup('/root/.ssh/id_dsa')
455
  if os.path.exists('/root/.ssh/id_dsa.pub'):
456
    utils.CreateBackup('/root/.ssh/id_dsa.pub')
457

    
458
  utils.RemoveFile('/root/.ssh/id_dsa')
459
  utils.RemoveFile('/root/.ssh/id_dsa.pub')
460

    
461
  result = utils.RunCmd(["ssh-keygen", "-t", "dsa",
462
                         "-f", "/root/.ssh/id_dsa",
463
                         "-q", "-N", ""])
464
  if result.failed:
465
    raise errors.OpExecError, ("could not generate ssh keypair, error %s" %
466
                               result.output)
467

    
468
  f = open('/root/.ssh/id_dsa.pub', 'r')
469
  try:
470
    utils.AddAuthorizedKey('/root/.ssh/authorized_keys', f.read(8192))
471
  finally:
472
    f.close()
473

    
474

    
475
def _InitGanetiServerSetup(ss):
476
  """Setup the necessary configuration for the initial node daemon.
477

478
  This creates the nodepass file containing the shared password for
479
  the cluster and also generates the SSL certificate.
480

481
  """
482
  # Create pseudo random password
483
  randpass = sha.new(os.urandom(64)).hexdigest()
484
  # and write it into sstore
485
  ss.SetKey(ss.SS_NODED_PASS, randpass)
486

    
487
  result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
488
                         "-days", str(365*5), "-nodes", "-x509",
489
                         "-keyout", constants.SSL_CERT_FILE,
490
                         "-out", constants.SSL_CERT_FILE, "-batch"])
491
  if result.failed:
492
    raise errors.OpExecError, ("could not generate server ssl cert, command"
493
                               " %s had exitcode %s and error message %s" %
494
                               (result.cmd, result.exit_code, result.output))
495

    
496
  os.chmod(constants.SSL_CERT_FILE, 0400)
497

    
498
  result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
499

    
500
  if result.failed:
501
    raise errors.OpExecError, ("could not start the node daemon, command %s"
502
                               " had exitcode %s and error %s" %
503
                               (result.cmd, result.exit_code, result.output))
504

    
505

    
506
class LUInitCluster(LogicalUnit):
507
  """Initialise the cluster.
508

509
  """
510
  HPATH = "cluster-init"
511
  HTYPE = constants.HTYPE_CLUSTER
512
  _OP_REQP = ["cluster_name", "hypervisor_type", "vg_name", "mac_prefix",
513
              "def_bridge", "master_netdev"]
514
  REQ_CLUSTER = False
515

    
516
  def BuildHooksEnv(self):
517
    """Build hooks env.
518

519
    Notes: Since we don't require a cluster, we must manually add
520
    ourselves in the post-run node list.
521

522
    """
523
    env = {
524
      "CLUSTER": self.op.cluster_name,
525
      "MASTER": self.hostname['hostname_full'],
526
      }
527
    return env, [], [self.hostname['hostname_full']]
528

    
529
  def CheckPrereq(self):
530
    """Verify that the passed name is a valid one.
531

532
    """
533
    if config.ConfigWriter.IsCluster():
534
      raise errors.OpPrereqError, ("Cluster is already initialised")
535

    
536
    hostname_local = socket.gethostname()
537
    self.hostname = hostname = utils.LookupHostname(hostname_local)
538
    if not hostname:
539
      raise errors.OpPrereqError, ("Cannot resolve my own hostname ('%s')" %
540
                                   hostname_local)
541

    
542
    self.clustername = clustername = utils.LookupHostname(self.op.cluster_name)
543
    if not clustername:
544
      raise errors.OpPrereqError, ("Cannot resolve given cluster name ('%s')"
545
                                   % self.op.cluster_name)
546

    
547
    result = utils.RunCmd(["fping", "-S127.0.0.1", "-q", hostname['ip']])
548
    if result.failed:
549
      raise errors.OpPrereqError, ("Inconsistency: this host's name resolves"
550
                                   " to %s,\nbut this ip address does not"
551
                                   " belong to this host."
552
                                   " Aborting." % hostname['ip'])
553

    
554
    secondary_ip = getattr(self.op, "secondary_ip", None)
555
    if secondary_ip and not utils.IsValidIP(secondary_ip):
556
      raise errors.OpPrereqError, ("Invalid secondary ip given")
557
    if secondary_ip and secondary_ip != hostname['ip']:
558
      result = utils.RunCmd(["fping", "-S127.0.0.1", "-q", secondary_ip])
559
      if result.failed:
560
        raise errors.OpPrereqError, ("You gave %s as secondary IP,\n"
561
                                     "but it does not belong to this host." %
562
                                     secondary_ip)
563
    self.secondary_ip = secondary_ip
564

    
565
    # checks presence of the volume group given
566
    vgstatus = _HasValidVG(utils.ListVolumeGroups(), self.op.vg_name)
567

    
568
    if vgstatus:
569
      raise errors.OpPrereqError, ("Error: %s" % vgstatus)
570

    
571
    if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$",
572
                    self.op.mac_prefix):
573
      raise errors.OpPrereqError, ("Invalid mac prefix given '%s'" %
574
                                   self.op.mac_prefix)
575

    
576
    if self.op.hypervisor_type not in hypervisor.VALID_HTYPES:
577
      raise errors.OpPrereqError, ("Invalid hypervisor type given '%s'" %
578
                                   self.op.hypervisor_type)
579

    
580
    result = utils.RunCmd(["ip", "link", "show", "dev", self.op.master_netdev])
581
    if result.failed:
582
      raise errors.OpPrereqError, ("Invalid master netdev given (%s): '%s'" %
583
                                   (self.op.master_netdev, result.output))
584

    
585
  def Exec(self, feedback_fn):
586
    """Initialize the cluster.
587

588
    """
589
    clustername = self.clustername
590
    hostname = self.hostname
591

    
592
    # set up the simple store
593
    ss = ssconf.SimpleStore()
594
    ss.SetKey(ss.SS_HYPERVISOR, self.op.hypervisor_type)
595
    ss.SetKey(ss.SS_MASTER_NODE, hostname['hostname_full'])
596
    ss.SetKey(ss.SS_MASTER_IP, clustername['ip'])
597
    ss.SetKey(ss.SS_MASTER_NETDEV, self.op.master_netdev)
598
    ss.SetKey(ss.SS_CLUSTER_NAME, clustername['hostname'])
599

    
600
    # set up the inter-node password and certificate
601
    _InitGanetiServerSetup(ss)
602

    
603
    # start the master ip
604
    rpc.call_node_start_master(hostname['hostname_full'])
605

    
606
    # set up ssh config and /etc/hosts
607
    f = open('/etc/ssh/ssh_host_rsa_key.pub', 'r')
608
    try:
609
      sshline = f.read()
610
    finally:
611
      f.close()
612
    sshkey = sshline.split(" ")[1]
613

    
614
    _UpdateEtcHosts(hostname['hostname_full'],
615
                    hostname['ip'],
616
                    )
617

    
618
    _UpdateKnownHosts(hostname['hostname_full'],
619
                      hostname['ip'],
620
                      sshkey,
621
                      )
622

    
623
    _InitSSHSetup(hostname['hostname'])
624

    
625
    # init of cluster config file
626
    cfgw = config.ConfigWriter()
627
    cfgw.InitConfig(hostname['hostname'], hostname['ip'], self.secondary_ip,
628
                    sshkey, self.op.mac_prefix,
629
                    self.op.vg_name, self.op.def_bridge)
630

    
631

    
632
class LUDestroyCluster(NoHooksLU):
633
  """Logical unit for destroying the cluster.
634

635
  """
636
  _OP_REQP = []
637

    
638
  def CheckPrereq(self):
639
    """Check prerequisites.
640

641
    This checks whether the cluster is empty.
642

643
    Any errors are signalled by raising errors.OpPrereqError.
644

645
    """
646
    master = self.sstore.GetMasterNode()
647

    
648
    nodelist = self.cfg.GetNodeList()
649
    if len(nodelist) != 1 or nodelist[0] != master:
650
      raise errors.OpPrereqError, ("There are still %d node(s) in "
651
                                   "this cluster." % (len(nodelist) - 1))
652
    instancelist = self.cfg.GetInstanceList()
653
    if instancelist:
654
      raise errors.OpPrereqError, ("There are still %d instance(s) in "
655
                                   "this cluster." % len(instancelist))
656

    
657
  def Exec(self, feedback_fn):
658
    """Destroys the cluster.
659

660
    """
661
    utils.CreateBackup('/root/.ssh/id_dsa')
662
    utils.CreateBackup('/root/.ssh/id_dsa.pub')
663
    rpc.call_node_leave_cluster(self.sstore.GetMasterNode())
664

    
665

    
666
class LUVerifyCluster(NoHooksLU):
667
  """Verifies the cluster status.
668

669
  """
670
  _OP_REQP = []
671

    
672
  def _VerifyNode(self, node, file_list, local_cksum, vglist, node_result,
673
                  remote_version, feedback_fn):
674
    """Run multiple tests against a node.
675

676
    Test list:
677
      - compares ganeti version
678
      - checks vg existance and size > 20G
679
      - checks config file checksum
680
      - checks ssh to other nodes
681

682
    Args:
683
      node: name of the node to check
684
      file_list: required list of files
685
      local_cksum: dictionary of local files and their checksums
686

687
    """
688
    # compares ganeti version
689
    local_version = constants.PROTOCOL_VERSION
690
    if not remote_version:
691
      feedback_fn(" - ERROR: connection to %s failed" % (node))
692
      return True
693

    
694
    if local_version != remote_version:
695
      feedback_fn("  - ERROR: sw version mismatch: master %s, node(%s) %s" %
696
                      (local_version, node, remote_version))
697
      return True
698

    
699
    # checks vg existance and size > 20G
700

    
701
    bad = False
702
    if not vglist:
703
      feedback_fn("  - ERROR: unable to check volume groups on node %s." %
704
                      (node,))
705
      bad = True
706
    else:
707
      vgstatus = _HasValidVG(vglist, self.cfg.GetVGName())
708
      if vgstatus:
709
        feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
710
        bad = True
711

    
712
    # checks config file checksum
713
    # checks ssh to any
714

    
715
    if 'filelist' not in node_result:
716
      bad = True
717
      feedback_fn("  - ERROR: node hasn't returned file checksum data")
718
    else:
719
      remote_cksum = node_result['filelist']
720
      for file_name in file_list:
721
        if file_name not in remote_cksum:
722
          bad = True
723
          feedback_fn("  - ERROR: file '%s' missing" % file_name)
724
        elif remote_cksum[file_name] != local_cksum[file_name]:
725
          bad = True
726
          feedback_fn("  - ERROR: file '%s' has wrong checksum" % file_name)
727

    
728
    if 'nodelist' not in node_result:
729
      bad = True
730
      feedback_fn("  - ERROR: node hasn't returned node connectivity data")
731
    else:
732
      if node_result['nodelist']:
733
        bad = True
734
        for node in node_result['nodelist']:
735
          feedback_fn("  - ERROR: communication with node '%s': %s" %
736
                          (node, node_result['nodelist'][node]))
737
    hyp_result = node_result.get('hypervisor', None)
738
    if hyp_result is not None:
739
      feedback_fn("  - ERROR: hypervisor verify failure: '%s'" % hyp_result)
740
    return bad
741

    
742
  def _VerifyInstance(self, instance, node_vol_is, node_instance, feedback_fn):
743
    """Verify an instance.
744

745
    This function checks to see if the required block devices are
746
    available on the instance's node.
747

748
    """
749
    bad = False
750

    
751
    instancelist = self.cfg.GetInstanceList()
752
    if not instance in instancelist:
753
      feedback_fn("  - ERROR: instance %s not in instance list %s" %
754
                      (instance, instancelist))
755
      bad = True
756

    
757
    instanceconfig = self.cfg.GetInstanceInfo(instance)
758
    node_current = instanceconfig.primary_node
759

    
760
    node_vol_should = {}
761
    instanceconfig.MapLVsByNode(node_vol_should)
762

    
763
    for node in node_vol_should:
764
      for volume in node_vol_should[node]:
765
        if node not in node_vol_is or volume not in node_vol_is[node]:
766
          feedback_fn("  - ERROR: volume %s missing on node %s" %
767
                          (volume, node))
768
          bad = True
769

    
770
    if not instanceconfig.status == 'down':
771
      if not instance in node_instance[node_current]:
772
        feedback_fn("  - ERROR: instance %s not running on node %s" %
773
                        (instance, node_current))
774
        bad = True
775

    
776
    for node in node_instance:
777
      if (not node == node_current):
778
        if instance in node_instance[node]:
779
          feedback_fn("  - ERROR: instance %s should not run on node %s" %
780
                          (instance, node))
781
          bad = True
782

    
783
    return not bad
784

    
785
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
786
    """Verify if there are any unknown volumes in the cluster.
787

788
    The .os, .swap and backup volumes are ignored. All other volumes are
789
    reported as unknown.
790

791
    """
792
    bad = False
793

    
794
    for node in node_vol_is:
795
      for volume in node_vol_is[node]:
796
        if node not in node_vol_should or volume not in node_vol_should[node]:
797
          feedback_fn("  - ERROR: volume %s on node %s should not exist" %
798
                      (volume, node))
799
          bad = True
800
    return bad
801

    
802
  def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
803
    """Verify the list of running instances.
804

805
    This checks what instances are running but unknown to the cluster.
806

807
    """
808
    bad = False
809
    for node in node_instance:
810
      for runninginstance in node_instance[node]:
811
        if runninginstance not in instancelist:
812
          feedback_fn("  - ERROR: instance %s on node %s should not exist" %
813
                          (runninginstance, node))
814
          bad = True
815
    return bad
816

    
817
  def CheckPrereq(self):
818
    """Check prerequisites.
819

820
    This has no prerequisites.
821

822
    """
823
    pass
824

    
825
  def Exec(self, feedback_fn):
826
    """Verify integrity of cluster, performing various test on nodes.
827

828
    """
829
    bad = False
830
    feedback_fn("* Verifying global settings")
831
    self.cfg.VerifyConfig()
832

    
833
    master = self.sstore.GetMasterNode()
834
    vg_name = self.cfg.GetVGName()
835
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
836
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
837
    node_volume = {}
838
    node_instance = {}
839

    
840
    # FIXME: verify OS list
841
    # do local checksums
842
    file_names = list(self.sstore.GetFileList())
843
    file_names.append(constants.SSL_CERT_FILE)
844
    file_names.append(constants.CLUSTER_CONF_FILE)
845
    local_checksums = utils.FingerprintFiles(file_names)
846

    
847
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
848
    all_volumeinfo = rpc.call_volume_list(nodelist, vg_name)
849
    all_instanceinfo = rpc.call_instance_list(nodelist)
850
    all_vglist = rpc.call_vg_list(nodelist)
851
    node_verify_param = {
852
      'filelist': file_names,
853
      'nodelist': nodelist,
854
      'hypervisor': None,
855
      }
856
    all_nvinfo = rpc.call_node_verify(nodelist, node_verify_param)
857
    all_rversion = rpc.call_version(nodelist)
858

    
859
    for node in nodelist:
860
      feedback_fn("* Verifying node %s" % node)
861
      result = self._VerifyNode(node, file_names, local_checksums,
862
                                all_vglist[node], all_nvinfo[node],
863
                                all_rversion[node], feedback_fn)
864
      bad = bad or result
865

    
866
      # node_volume
867
      volumeinfo = all_volumeinfo[node]
868

    
869
      if type(volumeinfo) != dict:
870
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
871
        bad = True
872
        continue
873

    
874
      node_volume[node] = volumeinfo
875

    
876
      # node_instance
877
      nodeinstance = all_instanceinfo[node]
878
      if type(nodeinstance) != list:
879
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
880
        bad = True
881
        continue
882

    
883
      node_instance[node] = nodeinstance
884

    
885
    node_vol_should = {}
886

    
887
    for instance in instancelist:
888
      feedback_fn("* Verifying instance %s" % instance)
889
      result =  self._VerifyInstance(instance, node_volume, node_instance,
890
                                     feedback_fn)
891
      bad = bad or result
892

    
893
      inst_config = self.cfg.GetInstanceInfo(instance)
894

    
895
      inst_config.MapLVsByNode(node_vol_should)
896

    
897
    feedback_fn("* Verifying orphan volumes")
898
    result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
899
                                       feedback_fn)
900
    bad = bad or result
901

    
902
    feedback_fn("* Verifying remaining instances")
903
    result = self._VerifyOrphanInstances(instancelist, node_instance,
904
                                         feedback_fn)
905
    bad = bad or result
906

    
907
    return int(bad)
908

    
909

    
910
def _WaitForSync(cfgw, instance, oneshot=False, unlock=False):
911
  """Sleep and poll for an instance's disk to sync.
912

913
  """
914
  if not instance.disks:
915
    return True
916

    
917
  if not oneshot:
918
    logger.ToStdout("Waiting for instance %s to sync disks." % instance.name)
919

    
920
  node = instance.primary_node
921

    
922
  for dev in instance.disks:
923
    cfgw.SetDiskID(dev, node)
924

    
925
  retries = 0
926
  while True:
927
    max_time = 0
928
    done = True
929
    cumul_degraded = False
930
    rstats = rpc.call_blockdev_getmirrorstatus(node, instance.disks)
931
    if not rstats:
932
      logger.ToStderr("Can't get any data from node %s" % node)
933
      retries += 1
934
      if retries >= 10:
935
        raise errors.RemoteError, ("Can't contact node %s for mirror data,"
936
                                   " aborting." % node)
937
      time.sleep(6)
938
      continue
939
    retries = 0
940
    for i in range(len(rstats)):
941
      mstat = rstats[i]
942
      if mstat is None:
943
        logger.ToStderr("Can't compute data for node %s/%s" %
944
                        (node, instance.disks[i].iv_name))
945
        continue
946
      perc_done, est_time, is_degraded = mstat
947
      cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
948
      if perc_done is not None:
949
        done = False
950
        if est_time is not None:
951
          rem_time = "%d estimated seconds remaining" % est_time
952
          max_time = est_time
953
        else:
954
          rem_time = "no time estimate"
955
        logger.ToStdout("- device %s: %5.2f%% done, %s" %
956
                        (instance.disks[i].iv_name, perc_done, rem_time))
957
    if done or oneshot:
958
      break
959

    
960
    if unlock:
961
      utils.Unlock('cmd')
962
    try:
963
      time.sleep(min(60, max_time))
964
    finally:
965
      if unlock:
966
        utils.Lock('cmd')
967

    
968
  if done:
969
    logger.ToStdout("Instance %s's disks are in sync." % instance.name)
970
  return not cumul_degraded
971

    
972

    
973
def _CheckDiskConsistency(cfgw, dev, node, on_primary):
974
  """Check that mirrors are not degraded.
975

976
  """
977
  cfgw.SetDiskID(dev, node)
978

    
979
  result = True
980
  if on_primary or dev.AssembleOnSecondary():
981
    rstats = rpc.call_blockdev_find(node, dev)
982
    if not rstats:
983
      logger.ToStderr("Can't get any data from node %s" % node)
984
      result = False
985
    else:
986
      result = result and (not rstats[5])
987
  if dev.children:
988
    for child in dev.children:
989
      result = result and _CheckDiskConsistency(cfgw, child, node, on_primary)
990

    
991
  return result
992

    
993

    
994
class LUDiagnoseOS(NoHooksLU):
995
  """Logical unit for OS diagnose/query.
996

997
  """
998
  _OP_REQP = []
999

    
1000
  def CheckPrereq(self):
1001
    """Check prerequisites.
1002

1003
    This always succeeds, since this is a pure query LU.
1004

1005
    """
1006
    return
1007

    
1008
  def Exec(self, feedback_fn):
1009
    """Compute the list of OSes.
1010

1011
    """
1012
    node_list = self.cfg.GetNodeList()
1013
    node_data = rpc.call_os_diagnose(node_list)
1014
    if node_data == False:
1015
      raise errors.OpExecError, "Can't gather the list of OSes"
1016
    return node_data
1017

    
1018

    
1019
class LURemoveNode(LogicalUnit):
1020
  """Logical unit for removing a node.
1021

1022
  """
1023
  HPATH = "node-remove"
1024
  HTYPE = constants.HTYPE_NODE
1025
  _OP_REQP = ["node_name"]
1026

    
1027
  def BuildHooksEnv(self):
1028
    """Build hooks env.
1029

1030
    This doesn't run on the target node in the pre phase as a failed
1031
    node would not allows itself to run.
1032

1033
    """
1034
    env = {
1035
      "NODE_NAME": self.op.node_name,
1036
      }
1037
    all_nodes = self.cfg.GetNodeList()
1038
    all_nodes.remove(self.op.node_name)
1039
    return env, all_nodes, all_nodes
1040

    
1041
  def CheckPrereq(self):
1042
    """Check prerequisites.
1043

1044
    This checks:
1045
     - the node exists in the configuration
1046
     - it does not have primary or secondary instances
1047
     - it's not the master
1048

1049
    Any errors are signalled by raising errors.OpPrereqError.
1050

1051
    """
1052
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1053
    if node is None:
1054
      logger.Error("Error: Node '%s' is unknown." % self.op.node_name)
1055
      return 1
1056

    
1057
    instance_list = self.cfg.GetInstanceList()
1058

    
1059
    masternode = self.sstore.GetMasterNode()
1060
    if node.name == masternode:
1061
      raise errors.OpPrereqError, ("Node is the master node,"
1062
                                   " you need to failover first.")
1063

    
1064
    for instance_name in instance_list:
1065
      instance = self.cfg.GetInstanceInfo(instance_name)
1066
      if node.name == instance.primary_node:
1067
        raise errors.OpPrereqError, ("Instance %s still running on the node,"
1068
                                     " please remove first." % instance_name)
1069
      if node.name in instance.secondary_nodes:
1070
        raise errors.OpPrereqError, ("Instance %s has node as a secondary,"
1071
                                     " please remove first." % instance_name)
1072
    self.op.node_name = node.name
1073
    self.node = node
1074

    
1075
  def Exec(self, feedback_fn):
1076
    """Removes the node from the cluster.
1077

1078
    """
1079
    node = self.node
1080
    logger.Info("stopping the node daemon and removing configs from node %s" %
1081
                node.name)
1082

    
1083
    rpc.call_node_leave_cluster(node.name)
1084

    
1085
    ssh.SSHCall(node.name, 'root', "%s stop" % constants.NODE_INITD_SCRIPT)
1086

    
1087
    logger.Info("Removing node %s from config" % node.name)
1088

    
1089
    self.cfg.RemoveNode(node.name)
1090

    
1091

    
1092
class LUQueryNodes(NoHooksLU):
1093
  """Logical unit for querying nodes.
1094

1095
  """
1096
  _OP_REQP = ["output_fields"]
1097

    
1098
  def CheckPrereq(self):
1099
    """Check prerequisites.
1100

1101
    This checks that the fields required are valid output fields.
1102

1103
    """
1104
    self.dynamic_fields = frozenset(["dtotal", "dfree",
1105
                                     "mtotal", "mnode", "mfree"])
1106

    
1107
    _CheckOutputFields(static=["name", "pinst", "sinst", "pip", "sip"],
1108
                       dynamic=self.dynamic_fields,
1109
                       selected=self.op.output_fields)
1110

    
1111

    
1112
  def Exec(self, feedback_fn):
1113
    """Computes the list of nodes and their attributes.
1114

1115
    """
1116
    nodenames = utils.NiceSort(self.cfg.GetNodeList())
1117
    nodelist = [self.cfg.GetNodeInfo(name) for name in nodenames]
1118

    
1119

    
1120
    # begin data gathering
1121

    
1122
    if self.dynamic_fields.intersection(self.op.output_fields):
1123
      live_data = {}
1124
      node_data = rpc.call_node_info(nodenames, self.cfg.GetVGName())
1125
      for name in nodenames:
1126
        nodeinfo = node_data.get(name, None)
1127
        if nodeinfo:
1128
          live_data[name] = {
1129
            "mtotal": utils.TryConvert(int, nodeinfo['memory_total']),
1130
            "mnode": utils.TryConvert(int, nodeinfo['memory_dom0']),
1131
            "mfree": utils.TryConvert(int, nodeinfo['memory_free']),
1132
            "dtotal": utils.TryConvert(int, nodeinfo['vg_size']),
1133
            "dfree": utils.TryConvert(int, nodeinfo['vg_free']),
1134
            }
1135
        else:
1136
          live_data[name] = {}
1137
    else:
1138
      live_data = dict.fromkeys(nodenames, {})
1139

    
1140
    node_to_primary = dict.fromkeys(nodenames, 0)
1141
    node_to_secondary = dict.fromkeys(nodenames, 0)
1142

    
1143
    if "pinst" in self.op.output_fields or "sinst" in self.op.output_fields:
1144
      instancelist = self.cfg.GetInstanceList()
1145

    
1146
      for instance in instancelist:
1147
        instanceinfo = self.cfg.GetInstanceInfo(instance)
1148
        node_to_primary[instanceinfo.primary_node] += 1
1149
        for secnode in instanceinfo.secondary_nodes:
1150
          node_to_secondary[secnode] += 1
1151

    
1152
    # end data gathering
1153

    
1154
    output = []
1155
    for node in nodelist:
1156
      node_output = []
1157
      for field in self.op.output_fields:
1158
        if field == "name":
1159
          val = node.name
1160
        elif field == "pinst":
1161
          val = node_to_primary[node.name]
1162
        elif field == "sinst":
1163
          val = node_to_secondary[node.name]
1164
        elif field == "pip":
1165
          val = node.primary_ip
1166
        elif field == "sip":
1167
          val = node.secondary_ip
1168
        elif field in self.dynamic_fields:
1169
          val = live_data[node.name].get(field, "?")
1170
        else:
1171
          raise errors.ParameterError, field
1172
        val = str(val)
1173
        node_output.append(val)
1174
      output.append(node_output)
1175

    
1176
    return output
1177

    
1178

    
1179
class LUQueryNodeVolumes(NoHooksLU):
1180
  """Logical unit for getting volumes on node(s).
1181

1182
  """
1183
  _OP_REQP = ["nodes", "output_fields"]
1184

    
1185
  def CheckPrereq(self):
1186
    """Check prerequisites.
1187

1188
    This checks that the fields required are valid output fields.
1189

1190
    """
1191
    self.nodes = _GetWantedNodes(self, self.op.nodes)
1192

    
1193
    _CheckOutputFields(static=["node"],
1194
                       dynamic=["phys", "vg", "name", "size", "instance"],
1195
                       selected=self.op.output_fields)
1196

    
1197

    
1198
  def Exec(self, feedback_fn):
1199
    """Computes the list of nodes and their attributes.
1200

1201
    """
1202
    nodenames = utils.NiceSort([node.name for node in self.nodes])
1203
    volumes = rpc.call_node_volumes(nodenames)
1204

    
1205
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1206
             in self.cfg.GetInstanceList()]
1207

    
1208
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1209

    
1210
    output = []
1211
    for node in nodenames:
1212
      node_vols = volumes[node][:]
1213
      node_vols.sort(key=lambda vol: vol['dev'])
1214

    
1215
      for vol in node_vols:
1216
        node_output = []
1217
        for field in self.op.output_fields:
1218
          if field == "node":
1219
            val = node
1220
          elif field == "phys":
1221
            val = vol['dev']
1222
          elif field == "vg":
1223
            val = vol['vg']
1224
          elif field == "name":
1225
            val = vol['name']
1226
          elif field == "size":
1227
            val = int(float(vol['size']))
1228
          elif field == "instance":
1229
            for inst in ilist:
1230
              if node not in lv_by_node[inst]:
1231
                continue
1232
              if vol['name'] in lv_by_node[inst][node]:
1233
                val = inst.name
1234
                break
1235
            else:
1236
              val = '-'
1237
          else:
1238
            raise errors.ParameterError, field
1239
          node_output.append(str(val))
1240

    
1241
        output.append(node_output)
1242

    
1243
    return output
1244

    
1245

    
1246
class LUAddNode(LogicalUnit):
1247
  """Logical unit for adding node to the cluster.
1248

1249
  """
1250
  HPATH = "node-add"
1251
  HTYPE = constants.HTYPE_NODE
1252
  _OP_REQP = ["node_name"]
1253

    
1254
  def BuildHooksEnv(self):
1255
    """Build hooks env.
1256

1257
    This will run on all nodes before, and on all nodes + the new node after.
1258

1259
    """
1260
    env = {
1261
      "NODE_NAME": self.op.node_name,
1262
      "NODE_PIP": self.op.primary_ip,
1263
      "NODE_SIP": self.op.secondary_ip,
1264
      }
1265
    nodes_0 = self.cfg.GetNodeList()
1266
    nodes_1 = nodes_0 + [self.op.node_name, ]
1267
    return env, nodes_0, nodes_1
1268

    
1269
  def CheckPrereq(self):
1270
    """Check prerequisites.
1271

1272
    This checks:
1273
     - the new node is not already in the config
1274
     - it is resolvable
1275
     - its parameters (single/dual homed) matches the cluster
1276

1277
    Any errors are signalled by raising errors.OpPrereqError.
1278

1279
    """
1280
    node_name = self.op.node_name
1281
    cfg = self.cfg
1282

    
1283
    dns_data = utils.LookupHostname(node_name)
1284
    if not dns_data:
1285
      raise errors.OpPrereqError, ("Node %s is not resolvable" % node_name)
1286

    
1287
    node = dns_data['hostname']
1288
    primary_ip = self.op.primary_ip = dns_data['ip']
1289
    secondary_ip = getattr(self.op, "secondary_ip", None)
1290
    if secondary_ip is None:
1291
      secondary_ip = primary_ip
1292
    if not utils.IsValidIP(secondary_ip):
1293
      raise errors.OpPrereqError, ("Invalid secondary IP given")
1294
    self.op.secondary_ip = secondary_ip
1295
    node_list = cfg.GetNodeList()
1296
    if node in node_list:
1297
      raise errors.OpPrereqError, ("Node %s is already in the configuration"
1298
                                   % node)
1299

    
1300
    for existing_node_name in node_list:
1301
      existing_node = cfg.GetNodeInfo(existing_node_name)
1302
      if (existing_node.primary_ip == primary_ip or
1303
          existing_node.secondary_ip == primary_ip or
1304
          existing_node.primary_ip == secondary_ip or
1305
          existing_node.secondary_ip == secondary_ip):
1306
        raise errors.OpPrereqError, ("New node ip address(es) conflict with"
1307
                                     " existing node %s" % existing_node.name)
1308

    
1309
    # check that the type of the node (single versus dual homed) is the
1310
    # same as for the master
1311
    myself = cfg.GetNodeInfo(self.sstore.GetMasterNode())
1312
    master_singlehomed = myself.secondary_ip == myself.primary_ip
1313
    newbie_singlehomed = secondary_ip == primary_ip
1314
    if master_singlehomed != newbie_singlehomed:
1315
      if master_singlehomed:
1316
        raise errors.OpPrereqError, ("The master has no private ip but the"
1317
                                     " new node has one")
1318
      else:
1319
        raise errors.OpPrereqError ("The master has a private ip but the"
1320
                                    " new node doesn't have one")
1321

    
1322
    # checks reachablity
1323
    command = ["fping", "-q", primary_ip]
1324
    result = utils.RunCmd(command)
1325
    if result.failed:
1326
      raise errors.OpPrereqError, ("Node not reachable by ping")
1327

    
1328
    if not newbie_singlehomed:
1329
      # check reachability from my secondary ip to newbie's secondary ip
1330
      command = ["fping", "-S%s" % myself.secondary_ip, "-q", secondary_ip]
1331
      result = utils.RunCmd(command)
1332
      if result.failed:
1333
        raise errors.OpPrereqError, ("Node secondary ip not reachable by ping")
1334

    
1335
    self.new_node = objects.Node(name=node,
1336
                                 primary_ip=primary_ip,
1337
                                 secondary_ip=secondary_ip)
1338

    
1339
  def Exec(self, feedback_fn):
1340
    """Adds the new node to the cluster.
1341

1342
    """
1343
    new_node = self.new_node
1344
    node = new_node.name
1345

    
1346
    # set up inter-node password and certificate and restarts the node daemon
1347
    gntpass = self.sstore.GetNodeDaemonPassword()
1348
    if not re.match('^[a-zA-Z0-9.]{1,64}$', gntpass):
1349
      raise errors.OpExecError, ("ganeti password corruption detected")
1350
    f = open(constants.SSL_CERT_FILE)
1351
    try:
1352
      gntpem = f.read(8192)
1353
    finally:
1354
      f.close()
1355
    # in the base64 pem encoding, neither '!' nor '.' are valid chars,
1356
    # so we use this to detect an invalid certificate; as long as the
1357
    # cert doesn't contain this, the here-document will be correctly
1358
    # parsed by the shell sequence below
1359
    if re.search('^!EOF\.', gntpem, re.MULTILINE):
1360
      raise errors.OpExecError, ("invalid PEM encoding in the SSL certificate")
1361
    if not gntpem.endswith("\n"):
1362
      raise errors.OpExecError, ("PEM must end with newline")
1363
    logger.Info("copy cluster pass to %s and starting the node daemon" % node)
1364

    
1365
    # remove first the root's known_hosts file
1366
    utils.RemoveFile("/root/.ssh/known_hosts")
1367
    # and then connect with ssh to set password and start ganeti-noded
1368
    # note that all the below variables are sanitized at this point,
1369
    # either by being constants or by the checks above
1370
    ss = self.sstore
1371
    mycommand = ("umask 077 && "
1372
                 "echo '%s' > '%s' && "
1373
                 "cat > '%s' << '!EOF.' && \n"
1374
                 "%s!EOF.\n%s restart" %
1375
                 (gntpass, ss.KeyToFilename(ss.SS_NODED_PASS),
1376
                  constants.SSL_CERT_FILE, gntpem,
1377
                  constants.NODE_INITD_SCRIPT))
1378

    
1379
    result = ssh.SSHCall(node, 'root', mycommand, batch=False, ask_key=True)
1380
    if result.failed:
1381
      raise errors.OpExecError, ("Remote command on node %s, error: %s,"
1382
                                 " output: %s" %
1383
                                 (node, result.fail_reason, result.output))
1384

    
1385
    # check connectivity
1386
    time.sleep(4)
1387

    
1388
    result = rpc.call_version([node])[node]
1389
    if result:
1390
      if constants.PROTOCOL_VERSION == result:
1391
        logger.Info("communication to node %s fine, sw version %s match" %
1392
                    (node, result))
1393
      else:
1394
        raise errors.OpExecError, ("Version mismatch master version %s,"
1395
                                   " node version %s" %
1396
                                   (constants.PROTOCOL_VERSION, result))
1397
    else:
1398
      raise errors.OpExecError, ("Cannot get version from the new node")
1399

    
1400
    # setup ssh on node
1401
    logger.Info("copy ssh key to node %s" % node)
1402
    keyarray = []
1403
    keyfiles = ["/etc/ssh/ssh_host_dsa_key", "/etc/ssh/ssh_host_dsa_key.pub",
1404
                "/etc/ssh/ssh_host_rsa_key", "/etc/ssh/ssh_host_rsa_key.pub",
1405
                "/root/.ssh/id_dsa", "/root/.ssh/id_dsa.pub"]
1406

    
1407
    for i in keyfiles:
1408
      f = open(i, 'r')
1409
      try:
1410
        keyarray.append(f.read())
1411
      finally:
1412
        f.close()
1413

    
1414
    result = rpc.call_node_add(node, keyarray[0], keyarray[1], keyarray[2],
1415
                               keyarray[3], keyarray[4], keyarray[5])
1416

    
1417
    if not result:
1418
      raise errors.OpExecError, ("Cannot transfer ssh keys to the new node")
1419

    
1420
    # Add node to our /etc/hosts, and add key to known_hosts
1421
    _UpdateEtcHosts(new_node.name, new_node.primary_ip)
1422
    _UpdateKnownHosts(new_node.name, new_node.primary_ip,
1423
                      self.cfg.GetHostKey())
1424

    
1425
    if new_node.secondary_ip != new_node.primary_ip:
1426
      result = ssh.SSHCall(node, "root",
1427
                           "fping -S 127.0.0.1 -q %s" % new_node.secondary_ip)
1428
      if result.failed:
1429
        raise errors.OpExecError, ("Node claims it doesn't have the"
1430
                                   " secondary ip you gave (%s).\n"
1431
                                   "Please fix and re-run this command." %
1432
                                   new_node.secondary_ip)
1433

    
1434
    # Distribute updated /etc/hosts and known_hosts to all nodes,
1435
    # including the node just added
1436
    myself = self.cfg.GetNodeInfo(self.sstore.GetMasterNode())
1437
    dist_nodes = self.cfg.GetNodeList() + [node]
1438
    if myself.name in dist_nodes:
1439
      dist_nodes.remove(myself.name)
1440

    
1441
    logger.Debug("Copying hosts and known_hosts to all nodes")
1442
    for fname in ("/etc/hosts", "/etc/ssh/ssh_known_hosts"):
1443
      result = rpc.call_upload_file(dist_nodes, fname)
1444
      for to_node in dist_nodes:
1445
        if not result[to_node]:
1446
          logger.Error("copy of file %s to node %s failed" %
1447
                       (fname, to_node))
1448

    
1449
    to_copy = ss.GetFileList()
1450
    for fname in to_copy:
1451
      if not ssh.CopyFileToNode(node, fname):
1452
        logger.Error("could not copy file %s to node %s" % (fname, node))
1453

    
1454
    logger.Info("adding node %s to cluster.conf" % node)
1455
    self.cfg.AddNode(new_node)
1456

    
1457

    
1458
class LUMasterFailover(LogicalUnit):
1459
  """Failover the master node to the current node.
1460

1461
  This is a special LU in that it must run on a non-master node.
1462

1463
  """
1464
  HPATH = "master-failover"
1465
  HTYPE = constants.HTYPE_CLUSTER
1466
  REQ_MASTER = False
1467
  _OP_REQP = []
1468

    
1469
  def BuildHooksEnv(self):
1470
    """Build hooks env.
1471

1472
    This will run on the new master only in the pre phase, and on all
1473
    the nodes in the post phase.
1474

1475
    """
1476
    env = {
1477
      "NEW_MASTER": self.new_master,
1478
      "OLD_MASTER": self.old_master,
1479
      }
1480
    return env, [self.new_master], self.cfg.GetNodeList()
1481

    
1482
  def CheckPrereq(self):
1483
    """Check prerequisites.
1484

1485
    This checks that we are not already the master.
1486

1487
    """
1488
    self.new_master = socket.gethostname()
1489

    
1490
    self.old_master = self.sstore.GetMasterNode()
1491

    
1492
    if self.old_master == self.new_master:
1493
      raise errors.OpPrereqError, ("This commands must be run on the node"
1494
                                   " where you want the new master to be.\n"
1495
                                   "%s is already the master" %
1496
                                   self.old_master)
1497

    
1498
  def Exec(self, feedback_fn):
1499
    """Failover the master node.
1500

1501
    This command, when run on a non-master node, will cause the current
1502
    master to cease being master, and the non-master to become new
1503
    master.
1504

1505
    """
1506
    #TODO: do not rely on gethostname returning the FQDN
1507
    logger.Info("setting master to %s, old master: %s" %
1508
                (self.new_master, self.old_master))
1509

    
1510
    if not rpc.call_node_stop_master(self.old_master):
1511
      logger.Error("could disable the master role on the old master"
1512
                   " %s, please disable manually" % self.old_master)
1513

    
1514
    ss = self.sstore
1515
    ss.SetKey(ss.SS_MASTER_NODE, self.new_master)
1516
    if not rpc.call_upload_file(self.cfg.GetNodeList(),
1517
                                ss.KeyToFilename(ss.SS_MASTER_NODE)):
1518
      logger.Error("could not distribute the new simple store master file"
1519
                   " to the other nodes, please check.")
1520

    
1521
    if not rpc.call_node_start_master(self.new_master):
1522
      logger.Error("could not start the master role on the new master"
1523
                   " %s, please check" % self.new_master)
1524
      feedback_fn("Error in activating the master IP on the new master,\n"
1525
                  "please fix manually.")
1526

    
1527

    
1528

    
1529
class LUQueryClusterInfo(NoHooksLU):
1530
  """Query cluster configuration.
1531

1532
  """
1533
  _OP_REQP = []
1534
  REQ_MASTER = False
1535

    
1536
  def CheckPrereq(self):
1537
    """No prerequsites needed for this LU.
1538

1539
    """
1540
    pass
1541

    
1542
  def Exec(self, feedback_fn):
1543
    """Return cluster config.
1544

1545
    """
1546
    result = {
1547
      "name": self.sstore.GetClusterName(),
1548
      "software_version": constants.RELEASE_VERSION,
1549
      "protocol_version": constants.PROTOCOL_VERSION,
1550
      "config_version": constants.CONFIG_VERSION,
1551
      "os_api_version": constants.OS_API_VERSION,
1552
      "export_version": constants.EXPORT_VERSION,
1553
      "master": self.sstore.GetMasterNode(),
1554
      "architecture": (platform.architecture()[0], platform.machine()),
1555
      }
1556

    
1557
    return result
1558

    
1559

    
1560
class LUClusterCopyFile(NoHooksLU):
1561
  """Copy file to cluster.
1562

1563
  """
1564
  _OP_REQP = ["nodes", "filename"]
1565

    
1566
  def CheckPrereq(self):
1567
    """Check prerequisites.
1568

1569
    It should check that the named file exists and that the given list
1570
    of nodes is valid.
1571

1572
    """
1573
    if not os.path.exists(self.op.filename):
1574
      raise errors.OpPrereqError("No such filename '%s'" % self.op.filename)
1575

    
1576
    self.nodes = _GetWantedNodes(self, self.op.nodes)
1577

    
1578
  def Exec(self, feedback_fn):
1579
    """Copy a file from master to some nodes.
1580

1581
    Args:
1582
      opts - class with options as members
1583
      args - list containing a single element, the file name
1584
    Opts used:
1585
      nodes - list containing the name of target nodes; if empty, all nodes
1586

1587
    """
1588
    filename = self.op.filename
1589

    
1590
    myname = socket.gethostname()
1591

    
1592
    for node in self.nodes:
1593
      if node == myname:
1594
        continue
1595
      if not ssh.CopyFileToNode(node, filename):
1596
        logger.Error("Copy of file %s to node %s failed" % (filename, node))
1597

    
1598

    
1599
class LUDumpClusterConfig(NoHooksLU):
1600
  """Return a text-representation of the cluster-config.
1601

1602
  """
1603
  _OP_REQP = []
1604

    
1605
  def CheckPrereq(self):
1606
    """No prerequisites.
1607

1608
    """
1609
    pass
1610

    
1611
  def Exec(self, feedback_fn):
1612
    """Dump a representation of the cluster config to the standard output.
1613

1614
    """
1615
    return self.cfg.DumpConfig()
1616

    
1617

    
1618
class LURunClusterCommand(NoHooksLU):
1619
  """Run a command on some nodes.
1620

1621
  """
1622
  _OP_REQP = ["command", "nodes"]
1623

    
1624
  def CheckPrereq(self):
1625
    """Check prerequisites.
1626

1627
    It checks that the given list of nodes is valid.
1628

1629
    """
1630
    self.nodes = _GetWantedNodes(self, self.op.nodes)
1631

    
1632
  def Exec(self, feedback_fn):
1633
    """Run a command on some nodes.
1634

1635
    """
1636
    data = []
1637
    for node in self.nodes:
1638
      result = utils.RunCmd(["ssh", node.name, self.op.command])
1639
      data.append((node.name, result.cmd, result.output, result.exit_code))
1640

    
1641
    return data
1642

    
1643

    
1644
class LUActivateInstanceDisks(NoHooksLU):
1645
  """Bring up an instance's disks.
1646

1647
  """
1648
  _OP_REQP = ["instance_name"]
1649

    
1650
  def CheckPrereq(self):
1651
    """Check prerequisites.
1652

1653
    This checks that the instance is in the cluster.
1654

1655
    """
1656
    instance = self.cfg.GetInstanceInfo(
1657
      self.cfg.ExpandInstanceName(self.op.instance_name))
1658
    if instance is None:
1659
      raise errors.OpPrereqError, ("Instance '%s' not known" %
1660
                                   self.op.instance_name)
1661
    self.instance = instance
1662

    
1663

    
1664
  def Exec(self, feedback_fn):
1665
    """Activate the disks.
1666

1667
    """
1668
    disks_ok, disks_info = _AssembleInstanceDisks(self.instance, self.cfg)
1669
    if not disks_ok:
1670
      raise errors.OpExecError, ("Cannot activate block devices")
1671

    
1672
    return disks_info
1673

    
1674

    
1675
def _AssembleInstanceDisks(instance, cfg, ignore_secondaries=False):
1676
  """Prepare the block devices for an instance.
1677

1678
  This sets up the block devices on all nodes.
1679

1680
  Args:
1681
    instance: a ganeti.objects.Instance object
1682
    ignore_secondaries: if true, errors on secondary nodes won't result
1683
                        in an error return from the function
1684

1685
  Returns:
1686
    false if the operation failed
1687
    list of (host, instance_visible_name, node_visible_name) if the operation
1688
         suceeded with the mapping from node devices to instance devices
1689
  """
1690
  device_info = []
1691
  disks_ok = True
1692
  for inst_disk in instance.disks:
1693
    master_result = None
1694
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
1695
      cfg.SetDiskID(node_disk, node)
1696
      is_primary = node == instance.primary_node
1697
      result = rpc.call_blockdev_assemble(node, node_disk, is_primary)
1698
      if not result:
1699
        logger.Error("could not prepare block device %s on node %s (is_pri"
1700
                     "mary=%s)" % (inst_disk.iv_name, node, is_primary))
1701
        if is_primary or not ignore_secondaries:
1702
          disks_ok = False
1703
      if is_primary:
1704
        master_result = result
1705
    device_info.append((instance.primary_node, inst_disk.iv_name,
1706
                        master_result))
1707

    
1708
  return disks_ok, device_info
1709

    
1710

    
1711
def _StartInstanceDisks(cfg, instance, force):
1712
  disks_ok, dummy = _AssembleInstanceDisks(instance, cfg,
1713
                                           ignore_secondaries=force)
1714
  if not disks_ok:
1715
    _ShutdownInstanceDisks(instance, cfg)
1716
    if force is not None and not force:
1717
      logger.Error("If the message above refers to a secondary node,"
1718
                   " you can retry the operation using '--force'.")
1719
    raise errors.OpExecError, ("Disk consistency error")
1720

    
1721

    
1722
class LUDeactivateInstanceDisks(NoHooksLU):
1723
  """Shutdown an instance's disks.
1724

1725
  """
1726
  _OP_REQP = ["instance_name"]
1727

    
1728
  def CheckPrereq(self):
1729
    """Check prerequisites.
1730

1731
    This checks that the instance is in the cluster.
1732

1733
    """
1734
    instance = self.cfg.GetInstanceInfo(
1735
      self.cfg.ExpandInstanceName(self.op.instance_name))
1736
    if instance is None:
1737
      raise errors.OpPrereqError, ("Instance '%s' not known" %
1738
                                   self.op.instance_name)
1739
    self.instance = instance
1740

    
1741
  def Exec(self, feedback_fn):
1742
    """Deactivate the disks
1743

1744
    """
1745
    instance = self.instance
1746
    ins_l = rpc.call_instance_list([instance.primary_node])
1747
    ins_l = ins_l[instance.primary_node]
1748
    if not type(ins_l) is list:
1749
      raise errors.OpExecError, ("Can't contact node '%s'" %
1750
                                 instance.primary_node)
1751

    
1752
    if self.instance.name in ins_l:
1753
      raise errors.OpExecError, ("Instance is running, can't shutdown"
1754
                                 " block devices.")
1755

    
1756
    _ShutdownInstanceDisks(instance, self.cfg)
1757

    
1758

    
1759
def _ShutdownInstanceDisks(instance, cfg, ignore_primary=False):
1760
  """Shutdown block devices of an instance.
1761

1762
  This does the shutdown on all nodes of the instance.
1763

1764
  If the ignore_primary is false, errors on the primary node are
1765
  ignored.
1766

1767
  """
1768
  result = True
1769
  for disk in instance.disks:
1770
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
1771
      cfg.SetDiskID(top_disk, node)
1772
      if not rpc.call_blockdev_shutdown(node, top_disk):
1773
        logger.Error("could not shutdown block device %s on node %s" %
1774
                     (disk.iv_name, node))
1775
        if not ignore_primary or node != instance.primary_node:
1776
          result = False
1777
  return result
1778

    
1779

    
1780
class LUStartupInstance(LogicalUnit):
1781
  """Starts an instance.
1782

1783
  """
1784
  HPATH = "instance-start"
1785
  HTYPE = constants.HTYPE_INSTANCE
1786
  _OP_REQP = ["instance_name", "force"]
1787

    
1788
  def BuildHooksEnv(self):
1789
    """Build hooks env.
1790

1791
    This runs on master, primary and secondary nodes of the instance.
1792

1793
    """
1794
    env = {
1795
      "FORCE": self.op.force,
1796
      }
1797
    env.update(_BuildInstanceHookEnvByObject(self.instance))
1798
    nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] +
1799
          list(self.instance.secondary_nodes))
1800
    return env, nl, nl
1801

    
1802
  def CheckPrereq(self):
1803
    """Check prerequisites.
1804

1805
    This checks that the instance is in the cluster.
1806

1807
    """
1808
    instance = self.cfg.GetInstanceInfo(
1809
      self.cfg.ExpandInstanceName(self.op.instance_name))
1810
    if instance is None:
1811
      raise errors.OpPrereqError, ("Instance '%s' not known" %
1812
                                   self.op.instance_name)
1813

    
1814
    # check bridges existance
1815
    brlist = [nic.bridge for nic in instance.nics]
1816
    if not rpc.call_bridges_exist(instance.primary_node, brlist):
1817
      raise errors.OpPrereqError, ("one or more target bridges %s does not"
1818
                                   " exist on destination node '%s'" %
1819
                                   (brlist, instance.primary_node))
1820

    
1821
    self.instance = instance
1822
    self.op.instance_name = instance.name
1823

    
1824
  def Exec(self, feedback_fn):
1825
    """Start the instance.
1826

1827
    """
1828
    instance = self.instance
1829
    force = self.op.force
1830
    extra_args = getattr(self.op, "extra_args", "")
1831

    
1832
    node_current = instance.primary_node
1833

    
1834
    nodeinfo = rpc.call_node_info([node_current], self.cfg.GetVGName())
1835
    if not nodeinfo:
1836
      raise errors.OpExecError, ("Could not contact node %s for infos" %
1837
                                 (node_current))
1838

    
1839
    freememory = nodeinfo[node_current]['memory_free']
1840
    memory = instance.memory
1841
    if memory > freememory:
1842
      raise errors.OpExecError, ("Not enough memory to start instance"
1843
                                 " %s on node %s"
1844
                                 " needed %s MiB, available %s MiB" %
1845
                                 (instance.name, node_current, memory,
1846
                                  freememory))
1847

    
1848
    _StartInstanceDisks(self.cfg, instance, force)
1849

    
1850
    if not rpc.call_instance_start(node_current, instance, extra_args):
1851
      _ShutdownInstanceDisks(instance, self.cfg)
1852
      raise errors.OpExecError, ("Could not start instance")
1853

    
1854
    self.cfg.MarkInstanceUp(instance.name)
1855

    
1856

    
1857
class LUShutdownInstance(LogicalUnit):
1858
  """Shutdown an instance.
1859

1860
  """
1861
  HPATH = "instance-stop"
1862
  HTYPE = constants.HTYPE_INSTANCE
1863
  _OP_REQP = ["instance_name"]
1864

    
1865
  def BuildHooksEnv(self):
1866
    """Build hooks env.
1867

1868
    This runs on master, primary and secondary nodes of the instance.
1869

1870
    """
1871
    env = _BuildInstanceHookEnvByObject(self.instance)
1872
    nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] +
1873
          list(self.instance.secondary_nodes))
1874
    return env, nl, nl
1875

    
1876
  def CheckPrereq(self):
1877
    """Check prerequisites.
1878

1879
    This checks that the instance is in the cluster.
1880

1881
    """
1882
    instance = self.cfg.GetInstanceInfo(
1883
      self.cfg.ExpandInstanceName(self.op.instance_name))
1884
    if instance is None:
1885
      raise errors.OpPrereqError, ("Instance '%s' not known" %
1886
                                   self.op.instance_name)
1887
    self.instance = instance
1888

    
1889
  def Exec(self, feedback_fn):
1890
    """Shutdown the instance.
1891

1892
    """
1893
    instance = self.instance
1894
    node_current = instance.primary_node
1895
    if not rpc.call_instance_shutdown(node_current, instance):
1896
      logger.Error("could not shutdown instance")
1897

    
1898
    self.cfg.MarkInstanceDown(instance.name)
1899
    _ShutdownInstanceDisks(instance, self.cfg)
1900

    
1901

    
1902
class LUReinstallInstance(LogicalUnit):
1903
  """Reinstall an instance.
1904

1905
  """
1906
  HPATH = "instance-reinstall"
1907
  HTYPE = constants.HTYPE_INSTANCE
1908
  _OP_REQP = ["instance_name"]
1909

    
1910
  def BuildHooksEnv(self):
1911
    """Build hooks env.
1912

1913
    This runs on master, primary and secondary nodes of the instance.
1914

1915
    """
1916
    env = _BuildInstanceHookEnvByObject(self.instance)
1917
    nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] +
1918
          list(self.instance.secondary_nodes))
1919
    return env, nl, nl
1920

    
1921
  def CheckPrereq(self):
1922
    """Check prerequisites.
1923

1924
    This checks that the instance is in the cluster and is not running.
1925

1926
    """
1927
    instance = self.cfg.GetInstanceInfo(
1928
      self.cfg.ExpandInstanceName(self.op.instance_name))
1929
    if instance is None:
1930
      raise errors.OpPrereqError, ("Instance '%s' not known" %
1931
                                   self.op.instance_name)
1932
    if instance.disk_template == constants.DT_DISKLESS:
1933
      raise errors.OpPrereqError, ("Instance '%s' has no disks" %
1934
                                   self.op.instance_name)
1935
    if instance.status != "down":
1936
      raise errors.OpPrereqError, ("Instance '%s' is marked to be up" %
1937
                                   self.op.instance_name)
1938
    remote_info = rpc.call_instance_info(instance.primary_node, instance.name)
1939
    if remote_info:
1940
      raise errors.OpPrereqError, ("Instance '%s' is running on the node %s" %
1941
                                   (self.op.instance_name,
1942
                                    instance.primary_node))
1943

    
1944
    self.op.os_type = getattr(self.op, "os_type", None)
1945
    if self.op.os_type is not None:
1946
      # OS verification
1947
      pnode = self.cfg.GetNodeInfo(
1948
        self.cfg.ExpandNodeName(instance.primary_node))
1949
      if pnode is None:
1950
        raise errors.OpPrereqError, ("Primary node '%s' is unknown" %
1951
                                     self.op.pnode)
1952
      os_obj = rpc.call_os_get([pnode.name], self.op.os_type)[pnode.name]
1953
      if not isinstance(os_obj, objects.OS):
1954
        raise errors.OpPrereqError, ("OS '%s' not in supported OS list for"
1955
                                     " primary node"  % self.op.os_type)
1956

    
1957
    self.instance = instance
1958

    
1959
  def Exec(self, feedback_fn):
1960
    """Reinstall the instance.
1961

1962
    """
1963
    inst = self.instance
1964

    
1965
    if self.op.os_type is not None:
1966
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
1967
      inst.os = self.op.os_type
1968
      self.cfg.AddInstance(inst)
1969

    
1970
    _StartInstanceDisks(self.cfg, inst, None)
1971
    try:
1972
      feedback_fn("Running the instance OS create scripts...")
1973
      if not rpc.call_instance_os_add(inst.primary_node, inst, "sda", "sdb"):
1974
        raise errors.OpExecError, ("Could not install OS for instance %s "
1975
                                   "on node %s" %
1976
                                   (inst.name, inst.primary_node))
1977
    finally:
1978
      _ShutdownInstanceDisks(inst, self.cfg)
1979

    
1980

    
1981
class LURemoveInstance(LogicalUnit):
1982
  """Remove an instance.
1983

1984
  """
1985
  HPATH = "instance-remove"
1986
  HTYPE = constants.HTYPE_INSTANCE
1987
  _OP_REQP = ["instance_name"]
1988

    
1989
  def BuildHooksEnv(self):
1990
    """Build hooks env.
1991

1992
    This runs on master, primary and secondary nodes of the instance.
1993

1994
    """
1995
    env = _BuildInstanceHookEnvByObject(self.instance)
1996
    nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] +
1997
          list(self.instance.secondary_nodes))
1998
    return env, nl, nl
1999

    
2000
  def CheckPrereq(self):
2001
    """Check prerequisites.
2002

2003
    This checks that the instance is in the cluster.
2004

2005
    """
2006
    instance = self.cfg.GetInstanceInfo(
2007
      self.cfg.ExpandInstanceName(self.op.instance_name))
2008
    if instance is None:
2009
      raise errors.OpPrereqError, ("Instance '%s' not known" %
2010
                                   self.op.instance_name)
2011
    self.instance = instance
2012

    
2013
  def Exec(self, feedback_fn):
2014
    """Remove the instance.
2015

2016
    """
2017
    instance = self.instance
2018
    logger.Info("shutting down instance %s on node %s" %
2019
                (instance.name, instance.primary_node))
2020

    
2021
    if not rpc.call_instance_shutdown(instance.primary_node, instance):
2022
      raise errors.OpExecError, ("Could not shutdown instance %s on node %s" %
2023
                                 (instance.name, instance.primary_node))
2024

    
2025
    logger.Info("removing block devices for instance %s" % instance.name)
2026

    
2027
    _RemoveDisks(instance, self.cfg)
2028

    
2029
    logger.Info("removing instance %s out of cluster config" % instance.name)
2030

    
2031
    self.cfg.RemoveInstance(instance.name)
2032

    
2033

    
2034
class LUQueryInstances(NoHooksLU):
2035
  """Logical unit for querying instances.
2036

2037
  """
2038
  _OP_REQP = ["output_fields"]
2039

    
2040
  def CheckPrereq(self):
2041
    """Check prerequisites.
2042

2043
    This checks that the fields required are valid output fields.
2044

2045
    """
2046
    self.dynamic_fields = frozenset(["oper_state", "oper_ram"])
2047
    _CheckOutputFields(static=["name", "os", "pnode", "snodes",
2048
                               "admin_state", "admin_ram",
2049
                               "disk_template", "ip", "mac", "bridge"],
2050
                       dynamic=self.dynamic_fields,
2051
                       selected=self.op.output_fields)
2052

    
2053
  def Exec(self, feedback_fn):
2054
    """Computes the list of nodes and their attributes.
2055

2056
    """
2057
    instance_names = utils.NiceSort(self.cfg.GetInstanceList())
2058
    instance_list = [self.cfg.GetInstanceInfo(iname) for iname
2059
                     in instance_names]
2060

    
2061
    # begin data gathering
2062

    
2063
    nodes = frozenset([inst.primary_node for inst in instance_list])
2064

    
2065
    bad_nodes = []
2066
    if self.dynamic_fields.intersection(self.op.output_fields):
2067
      live_data = {}
2068
      node_data = rpc.call_all_instances_info(nodes)
2069
      for name in nodes:
2070
        result = node_data[name]
2071
        if result:
2072
          live_data.update(result)
2073
        elif result == False:
2074
          bad_nodes.append(name)
2075
        # else no instance is alive
2076
    else:
2077
      live_data = dict([(name, {}) for name in instance_names])
2078

    
2079
    # end data gathering
2080

    
2081
    output = []
2082
    for instance in instance_list:
2083
      iout = []
2084
      for field in self.op.output_fields:
2085
        if field == "name":
2086
          val = instance.name
2087
        elif field == "os":
2088
          val = instance.os
2089
        elif field == "pnode":
2090
          val = instance.primary_node
2091
        elif field == "snodes":
2092
          val = ",".join(instance.secondary_nodes) or "-"
2093
        elif field == "admin_state":
2094
          if instance.status == "down":
2095
            val = "no"
2096
          else:
2097
            val = "yes"
2098
        elif field == "oper_state":
2099
          if instance.primary_node in bad_nodes:
2100
            val = "(node down)"
2101
          else:
2102
            if live_data.get(instance.name):
2103
              val = "running"
2104
            else:
2105
              val = "stopped"
2106
        elif field == "admin_ram":
2107
          val = instance.memory
2108
        elif field == "oper_ram":
2109
          if instance.primary_node in bad_nodes:
2110
            val = "(node down)"
2111
          elif instance.name in live_data:
2112
            val = live_data[instance.name].get("memory", "?")
2113
          else:
2114
            val = "-"
2115
        elif field == "disk_template":
2116
          val = instance.disk_template
2117
        elif field == "ip":
2118
          val = instance.nics[0].ip
2119
        elif field == "bridge":
2120
          val = instance.nics[0].bridge
2121
        elif field == "mac":
2122
          val = instance.nics[0].mac
2123
        else:
2124
          raise errors.ParameterError, field
2125
        val = str(val)
2126
        iout.append(val)
2127
      output.append(iout)
2128

    
2129
    return output
2130

    
2131

    
2132
class LUFailoverInstance(LogicalUnit):
2133
  """Failover an instance.
2134

2135
  """
2136
  HPATH = "instance-failover"
2137
  HTYPE = constants.HTYPE_INSTANCE
2138
  _OP_REQP = ["instance_name", "ignore_consistency"]
2139

    
2140
  def BuildHooksEnv(self):
2141
    """Build hooks env.
2142

2143
    This runs on master, primary and secondary nodes of the instance.
2144

2145
    """
2146
    env = {
2147
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
2148
      }
2149
    env.update(_BuildInstanceHookEnvByObject(self.instance))
2150
    nl = [self.sstore.GetMasterNode()] + list(self.instance.secondary_nodes)
2151
    return env, nl, nl
2152

    
2153
  def CheckPrereq(self):
2154
    """Check prerequisites.
2155

2156
    This checks that the instance is in the cluster.
2157

2158
    """
2159
    instance = self.cfg.GetInstanceInfo(
2160
      self.cfg.ExpandInstanceName(self.op.instance_name))
2161
    if instance is None:
2162
      raise errors.OpPrereqError, ("Instance '%s' not known" %
2163
                                   self.op.instance_name)
2164

    
2165
    # check memory requirements on the secondary node
2166
    target_node = instance.secondary_nodes[0]
2167
    nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName())
2168
    info = nodeinfo.get(target_node, None)
2169
    if not info:
2170
      raise errors.OpPrereqError, ("Cannot get current information"
2171
                                   " from node '%s'" % nodeinfo)
2172
    if instance.memory > info['memory_free']:
2173
      raise errors.OpPrereqError, ("Not enough memory on target node %s."
2174
                                   " %d MB available, %d MB required" %
2175
                                   (target_node, info['memory_free'],
2176
                                    instance.memory))
2177

    
2178
    # check bridge existance
2179
    brlist = [nic.bridge for nic in instance.nics]
2180
    if not rpc.call_bridges_exist(instance.primary_node, brlist):
2181
      raise errors.OpPrereqError, ("one or more target bridges %s does not"
2182
                                   " exist on destination node '%s'" %
2183
                                   (brlist, instance.primary_node))
2184

    
2185
    self.instance = instance
2186

    
2187
  def Exec(self, feedback_fn):
2188
    """Failover an instance.
2189

2190
    The failover is done by shutting it down on its present node and
2191
    starting it on the secondary.
2192

2193
    """
2194
    instance = self.instance
2195

    
2196
    source_node = instance.primary_node
2197
    target_node = instance.secondary_nodes[0]
2198

    
2199
    feedback_fn("* checking disk consistency between source and target")
2200
    for dev in instance.disks:
2201
      # for remote_raid1, these are md over drbd
2202
      if not _CheckDiskConsistency(self.cfg, dev, target_node, False):
2203
        if not self.op.ignore_consistency:
2204
          raise errors.OpExecError, ("Disk %s is degraded on target node,"
2205
                                     " aborting failover." % dev.iv_name)
2206

    
2207
    feedback_fn("* checking target node resource availability")
2208
    nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName())
2209

    
2210
    if not nodeinfo:
2211
      raise errors.OpExecError, ("Could not contact target node %s." %
2212
                                 target_node)
2213

    
2214
    free_memory = int(nodeinfo[target_node]['memory_free'])
2215
    memory = instance.memory
2216
    if memory > free_memory:
2217
      raise errors.OpExecError, ("Not enough memory to create instance %s on"
2218
                                 " node %s. needed %s MiB, available %s MiB" %
2219
                                 (instance.name, target_node, memory,
2220
                                  free_memory))
2221

    
2222
    feedback_fn("* shutting down instance on source node")
2223
    logger.Info("Shutting down instance %s on node %s" %
2224
                (instance.name, source_node))
2225

    
2226
    if not rpc.call_instance_shutdown(source_node, instance):
2227
      logger.Error("Could not shutdown instance %s on node %s. Proceeding"
2228
                   " anyway. Please make sure node %s is down"  %
2229
                   (instance.name, source_node, source_node))
2230

    
2231
    feedback_fn("* deactivating the instance's disks on source node")
2232
    if not _ShutdownInstanceDisks(instance, self.cfg, ignore_primary=True):
2233
      raise errors.OpExecError, ("Can't shut down the instance's disks.")
2234

    
2235
    instance.primary_node = target_node
2236
    # distribute new instance config to the other nodes
2237
    self.cfg.AddInstance(instance)
2238

    
2239
    feedback_fn("* activating the instance's disks on target node")
2240
    logger.Info("Starting instance %s on node %s" %
2241
                (instance.name, target_node))
2242

    
2243
    disks_ok, dummy = _AssembleInstanceDisks(instance, self.cfg,
2244
                                             ignore_secondaries=True)
2245
    if not disks_ok:
2246
      _ShutdownInstanceDisks(instance, self.cfg)
2247
      raise errors.OpExecError, ("Can't activate the instance's disks")
2248

    
2249
    feedback_fn("* starting the instance on the target node")
2250
    if not rpc.call_instance_start(target_node, instance, None):
2251
      _ShutdownInstanceDisks(instance, self.cfg)
2252
      raise errors.OpExecError("Could not start instance %s on node %s." %
2253
                               (instance.name, target_node))
2254

    
2255

    
2256
def _CreateBlockDevOnPrimary(cfg, node, device, info):
2257
  """Create a tree of block devices on the primary node.
2258

2259
  This always creates all devices.
2260

2261
  """
2262
  if device.children:
2263
    for child in device.children:
2264
      if not _CreateBlockDevOnPrimary(cfg, node, child, info):
2265
        return False
2266

    
2267
  cfg.SetDiskID(device, node)
2268
  new_id = rpc.call_blockdev_create(node, device, device.size, True, info)
2269
  if not new_id:
2270
    return False
2271
  if device.physical_id is None:
2272
    device.physical_id = new_id
2273
  return True
2274

    
2275

    
2276
def _CreateBlockDevOnSecondary(cfg, node, device, force, info):
2277
  """Create a tree of block devices on a secondary node.
2278

2279
  If this device type has to be created on secondaries, create it and
2280
  all its children.
2281

2282
  If not, just recurse to children keeping the same 'force' value.
2283

2284
  """
2285
  if device.CreateOnSecondary():
2286
    force = True
2287
  if device.children:
2288
    for child in device.children:
2289
      if not _CreateBlockDevOnSecondary(cfg, node, child, force, info):
2290
        return False
2291

    
2292
  if not force:
2293
    return True
2294
  cfg.SetDiskID(device, node)
2295
  new_id = rpc.call_blockdev_create(node, device, device.size, False, info)
2296
  if not new_id:
2297
    return False
2298
  if device.physical_id is None:
2299
    device.physical_id = new_id
2300
  return True
2301

    
2302

    
2303
def _GenerateUniqueNames(cfg, exts):
2304
  """Generate a suitable LV name.
2305

2306
  This will generate a logical volume name for the given instance.
2307

2308
  """
2309
  results = []
2310
  for val in exts:
2311
    new_id = cfg.GenerateUniqueID()
2312
    results.append("%s%s" % (new_id, val))
2313
  return results
2314

    
2315

    
2316
def _GenerateMDDRBDBranch(cfg, primary, secondary, size, names):
2317
  """Generate a drbd device complete with its children.
2318

2319
  """
2320
  port = cfg.AllocatePort()
2321
  vgname = cfg.GetVGName()
2322
  dev_data = objects.Disk(dev_type="lvm", size=size,
2323
                          logical_id=(vgname, names[0]))
2324
  dev_meta = objects.Disk(dev_type="lvm", size=128,
2325
                          logical_id=(vgname, names[1]))
2326
  drbd_dev = objects.Disk(dev_type="drbd", size=size,
2327
                          logical_id = (primary, secondary, port),
2328
                          children = [dev_data, dev_meta])
2329
  return drbd_dev
2330

    
2331

    
2332
def _GenerateDiskTemplate(cfg, template_name,
2333
                          instance_name, primary_node,
2334
                          secondary_nodes, disk_sz, swap_sz):
2335
  """Generate the entire disk layout for a given template type.
2336

2337
  """
2338
  #TODO: compute space requirements
2339

    
2340
  vgname = cfg.GetVGName()
2341
  if template_name == "diskless":
2342
    disks = []
2343
  elif template_name == "plain":
2344
    if len(secondary_nodes) != 0:
2345
      raise errors.ProgrammerError("Wrong template configuration")
2346

    
2347
    names = _GenerateUniqueNames(cfg, [".sda", ".sdb"])
2348
    sda_dev = objects.Disk(dev_type="lvm", size=disk_sz,
2349
                           logical_id=(vgname, names[0]),
2350
                           iv_name = "sda")
2351
    sdb_dev = objects.Disk(dev_type="lvm", size=swap_sz,
2352
                           logical_id=(vgname, names[1]),
2353
                           iv_name = "sdb")
2354
    disks = [sda_dev, sdb_dev]
2355
  elif template_name == "local_raid1":
2356
    if len(secondary_nodes) != 0:
2357
      raise errors.ProgrammerError("Wrong template configuration")
2358

    
2359

    
2360
    names = _GenerateUniqueNames(cfg, [".sda_m1", ".sda_m2",
2361
                                       ".sdb_m1", ".sdb_m2"])
2362
    sda_dev_m1 = objects.Disk(dev_type="lvm", size=disk_sz,
2363
                              logical_id=(vgname, names[0]))
2364
    sda_dev_m2 = objects.Disk(dev_type="lvm", size=disk_sz,
2365
                              logical_id=(vgname, names[1]))
2366
    md_sda_dev = objects.Disk(dev_type="md_raid1", iv_name = "sda",
2367
                              size=disk_sz,
2368
                              children = [sda_dev_m1, sda_dev_m2])
2369
    sdb_dev_m1 = objects.Disk(dev_type="lvm", size=swap_sz,
2370
                              logical_id=(vgname, names[2]))
2371
    sdb_dev_m2 = objects.Disk(dev_type="lvm", size=swap_sz,
2372
                              logical_id=(vgname, names[3]))
2373
    md_sdb_dev = objects.Disk(dev_type="md_raid1", iv_name = "sdb",
2374
                              size=swap_sz,
2375
                              children = [sdb_dev_m1, sdb_dev_m2])
2376
    disks = [md_sda_dev, md_sdb_dev]
2377
  elif template_name == "remote_raid1":
2378
    if len(secondary_nodes) != 1:
2379
      raise errors.ProgrammerError("Wrong template configuration")
2380
    remote_node = secondary_nodes[0]
2381
    names = _GenerateUniqueNames(cfg, [".sda_data", ".sda_meta",
2382
                                       ".sdb_data", ".sdb_meta"])
2383
    drbd_sda_dev = _GenerateMDDRBDBranch(cfg, primary_node, remote_node,
2384
                                         disk_sz, names[0:2])
2385
    md_sda_dev = objects.Disk(dev_type="md_raid1", iv_name="sda",
2386
                              children = [drbd_sda_dev], size=disk_sz)
2387
    drbd_sdb_dev = _GenerateMDDRBDBranch(cfg, primary_node, remote_node,
2388
                                         swap_sz, names[2:4])
2389
    md_sdb_dev = objects.Disk(dev_type="md_raid1", iv_name="sdb",
2390
                              children = [drbd_sdb_dev], size=swap_sz)
2391
    disks = [md_sda_dev, md_sdb_dev]
2392
  else:
2393
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
2394
  return disks
2395

    
2396

    
2397
def _GetInstanceInfoText(instance):
2398
  return "originstname+%s" % instance.name
2399

    
2400

    
2401
def _CreateDisks(cfg, instance):
2402
  """Create all disks for an instance.
2403

2404
  This abstracts away some work from AddInstance.
2405

2406
  Args:
2407
    instance: the instance object
2408

2409
  Returns:
2410
    True or False showing the success of the creation process
2411

2412
  """
2413
  info = _GetInstanceInfoText(instance)
2414

    
2415
  for device in instance.disks:
2416
    logger.Info("creating volume %s for instance %s" %
2417
              (device.iv_name, instance.name))
2418
    #HARDCODE
2419
    for secondary_node in instance.secondary_nodes:
2420
      if not _CreateBlockDevOnSecondary(cfg, secondary_node, device, False,
2421
                                        info):
2422
        logger.Error("failed to create volume %s (%s) on secondary node %s!" %
2423
                     (device.iv_name, device, secondary_node))
2424
        return False
2425
    #HARDCODE
2426
    if not _CreateBlockDevOnPrimary(cfg, instance.primary_node, device, info):
2427
      logger.Error("failed to create volume %s on primary!" %
2428
                   device.iv_name)
2429
      return False
2430
  return True
2431

    
2432

    
2433
def _RemoveDisks(instance, cfg):
2434
  """Remove all disks for an instance.
2435

2436
  This abstracts away some work from `AddInstance()` and
2437
  `RemoveInstance()`. Note that in case some of the devices couldn't
2438
  be remove, the removal will continue with the other ones (compare
2439
  with `_CreateDisks()`).
2440

2441
  Args:
2442
    instance: the instance object
2443

2444
  Returns:
2445
    True or False showing the success of the removal proces
2446

2447
  """
2448
  logger.Info("removing block devices for instance %s" % instance.name)
2449

    
2450
  result = True
2451
  for device in instance.disks:
2452
    for node, disk in device.ComputeNodeTree(instance.primary_node):
2453
      cfg.SetDiskID(disk, node)
2454
      if not rpc.call_blockdev_remove(node, disk):
2455
        logger.Error("could not remove block device %s on node %s,"
2456
                     " continuing anyway" %
2457
                     (device.iv_name, node))
2458
        result = False
2459
  return result
2460

    
2461

    
2462
class LUCreateInstance(LogicalUnit):
2463
  """Create an instance.
2464

2465
  """
2466
  HPATH = "instance-add"
2467
  HTYPE = constants.HTYPE_INSTANCE
2468
  _OP_REQP = ["instance_name", "mem_size", "disk_size", "pnode",
2469
              "disk_template", "swap_size", "mode", "start", "vcpus",
2470
              "wait_for_sync"]
2471

    
2472
  def BuildHooksEnv(self):
2473
    """Build hooks env.
2474

2475
    This runs on master, primary and secondary nodes of the instance.
2476

2477
    """
2478
    env = {
2479
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
2480
      "INSTANCE_DISK_SIZE": self.op.disk_size,
2481
      "INSTANCE_SWAP_SIZE": self.op.swap_size,
2482
      "INSTANCE_ADD_MODE": self.op.mode,
2483
      }
2484
    if self.op.mode == constants.INSTANCE_IMPORT:
2485
      env["INSTANCE_SRC_NODE"] = self.op.src_node
2486
      env["INSTANCE_SRC_PATH"] = self.op.src_path
2487
      env["INSTANCE_SRC_IMAGE"] = self.src_image
2488

    
2489
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
2490
      primary_node=self.op.pnode,
2491
      secondary_nodes=self.secondaries,
2492
      status=self.instance_status,
2493
      os_type=self.op.os_type,
2494
      memory=self.op.mem_size,
2495
      vcpus=self.op.vcpus,
2496
      nics=[(self.inst_ip, self.op.bridge)],
2497
    ))
2498

    
2499
    nl = ([self.sstore.GetMasterNode(), self.op.pnode] +
2500
          self.secondaries)
2501
    return env, nl, nl
2502

    
2503

    
2504
  def CheckPrereq(self):
2505
    """Check prerequisites.
2506

2507
    """
2508
    if self.op.mode not in (constants.INSTANCE_CREATE,
2509
                            constants.INSTANCE_IMPORT):
2510
      raise errors.OpPrereqError, ("Invalid instance creation mode '%s'" %
2511
                                   self.op.mode)
2512

    
2513
    if self.op.mode == constants.INSTANCE_IMPORT:
2514
      src_node = getattr(self.op, "src_node", None)
2515
      src_path = getattr(self.op, "src_path", None)
2516
      if src_node is None or src_path is None:
2517
        raise errors.OpPrereqError, ("Importing an instance requires source"
2518
                                     " node and path options")
2519
      src_node_full = self.cfg.ExpandNodeName(src_node)
2520
      if src_node_full is None:
2521
        raise errors.OpPrereqError, ("Unknown source node '%s'" % src_node)
2522
      self.op.src_node = src_node = src_node_full
2523

    
2524
      if not os.path.isabs(src_path):
2525
        raise errors.OpPrereqError, ("The source path must be absolute")
2526

    
2527
      export_info = rpc.call_export_info(src_node, src_path)
2528

    
2529
      if not export_info:
2530
        raise errors.OpPrereqError, ("No export found in dir %s" % src_path)
2531

    
2532
      if not export_info.has_section(constants.INISECT_EXP):
2533
        raise errors.ProgrammerError, ("Corrupted export config")
2534

    
2535
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
2536
      if (int(ei_version) != constants.EXPORT_VERSION):
2537
        raise errors.OpPrereqError, ("Wrong export version %s (wanted %d)" %
2538
                                     (ei_version, constants.EXPORT_VERSION))
2539

    
2540
      if int(export_info.get(constants.INISECT_INS, 'disk_count')) > 1:
2541
        raise errors.OpPrereqError, ("Can't import instance with more than"
2542
                                     " one data disk")
2543

    
2544
      # FIXME: are the old os-es, disk sizes, etc. useful?
2545
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
2546
      diskimage = os.path.join(src_path, export_info.get(constants.INISECT_INS,
2547
                                                         'disk0_dump'))
2548
      self.src_image = diskimage
2549
    else: # INSTANCE_CREATE
2550
      if getattr(self.op, "os_type", None) is None:
2551
        raise errors.OpPrereqError, ("No guest OS specified")
2552

    
2553
    # check primary node
2554
    pnode = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.pnode))
2555
    if pnode is None:
2556
      raise errors.OpPrereqError, ("Primary node '%s' is unknown" %
2557
                                   self.op.pnode)
2558
    self.op.pnode = pnode.name
2559
    self.pnode = pnode
2560
    self.secondaries = []
2561
    # disk template and mirror node verification
2562
    if self.op.disk_template not in constants.DISK_TEMPLATES:
2563
      raise errors.OpPrereqError, ("Invalid disk template name")
2564

    
2565
    if self.op.disk_template == constants.DT_REMOTE_RAID1:
2566
      if getattr(self.op, "snode", None) is None:
2567
        raise errors.OpPrereqError, ("The 'remote_raid1' disk template needs"
2568
                                     " a mirror node")
2569

    
2570
      snode_name = self.cfg.ExpandNodeName(self.op.snode)
2571
      if snode_name is None:
2572
        raise errors.OpPrereqError, ("Unknown secondary node '%s'" %
2573
                                     self.op.snode)
2574
      elif snode_name == pnode.name:
2575
        raise errors.OpPrereqError, ("The secondary node cannot be"
2576
                                     " the primary node.")
2577
      self.secondaries.append(snode_name)
2578

    
2579
    # Check lv size requirements
2580
    nodenames = [pnode.name] + self.secondaries
2581
    nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName())
2582

    
2583
    # Required free disk space as a function of disk and swap space
2584
    req_size_dict = {
2585
      constants.DT_DISKLESS: 0,
2586
      constants.DT_PLAIN: self.op.disk_size + self.op.swap_size,
2587
      constants.DT_LOCAL_RAID1: (self.op.disk_size + self.op.swap_size) * 2,
2588
      # 256 MB are added for drbd metadata, 128MB for each drbd device
2589
      constants.DT_REMOTE_RAID1: self.op.disk_size + self.op.swap_size + 256,
2590
    }
2591

    
2592
    if self.op.disk_template not in req_size_dict:
2593
      raise errors.ProgrammerError, ("Disk template '%s' size requirement"
2594
                                     " is unknown" %  self.op.disk_template)
2595

    
2596
    req_size = req_size_dict[self.op.disk_template]
2597

    
2598
    for node in nodenames:
2599
      info = nodeinfo.get(node, None)
2600
      if not info:
2601
        raise errors.OpPrereqError, ("Cannot get current information"
2602
                                     " from node '%s'" % nodeinfo)
2603
      if req_size > info['vg_free']:
2604
        raise errors.OpPrereqError, ("Not enough disk space on target node %s."
2605
                                     " %d MB available, %d MB required" %
2606
                                     (node, info['vg_free'], req_size))
2607

    
2608
    # os verification
2609
    os_obj = rpc.call_os_get([pnode.name], self.op.os_type)[pnode.name]
2610
    if not isinstance(os_obj, objects.OS):
2611
      raise errors.OpPrereqError, ("OS '%s' not in supported os list for"
2612
                                   " primary node"  % self.op.os_type)
2613

    
2614
    # instance verification
2615
    hostname1 = utils.LookupHostname(self.op.instance_name)
2616
    if not hostname1:
2617
      raise errors.OpPrereqError, ("Instance name '%s' not found in dns" %
2618
                                   self.op.instance_name)
2619

    
2620
    self.op.instance_name = instance_name = hostname1['hostname']
2621
    instance_list = self.cfg.GetInstanceList()
2622
    if instance_name in instance_list:
2623
      raise errors.OpPrereqError, ("Instance '%s' is already in the cluster" %
2624
                                   instance_name)
2625

    
2626
    ip = getattr(self.op, "ip", None)
2627
    if ip is None or ip.lower() == "none":
2628
      inst_ip = None
2629
    elif ip.lower() == "auto":
2630
      inst_ip = hostname1['ip']
2631
    else:
2632
      if not utils.IsValidIP(ip):
2633
        raise errors.OpPrereqError, ("given IP address '%s' doesn't look"
2634
                                     " like a valid IP" % ip)
2635
      inst_ip = ip
2636
    self.inst_ip = inst_ip
2637

    
2638
    command = ["fping", "-q", hostname1['ip']]
2639
    result = utils.RunCmd(command)
2640
    if not result.failed:
2641
      raise errors.OpPrereqError, ("IP %s of instance %s already in use" %
2642
                                   (hostname1['ip'], instance_name))
2643

    
2644
    # bridge verification
2645
    bridge = getattr(self.op, "bridge", None)
2646
    if bridge is None:
2647
      self.op.bridge = self.cfg.GetDefBridge()
2648
    else:
2649
      self.op.bridge = bridge
2650

    
2651
    if not rpc.call_bridges_exist(self.pnode.name, [self.op.bridge]):
2652
      raise errors.OpPrereqError, ("target bridge '%s' does not exist on"
2653
                                   " destination node '%s'" %
2654
                                   (self.op.bridge, pnode.name))
2655

    
2656
    if self.op.start:
2657
      self.instance_status = 'up'
2658
    else:
2659
      self.instance_status = 'down'
2660

    
2661
  def Exec(self, feedback_fn):
2662
    """Create and add the instance to the cluster.
2663

2664
    """
2665
    instance = self.op.instance_name
2666
    pnode_name = self.pnode.name
2667

    
2668
    nic = objects.NIC(bridge=self.op.bridge, mac=self.cfg.GenerateMAC())
2669
    if self.inst_ip is not None:
2670
      nic.ip = self.inst_ip
2671

    
2672
    disks = _GenerateDiskTemplate(self.cfg,
2673
                                  self.op.disk_template,
2674
                                  instance, pnode_name,
2675
                                  self.secondaries, self.op.disk_size,
2676
                                  self.op.swap_size)
2677

    
2678
    iobj = objects.Instance(name=instance, os=self.op.os_type,
2679
                            primary_node=pnode_name,
2680
                            memory=self.op.mem_size,
2681
                            vcpus=self.op.vcpus,
2682
                            nics=[nic], disks=disks,
2683
                            disk_template=self.op.disk_template,
2684
                            status=self.instance_status,
2685
                            )
2686

    
2687
    feedback_fn("* creating instance disks...")
2688
    if not _CreateDisks(self.cfg, iobj):
2689
      _RemoveDisks(iobj, self.cfg)
2690
      raise errors.OpExecError, ("Device creation failed, reverting...")
2691

    
2692
    feedback_fn("adding instance %s to cluster config" % instance)
2693

    
2694
    self.cfg.AddInstance(iobj)
2695

    
2696
    if self.op.wait_for_sync:
2697
      disk_abort = not _WaitForSync(self.cfg, iobj)
2698
    elif iobj.disk_template == "remote_raid1":
2699
      # make sure the disks are not degraded (still sync-ing is ok)
2700
      time.sleep(15)
2701
      feedback_fn("* checking mirrors status")
2702
      disk_abort = not _WaitForSync(self.cfg, iobj, oneshot=True)
2703
    else:
2704
      disk_abort = False
2705

    
2706
    if disk_abort:
2707
      _RemoveDisks(iobj, self.cfg)
2708
      self.cfg.RemoveInstance(iobj.name)
2709
      raise errors.OpExecError, ("There are some degraded disks for"
2710
                                      " this instance")
2711

    
2712
    feedback_fn("creating os for instance %s on node %s" %
2713
                (instance, pnode_name))
2714

    
2715
    if iobj.disk_template != constants.DT_DISKLESS:
2716
      if self.op.mode == constants.INSTANCE_CREATE:
2717
        feedback_fn("* running the instance OS create scripts...")
2718
        if not rpc.call_instance_os_add(pnode_name, iobj, "sda", "sdb"):
2719
          raise errors.OpExecError, ("could not add os for instance %s"
2720
                                          " on node %s" %
2721
                                          (instance, pnode_name))
2722

    
2723
      elif self.op.mode == constants.INSTANCE_IMPORT:
2724
        feedback_fn("* running the instance OS import scripts...")
2725
        src_node = self.op.src_node
2726
        src_image = self.src_image
2727
        if not rpc.call_instance_os_import(pnode_name, iobj, "sda", "sdb",
2728
                                                src_node, src_image):
2729
          raise errors.OpExecError, ("Could not import os for instance"
2730
                                          " %s on node %s" %
2731
                                          (instance, pnode_name))
2732
      else:
2733
        # also checked in the prereq part
2734
        raise errors.ProgrammerError, ("Unknown OS initialization mode '%s'"
2735
                                       % self.op.mode)
2736

    
2737
    if self.op.start:
2738
      logger.Info("starting instance %s on node %s" % (instance, pnode_name))
2739
      feedback_fn("* starting instance...")
2740
      if not rpc.call_instance_start(pnode_name, iobj, None):
2741
        raise errors.OpExecError, ("Could not start instance")
2742

    
2743

    
2744
class LUConnectConsole(NoHooksLU):
2745
  """Connect to an instance's console.
2746

2747
  This is somewhat special in that it returns the command line that
2748
  you need to run on the master node in order to connect to the
2749
  console.
2750

2751
  """
2752
  _OP_REQP = ["instance_name"]
2753

    
2754
  def CheckPrereq(self):
2755
    """Check prerequisites.
2756

2757
    This checks that the instance is in the cluster.
2758

2759
    """
2760
    instance = self.cfg.GetInstanceInfo(
2761
      self.cfg.ExpandInstanceName(self.op.instance_name))
2762
    if instance is None:
2763
      raise errors.OpPrereqError, ("Instance '%s' not known" %
2764
                                   self.op.instance_name)
2765
    self.instance = instance
2766

    
2767
  def Exec(self, feedback_fn):
2768
    """Connect to the console of an instance
2769

2770
    """
2771
    instance = self.instance
2772
    node = instance.primary_node
2773

    
2774
    node_insts = rpc.call_instance_list([node])[node]
2775
    if node_insts is False:
2776
      raise errors.OpExecError, ("Can't connect to node %s." % node)
2777

    
2778
    if instance.name not in node_insts:
2779
      raise errors.OpExecError, ("Instance %s is not running." % instance.name)
2780

    
2781
    logger.Debug("connecting to console of %s on %s" % (instance.name, node))
2782

    
2783
    hyper = hypervisor.GetHypervisor()
2784
    console_cmd = hyper.GetShellCommandForConsole(instance.name)
2785
    return node, console_cmd
2786

    
2787

    
2788
class LUAddMDDRBDComponent(LogicalUnit):
2789
  """Adda new mirror member to an instance's disk.
2790

2791
  """
2792
  HPATH = "mirror-add"
2793
  HTYPE = constants.HTYPE_INSTANCE
2794
  _OP_REQP = ["instance_name", "remote_node", "disk_name"]
2795

    
2796
  def BuildHooksEnv(self):
2797
    """Build hooks env.
2798

2799
    This runs on the master, the primary and all the secondaries.
2800

2801
    """
2802
    env = {
2803
      "NEW_SECONDARY": self.op.remote_node,
2804
      "DISK_NAME": self.op.disk_name,
2805
      }
2806
    env.update(_BuildInstanceHookEnvByObject(self.instance))
2807
    nl = [self.sstore.GetMasterNode(), self.instance.primary_node,
2808
          self.op.remote_node,] + list(self.instance.secondary_nodes)
2809
    return env, nl, nl
2810

    
2811
  def CheckPrereq(self):
2812
    """Check prerequisites.
2813

2814
    This checks that the instance is in the cluster.
2815

2816
    """
2817
    instance = self.cfg.GetInstanceInfo(
2818
      self.cfg.ExpandInstanceName(self.op.instance_name))
2819
    if instance is None:
2820
      raise errors.OpPrereqError, ("Instance '%s' not known" %
2821
                                   self.op.instance_name)
2822
    self.instance = instance
2823

    
2824
    remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
2825
    if remote_node is None:
2826
      raise errors.OpPrereqError, ("Node '%s' not known" % self.op.remote_node)
2827
    self.remote_node = remote_node
2828

    
2829
    if remote_node == instance.primary_node:
2830
      raise errors.OpPrereqError, ("The specified node is the primary node of"
2831
                                   " the instance.")
2832

    
2833
    if instance.disk_template != constants.DT_REMOTE_RAID1:
2834
      raise errors.OpPrereqError, ("Instance's disk layout is not"
2835
                                   " remote_raid1.")
2836
    for disk in instance.disks:
2837
      if disk.iv_name == self.op.disk_name:
2838
        break
2839
    else:
2840
      raise errors.OpPrereqError, ("Can't find this device ('%s') in the"
2841
                                   " instance." % self.op.disk_name)
2842
    if len(disk.children) > 1:
2843
      raise errors.OpPrereqError, ("The device already has two slave"
2844
                                   " devices.\n"
2845
                                   "This would create a 3-disk raid1"
2846
                                   " which we don't allow.")
2847
    self.disk = disk
2848

    
2849
  def Exec(self, feedback_fn):
2850
    """Add the mirror component
2851

2852
    """
2853
    disk = self.disk
2854
    instance = self.instance
2855

    
2856
    remote_node = self.remote_node
2857
    lv_names = [".%s_%s" % (disk.iv_name, suf) for suf in ["data", "meta"]]
2858
    names = _GenerateUniqueNames(self.cfg, lv_names)
2859
    new_drbd = _GenerateMDDRBDBranch(self.cfg, instance.primary_node,
2860
                                     remote_node, disk.size, names)
2861

    
2862
    logger.Info("adding new mirror component on secondary")
2863
    #HARDCODE
2864
    if not _CreateBlockDevOnSecondary(self.cfg, remote_node, new_drbd, False,
2865
                                      _GetInstanceInfoText(instance)):
2866
      raise errors.OpExecError, ("Failed to create new component on secondary"
2867
                                 " node %s" % remote_node)
2868

    
2869
    logger.Info("adding new mirror component on primary")
2870
    #HARDCODE
2871
    if not _CreateBlockDevOnPrimary(self.cfg, instance.primary_node, new_drbd,
2872
                                    _GetInstanceInfoText(instance)):
2873
      # remove secondary dev
2874
      self.cfg.SetDiskID(new_drbd, remote_node)
2875
      rpc.call_blockdev_remove(remote_node, new_drbd)
2876
      raise errors.OpExecError, ("Failed to create volume on primary")
2877

    
2878
    # the device exists now
2879
    # call the primary node to add the mirror to md
2880
    logger.Info("adding new mirror component to md")
2881
    if not rpc.call_blockdev_addchild(instance.primary_node,
2882
                                           disk, new_drbd):
2883
      logger.Error("Can't add mirror compoment to md!")
2884
      self.cfg.SetDiskID(new_drbd, remote_node)
2885
      if not rpc.call_blockdev_remove(remote_node, new_drbd):
2886
        logger.Error("Can't rollback on secondary")
2887
      self.cfg.SetDiskID(new_drbd, instance.primary_node)
2888
      if not rpc.call_blockdev_remove(instance.primary_node, new_drbd):
2889
        logger.Error("Can't rollback on primary")
2890
      raise errors.OpExecError, "Can't add mirror component to md array"
2891

    
2892
    disk.children.append(new_drbd)
2893

    
2894
    self.cfg.AddInstance(instance)
2895

    
2896
    _WaitForSync(self.cfg, instance)
2897

    
2898
    return 0
2899

    
2900

    
2901
class LURemoveMDDRBDComponent(LogicalUnit):
2902
  """Remove a component from a remote_raid1 disk.
2903

2904
  """
2905
  HPATH = "mirror-remove"
2906
  HTYPE = constants.HTYPE_INSTANCE
2907
  _OP_REQP = ["instance_name", "disk_name", "disk_id"]
2908

    
2909
  def BuildHooksEnv(self):
2910
    """Build hooks env.
2911

2912
    This runs on the master, the primary and all the secondaries.
2913

2914
    """
2915
    env = {
2916
      "DISK_NAME": self.op.disk_name,
2917
      "DISK_ID": self.op.disk_id,
2918
      "OLD_SECONDARY": self.old_secondary,
2919
      }
2920
    env.update(_BuildInstanceHookEnvByObject(self.instance))
2921
    nl = [self.sstore.GetMasterNode(),
2922
          self.instance.primary_node] + list(self.instance.secondary_nodes)
2923
    return env, nl, nl
2924

    
2925
  def CheckPrereq(self):
2926
    """Check prerequisites.
2927

2928
    This checks that the instance is in the cluster.
2929

2930
    """
2931
    instance = self.cfg.GetInstanceInfo(
2932
      self.cfg.ExpandInstanceName(self.op.instance_name))
2933
    if instance is None:
2934
      raise errors.OpPrereqError, ("Instance '%s' not known" %
2935
                                   self.op.instance_name)
2936
    self.instance = instance
2937

    
2938
    if instance.disk_template != constants.DT_REMOTE_RAID1:
2939
      raise errors.OpPrereqError, ("Instance's disk layout is not"
2940
                                   " remote_raid1.")
2941
    for disk in instance.disks:
2942
      if disk.iv_name == self.op.disk_name:
2943
        break
2944
    else:
2945
      raise errors.OpPrereqError, ("Can't find this device ('%s') in the"
2946
                                   " instance." % self.op.disk_name)
2947
    for child in disk.children:
2948
      if child.dev_type == "drbd" and child.logical_id[2] == self.op.disk_id:
2949
        break
2950
    else:
2951
      raise errors.OpPrereqError, ("Can't find the device with this port.")
2952

    
2953
    if len(disk.children) < 2:
2954
      raise errors.OpPrereqError, ("Cannot remove the last component from"
2955
                                   " a mirror.")
2956
    self.disk = disk
2957
    self.child = child
2958
    if self.child.logical_id[0] == instance.primary_node:
2959
      oid = 1
2960
    else:
2961
      oid = 0
2962
    self.old_secondary = self.child.logical_id[oid]
2963

    
2964
  def Exec(self, feedback_fn):
2965
    """Remove the mirror component
2966

2967
    """
2968
    instance = self.instance
2969
    disk = self.disk
2970
    child = self.child
2971
    logger.Info("remove mirror component")
2972
    self.cfg.SetDiskID(disk, instance.primary_node)
2973
    if not rpc.call_blockdev_removechild(instance.primary_node,
2974
                                              disk, child):
2975
      raise errors.OpExecError, ("Can't remove child from mirror.")
2976

    
2977
    for node in child.logical_id[:2]:
2978
      self.cfg.SetDiskID(child, node)
2979
      if not rpc.call_blockdev_remove(node, child):
2980
        logger.Error("Warning: failed to remove device from node %s,"
2981
                     " continuing operation." % node)
2982

    
2983
    disk.children.remove(child)
2984
    self.cfg.AddInstance(instance)
2985

    
2986

    
2987
class LUReplaceDisks(LogicalUnit):
2988
  """Replace the disks of an instance.
2989

2990
  """
2991
  HPATH = "mirrors-replace"
2992
  HTYPE = constants.HTYPE_INSTANCE
2993
  _OP_REQP = ["instance_name"]
2994

    
2995
  def BuildHooksEnv(self):
2996
    """Build hooks env.
2997

2998
    This runs on the master, the primary and all the secondaries.
2999

3000
    """
3001
    env = {
3002
      "NEW_SECONDARY": self.op.remote_node,
3003
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
3004
      }
3005
    env.update(_BuildInstanceHookEnvByObject(self.instance))
3006
    nl = [self.sstore.GetMasterNode(),
3007
          self.instance.primary_node] + list(self.instance.secondary_nodes)
3008
    return env, nl, nl
3009

    
3010
  def CheckPrereq(self):
3011
    """Check prerequisites.
3012

3013
    This checks that the instance is in the cluster.
3014

3015
    """
3016
    instance = self.cfg.GetInstanceInfo(
3017
      self.cfg.ExpandInstanceName(self.op.instance_name))
3018
    if instance is None:
3019
      raise errors.OpPrereqError, ("Instance '%s' not known" %
3020
                                   self.op.instance_name)
3021
    self.instance = instance
3022

    
3023
    if instance.disk_template != constants.DT_REMOTE_RAID1:
3024
      raise errors.OpPrereqError, ("Instance's disk layout is not"
3025
                                   " remote_raid1.")
3026

    
3027
    if len(instance.secondary_nodes) != 1:
3028
      raise errors.OpPrereqError, ("The instance has a strange layout,"
3029
                                   " expected one secondary but found %d" %
3030
                                   len(instance.secondary_nodes))
3031

    
3032
    remote_node = getattr(self.op, "remote_node", None)
3033
    if remote_node is None:
3034
      remote_node = instance.secondary_nodes[0]
3035
    else:
3036
      remote_node = self.cfg.ExpandNodeName(remote_node)
3037
      if remote_node is None:
3038
        raise errors.OpPrereqError, ("Node '%s' not known" %
3039
                                     self.op.remote_node)
3040
    if remote_node == instance.primary_node:
3041
      raise errors.OpPrereqError, ("The specified node is the primary node of"
3042
                                   " the instance.")
3043
    self.op.remote_node = remote_node
3044

    
3045
  def Exec(self, feedback_fn):
3046
    """Replace the disks of an instance.
3047

3048
    """
3049
    instance = self.instance
3050
    iv_names = {}
3051
    # start of work
3052
    remote_node = self.op.remote_node
3053
    cfg = self.cfg
3054
    vgname = cfg.GetVGName()
3055
    for dev in instance.disks:
3056
      size = dev.size
3057
      lv_names = [".%s_%s" % (dev.iv_name, suf) for suf in ["data", "meta"]]
3058
      names = _GenerateUniqueNames(cfg, lv_names)
3059
      new_drbd = _GenerateMDDRBDBranch(cfg, instance.primary_node,
3060
                                       remote_node, size, names)
3061
      iv_names[dev.iv_name] = (dev, dev.children[0], new_drbd)
3062
      logger.Info("adding new mirror component on secondary for %s" %
3063
                  dev.iv_name)
3064
      #HARDCODE
3065
      if not _CreateBlockDevOnSecondary(cfg, remote_node, new_drbd, False,
3066
                                        _GetInstanceInfoText(instance)):
3067
        raise errors.OpExecError, ("Failed to create new component on"
3068
                                   " secondary node %s\n"
3069
                                   "Full abort, cleanup manually!" %
3070
                                   remote_node)
3071

    
3072
      logger.Info("adding new mirror component on primary")
3073
      #HARDCODE
3074
      if not _CreateBlockDevOnPrimary(cfg, instance.primary_node, new_drbd,
3075
                                      _GetInstanceInfoText(instance)):
3076
        # remove secondary dev
3077
        cfg.SetDiskID(new_drbd, remote_node)
3078
        rpc.call_blockdev_remove(remote_node, new_drbd)
3079
        raise errors.OpExecError("Failed to create volume on primary!\n"
3080
                                 "Full abort, cleanup manually!!")
3081

    
3082
      # the device exists now
3083
      # call the primary node to add the mirror to md
3084
      logger.Info("adding new mirror component to md")
3085
      if not rpc.call_blockdev_addchild(instance.primary_node, dev,
3086
                                        new_drbd):
3087
        logger.Error("Can't add mirror compoment to md!")
3088
        cfg.SetDiskID(new_drbd, remote_node)
3089
        if not rpc.call_blockdev_remove(remote_node, new_drbd):
3090
          logger.Error("Can't rollback on secondary")
3091
        cfg.SetDiskID(new_drbd, instance.primary_node)
3092
        if not rpc.call_blockdev_remove(instance.primary_node, new_drbd):
3093
          logger.Error("Can't rollback on primary")
3094
        raise errors.OpExecError, ("Full abort, cleanup manually!!")
3095

    
3096
      dev.children.append(new_drbd)
3097
      cfg.AddInstance(instance)
3098

    
3099
    # this can fail as the old devices are degraded and _WaitForSync
3100
    # does a combined result over all disks, so we don't check its
3101
    # return value
3102
    _WaitForSync(cfg, instance, unlock=True)
3103

    
3104
    # so check manually all the devices
3105
    for name in iv_names:
3106
      dev, child, new_drbd = iv_names[name]
3107
      cfg.SetDiskID(dev, instance.primary_node)
3108
      is_degr = rpc.call_blockdev_find(instance.primary_node, dev)[5]
3109
      if is_degr:
3110
        raise errors.OpExecError, ("MD device %s is degraded!" % name)
3111
      cfg.SetDiskID(new_drbd, instance.primary_node)
3112
      is_degr = rpc.call_blockdev_find(instance.primary_node, new_drbd)[5]
3113
      if is_degr:
3114
        raise errors.OpExecError, ("New drbd device %s is degraded!" % name)
3115

    
3116
    for name in iv_names:
3117
      dev, child, new_drbd = iv_names[name]
3118
      logger.Info("remove mirror %s component" % name)
3119
      cfg.SetDiskID(dev, instance.primary_node)
3120
      if not rpc.call_blockdev_removechild(instance.primary_node,
3121
                                                dev, child):
3122
        logger.Error("Can't remove child from mirror, aborting"
3123
                     " *this device cleanup*.\nYou need to cleanup manually!!")
3124
        continue
3125

    
3126
      for node in child.logical_id[:2]:
3127
        logger.Info("remove child device on %s" % node)
3128
        cfg.SetDiskID(child, node)
3129
        if not rpc.call_blockdev_remove(node, child):
3130
          logger.Error("Warning: failed to remove device from node %s,"
3131
                       " continuing operation." % node)
3132

    
3133
      dev.children.remove(child)
3134

    
3135
      cfg.AddInstance(instance)
3136

    
3137

    
3138
class LUQueryInstanceData(NoHooksLU):
3139
  """Query runtime instance data.
3140

3141
  """
3142
  _OP_REQP = ["instances"]
3143

    
3144
  def CheckPrereq(self):
3145
    """Check prerequisites.
3146

3147
    This only checks the optional instance list against the existing names.
3148

3149
    """
3150
    if not isinstance(self.op.instances, list):
3151
      raise errors.OpPrereqError, "Invalid argument type 'instances'"
3152
    if self.op.instances:
3153
      self.wanted_instances = []
3154
      names = self.op.instances
3155
      for name in names:
3156
        instance = self.cfg.GetInstanceInfo(self.cfg.ExpandInstanceName(name))
3157
        if instance is None:
3158
          raise errors.OpPrereqError, ("No such instance name '%s'" % name)
3159
      self.wanted_instances.append(instance)
3160
    else:
3161
      self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
3162
                               in self.cfg.GetInstanceList()]
3163
    return
3164

    
3165

    
3166
  def _ComputeDiskStatus(self, instance, snode, dev):
3167
    """Compute block device status.
3168

3169
    """
3170
    self.cfg.SetDiskID(dev, instance.primary_node)
3171
    dev_pstatus = rpc.call_blockdev_find(instance.primary_node, dev)
3172
    if dev.dev_type == "drbd":
3173
      # we change the snode then (otherwise we use the one passed in)
3174
      if dev.logical_id[0] == instance.primary_node:
3175
        snode = dev.logical_id[1]
3176
      else:
3177
        snode = dev.logical_id[0]
3178

    
3179
    if snode:
3180
      self.cfg.SetDiskID(dev, snode)
3181
      dev_sstatus = rpc.call_blockdev_find(snode, dev)
3182
    else:
3183
      dev_sstatus = None
3184

    
3185
    if dev.children:
3186
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
3187
                      for child in dev.children]
3188
    else:
3189
      dev_children = []
3190

    
3191
    data = {
3192
      "iv_name": dev.iv_name,
3193
      "dev_type": dev.dev_type,
3194
      "logical_id": dev.logical_id,
3195
      "physical_id": dev.physical_id,
3196
      "pstatus": dev_pstatus,
3197
      "sstatus": dev_sstatus,
3198
      "children": dev_children,
3199
      }
3200

    
3201
    return data
3202

    
3203
  def Exec(self, feedback_fn):
3204
    """Gather and return data"""
3205
    result = {}
3206
    for instance in self.wanted_instances:
3207
      remote_info = rpc.call_instance_info(instance.primary_node,
3208
                                                instance.name)
3209
      if remote_info and "state" in remote_info:
3210
        remote_state = "up"
3211
      else:
3212
        remote_state = "down"
3213
      if instance.status == "down":
3214
        config_state = "down"
3215
      else:
3216
        config_state = "up"
3217

    
3218
      disks = [self._ComputeDiskStatus(instance, None, device)
3219
               for device in instance.disks]
3220

    
3221
      idict = {
3222
        "name": instance.name,
3223
        "config_state": config_state,
3224
        "run_state": remote_state,
3225
        "pnode": instance.primary_node,
3226
        "snodes": instance.secondary_nodes,
3227
        "os": instance.os,
3228
        "memory": instance.memory,
3229
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
3230
        "disks": disks,
3231
        }
3232

    
3233
      result[instance.name] = idict
3234

    
3235
    return result
3236

    
3237

    
3238
class LUQueryNodeData(NoHooksLU):
3239
  """Logical unit for querying node data.
3240

3241
  """
3242
  _OP_REQP = ["nodes"]
3243

    
3244
  def CheckPrereq(self):
3245
    """Check prerequisites.
3246

3247
    This only checks the optional node list against the existing names.
3248

3249
    """
3250
    self.wanted_nodes = _GetWantedNodes(self, self.op.nodes)
3251

    
3252
  def Exec(self, feedback_fn):
3253
    """Compute and return the list of nodes.
3254

3255
    """
3256
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3257
             in self.cfg.GetInstanceList()]
3258
    result = []
3259
    for node in self.wanted_nodes:
3260
      result.append((node.name, node.primary_ip, node.secondary_ip,
3261
                     [inst.name for inst in ilist
3262
                      if inst.primary_node == node.name],
3263
                     [inst.name for inst in ilist
3264
                      if node.name in inst.secondary_nodes],
3265
                     ))
3266
    return result
3267

    
3268

    
3269
class LUSetInstanceParms(LogicalUnit):
3270
  """Modifies an instances's parameters.
3271

3272
  """
3273
  HPATH = "instance-modify"
3274
  HTYPE = constants.HTYPE_INSTANCE
3275
  _OP_REQP = ["instance_name"]
3276

    
3277
  def BuildHooksEnv(self):
3278
    """Build hooks env.
3279

3280
    This runs on the master, primary and secondaries.
3281

3282
    """
3283
    args = dict()
3284
    if self.mem:
3285
      args['memory'] = self.mem
3286
    if self.vcpus:
3287
      args['vcpus'] = self.vcpus
3288
    if self.do_ip or self.do_bridge:
3289
      if self.do_ip:
3290
        ip = self.ip
3291
      else:
3292
        ip = self.instance.nics[0].ip
3293
      if self.bridge:
3294
        bridge = self.bridge
3295
      else:
3296
        bridge = self.instance.nics[0].bridge
3297
      args['nics'] = [(ip, bridge)]
3298
    env = _BuildInstanceHookEnvByObject(self.instance, override=args)
3299
    nl = [self.sstore.GetMasterNode(),
3300
          self.instance.primary_node] + list(self.instance.secondary_nodes)
3301
    return env, nl, nl
3302

    
3303
  def CheckPrereq(self):
3304
    """Check prerequisites.
3305

3306
    This only checks the instance list against the existing names.
3307

3308
    """
3309
    self.mem = getattr(self.op, "mem", None)
3310
    self.vcpus = getattr(self.op, "vcpus", None)
3311
    self.ip = getattr(self.op, "ip", None)
3312
    self.bridge = getattr(self.op, "bridge", None)
3313
    if [self.mem, self.vcpus, self.ip, self.bridge].count(None) == 4:
3314
      raise errors.OpPrereqError, ("No changes submitted")
3315
    if self.mem is not None:
3316
      try:
3317
        self.mem = int(self.mem)
3318
      except ValueError, err:
3319
        raise errors.OpPrereqError, ("Invalid memory size: %s" % str(err))
3320
    if self.vcpus is not None:
3321
      try:
3322
        self.vcpus = int(self.vcpus)
3323
      except ValueError, err:
3324
        raise errors.OpPrereqError, ("Invalid vcpus number: %s" % str(err))
3325
    if self.ip is not None:
3326
      self.do_ip = True
3327
      if self.ip.lower() == "none":
3328
        self.ip = None
3329
      else:
3330
        if not utils.IsValidIP(self.ip):
3331
          raise errors.OpPrereqError, ("Invalid IP address '%s'." % self.ip)
3332
    else:
3333
      self.do_ip = False
3334
    self.do_bridge = (self.bridge is not None)
3335

    
3336
    instance = self.cfg.GetInstanceInfo(
3337
      self.cfg.ExpandInstanceName(self.op.instance_name))
3338
    if instance is None:
3339
      raise errors.OpPrereqError, ("No such instance name '%s'" %
3340
                                   self.op.instance_name)
3341
    self.op.instance_name = instance.name
3342
    self.instance = instance
3343
    return
3344

    
3345
  def Exec(self, feedback_fn):
3346
    """Modifies an instance.
3347

3348
    All parameters take effect only at the next restart of the instance.
3349
    """
3350
    result = []
3351
    instance = self.instance
3352
    if self.mem:
3353
      instance.memory = self.mem
3354
      result.append(("mem", self.mem))
3355
    if self.vcpus:
3356
      instance.vcpus = self.vcpus
3357
      result.append(("vcpus",  self.vcpus))
3358
    if self.do_ip:
3359
      instance.nics[0].ip = self.ip
3360
      result.append(("ip", self.ip))
3361
    if self.bridge:
3362
      instance.nics[0].bridge = self.bridge
3363
      result.append(("bridge", self.bridge))
3364

    
3365
    self.cfg.AddInstance(instance)
3366

    
3367
    return result
3368

    
3369

    
3370
class LUQueryExports(NoHooksLU):
3371
  """Query the exports list
3372

3373
  """
3374
  _OP_REQP = []
3375

    
3376
  def CheckPrereq(self):
3377
    """Check that the nodelist contains only existing nodes.
3378

3379
    """
3380
    self.nodes = _GetWantedNodes(self, getattr(self.op, "nodes", None))
3381

    
3382
  def Exec(self, feedback_fn):
3383
    """Compute the list of all the exported system images.
3384

3385
    Returns:
3386
      a dictionary with the structure node->(export-list)
3387
      where export-list is a list of the instances exported on
3388
      that node.
3389

3390
    """
3391
    return rpc.call_export_list([node.name for node in self.nodes])
3392

    
3393

    
3394
class LUExportInstance(LogicalUnit):
3395
  """Export an instance to an image in the cluster.
3396

3397
  """
3398
  HPATH = "instance-export"
3399
  HTYPE = constants.HTYPE_INSTANCE
3400
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
3401

    
3402
  def BuildHooksEnv(self):
3403
    """Build hooks env.
3404

3405
    This will run on the master, primary node and target node.
3406

3407
    """
3408
    env = {
3409
      "EXPORT_NODE": self.op.target_node,
3410
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
3411
      }
3412
    env.update(_BuildInstanceHookEnvByObject(self.instance))
3413
    nl = [self.sstore.GetMasterNode(), self.instance.primary_node,
3414
          self.op.target_node]
3415
    return env, nl, nl
3416

    
3417
  def CheckPrereq(self):
3418
    """Check prerequisites.
3419

3420
    This checks that the instance name is a valid one.
3421

3422
    """
3423
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
3424
    self.instance = self.cfg.GetInstanceInfo(instance_name)
3425
    if self.instance is None:
3426
      raise errors.OpPrereqError, ("Instance '%s' not found" %
3427
                                   self.op.instance_name)
3428

    
3429
    # node verification
3430
    dst_node_short = self.cfg.ExpandNodeName(self.op.target_node)
3431
    self.dst_node = self.cfg.GetNodeInfo(dst_node_short)
3432

    
3433
    if self.dst_node is None:
3434
      raise errors.OpPrereqError, ("Destination node '%s' is unknown." %
3435
                                   self.op.target_node)
3436
    self.op.target_node = self.dst_node.name
3437

    
3438
  def Exec(self, feedback_fn):
3439
    """Export an instance to an image in the cluster.
3440

3441
    """
3442
    instance = self.instance
3443
    dst_node = self.dst_node
3444
    src_node = instance.primary_node
3445
    # shutdown the instance, unless requested not to do so
3446
    if self.op.shutdown:
3447
      op = opcodes.OpShutdownInstance(instance_name=instance.name)
3448
      self.processor.ChainOpCode(op, feedback_fn)
3449

    
3450
    vgname = self.cfg.GetVGName()
3451

    
3452
    snap_disks = []
3453

    
3454
    try:
3455
      for disk in instance.disks:
3456
        if disk.iv_name == "sda":
3457
          # new_dev_name will be a snapshot of an lvm leaf of the one we passed
3458
          new_dev_name = rpc.call_blockdev_snapshot(src_node, disk)
3459

    
3460
          if not new_dev_name:
3461
            logger.Error("could not snapshot block device %s on node %s" %
3462
                         (disk.logical_id[1], src_node))
3463
          else:
3464
            new_dev = objects.Disk(dev_type="lvm", size=disk.size,
3465
                                      logical_id=(vgname, new_dev_name),
3466
                                      physical_id=(vgname, new_dev_name),
3467
                                      iv_name=disk.iv_name)
3468
            snap_disks.append(new_dev)
3469

    
3470
    finally:
3471
      if self.op.shutdown:
3472
        op = opcodes.OpStartupInstance(instance_name=instance.name,
3473
                                       force=False)
3474
        self.processor.ChainOpCode(op, feedback_fn)
3475

    
3476
    # TODO: check for size
3477

    
3478
    for dev in snap_disks:
3479
      if not rpc.call_snapshot_export(src_node, dev, dst_node.name,
3480
                                           instance):
3481
        logger.Error("could not export block device %s from node"
3482
                     " %s to node %s" %
3483
                     (dev.logical_id[1], src_node, dst_node.name))
3484
      if not rpc.call_blockdev_remove(src_node, dev):
3485
        logger.Error("could not remove snapshot block device %s from"
3486
                     " node %s" % (dev.logical_id[1], src_node))
3487

    
3488
    if not rpc.call_finalize_export(dst_node.name, instance, snap_disks):
3489
      logger.Error("could not finalize export for instance %s on node %s" %
3490
                   (instance.name, dst_node.name))
3491

    
3492
    nodelist = self.cfg.GetNodeList()
3493
    nodelist.remove(dst_node.name)
3494

    
3495
    # on one-node clusters nodelist will be empty after the removal
3496
    # if we proceed the backup would be removed because OpQueryExports
3497
    # substitutes an empty list with the full cluster node list.
3498
    if nodelist:
3499
      op = opcodes.OpQueryExports(nodes=nodelist)
3500
      exportlist = self.processor.ChainOpCode(op, feedback_fn)
3501
      for node in exportlist:
3502
        if instance.name in exportlist[node]:
3503
          if not rpc.call_export_remove(node, instance.name):
3504
            logger.Error("could not remove older export for instance %s"
3505
                         " on node %s" % (instance.name, node))