4 # Copyright (C) 2006, 2007 Google Inc.
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.
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.
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
22 """Functions used by the node daemon
24 @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
25 the L{UploadFile} function
26 @var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted
27 in the L{_CleanDirectory} function
31 # pylint: disable-msg=E1103
33 # E1103: %s %r has no %r member (but some types could not be
34 # inferred), because the _TryOSFromDisk returns either (True, os_obj)
35 # or (False, "string") which confuses pylint
51 from ganeti import errors
52 from ganeti import utils
53 from ganeti import ssh
54 from ganeti import hypervisor
55 from ganeti import constants
56 from ganeti import bdev
57 from ganeti import objects
58 from ganeti import ssconf
61 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
62 _ALLOWED_CLEAN_DIRS = frozenset([
64 constants.JOB_QUEUE_ARCHIVE_DIR,
69 class RPCFail(Exception):
70 """Class denoting RPC failure.
72 Its argument is the error message.
77 def _Fail(msg, *args, **kwargs):
78 """Log an error and the raise an RPCFail exception.
80 This exception is then handled specially in the ganeti daemon and
81 turned into a 'failed' return type. As such, this function is a
82 useful shortcut for logging the error and returning it to the master
86 @param msg: the text of the exception
92 if "log" not in kwargs or kwargs["log"]: # if we should log this error
93 if "exc" in kwargs and kwargs["exc"]:
94 logging.exception(msg)
101 """Simple wrapper to return a SimpleStore.
103 @rtype: L{ssconf.SimpleStore}
104 @return: a SimpleStore instance
107 return ssconf.SimpleStore()
110 def _GetSshRunner(cluster_name):
111 """Simple wrapper to return an SshRunner.
113 @type cluster_name: str
114 @param cluster_name: the cluster name, which is needed
115 by the SshRunner constructor
116 @rtype: L{ssh.SshRunner}
117 @return: an SshRunner instance
120 return ssh.SshRunner(cluster_name)
123 def _Decompress(data):
124 """Unpacks data compressed by the RPC client.
126 @type data: list or tuple
127 @param data: Data sent by RPC client
129 @return: Decompressed data
132 assert isinstance(data, (list, tuple))
133 assert len(data) == 2
134 (encoding, content) = data
135 if encoding == constants.RPC_ENCODING_NONE:
137 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
138 return zlib.decompress(base64.b64decode(content))
140 raise AssertionError("Unknown data encoding")
143 def _CleanDirectory(path, exclude=None):
144 """Removes all regular files in a directory.
147 @param path: the directory to clean
149 @param exclude: list of files to be excluded, defaults
153 if path not in _ALLOWED_CLEAN_DIRS:
154 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
157 if not os.path.isdir(path):
162 # Normalize excluded paths
163 exclude = [os.path.normpath(i) for i in exclude]
165 for rel_name in utils.ListVisibleFiles(path):
166 full_name = utils.PathJoin(path, rel_name)
167 if full_name in exclude:
169 if os.path.isfile(full_name) and not os.path.islink(full_name):
170 utils.RemoveFile(full_name)
173 def _BuildUploadFileList():
174 """Build the list of allowed upload files.
176 This is abstracted so that it's built only once at module import time.
179 allowed_files = set([
180 constants.CLUSTER_CONF_FILE,
182 constants.SSH_KNOWN_HOSTS_FILE,
183 constants.VNC_PASSWORD_FILE,
184 constants.RAPI_CERT_FILE,
185 constants.RAPI_USERS_FILE,
186 constants.HMAC_CLUSTER_KEY,
189 for hv_name in constants.HYPER_TYPES:
190 hv_class = hypervisor.GetHypervisorClass(hv_name)
191 allowed_files.update(hv_class.GetAncillaryFiles())
193 return frozenset(allowed_files)
196 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
200 """Removes job queue files and archived jobs.
206 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
207 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
211 """Returns master information.
213 This is an utility function to compute master information, either
214 for consumption here or from the node daemon.
217 @return: master_netdev, master_ip, master_name
218 @raise RPCFail: in case of errors
223 master_netdev = cfg.GetMasterNetdev()
224 master_ip = cfg.GetMasterIP()
225 master_node = cfg.GetMasterNode()
226 except errors.ConfigurationError, err:
227 _Fail("Cluster configuration incomplete: %s", err, exc=True)
228 return (master_netdev, master_ip, master_node)
231 def StartMaster(start_daemons, no_voting):
232 """Activate local node as master node.
234 The function will always try activate the IP address of the master
235 (unless someone else has it). It will also start the master daemons,
236 based on the start_daemons parameter.
238 @type start_daemons: boolean
239 @param start_daemons: whether to also start the master
240 daemons (ganeti-masterd and ganeti-rapi)
241 @type no_voting: boolean
242 @param no_voting: whether to start ganeti-masterd without a node vote
243 (if start_daemons is True), but still non-interactively
247 # GetMasterInfo will raise an exception if not able to return data
248 master_netdev, master_ip, _ = GetMasterInfo()
251 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
252 if utils.OwnIpAddress(master_ip):
253 # we already have the ip:
254 logging.debug("Master IP already configured, doing nothing")
256 msg = "Someone else has the master ip, not activating"
260 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
261 "dev", master_netdev, "label",
262 "%s:0" % master_netdev])
264 msg = "Can't activate master IP: %s" % result.output
268 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
269 "-s", master_ip, master_ip])
270 # we'll ignore the exit code of arping
272 # and now start the master and rapi daemons
275 masterd_args = "--no-voting --yes-do-it"
280 "EXTRA_MASTERD_ARGS": masterd_args,
283 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
285 msg = "Can't start Ganeti master: %s" % result.output
290 _Fail("; ".join(err_msgs))
293 def StopMaster(stop_daemons):
294 """Deactivate this node as master.
296 The function will always try to deactivate the IP address of the
297 master. It will also stop the master daemons depending on the
298 stop_daemons parameter.
300 @type stop_daemons: boolean
301 @param stop_daemons: whether to also stop the master daemons
302 (ganeti-masterd and ganeti-rapi)
306 # TODO: log and report back to the caller the error failures; we
307 # need to decide in which case we fail the RPC for this
309 # GetMasterInfo will raise an exception if not able to return data
310 master_netdev, master_ip, _ = GetMasterInfo()
312 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
313 "dev", master_netdev])
315 logging.error("Can't remove the master IP, error: %s", result.output)
316 # but otherwise ignore the failure
319 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
321 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
323 result.cmd, result.exit_code, result.output)
326 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
327 """Joins this node to the cluster.
329 This does the following:
330 - updates the hostkeys of the machine (rsa and dsa)
331 - adds the ssh private key to the user
332 - adds the ssh public key to the users' authorized_keys file
335 @param dsa: the DSA private key to write
337 @param dsapub: the DSA public key to write
339 @param rsa: the RSA private key to write
341 @param rsapub: the RSA public key to write
343 @param sshkey: the SSH private key to write
345 @param sshpub: the SSH public key to write
347 @return: the success of the operation
350 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
351 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
352 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
353 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
354 for name, content, mode in sshd_keys:
355 utils.WriteFile(name, data=content, mode=mode)
358 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
360 except errors.OpExecError, err:
361 _Fail("Error while processing user ssh files: %s", err, exc=True)
363 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
364 utils.WriteFile(name, data=content, mode=0600)
366 utils.AddAuthorizedKey(auth_keys, sshpub)
368 result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"])
370 _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)",
371 result.cmd, result.exit_code, result.output)
374 def LeaveCluster(modify_ssh_setup):
375 """Cleans up and remove the current node.
377 This function cleans up and prepares the current node to be removed
380 If processing is successful, then it raises an
381 L{errors.QuitGanetiException} which is used as a special case to
382 shutdown the node daemon.
384 @param modify_ssh_setup: boolean
387 _CleanDirectory(constants.DATA_DIR)
392 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
394 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
396 utils.RemoveFile(priv_key)
397 utils.RemoveFile(pub_key)
398 except errors.OpExecError:
399 logging.exception("Error while processing ssh files")
402 utils.RemoveFile(constants.HMAC_CLUSTER_KEY)
403 utils.RemoveFile(constants.RAPI_CERT_FILE)
404 utils.RemoveFile(constants.SSL_CERT_FILE)
405 except: # pylint: disable-msg=W0702
406 logging.exception("Error while removing cluster secrets")
408 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
410 logging.error("Command %s failed with exitcode %s and error %s",
411 result.cmd, result.exit_code, result.output)
413 # Raise a custom exception (handled in ganeti-noded)
414 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
417 def GetNodeInfo(vgname, hypervisor_type):
418 """Gives back a hash with different information about the node.
420 @type vgname: C{string}
421 @param vgname: the name of the volume group to ask for disk space information
422 @type hypervisor_type: C{str}
423 @param hypervisor_type: the name of the hypervisor to ask for
426 @return: dictionary with the following keys:
427 - vg_size is the size of the configured volume group in MiB
428 - vg_free is the free size of the volume group in MiB
429 - memory_dom0 is the memory allocated for domain0 in MiB
430 - memory_free is the currently available (free) ram in MiB
431 - memory_total is the total number of ram in MiB
435 vginfo = _GetVGInfo(vgname)
436 outputarray['vg_size'] = vginfo['vg_size']
437 outputarray['vg_free'] = vginfo['vg_free']
439 hyper = hypervisor.GetHypervisor(hypervisor_type)
440 hyp_info = hyper.GetNodeInfo()
441 if hyp_info is not None:
442 outputarray.update(hyp_info)
444 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
449 def VerifyNode(what, cluster_name):
450 """Verify the status of the local node.
452 Based on the input L{what} parameter, various checks are done on the
455 If the I{filelist} key is present, this list of
456 files is checksummed and the file/checksum pairs are returned.
458 If the I{nodelist} key is present, we check that we have
459 connectivity via ssh with the target nodes (and check the hostname
462 If the I{node-net-test} key is present, we check that we have
463 connectivity to the given nodes via both primary IP and, if
464 applicable, secondary IPs.
467 @param what: a dictionary of things to check:
468 - filelist: list of files for which to compute checksums
469 - nodelist: list of nodes we should check ssh communication with
470 - node-net-test: list of nodes we should check node daemon port
472 - hypervisor: list with hypervisors to run the verify for
474 @return: a dictionary with the same keys as the input dict, and
475 values representing the result of the checks
480 if constants.NV_HYPERVISOR in what:
481 result[constants.NV_HYPERVISOR] = tmp = {}
482 for hv_name in what[constants.NV_HYPERVISOR]:
483 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
485 if constants.NV_FILELIST in what:
486 result[constants.NV_FILELIST] = utils.FingerprintFiles(
487 what[constants.NV_FILELIST])
489 if constants.NV_NODELIST in what:
490 result[constants.NV_NODELIST] = tmp = {}
491 random.shuffle(what[constants.NV_NODELIST])
492 for node in what[constants.NV_NODELIST]:
493 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
497 if constants.NV_NODENETTEST in what:
498 result[constants.NV_NODENETTEST] = tmp = {}
499 my_name = utils.HostInfo().name
500 my_pip = my_sip = None
501 for name, pip, sip in what[constants.NV_NODENETTEST]:
507 tmp[my_name] = ("Can't find my own primary/secondary IP"
510 port = utils.GetDaemonPort(constants.NODED)
511 for name, pip, sip in what[constants.NV_NODENETTEST]:
513 if not utils.TcpPing(pip, port, source=my_pip):
514 fail.append("primary")
516 if not utils.TcpPing(sip, port, source=my_sip):
517 fail.append("secondary")
519 tmp[name] = ("failure using the %s interface(s)" %
522 if constants.NV_LVLIST in what:
523 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
525 if constants.NV_INSTANCELIST in what:
526 result[constants.NV_INSTANCELIST] = GetInstanceList(
527 what[constants.NV_INSTANCELIST])
529 if constants.NV_VGLIST in what:
530 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
532 if constants.NV_PVLIST in what:
533 result[constants.NV_PVLIST] = \
534 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
535 filter_allocatable=False)
537 if constants.NV_VERSION in what:
538 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
539 constants.RELEASE_VERSION)
541 if constants.NV_HVINFO in what:
542 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
543 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
545 if constants.NV_DRBDLIST in what:
547 used_minors = bdev.DRBD8.GetUsedDevs().keys()
548 except errors.BlockDeviceError, err:
549 logging.warning("Can't get used minors list", exc_info=True)
550 used_minors = str(err)
551 result[constants.NV_DRBDLIST] = used_minors
553 if constants.NV_NODESETUP in what:
554 result[constants.NV_NODESETUP] = tmpr = []
555 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
556 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
557 " under /sys, missing required directories /sys/block"
558 " and /sys/class/net")
559 if (not os.path.isdir("/proc/sys") or
560 not os.path.isfile("/proc/sysrq-trigger")):
561 tmpr.append("The procfs filesystem doesn't seem to be mounted"
562 " under /proc, missing required directory /proc/sys and"
563 " the file /proc/sysrq-trigger")
565 if constants.NV_TIME in what:
566 result[constants.NV_TIME] = utils.SplitTime(time.time())
571 def GetVolumeList(vg_name):
572 """Compute list of logical volumes and their size.
575 @param vg_name: the volume group whose LVs we should list
578 dictionary of all partions (key) with value being a tuple of
579 their size (in MiB), inactive and online status::
581 {'test1': ('20.06', True, True)}
583 in case of errors, a string is returned with the error
589 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
590 "--separator=%s" % sep,
591 "-olv_name,lv_size,lv_attr", vg_name])
593 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
595 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
596 for line in result.stdout.splitlines():
598 match = valid_line_re.match(line)
600 logging.error("Invalid line returned from lvs output: '%s'", line)
602 name, size, attr = match.groups()
603 inactive = attr[4] == '-'
604 online = attr[5] == 'o'
605 virtual = attr[0] == 'v'
607 # we don't want to report such volumes as existing, since they
608 # don't really hold data
610 lvs[name] = (size, inactive, online)
615 def ListVolumeGroups():
616 """List the volume groups and their size.
619 @return: dictionary with keys volume name and values the
623 return utils.ListVolumeGroups()
627 """List all volumes on this node.
631 A list of dictionaries, each having four keys:
632 - name: the logical volume name,
633 - size: the size of the logical volume
634 - dev: the physical device on which the LV lives
635 - vg: the volume group to which it belongs
637 In case of errors, we return an empty list and log the
640 Note that since a logical volume can live on multiple physical
641 volumes, the resulting list might include a logical volume
645 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
647 "--options=lv_name,lv_size,devices,vg_name"])
649 _Fail("Failed to list logical volumes, lvs output: %s",
654 return dev.split('(')[0]
660 'name': line[0].strip(),
661 'size': line[1].strip(),
662 'dev': parse_dev(line[2].strip()),
663 'vg': line[3].strip(),
666 return [map_line(line.split('|')) for line in result.stdout.splitlines()
667 if line.count('|') >= 3]
670 def BridgesExist(bridges_list):
671 """Check if a list of bridges exist on the current node.
674 @return: C{True} if all of them exist, C{False} otherwise
678 for bridge in bridges_list:
679 if not utils.BridgeExists(bridge):
680 missing.append(bridge)
683 _Fail("Missing bridges %s", utils.CommaJoin(missing))
686 def GetInstanceList(hypervisor_list):
687 """Provides a list of instances.
689 @type hypervisor_list: list
690 @param hypervisor_list: the list of hypervisors to query information
693 @return: a list of all running instances on the current node
694 - instance1.example.com
695 - instance2.example.com
699 for hname in hypervisor_list:
701 names = hypervisor.GetHypervisor(hname).ListInstances()
702 results.extend(names)
703 except errors.HypervisorError, err:
704 _Fail("Error enumerating instances (hypervisor %s): %s",
705 hname, err, exc=True)
710 def GetInstanceInfo(instance, hname):
711 """Gives back the information about an instance as a dictionary.
713 @type instance: string
714 @param instance: the instance name
716 @param hname: the hypervisor type of the instance
719 @return: dictionary with the following keys:
720 - memory: memory size of instance (int)
721 - state: xen state of instance (string)
722 - time: cpu time of instance (float)
727 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
728 if iinfo is not None:
729 output['memory'] = iinfo[2]
730 output['state'] = iinfo[4]
731 output['time'] = iinfo[5]
736 def GetInstanceMigratable(instance):
737 """Gives whether an instance can be migrated.
739 @type instance: L{objects.Instance}
740 @param instance: object representing the instance to be checked.
743 @return: tuple of (result, description) where:
744 - result: whether the instance can be migrated or not
745 - description: a description of the issue, if relevant
748 hyper = hypervisor.GetHypervisor(instance.hypervisor)
749 iname = instance.name
750 if iname not in hyper.ListInstances():
751 _Fail("Instance %s is not running", iname)
753 for idx in range(len(instance.disks)):
754 link_name = _GetBlockDevSymlinkPath(iname, idx)
755 if not os.path.islink(link_name):
756 _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
759 def GetAllInstancesInfo(hypervisor_list):
760 """Gather data about all instances.
762 This is the equivalent of L{GetInstanceInfo}, except that it
763 computes data for all instances at once, thus being faster if one
764 needs data about more than one instance.
766 @type hypervisor_list: list
767 @param hypervisor_list: list of hypervisors to query for instance data
770 @return: dictionary of instance: data, with data having the following keys:
771 - memory: memory size of instance (int)
772 - state: xen state of instance (string)
773 - time: cpu time of instance (float)
774 - vcpus: the number of vcpus
779 for hname in hypervisor_list:
780 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
782 for name, _, memory, vcpus, state, times in iinfo:
790 # we only check static parameters, like memory and vcpus,
791 # and not state and time which can change between the
792 # invocations of the different hypervisors
793 for key in 'memory', 'vcpus':
794 if value[key] != output[name][key]:
795 _Fail("Instance %s is running twice"
796 " with different parameters", name)
802 def _InstanceLogName(kind, os_name, instance):
803 """Compute the OS log filename for a given instance and operation.
805 The instance name and os name are passed in as strings since not all
806 operations have these as part of an instance object.
809 @param kind: the operation type (e.g. add, import, etc.)
810 @type os_name: string
811 @param os_name: the os name
812 @type instance: string
813 @param instance: the name of the instance being imported/added/etc.
816 base = "%s-%s-%s-%d.log" % (kind, os_name, instance, int(time.time()))
817 return utils.PathJoin(constants.LOG_OS_DIR, base)
820 def InstanceOsAdd(instance, reinstall, debug):
821 """Add an OS to an instance.
823 @type instance: L{objects.Instance}
824 @param instance: Instance whose OS is to be installed
825 @type reinstall: boolean
826 @param reinstall: whether this is an instance reinstall
828 @param debug: debug level, passed to the OS scripts
832 inst_os = OSFromDisk(instance.os)
834 create_env = OSEnvironment(instance, inst_os, debug)
836 create_env['INSTANCE_REINSTALL'] = "1"
838 logfile = _InstanceLogName("add", instance.os, instance.name)
840 result = utils.RunCmd([inst_os.create_script], env=create_env,
841 cwd=inst_os.path, output=logfile,)
843 logging.error("os create command '%s' returned error: %s, logfile: %s,"
844 " output: %s", result.cmd, result.fail_reason, logfile,
846 lines = [utils.SafeEncode(val)
847 for val in utils.TailFile(logfile, lines=20)]
848 _Fail("OS create script failed (%s), last lines in the"
849 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
852 def RunRenameInstance(instance, old_name, debug):
853 """Run the OS rename script for an instance.
855 @type instance: L{objects.Instance}
856 @param instance: Instance whose OS is to be installed
857 @type old_name: string
858 @param old_name: previous instance name
860 @param debug: debug level, passed to the OS scripts
862 @return: the success of the operation
865 inst_os = OSFromDisk(instance.os)
867 rename_env = OSEnvironment(instance, inst_os, debug)
868 rename_env['OLD_INSTANCE_NAME'] = old_name
870 logfile = _InstanceLogName("rename", instance.os,
871 "%s-%s" % (old_name, instance.name))
873 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
874 cwd=inst_os.path, output=logfile)
877 logging.error("os create command '%s' returned error: %s output: %s",
878 result.cmd, result.fail_reason, result.output)
879 lines = [utils.SafeEncode(val)
880 for val in utils.TailFile(logfile, lines=20)]
881 _Fail("OS rename script failed (%s), last lines in the"
882 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
885 def _GetVGInfo(vg_name):
886 """Get information about the volume group.
889 @param vg_name: the volume group which we query
892 A dictionary with the following keys:
893 - C{vg_size} is the total size of the volume group in MiB
894 - C{vg_free} is the free size of the volume group in MiB
895 - C{pv_count} are the number of physical disks in that VG
897 If an error occurs during gathering of data, we return the same dict
898 with keys all set to None.
901 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
903 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
904 "--nosuffix", "--units=m", "--separator=:", vg_name])
907 logging.error("volume group %s not present", vg_name)
909 valarr = retval.stdout.strip().rstrip(':').split(':')
913 "vg_size": int(round(float(valarr[0]), 0)),
914 "vg_free": int(round(float(valarr[1]), 0)),
915 "pv_count": int(valarr[2]),
917 except (TypeError, ValueError), err:
918 logging.exception("Fail to parse vgs output: %s", err)
920 logging.error("vgs output has the wrong number of fields (expected"
921 " three): %s", str(valarr))
925 def _GetBlockDevSymlinkPath(instance_name, idx):
926 return utils.PathJoin(constants.DISK_LINKS_DIR,
927 "%s:%d" % (instance_name, idx))
930 def _SymlinkBlockDev(instance_name, device_path, idx):
931 """Set up symlinks to a instance's block device.
933 This is an auxiliary function run when an instance is start (on the primary
934 node) or when an instance is migrated (on the target node).
937 @param instance_name: the name of the target instance
938 @param device_path: path of the physical block device, on the node
939 @param idx: the disk index
940 @return: absolute path to the disk's symlink
943 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
945 os.symlink(device_path, link_name)
947 if err.errno == errno.EEXIST:
948 if (not os.path.islink(link_name) or
949 os.readlink(link_name) != device_path):
951 os.symlink(device_path, link_name)
958 def _RemoveBlockDevLinks(instance_name, disks):
959 """Remove the block device symlinks belonging to the given instance.
962 for idx, _ in enumerate(disks):
963 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
964 if os.path.islink(link_name):
968 logging.exception("Can't remove symlink '%s'", link_name)
971 def _GatherAndLinkBlockDevs(instance):
972 """Set up an instance's block device(s).
974 This is run on the primary node at instance startup. The block
975 devices must be already assembled.
977 @type instance: L{objects.Instance}
978 @param instance: the instance whose disks we shoul assemble
980 @return: list of (disk_object, device_path)
984 for idx, disk in enumerate(instance.disks):
985 device = _RecursiveFindBD(disk)
987 raise errors.BlockDeviceError("Block device '%s' is not set up." %
991 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
993 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
996 block_devices.append((disk, link_name))
1001 def StartInstance(instance):
1002 """Start an instance.
1004 @type instance: L{objects.Instance}
1005 @param instance: the instance object
1009 running_instances = GetInstanceList([instance.hypervisor])
1011 if instance.name in running_instances:
1012 logging.info("Instance %s already running, not starting", instance.name)
1016 block_devices = _GatherAndLinkBlockDevs(instance)
1017 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1018 hyper.StartInstance(instance, block_devices)
1019 except errors.BlockDeviceError, err:
1020 _Fail("Block device error: %s", err, exc=True)
1021 except errors.HypervisorError, err:
1022 _RemoveBlockDevLinks(instance.name, instance.disks)
1023 _Fail("Hypervisor error: %s", err, exc=True)
1026 def InstanceShutdown(instance, timeout):
1027 """Shut an instance down.
1029 @note: this functions uses polling with a hardcoded timeout.
1031 @type instance: L{objects.Instance}
1032 @param instance: the instance object
1033 @type timeout: integer
1034 @param timeout: maximum timeout for soft shutdown
1038 hv_name = instance.hypervisor
1039 hyper = hypervisor.GetHypervisor(hv_name)
1040 iname = instance.name
1042 if instance.name not in hyper.ListInstances():
1043 logging.info("Instance %s not running, doing nothing", iname)
1048 self.tried_once = False
1051 if iname not in hyper.ListInstances():
1055 hyper.StopInstance(instance, retry=self.tried_once)
1056 except errors.HypervisorError, err:
1057 if iname not in hyper.ListInstances():
1058 # if the instance is no longer existing, consider this a
1059 # success and go to cleanup
1062 _Fail("Failed to stop instance %s: %s", iname, err)
1064 self.tried_once = True
1066 raise utils.RetryAgain()
1069 utils.Retry(_TryShutdown(), 5, timeout)
1070 except utils.RetryTimeout:
1071 # the shutdown did not succeed
1072 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1075 hyper.StopInstance(instance, force=True)
1076 except errors.HypervisorError, err:
1077 if iname in hyper.ListInstances():
1078 # only raise an error if the instance still exists, otherwise
1079 # the error could simply be "instance ... unknown"!
1080 _Fail("Failed to force stop instance %s: %s", iname, err)
1084 if iname in hyper.ListInstances():
1085 _Fail("Could not shutdown instance %s even by destroy", iname)
1087 _RemoveBlockDevLinks(iname, instance.disks)
1090 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1091 """Reboot an instance.
1093 @type instance: L{objects.Instance}
1094 @param instance: the instance object to reboot
1095 @type reboot_type: str
1096 @param reboot_type: the type of reboot, one the following
1098 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1099 instance OS, do not recreate the VM
1100 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1101 restart the VM (at the hypervisor level)
1102 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1103 not accepted here, since that mode is handled differently, in
1104 cmdlib, and translates into full stop and start of the
1105 instance (instead of a call_instance_reboot RPC)
1106 @type shutdown_timeout: integer
1107 @param shutdown_timeout: maximum timeout for soft shutdown
1111 running_instances = GetInstanceList([instance.hypervisor])
1113 if instance.name not in running_instances:
1114 _Fail("Cannot reboot instance %s that is not running", instance.name)
1116 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1117 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1119 hyper.RebootInstance(instance)
1120 except errors.HypervisorError, err:
1121 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1122 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1124 InstanceShutdown(instance, shutdown_timeout)
1125 return StartInstance(instance)
1126 except errors.HypervisorError, err:
1127 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1129 _Fail("Invalid reboot_type received: %s", reboot_type)
1132 def MigrationInfo(instance):
1133 """Gather information about an instance to be migrated.
1135 @type instance: L{objects.Instance}
1136 @param instance: the instance definition
1139 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1141 info = hyper.MigrationInfo(instance)
1142 except errors.HypervisorError, err:
1143 _Fail("Failed to fetch migration information: %s", err, exc=True)
1147 def AcceptInstance(instance, info, target):
1148 """Prepare the node to accept an instance.
1150 @type instance: L{objects.Instance}
1151 @param instance: the instance definition
1152 @type info: string/data (opaque)
1153 @param info: migration information, from the source node
1154 @type target: string
1155 @param target: target host (usually ip), on this node
1158 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1160 hyper.AcceptInstance(instance, info, target)
1161 except errors.HypervisorError, err:
1162 _Fail("Failed to accept instance: %s", err, exc=True)
1165 def FinalizeMigration(instance, info, success):
1166 """Finalize any preparation to accept an instance.
1168 @type instance: L{objects.Instance}
1169 @param instance: the instance definition
1170 @type info: string/data (opaque)
1171 @param info: migration information, from the source node
1172 @type success: boolean
1173 @param success: whether the migration was a success or a failure
1176 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1178 hyper.FinalizeMigration(instance, info, success)
1179 except errors.HypervisorError, err:
1180 _Fail("Failed to finalize migration: %s", err, exc=True)
1183 def MigrateInstance(instance, target, live):
1184 """Migrates an instance to another node.
1186 @type instance: L{objects.Instance}
1187 @param instance: the instance definition
1188 @type target: string
1189 @param target: the target node name
1191 @param live: whether the migration should be done live or not (the
1192 interpretation of this parameter is left to the hypervisor)
1194 @return: a tuple of (success, msg) where:
1195 - succes is a boolean denoting the success/failure of the operation
1196 - msg is a string with details in case of failure
1199 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1202 hyper.MigrateInstance(instance, target, live)
1203 except errors.HypervisorError, err:
1204 _Fail("Failed to migrate instance: %s", err, exc=True)
1207 def BlockdevCreate(disk, size, owner, on_primary, info):
1208 """Creates a block device for an instance.
1210 @type disk: L{objects.Disk}
1211 @param disk: the object describing the disk we should create
1213 @param size: the size of the physical underlying device, in MiB
1215 @param owner: the name of the instance for which disk is created,
1216 used for device cache data
1217 @type on_primary: boolean
1218 @param on_primary: indicates if it is the primary node or not
1220 @param info: string that will be sent to the physical device
1221 creation, used for example to set (LVM) tags on LVs
1223 @return: the new unique_id of the device (this can sometime be
1224 computed only after creation), or None. On secondary nodes,
1225 it's not required to return anything.
1228 # TODO: remove the obsolete 'size' argument
1229 # pylint: disable-msg=W0613
1232 for child in disk.children:
1234 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1235 except errors.BlockDeviceError, err:
1236 _Fail("Can't assemble device %s: %s", child, err)
1237 if on_primary or disk.AssembleOnSecondary():
1238 # we need the children open in case the device itself has to
1241 # pylint: disable-msg=E1103
1243 except errors.BlockDeviceError, err:
1244 _Fail("Can't make child '%s' read-write: %s", child, err)
1248 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1249 except errors.BlockDeviceError, err:
1250 _Fail("Can't create block device: %s", err)
1252 if on_primary or disk.AssembleOnSecondary():
1255 except errors.BlockDeviceError, err:
1256 _Fail("Can't assemble device after creation, unusual event: %s", err)
1257 device.SetSyncSpeed(constants.SYNC_SPEED)
1258 if on_primary or disk.OpenOnSecondary():
1260 device.Open(force=True)
1261 except errors.BlockDeviceError, err:
1262 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1263 DevCacheManager.UpdateCache(device.dev_path, owner,
1264 on_primary, disk.iv_name)
1266 device.SetInfo(info)
1268 return device.unique_id
1271 def BlockdevRemove(disk):
1272 """Remove a block device.
1274 @note: This is intended to be called recursively.
1276 @type disk: L{objects.Disk}
1277 @param disk: the disk object we should remove
1279 @return: the success of the operation
1284 rdev = _RecursiveFindBD(disk)
1285 except errors.BlockDeviceError, err:
1286 # probably can't attach
1287 logging.info("Can't attach to device %s in remove", disk)
1289 if rdev is not None:
1290 r_path = rdev.dev_path
1293 except errors.BlockDeviceError, err:
1294 msgs.append(str(err))
1296 DevCacheManager.RemoveCache(r_path)
1299 for child in disk.children:
1301 BlockdevRemove(child)
1302 except RPCFail, err:
1303 msgs.append(str(err))
1306 _Fail("; ".join(msgs))
1309 def _RecursiveAssembleBD(disk, owner, as_primary):
1310 """Activate a block device for an instance.
1312 This is run on the primary and secondary nodes for an instance.
1314 @note: this function is called recursively.
1316 @type disk: L{objects.Disk}
1317 @param disk: the disk we try to assemble
1319 @param owner: the name of the instance which owns the disk
1320 @type as_primary: boolean
1321 @param as_primary: if we should make the block device
1324 @return: the assembled device or None (in case no device
1326 @raise errors.BlockDeviceError: in case there is an error
1327 during the activation of the children or the device
1333 mcn = disk.ChildrenNeeded()
1335 mcn = 0 # max number of Nones allowed
1337 mcn = len(disk.children) - mcn # max number of Nones
1338 for chld_disk in disk.children:
1340 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1341 except errors.BlockDeviceError, err:
1342 if children.count(None) >= mcn:
1345 logging.error("Error in child activation (but continuing): %s",
1347 children.append(cdev)
1349 if as_primary or disk.AssembleOnSecondary():
1350 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1351 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1353 if as_primary or disk.OpenOnSecondary():
1355 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1356 as_primary, disk.iv_name)
1363 def BlockdevAssemble(disk, owner, as_primary):
1364 """Activate a block device for an instance.
1366 This is a wrapper over _RecursiveAssembleBD.
1368 @rtype: str or boolean
1369 @return: a C{/dev/...} path for primary nodes, and
1370 C{True} for secondary nodes
1374 result = _RecursiveAssembleBD(disk, owner, as_primary)
1375 if isinstance(result, bdev.BlockDev):
1376 # pylint: disable-msg=E1103
1377 result = result.dev_path
1378 except errors.BlockDeviceError, err:
1379 _Fail("Error while assembling disk: %s", err, exc=True)
1384 def BlockdevShutdown(disk):
1385 """Shut down a block device.
1387 First, if the device is assembled (Attach() is successful), then
1388 the device is shutdown. Then the children of the device are
1391 This function is called recursively. Note that we don't cache the
1392 children or such, as oppossed to assemble, shutdown of different
1393 devices doesn't require that the upper device was active.
1395 @type disk: L{objects.Disk}
1396 @param disk: the description of the disk we should
1402 r_dev = _RecursiveFindBD(disk)
1403 if r_dev is not None:
1404 r_path = r_dev.dev_path
1407 DevCacheManager.RemoveCache(r_path)
1408 except errors.BlockDeviceError, err:
1409 msgs.append(str(err))
1412 for child in disk.children:
1414 BlockdevShutdown(child)
1415 except RPCFail, err:
1416 msgs.append(str(err))
1419 _Fail("; ".join(msgs))
1422 def BlockdevAddchildren(parent_cdev, new_cdevs):
1423 """Extend a mirrored block device.
1425 @type parent_cdev: L{objects.Disk}
1426 @param parent_cdev: the disk to which we should add children
1427 @type new_cdevs: list of L{objects.Disk}
1428 @param new_cdevs: the list of children which we should add
1432 parent_bdev = _RecursiveFindBD(parent_cdev)
1433 if parent_bdev is None:
1434 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1435 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1436 if new_bdevs.count(None) > 0:
1437 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1438 parent_bdev.AddChildren(new_bdevs)
1441 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1442 """Shrink a mirrored block device.
1444 @type parent_cdev: L{objects.Disk}
1445 @param parent_cdev: the disk from which we should remove children
1446 @type new_cdevs: list of L{objects.Disk}
1447 @param new_cdevs: the list of children which we should remove
1451 parent_bdev = _RecursiveFindBD(parent_cdev)
1452 if parent_bdev is None:
1453 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1455 for disk in new_cdevs:
1456 rpath = disk.StaticDevPath()
1458 bd = _RecursiveFindBD(disk)
1460 _Fail("Can't find device %s while removing children", disk)
1462 devs.append(bd.dev_path)
1465 parent_bdev.RemoveChildren(devs)
1468 def BlockdevGetmirrorstatus(disks):
1469 """Get the mirroring status of a list of devices.
1471 @type disks: list of L{objects.Disk}
1472 @param disks: the list of disks which we should query
1475 a list of (mirror_done, estimated_time) tuples, which
1476 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1477 @raise errors.BlockDeviceError: if any of the disks cannot be
1483 rbd = _RecursiveFindBD(dsk)
1485 _Fail("Can't find device %s", dsk)
1487 stats.append(rbd.CombinedSyncStatus())
1492 def _RecursiveFindBD(disk):
1493 """Check if a device is activated.
1495 If so, return information about the real device.
1497 @type disk: L{objects.Disk}
1498 @param disk: the disk object we need to find
1500 @return: None if the device can't be found,
1501 otherwise the device instance
1506 for chdisk in disk.children:
1507 children.append(_RecursiveFindBD(chdisk))
1509 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1512 def BlockdevFind(disk):
1513 """Check if a device is activated.
1515 If it is, return information about the real device.
1517 @type disk: L{objects.Disk}
1518 @param disk: the disk to find
1519 @rtype: None or objects.BlockDevStatus
1520 @return: None if the disk cannot be found, otherwise a the current
1525 rbd = _RecursiveFindBD(disk)
1526 except errors.BlockDeviceError, err:
1527 _Fail("Failed to find device: %s", err, exc=True)
1532 return rbd.GetSyncStatus()
1535 def BlockdevGetsize(disks):
1536 """Computes the size of the given disks.
1538 If a disk is not found, returns None instead.
1540 @type disks: list of L{objects.Disk}
1541 @param disks: the list of disk to compute the size for
1543 @return: list with elements None if the disk cannot be found,
1550 rbd = _RecursiveFindBD(cf)
1551 except errors.BlockDeviceError:
1557 result.append(rbd.GetActualSize())
1561 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1562 """Export a block device to a remote node.
1564 @type disk: L{objects.Disk}
1565 @param disk: the description of the disk to export
1566 @type dest_node: str
1567 @param dest_node: the destination node to export to
1568 @type dest_path: str
1569 @param dest_path: the destination path on the target node
1570 @type cluster_name: str
1571 @param cluster_name: the cluster name, needed for SSH hostalias
1575 real_disk = _RecursiveFindBD(disk)
1576 if real_disk is None:
1577 _Fail("Block device '%s' is not set up", disk)
1581 # the block size on the read dd is 1MiB to match our units
1582 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1583 "dd if=%s bs=1048576 count=%s",
1584 real_disk.dev_path, str(disk.size))
1586 # we set here a smaller block size as, due to ssh buffering, more
1587 # than 64-128k will mostly ignored; we use nocreat to fail if the
1588 # device is not already there or we pass a wrong path; we use
1589 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1590 # to not buffer too much memory; this means that at best, we flush
1591 # every 64k, which will not be very fast
1592 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1593 " oflag=dsync", dest_path)
1595 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1596 constants.GANETI_RUNAS,
1599 # all commands have been checked, so we're safe to combine them
1600 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1602 result = utils.RunCmd(["bash", "-c", command])
1605 _Fail("Disk copy command '%s' returned error: %s"
1606 " output: %s", command, result.fail_reason, result.output)
1609 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1610 """Write a file to the filesystem.
1612 This allows the master to overwrite(!) a file. It will only perform
1613 the operation if the file belongs to a list of configuration files.
1615 @type file_name: str
1616 @param file_name: the target file name
1618 @param data: the new contents of the file
1620 @param mode: the mode to give the file (can be None)
1622 @param uid: the owner of the file (can be -1 for default)
1624 @param gid: the group of the file (can be -1 for default)
1626 @param atime: the atime to set on the file (can be None)
1628 @param mtime: the mtime to set on the file (can be None)
1632 if not os.path.isabs(file_name):
1633 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1635 if file_name not in _ALLOWED_UPLOAD_FILES:
1636 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1639 raw_data = _Decompress(data)
1641 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1642 atime=atime, mtime=mtime)
1645 def WriteSsconfFiles(values):
1646 """Update all ssconf files.
1648 Wrapper around the SimpleStore.WriteFiles.
1651 ssconf.SimpleStore().WriteFiles(values)
1654 def _ErrnoOrStr(err):
1655 """Format an EnvironmentError exception.
1657 If the L{err} argument has an errno attribute, it will be looked up
1658 and converted into a textual C{E...} description. Otherwise the
1659 string representation of the error will be returned.
1661 @type err: L{EnvironmentError}
1662 @param err: the exception to format
1665 if hasattr(err, 'errno'):
1666 detail = errno.errorcode[err.errno]
1672 def _OSOndiskAPIVersion(os_dir):
1673 """Compute and return the API version of a given OS.
1675 This function will try to read the API version of the OS residing in
1676 the 'os_dir' directory.
1679 @param os_dir: the directory in which we should look for the OS
1681 @return: tuple (status, data) with status denoting the validity and
1682 data holding either the vaid versions or an error message
1685 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1688 st = os.stat(api_file)
1689 except EnvironmentError, err:
1690 return False, ("Required file '%s' not found under path %s: %s" %
1691 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1693 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1694 return False, ("File '%s' in %s is not a regular file" %
1695 (constants.OS_API_FILE, os_dir))
1698 api_versions = utils.ReadFile(api_file).splitlines()
1699 except EnvironmentError, err:
1700 return False, ("Error while reading the API version file at %s: %s" %
1701 (api_file, _ErrnoOrStr(err)))
1704 api_versions = [int(version.strip()) for version in api_versions]
1705 except (TypeError, ValueError), err:
1706 return False, ("API version(s) can't be converted to integer: %s" %
1709 return True, api_versions
1712 def DiagnoseOS(top_dirs=None):
1713 """Compute the validity for all OSes.
1715 @type top_dirs: list
1716 @param top_dirs: the list of directories in which to
1717 search (if not given defaults to
1718 L{constants.OS_SEARCH_PATH})
1719 @rtype: list of L{objects.OS}
1720 @return: a list of tuples (name, path, status, diagnose, variants)
1721 for all (potential) OSes under all search paths, where:
1722 - name is the (potential) OS name
1723 - path is the full path to the OS
1724 - status True/False is the validity of the OS
1725 - diagnose is the error message for an invalid OS, otherwise empty
1726 - variants is a list of supported OS variants, if any
1729 if top_dirs is None:
1730 top_dirs = constants.OS_SEARCH_PATH
1733 for dir_name in top_dirs:
1734 if os.path.isdir(dir_name):
1736 f_names = utils.ListVisibleFiles(dir_name)
1737 except EnvironmentError, err:
1738 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1740 for name in f_names:
1741 os_path = utils.PathJoin(dir_name, name)
1742 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1745 variants = os_inst.supported_variants
1749 result.append((name, os_path, status, diagnose, variants))
1754 def _TryOSFromDisk(name, base_dir=None):
1755 """Create an OS instance from disk.
1757 This function will return an OS instance if the given name is a
1760 @type base_dir: string
1761 @keyword base_dir: Base directory containing OS installations.
1762 Defaults to a search in all the OS_SEARCH_PATH dirs.
1764 @return: success and either the OS instance if we find a valid one,
1768 if base_dir is None:
1769 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1771 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1774 return False, "Directory for OS %s not found in search path" % name
1776 status, api_versions = _OSOndiskAPIVersion(os_dir)
1779 return status, api_versions
1781 if not constants.OS_API_VERSIONS.intersection(api_versions):
1782 return False, ("API version mismatch for path '%s': found %s, want %s." %
1783 (os_dir, api_versions, constants.OS_API_VERSIONS))
1785 # OS Files dictionary, we will populate it with the absolute path names
1786 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1788 if max(api_versions) >= constants.OS_API_V15:
1789 os_files[constants.OS_VARIANTS_FILE] = ''
1791 for filename in os_files:
1792 os_files[filename] = utils.PathJoin(os_dir, filename)
1795 st = os.stat(os_files[filename])
1796 except EnvironmentError, err:
1797 return False, ("File '%s' under path '%s' is missing (%s)" %
1798 (filename, os_dir, _ErrnoOrStr(err)))
1800 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1801 return False, ("File '%s' under path '%s' is not a regular file" %
1804 if filename in constants.OS_SCRIPTS:
1805 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1806 return False, ("File '%s' under path '%s' is not executable" %
1810 if constants.OS_VARIANTS_FILE in os_files:
1811 variants_file = os_files[constants.OS_VARIANTS_FILE]
1813 variants = utils.ReadFile(variants_file).splitlines()
1814 except EnvironmentError, err:
1815 return False, ("Error while reading the OS variants file at %s: %s" %
1816 (variants_file, _ErrnoOrStr(err)))
1818 return False, ("No supported os variant found")
1820 os_obj = objects.OS(name=name, path=os_dir,
1821 create_script=os_files[constants.OS_SCRIPT_CREATE],
1822 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1823 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1824 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1825 supported_variants=variants,
1826 api_versions=api_versions)
1830 def OSFromDisk(name, base_dir=None):
1831 """Create an OS instance from disk.
1833 This function will return an OS instance if the given name is a
1834 valid OS name. Otherwise, it will raise an appropriate
1835 L{RPCFail} exception, detailing why this is not a valid OS.
1837 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1838 an exception but returns true/false status data.
1840 @type base_dir: string
1841 @keyword base_dir: Base directory containing OS installations.
1842 Defaults to a search in all the OS_SEARCH_PATH dirs.
1843 @rtype: L{objects.OS}
1844 @return: the OS instance if we find a valid one
1845 @raise RPCFail: if we don't find a valid OS
1848 name_only = name.split("+", 1)[0]
1849 status, payload = _TryOSFromDisk(name_only, base_dir)
1857 def OSEnvironment(instance, inst_os, debug=0):
1858 """Calculate the environment for an os script.
1860 @type instance: L{objects.Instance}
1861 @param instance: target instance for the os script run
1862 @type inst_os: L{objects.OS}
1863 @param inst_os: operating system for which the environment is being built
1864 @type debug: integer
1865 @param debug: debug level (0 or 1, for OS Api 10)
1867 @return: dict of environment variables
1868 @raise errors.BlockDeviceError: if the block device
1874 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1875 result['OS_API_VERSION'] = '%d' % api_version
1876 result['INSTANCE_NAME'] = instance.name
1877 result['INSTANCE_OS'] = instance.os
1878 result['HYPERVISOR'] = instance.hypervisor
1879 result['DISK_COUNT'] = '%d' % len(instance.disks)
1880 result['NIC_COUNT'] = '%d' % len(instance.nics)
1881 result['DEBUG_LEVEL'] = '%d' % debug
1882 if api_version >= constants.OS_API_V15:
1884 variant = instance.os.split('+', 1)[1]
1886 variant = inst_os.supported_variants[0]
1887 result['OS_VARIANT'] = variant
1888 for idx, disk in enumerate(instance.disks):
1889 real_disk = _RecursiveFindBD(disk)
1890 if real_disk is None:
1891 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1894 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1895 result['DISK_%d_ACCESS' % idx] = disk.mode
1896 if constants.HV_DISK_TYPE in instance.hvparams:
1897 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1898 instance.hvparams[constants.HV_DISK_TYPE]
1899 if disk.dev_type in constants.LDS_BLOCK:
1900 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1901 elif disk.dev_type == constants.LD_FILE:
1902 result['DISK_%d_BACKEND_TYPE' % idx] = \
1903 'file:%s' % disk.physical_id[0]
1904 for idx, nic in enumerate(instance.nics):
1905 result['NIC_%d_MAC' % idx] = nic.mac
1907 result['NIC_%d_IP' % idx] = nic.ip
1908 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1909 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1910 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1911 if nic.nicparams[constants.NIC_LINK]:
1912 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1913 if constants.HV_NIC_TYPE in instance.hvparams:
1914 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1915 instance.hvparams[constants.HV_NIC_TYPE]
1917 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1918 for key, value in source.items():
1919 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1923 def BlockdevGrow(disk, amount):
1924 """Grow a stack of block devices.
1926 This function is called recursively, with the childrens being the
1927 first ones to resize.
1929 @type disk: L{objects.Disk}
1930 @param disk: the disk to be grown
1931 @rtype: (status, result)
1932 @return: a tuple with the status of the operation
1933 (True/False), and the errors message if status
1937 r_dev = _RecursiveFindBD(disk)
1939 _Fail("Cannot find block device %s", disk)
1943 except errors.BlockDeviceError, err:
1944 _Fail("Failed to grow block device: %s", err, exc=True)
1947 def BlockdevSnapshot(disk):
1948 """Create a snapshot copy of a block device.
1950 This function is called recursively, and the snapshot is actually created
1951 just for the leaf lvm backend device.
1953 @type disk: L{objects.Disk}
1954 @param disk: the disk to be snapshotted
1956 @return: snapshot disk path
1959 if disk.dev_type == constants.LD_DRBD8:
1960 if not disk.children:
1961 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
1963 return BlockdevSnapshot(disk.children[0])
1964 elif disk.dev_type == constants.LD_LV:
1965 r_dev = _RecursiveFindBD(disk)
1966 if r_dev is not None:
1967 # FIXME: choose a saner value for the snapshot size
1968 # let's stay on the safe side and ask for the full size, for now
1969 return r_dev.Snapshot(disk.size)
1971 _Fail("Cannot find block device %s", disk)
1973 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1974 disk.unique_id, disk.dev_type)
1977 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx, debug):
1978 """Export a block device snapshot to a remote node.
1980 @type disk: L{objects.Disk}
1981 @param disk: the description of the disk to export
1982 @type dest_node: str
1983 @param dest_node: the destination node to export to
1984 @type instance: L{objects.Instance}
1985 @param instance: the instance object to whom the disk belongs
1986 @type cluster_name: str
1987 @param cluster_name: the cluster name, needed for SSH hostalias
1989 @param idx: the index of the disk in the instance's disk list,
1990 used to export to the OS scripts environment
1991 @type debug: integer
1992 @param debug: debug level, passed to the OS scripts
1996 inst_os = OSFromDisk(instance.os)
1997 export_env = OSEnvironment(instance, inst_os, debug)
1999 export_script = inst_os.export_script
2001 logfile = _InstanceLogName("export", inst_os.name, instance.name)
2002 if not os.path.exists(constants.LOG_OS_DIR):
2003 os.mkdir(constants.LOG_OS_DIR, 0750)
2004 real_disk = _RecursiveFindBD(disk)
2005 if real_disk is None:
2006 _Fail("Block device '%s' is not set up", disk)
2010 export_env['EXPORT_DEVICE'] = real_disk.dev_path
2011 export_env['EXPORT_INDEX'] = str(idx)
2013 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2014 destfile = disk.physical_id[1]
2016 # the target command is built out of three individual commands,
2017 # which are joined by pipes; we check each individual command for
2019 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
2020 inst_os.path, export_script, logfile)
2024 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
2025 destdir, destdir, destfile)
2026 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2027 constants.GANETI_RUNAS,
2030 # all commands have been checked, so we're safe to combine them
2031 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
2033 result = utils.RunCmd(["bash", "-c", command], env=export_env)
2036 _Fail("OS snapshot export command '%s' returned error: %s"
2037 " output: %s", command, result.fail_reason, result.output)
2040 def FinalizeExport(instance, snap_disks):
2041 """Write out the export configuration information.
2043 @type instance: L{objects.Instance}
2044 @param instance: the instance which we export, used for
2045 saving configuration
2046 @type snap_disks: list of L{objects.Disk}
2047 @param snap_disks: list of snapshot block devices, which
2048 will be used to get the actual name of the dump file
2053 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2054 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2056 config = objects.SerializableConfigParser()
2058 config.add_section(constants.INISECT_EXP)
2059 config.set(constants.INISECT_EXP, 'version', '0')
2060 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2061 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2062 config.set(constants.INISECT_EXP, 'os', instance.os)
2063 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2065 config.add_section(constants.INISECT_INS)
2066 config.set(constants.INISECT_INS, 'name', instance.name)
2067 config.set(constants.INISECT_INS, 'memory', '%d' %
2068 instance.beparams[constants.BE_MEMORY])
2069 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2070 instance.beparams[constants.BE_VCPUS])
2071 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2074 for nic_count, nic in enumerate(instance.nics):
2076 config.set(constants.INISECT_INS, 'nic%d_mac' %
2077 nic_count, '%s' % nic.mac)
2078 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2079 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
2081 # TODO: redundant: on load can read nics until it doesn't exist
2082 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2085 for disk_count, disk in enumerate(snap_disks):
2088 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2089 ('%s' % disk.iv_name))
2090 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2091 ('%s' % disk.physical_id[1]))
2092 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2095 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2097 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2098 data=config.Dumps())
2099 shutil.rmtree(finaldestdir, True)
2100 shutil.move(destdir, finaldestdir)
2103 def ExportInfo(dest):
2104 """Get export configuration information.
2107 @param dest: directory containing the export
2109 @rtype: L{objects.SerializableConfigParser}
2110 @return: a serializable config file containing the
2114 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2116 config = objects.SerializableConfigParser()
2119 if (not config.has_section(constants.INISECT_EXP) or
2120 not config.has_section(constants.INISECT_INS)):
2121 _Fail("Export info file doesn't have the required fields")
2123 return config.Dumps()
2126 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name, debug):
2127 """Import an os image into an instance.
2129 @type instance: L{objects.Instance}
2130 @param instance: instance to import the disks into
2131 @type src_node: string
2132 @param src_node: source node for the disk images
2133 @type src_images: list of string
2134 @param src_images: absolute paths of the disk images
2135 @type debug: integer
2136 @param debug: debug level, passed to the OS scripts
2137 @rtype: list of boolean
2138 @return: each boolean represent the success of importing the n-th disk
2141 inst_os = OSFromDisk(instance.os)
2142 import_env = OSEnvironment(instance, inst_os, debug)
2143 import_script = inst_os.import_script
2145 logfile = _InstanceLogName("import", instance.os, instance.name)
2146 if not os.path.exists(constants.LOG_OS_DIR):
2147 os.mkdir(constants.LOG_OS_DIR, 0750)
2150 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2151 import_script, logfile)
2154 for idx, image in enumerate(src_images):
2156 destcmd = utils.BuildShellCmd('cat %s', image)
2157 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2158 constants.GANETI_RUNAS,
2160 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2161 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2162 import_env['IMPORT_INDEX'] = str(idx)
2163 result = utils.RunCmd(command, env=import_env)
2165 logging.error("Disk import command '%s' returned error: %s"
2166 " output: %s", command, result.fail_reason,
2168 final_result.append("error importing disk %d: %s, %s" %
2169 (idx, result.fail_reason, result.output[-100]))
2172 _Fail("; ".join(final_result), log=False)
2176 """Return a list of exports currently available on this machine.
2179 @return: list of the exports
2182 if os.path.isdir(constants.EXPORT_DIR):
2183 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2185 _Fail("No exports directory")
2188 def RemoveExport(export):
2189 """Remove an existing export from the node.
2192 @param export: the name of the export to remove
2196 target = utils.PathJoin(constants.EXPORT_DIR, export)
2199 shutil.rmtree(target)
2200 except EnvironmentError, err:
2201 _Fail("Error while removing the export: %s", err, exc=True)
2204 def BlockdevRename(devlist):
2205 """Rename a list of block devices.
2207 @type devlist: list of tuples
2208 @param devlist: list of tuples of the form (disk,
2209 new_logical_id, new_physical_id); disk is an
2210 L{objects.Disk} object describing the current disk,
2211 and new logical_id/physical_id is the name we
2214 @return: True if all renames succeeded, False otherwise
2219 for disk, unique_id in devlist:
2220 dev = _RecursiveFindBD(disk)
2222 msgs.append("Can't find device %s in rename" % str(disk))
2226 old_rpath = dev.dev_path
2227 dev.Rename(unique_id)
2228 new_rpath = dev.dev_path
2229 if old_rpath != new_rpath:
2230 DevCacheManager.RemoveCache(old_rpath)
2231 # FIXME: we should add the new cache information here, like:
2232 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2233 # but we don't have the owner here - maybe parse from existing
2234 # cache? for now, we only lose lvm data when we rename, which
2235 # is less critical than DRBD or MD
2236 except errors.BlockDeviceError, err:
2237 msgs.append("Can't rename device '%s' to '%s': %s" %
2238 (dev, unique_id, err))
2239 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2242 _Fail("; ".join(msgs))
2245 def _TransformFileStorageDir(file_storage_dir):
2246 """Checks whether given file_storage_dir is valid.
2248 Checks wheter the given file_storage_dir is within the cluster-wide
2249 default file_storage_dir stored in SimpleStore. Only paths under that
2250 directory are allowed.
2252 @type file_storage_dir: str
2253 @param file_storage_dir: the path to check
2255 @return: the normalized path if valid, None otherwise
2258 if not constants.ENABLE_FILE_STORAGE:
2259 _Fail("File storage disabled at configure time")
2261 file_storage_dir = os.path.normpath(file_storage_dir)
2262 base_file_storage_dir = cfg.GetFileStorageDir()
2263 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2264 base_file_storage_dir):
2265 _Fail("File storage directory '%s' is not under base file"
2266 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2267 return file_storage_dir
2270 def CreateFileStorageDir(file_storage_dir):
2271 """Create file storage directory.
2273 @type file_storage_dir: str
2274 @param file_storage_dir: directory to create
2277 @return: tuple with first element a boolean indicating wheter dir
2278 creation was successful or not
2281 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2282 if os.path.exists(file_storage_dir):
2283 if not os.path.isdir(file_storage_dir):
2284 _Fail("Specified storage dir '%s' is not a directory",
2288 os.makedirs(file_storage_dir, 0750)
2289 except OSError, err:
2290 _Fail("Cannot create file storage directory '%s': %s",
2291 file_storage_dir, err, exc=True)
2294 def RemoveFileStorageDir(file_storage_dir):
2295 """Remove file storage directory.
2297 Remove it only if it's empty. If not log an error and return.
2299 @type file_storage_dir: str
2300 @param file_storage_dir: the directory we should cleanup
2301 @rtype: tuple (success,)
2302 @return: tuple of one element, C{success}, denoting
2303 whether the operation was successful
2306 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2307 if os.path.exists(file_storage_dir):
2308 if not os.path.isdir(file_storage_dir):
2309 _Fail("Specified Storage directory '%s' is not a directory",
2311 # deletes dir only if empty, otherwise we want to fail the rpc call
2313 os.rmdir(file_storage_dir)
2314 except OSError, err:
2315 _Fail("Cannot remove file storage directory '%s': %s",
2316 file_storage_dir, err)
2319 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2320 """Rename the file storage directory.
2322 @type old_file_storage_dir: str
2323 @param old_file_storage_dir: the current path
2324 @type new_file_storage_dir: str
2325 @param new_file_storage_dir: the name we should rename to
2326 @rtype: tuple (success,)
2327 @return: tuple of one element, C{success}, denoting
2328 whether the operation was successful
2331 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2332 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2333 if not os.path.exists(new_file_storage_dir):
2334 if os.path.isdir(old_file_storage_dir):
2336 os.rename(old_file_storage_dir, new_file_storage_dir)
2337 except OSError, err:
2338 _Fail("Cannot rename '%s' to '%s': %s",
2339 old_file_storage_dir, new_file_storage_dir, err)
2341 _Fail("Specified storage dir '%s' is not a directory",
2342 old_file_storage_dir)
2344 if os.path.exists(old_file_storage_dir):
2345 _Fail("Cannot rename '%s' to '%s': both locations exist",
2346 old_file_storage_dir, new_file_storage_dir)
2349 def _EnsureJobQueueFile(file_name):
2350 """Checks whether the given filename is in the queue directory.
2352 @type file_name: str
2353 @param file_name: the file name we should check
2355 @raises RPCFail: if the file is not valid
2358 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2359 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2362 _Fail("Passed job queue file '%s' does not belong to"
2363 " the queue directory '%s'", file_name, queue_dir)
2366 def JobQueueUpdate(file_name, content):
2367 """Updates a file in the queue directory.
2369 This is just a wrapper over L{utils.WriteFile}, with proper
2372 @type file_name: str
2373 @param file_name: the job file name
2375 @param content: the new job contents
2377 @return: the success of the operation
2380 _EnsureJobQueueFile(file_name)
2382 # Write and replace the file atomically
2383 utils.WriteFile(file_name, data=_Decompress(content))
2386 def JobQueueRename(old, new):
2387 """Renames a job queue file.
2389 This is just a wrapper over os.rename with proper checking.
2392 @param old: the old (actual) file name
2394 @param new: the desired file name
2396 @return: the success of the operation and payload
2399 _EnsureJobQueueFile(old)
2400 _EnsureJobQueueFile(new)
2402 utils.RenameFile(old, new, mkdir=True)
2405 def JobQueueSetDrainFlag(drain_flag):
2406 """Set the drain flag for the queue.
2408 This will set or unset the queue drain flag.
2410 @type drain_flag: boolean
2411 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2413 @return: always True, None
2414 @warning: the function always returns True
2418 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2420 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2423 def BlockdevClose(instance_name, disks):
2424 """Closes the given block devices.
2426 This means they will be switched to secondary mode (in case of
2429 @param instance_name: if the argument is not empty, the symlinks
2430 of this instance will be removed
2431 @type disks: list of L{objects.Disk}
2432 @param disks: the list of disks to be closed
2433 @rtype: tuple (success, message)
2434 @return: a tuple of success and message, where success
2435 indicates the succes of the operation, and message
2436 which will contain the error details in case we
2442 rd = _RecursiveFindBD(cf)
2444 _Fail("Can't find device %s", cf)
2451 except errors.BlockDeviceError, err:
2452 msg.append(str(err))
2454 _Fail("Can't make devices secondary: %s", ",".join(msg))
2457 _RemoveBlockDevLinks(instance_name, disks)
2460 def ValidateHVParams(hvname, hvparams):
2461 """Validates the given hypervisor parameters.
2463 @type hvname: string
2464 @param hvname: the hypervisor name
2465 @type hvparams: dict
2466 @param hvparams: the hypervisor parameters to be validated
2471 hv_type = hypervisor.GetHypervisor(hvname)
2472 hv_type.ValidateParameters(hvparams)
2473 except errors.HypervisorError, err:
2474 _Fail(str(err), log=False)
2478 """Demotes the current node from master candidate role.
2481 # try to ensure we're not the master by mistake
2482 master, myself = ssconf.GetMasterAndMyself()
2483 if master == myself:
2484 _Fail("ssconf status shows I'm the master node, will not demote")
2486 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2487 if not result.failed:
2488 _Fail("The master daemon is running, will not demote")
2491 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2492 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2493 except EnvironmentError, err:
2494 if err.errno != errno.ENOENT:
2495 _Fail("Error while backing up cluster file: %s", err, exc=True)
2497 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2500 def _FindDisks(nodes_ip, disks):
2501 """Sets the physical ID on disks and returns the block devices.
2504 # set the correct physical ID
2505 my_name = utils.HostInfo().name
2507 cf.SetPhysicalID(my_name, nodes_ip)
2512 rd = _RecursiveFindBD(cf)
2514 _Fail("Can't find device %s", cf)
2519 def DrbdDisconnectNet(nodes_ip, disks):
2520 """Disconnects the network on a list of drbd devices.
2523 bdevs = _FindDisks(nodes_ip, disks)
2529 except errors.BlockDeviceError, err:
2530 _Fail("Can't change network configuration to standalone mode: %s",
2534 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2535 """Attaches the network on a list of drbd devices.
2538 bdevs = _FindDisks(nodes_ip, disks)
2541 for idx, rd in enumerate(bdevs):
2543 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2544 except EnvironmentError, err:
2545 _Fail("Can't create symlink: %s", err)
2546 # reconnect disks, switch to new master configuration and if
2547 # needed primary mode
2550 rd.AttachNet(multimaster)
2551 except errors.BlockDeviceError, err:
2552 _Fail("Can't change network configuration: %s", err)
2554 # wait until the disks are connected; we need to retry the re-attach
2555 # if the device becomes standalone, as this might happen if the one
2556 # node disconnects and reconnects in a different mode before the
2557 # other node reconnects; in this case, one or both of the nodes will
2558 # decide it has wrong configuration and switch to standalone
2561 all_connected = True
2564 stats = rd.GetProcStatus()
2566 all_connected = (all_connected and
2567 (stats.is_connected or stats.is_in_resync))
2569 if stats.is_standalone:
2570 # peer had different config info and this node became
2571 # standalone, even though this should not happen with the
2572 # new staged way of changing disk configs
2574 rd.AttachNet(multimaster)
2575 except errors.BlockDeviceError, err:
2576 _Fail("Can't change network configuration: %s", err)
2578 if not all_connected:
2579 raise utils.RetryAgain()
2582 # Start with a delay of 100 miliseconds and go up to 5 seconds
2583 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2584 except utils.RetryTimeout:
2585 _Fail("Timeout in disk reconnecting")
2588 # change to primary mode
2592 except errors.BlockDeviceError, err:
2593 _Fail("Can't change to primary mode: %s", err)
2596 def DrbdWaitSync(nodes_ip, disks):
2597 """Wait until DRBDs have synchronized.
2601 stats = rd.GetProcStatus()
2602 if not (stats.is_connected or stats.is_in_resync):
2603 raise utils.RetryAgain()
2606 bdevs = _FindDisks(nodes_ip, disks)
2612 # poll each second for 15 seconds
2613 stats = utils.Retry(_helper, 1, 15, args=[rd])
2614 except utils.RetryTimeout:
2615 stats = rd.GetProcStatus()
2617 if not (stats.is_connected or stats.is_in_resync):
2618 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2619 alldone = alldone and (not stats.is_in_resync)
2620 if stats.sync_percent is not None:
2621 min_resync = min(min_resync, stats.sync_percent)
2623 return (alldone, min_resync)
2626 def PowercycleNode(hypervisor_type):
2627 """Hard-powercycle the node.
2629 Because we need to return first, and schedule the powercycle in the
2630 background, we won't be able to report failures nicely.
2633 hyper = hypervisor.GetHypervisor(hypervisor_type)
2637 # if we can't fork, we'll pretend that we're in the child process
2640 return "Reboot scheduled in 5 seconds"
2642 hyper.PowercycleNode()
2645 class HooksRunner(object):
2648 This class is instantiated on the node side (ganeti-noded) and not
2652 def __init__(self, hooks_base_dir=None):
2653 """Constructor for hooks runner.
2655 @type hooks_base_dir: str or None
2656 @param hooks_base_dir: if not None, this overrides the
2657 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2660 if hooks_base_dir is None:
2661 hooks_base_dir = constants.HOOKS_BASE_DIR
2662 # yeah, _BASE_DIR is not valid for attributes, we use it like a
2664 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
2666 def RunHooks(self, hpath, phase, env):
2667 """Run the scripts in the hooks directory.
2670 @param hpath: the path to the hooks directory which
2673 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2674 L{constants.HOOKS_PHASE_POST}
2676 @param env: dictionary with the environment for the hook
2678 @return: list of 3-element tuples:
2680 - script result, either L{constants.HKR_SUCCESS} or
2681 L{constants.HKR_FAIL}
2682 - output of the script
2684 @raise errors.ProgrammerError: for invalid input
2688 if phase == constants.HOOKS_PHASE_PRE:
2690 elif phase == constants.HOOKS_PHASE_POST:
2693 _Fail("Unknown hooks phase '%s'", phase)
2696 subdir = "%s-%s.d" % (hpath, suffix)
2697 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2701 if not os.path.isdir(dir_name):
2702 # for non-existing/non-dirs, we simply exit instead of logging a
2703 # warning at every operation
2706 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
2708 for (relname, relstatus, runresult) in runparts_results:
2709 if relstatus == constants.RUNPARTS_SKIP:
2710 rrval = constants.HKR_SKIP
2712 elif relstatus == constants.RUNPARTS_ERR:
2713 rrval = constants.HKR_FAIL
2714 output = "Hook script execution error: %s" % runresult
2715 elif relstatus == constants.RUNPARTS_RUN:
2716 if runresult.failed:
2717 rrval = constants.HKR_FAIL
2719 rrval = constants.HKR_SUCCESS
2720 output = utils.SafeEncode(runresult.output.strip())
2721 results.append(("%s/%s" % (subdir, relname), rrval, output))
2726 class IAllocatorRunner(object):
2727 """IAllocator runner.
2729 This class is instantiated on the node side (ganeti-noded) and not on
2734 def Run(name, idata):
2735 """Run an iallocator script.
2738 @param name: the iallocator script name
2740 @param idata: the allocator input data
2743 @return: two element tuple of:
2745 - either error message or stdout of allocator (for success)
2748 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2750 if alloc_script is None:
2751 _Fail("iallocator module '%s' not found in the search path", name)
2753 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2757 result = utils.RunCmd([alloc_script, fin_name])
2759 _Fail("iallocator module '%s' failed: %s, output '%s'",
2760 name, result.fail_reason, result.output)
2764 return result.stdout
2767 class DevCacheManager(object):
2768 """Simple class for managing a cache of block device information.
2771 _DEV_PREFIX = "/dev/"
2772 _ROOT_DIR = constants.BDEV_CACHE_DIR
2775 def _ConvertPath(cls, dev_path):
2776 """Converts a /dev/name path to the cache file name.
2778 This replaces slashes with underscores and strips the /dev
2779 prefix. It then returns the full path to the cache file.
2782 @param dev_path: the C{/dev/} path name
2784 @return: the converted path name
2787 if dev_path.startswith(cls._DEV_PREFIX):
2788 dev_path = dev_path[len(cls._DEV_PREFIX):]
2789 dev_path = dev_path.replace("/", "_")
2790 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2794 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2795 """Updates the cache information for a given device.
2798 @param dev_path: the pathname of the device
2800 @param owner: the owner (instance name) of the device
2801 @type on_primary: bool
2802 @param on_primary: whether this is the primary
2805 @param iv_name: the instance-visible name of the
2806 device, as in objects.Disk.iv_name
2811 if dev_path is None:
2812 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2814 fpath = cls._ConvertPath(dev_path)
2820 iv_name = "not_visible"
2821 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2823 utils.WriteFile(fpath, data=fdata)
2824 except EnvironmentError, err:
2825 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2828 def RemoveCache(cls, dev_path):
2829 """Remove data for a dev_path.
2831 This is just a wrapper over L{utils.RemoveFile} with a converted
2832 path name and logging.
2835 @param dev_path: the pathname of the device
2840 if dev_path is None:
2841 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2843 fpath = cls._ConvertPath(dev_path)
2845 utils.RemoveFile(fpath)
2846 except EnvironmentError, err:
2847 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)