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
29 # pylint: disable-msg=E1103
31 # E1103: %s %r has no %r member (but some types could not be
32 # inferred), because the _TryOSFromDisk returns either (True, os_obj)
33 # or (False, "string") which confuses pylint
50 from ganeti import errors
51 from ganeti import utils
52 from ganeti import ssh
53 from ganeti import hypervisor
54 from ganeti import constants
55 from ganeti import bdev
56 from ganeti import objects
57 from ganeti import ssconf
60 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
63 class RPCFail(Exception):
64 """Class denoting RPC failure.
66 Its argument is the error message.
71 def _Fail(msg, *args, **kwargs):
72 """Log an error and the raise an RPCFail exception.
74 This exception is then handled specially in the ganeti daemon and
75 turned into a 'failed' return type. As such, this function is a
76 useful shortcut for logging the error and returning it to the master
80 @param msg: the text of the exception
86 if "log" not in kwargs or kwargs["log"]: # if we should log this error
87 if "exc" in kwargs and kwargs["exc"]:
88 logging.exception(msg)
95 """Simple wrapper to return a SimpleStore.
97 @rtype: L{ssconf.SimpleStore}
98 @return: a SimpleStore instance
101 return ssconf.SimpleStore()
104 def _GetSshRunner(cluster_name):
105 """Simple wrapper to return an SshRunner.
107 @type cluster_name: str
108 @param cluster_name: the cluster name, which is needed
109 by the SshRunner constructor
110 @rtype: L{ssh.SshRunner}
111 @return: an SshRunner instance
114 return ssh.SshRunner(cluster_name)
117 def _Decompress(data):
118 """Unpacks data compressed by the RPC client.
120 @type data: list or tuple
121 @param data: Data sent by RPC client
123 @return: Decompressed data
126 assert isinstance(data, (list, tuple))
127 assert len(data) == 2
128 (encoding, content) = data
129 if encoding == constants.RPC_ENCODING_NONE:
131 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
132 return zlib.decompress(base64.b64decode(content))
134 raise AssertionError("Unknown data encoding")
137 def _CleanDirectory(path, exclude=None):
138 """Removes all regular files in a directory.
141 @param path: the directory to clean
143 @param exclude: list of files to be excluded, defaults
147 if not os.path.isdir(path):
152 # Normalize excluded paths
153 exclude = [os.path.normpath(i) for i in exclude]
155 for rel_name in utils.ListVisibleFiles(path):
156 full_name = os.path.normpath(os.path.join(path, rel_name))
157 if full_name in exclude:
159 if os.path.isfile(full_name) and not os.path.islink(full_name):
160 utils.RemoveFile(full_name)
163 def _BuildUploadFileList():
164 """Build the list of allowed upload files.
166 This is abstracted so that it's built only once at module import time.
169 allowed_files = set([
170 constants.CLUSTER_CONF_FILE,
172 constants.SSH_KNOWN_HOSTS_FILE,
173 constants.VNC_PASSWORD_FILE,
174 constants.RAPI_CERT_FILE,
175 constants.RAPI_USERS_FILE,
176 constants.HMAC_CLUSTER_KEY,
179 for hv_name in constants.HYPER_TYPES:
180 hv_class = hypervisor.GetHypervisorClass(hv_name)
181 allowed_files.update(hv_class.GetAncillaryFiles())
183 return frozenset(allowed_files)
186 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
190 """Removes job queue files and archived jobs.
196 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
197 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
201 """Returns master information.
203 This is an utility function to compute master information, either
204 for consumption here or from the node daemon.
207 @return: master_netdev, master_ip, master_name
208 @raise RPCFail: in case of errors
213 master_netdev = cfg.GetMasterNetdev()
214 master_ip = cfg.GetMasterIP()
215 master_node = cfg.GetMasterNode()
216 except errors.ConfigurationError, err:
217 _Fail("Cluster configuration incomplete: %s", err, exc=True)
218 return (master_netdev, master_ip, master_node)
221 def StartMaster(start_daemons, no_voting):
222 """Activate local node as master node.
224 The function will always try activate the IP address of the master
225 (unless someone else has it). It will also start the master daemons,
226 based on the start_daemons parameter.
228 @type start_daemons: boolean
229 @param start_daemons: whether to also start the master
230 daemons (ganeti-masterd and ganeti-rapi)
231 @type no_voting: boolean
232 @param no_voting: whether to start ganeti-masterd without a node vote
233 (if start_daemons is True), but still non-interactively
237 # GetMasterInfo will raise an exception if not able to return data
238 master_netdev, master_ip, _ = GetMasterInfo()
241 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
242 if utils.OwnIpAddress(master_ip):
243 # we already have the ip:
244 logging.debug("Master IP already configured, doing nothing")
246 msg = "Someone else has the master ip, not activating"
250 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
251 "dev", master_netdev, "label",
252 "%s:0" % master_netdev])
254 msg = "Can't activate master IP: %s" % result.output
258 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
259 "-s", master_ip, master_ip])
260 # we'll ignore the exit code of arping
262 # and now start the master and rapi daemons
265 masterd_args = "--no-voting --yes-do-it"
270 "EXTRA_MASTERD_ARGS": masterd_args,
273 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
275 msg = "Can't start Ganeti master: %s" % result.output
280 _Fail("; ".join(err_msgs))
283 def StopMaster(stop_daemons):
284 """Deactivate this node as master.
286 The function will always try to deactivate the IP address of the
287 master. It will also stop the master daemons depending on the
288 stop_daemons parameter.
290 @type stop_daemons: boolean
291 @param stop_daemons: whether to also stop the master daemons
292 (ganeti-masterd and ganeti-rapi)
296 # TODO: log and report back to the caller the error failures; we
297 # need to decide in which case we fail the RPC for this
299 # GetMasterInfo will raise an exception if not able to return data
300 master_netdev, master_ip, _ = GetMasterInfo()
302 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
303 "dev", master_netdev])
305 logging.error("Can't remove the master IP, error: %s", result.output)
306 # but otherwise ignore the failure
309 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
311 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
313 result.cmd, result.exit_code, result.output)
316 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
317 """Joins this node to the cluster.
319 This does the following:
320 - updates the hostkeys of the machine (rsa and dsa)
321 - adds the ssh private key to the user
322 - adds the ssh public key to the users' authorized_keys file
325 @param dsa: the DSA private key to write
327 @param dsapub: the DSA public key to write
329 @param rsa: the RSA private key to write
331 @param rsapub: the RSA public key to write
333 @param sshkey: the SSH private key to write
335 @param sshpub: the SSH public key to write
337 @return: the success of the operation
340 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
341 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
342 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
343 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
344 for name, content, mode in sshd_keys:
345 utils.WriteFile(name, data=content, mode=mode)
348 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
350 except errors.OpExecError, err:
351 _Fail("Error while processing user ssh files: %s", err, exc=True)
353 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
354 utils.WriteFile(name, data=content, mode=0600)
356 utils.AddAuthorizedKey(auth_keys, sshpub)
358 result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"])
360 _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)",
361 result.cmd, result.exit_code, result.output)
364 def LeaveCluster(modify_ssh_setup):
365 """Cleans up and remove the current node.
367 This function cleans up and prepares the current node to be removed
370 If processing is successful, then it raises an
371 L{errors.QuitGanetiException} which is used as a special case to
372 shutdown the node daemon.
374 @param modify_ssh_setup: boolean
377 _CleanDirectory(constants.DATA_DIR)
382 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
384 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
386 utils.RemoveFile(priv_key)
387 utils.RemoveFile(pub_key)
388 except errors.OpExecError:
389 logging.exception("Error while processing ssh files")
392 utils.RemoveFile(constants.HMAC_CLUSTER_KEY)
393 utils.RemoveFile(constants.RAPI_CERT_FILE)
394 utils.RemoveFile(constants.SSL_CERT_FILE)
395 except: # pylint: disable-msg=W0702
396 logging.exception("Error while removing cluster secrets")
398 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
400 logging.error("Command %s failed with exitcode %s and error %s",
401 result.cmd, result.exit_code, result.output)
403 # Raise a custom exception (handled in ganeti-noded)
404 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
407 def GetNodeInfo(vgname, hypervisor_type):
408 """Gives back a hash with different information about the node.
410 @type vgname: C{string}
411 @param vgname: the name of the volume group to ask for disk space information
412 @type hypervisor_type: C{str}
413 @param hypervisor_type: the name of the hypervisor to ask for
416 @return: dictionary with the following keys:
417 - vg_size is the size of the configured volume group in MiB
418 - vg_free is the free size of the volume group in MiB
419 - memory_dom0 is the memory allocated for domain0 in MiB
420 - memory_free is the currently available (free) ram in MiB
421 - memory_total is the total number of ram in MiB
425 vginfo = _GetVGInfo(vgname)
426 outputarray['vg_size'] = vginfo['vg_size']
427 outputarray['vg_free'] = vginfo['vg_free']
429 hyper = hypervisor.GetHypervisor(hypervisor_type)
430 hyp_info = hyper.GetNodeInfo()
431 if hyp_info is not None:
432 outputarray.update(hyp_info)
434 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
439 def VerifyNode(what, cluster_name):
440 """Verify the status of the local node.
442 Based on the input L{what} parameter, various checks are done on the
445 If the I{filelist} key is present, this list of
446 files is checksummed and the file/checksum pairs are returned.
448 If the I{nodelist} key is present, we check that we have
449 connectivity via ssh with the target nodes (and check the hostname
452 If the I{node-net-test} key is present, we check that we have
453 connectivity to the given nodes via both primary IP and, if
454 applicable, secondary IPs.
457 @param what: a dictionary of things to check:
458 - filelist: list of files for which to compute checksums
459 - nodelist: list of nodes we should check ssh communication with
460 - node-net-test: list of nodes we should check node daemon port
462 - hypervisor: list with hypervisors to run the verify for
464 @return: a dictionary with the same keys as the input dict, and
465 values representing the result of the checks
470 if constants.NV_HYPERVISOR in what:
471 result[constants.NV_HYPERVISOR] = tmp = {}
472 for hv_name in what[constants.NV_HYPERVISOR]:
473 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
475 if constants.NV_FILELIST in what:
476 result[constants.NV_FILELIST] = utils.FingerprintFiles(
477 what[constants.NV_FILELIST])
479 if constants.NV_NODELIST in what:
480 result[constants.NV_NODELIST] = tmp = {}
481 random.shuffle(what[constants.NV_NODELIST])
482 for node in what[constants.NV_NODELIST]:
483 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
487 if constants.NV_NODENETTEST in what:
488 result[constants.NV_NODENETTEST] = tmp = {}
489 my_name = utils.HostInfo().name
490 my_pip = my_sip = None
491 for name, pip, sip in what[constants.NV_NODENETTEST]:
497 tmp[my_name] = ("Can't find my own primary/secondary IP"
500 port = utils.GetDaemonPort(constants.NODED)
501 for name, pip, sip in what[constants.NV_NODENETTEST]:
503 if not utils.TcpPing(pip, port, source=my_pip):
504 fail.append("primary")
506 if not utils.TcpPing(sip, port, source=my_sip):
507 fail.append("secondary")
509 tmp[name] = ("failure using the %s interface(s)" %
512 if constants.NV_LVLIST in what:
513 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
515 if constants.NV_INSTANCELIST in what:
516 result[constants.NV_INSTANCELIST] = GetInstanceList(
517 what[constants.NV_INSTANCELIST])
519 if constants.NV_VGLIST in what:
520 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
522 if constants.NV_PVLIST in what:
523 result[constants.NV_PVLIST] = \
524 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
525 filter_allocatable=False)
527 if constants.NV_VERSION in what:
528 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
529 constants.RELEASE_VERSION)
531 if constants.NV_HVINFO in what:
532 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
533 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
535 if constants.NV_DRBDLIST in what:
537 used_minors = bdev.DRBD8.GetUsedDevs().keys()
538 except errors.BlockDeviceError, err:
539 logging.warning("Can't get used minors list", exc_info=True)
540 used_minors = str(err)
541 result[constants.NV_DRBDLIST] = used_minors
543 if constants.NV_NODESETUP in what:
544 result[constants.NV_NODESETUP] = tmpr = []
545 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
546 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
547 " under /sys, missing required directories /sys/block"
548 " and /sys/class/net")
549 if (not os.path.isdir("/proc/sys") or
550 not os.path.isfile("/proc/sysrq-trigger")):
551 tmpr.append("The procfs filesystem doesn't seem to be mounted"
552 " under /proc, missing required directory /proc/sys and"
553 " the file /proc/sysrq-trigger")
555 if constants.NV_TIME in what:
556 result[constants.NV_TIME] = utils.SplitTime(time.time())
561 def GetVolumeList(vg_name):
562 """Compute list of logical volumes and their size.
565 @param vg_name: the volume group whose LVs we should list
568 dictionary of all partions (key) with value being a tuple of
569 their size (in MiB), inactive and online status::
571 {'test1': ('20.06', True, True)}
573 in case of errors, a string is returned with the error
579 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
580 "--separator=%s" % sep,
581 "-olv_name,lv_size,lv_attr", vg_name])
583 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
585 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
586 for line in result.stdout.splitlines():
588 match = valid_line_re.match(line)
590 logging.error("Invalid line returned from lvs output: '%s'", line)
592 name, size, attr = match.groups()
593 inactive = attr[4] == '-'
594 online = attr[5] == 'o'
595 virtual = attr[0] == 'v'
597 # we don't want to report such volumes as existing, since they
598 # don't really hold data
600 lvs[name] = (size, inactive, online)
605 def ListVolumeGroups():
606 """List the volume groups and their size.
609 @return: dictionary with keys volume name and values the
613 return utils.ListVolumeGroups()
617 """List all volumes on this node.
621 A list of dictionaries, each having four keys:
622 - name: the logical volume name,
623 - size: the size of the logical volume
624 - dev: the physical device on which the LV lives
625 - vg: the volume group to which it belongs
627 In case of errors, we return an empty list and log the
630 Note that since a logical volume can live on multiple physical
631 volumes, the resulting list might include a logical volume
635 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
637 "--options=lv_name,lv_size,devices,vg_name"])
639 _Fail("Failed to list logical volumes, lvs output: %s",
644 return dev.split('(')[0]
650 'name': line[0].strip(),
651 'size': line[1].strip(),
652 'dev': parse_dev(line[2].strip()),
653 'vg': line[3].strip(),
656 return [map_line(line.split('|')) for line in result.stdout.splitlines()
657 if line.count('|') >= 3]
660 def BridgesExist(bridges_list):
661 """Check if a list of bridges exist on the current node.
664 @return: C{True} if all of them exist, C{False} otherwise
668 for bridge in bridges_list:
669 if not utils.BridgeExists(bridge):
670 missing.append(bridge)
673 _Fail("Missing bridges %s", utils.CommaJoin(missing))
676 def GetInstanceList(hypervisor_list):
677 """Provides a list of instances.
679 @type hypervisor_list: list
680 @param hypervisor_list: the list of hypervisors to query information
683 @return: a list of all running instances on the current node
684 - instance1.example.com
685 - instance2.example.com
689 for hname in hypervisor_list:
691 names = hypervisor.GetHypervisor(hname).ListInstances()
692 results.extend(names)
693 except errors.HypervisorError, err:
694 _Fail("Error enumerating instances (hypervisor %s): %s",
695 hname, err, exc=True)
700 def GetInstanceInfo(instance, hname):
701 """Gives back the information about an instance as a dictionary.
703 @type instance: string
704 @param instance: the instance name
706 @param hname: the hypervisor type of the instance
709 @return: dictionary with the following keys:
710 - memory: memory size of instance (int)
711 - state: xen state of instance (string)
712 - time: cpu time of instance (float)
717 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
718 if iinfo is not None:
719 output['memory'] = iinfo[2]
720 output['state'] = iinfo[4]
721 output['time'] = iinfo[5]
726 def GetInstanceMigratable(instance):
727 """Gives whether an instance can be migrated.
729 @type instance: L{objects.Instance}
730 @param instance: object representing the instance to be checked.
733 @return: tuple of (result, description) where:
734 - result: whether the instance can be migrated or not
735 - description: a description of the issue, if relevant
738 hyper = hypervisor.GetHypervisor(instance.hypervisor)
739 iname = instance.name
740 if iname not in hyper.ListInstances():
741 _Fail("Instance %s is not running", iname)
743 for idx in range(len(instance.disks)):
744 link_name = _GetBlockDevSymlinkPath(iname, idx)
745 if not os.path.islink(link_name):
746 _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
749 def GetAllInstancesInfo(hypervisor_list):
750 """Gather data about all instances.
752 This is the equivalent of L{GetInstanceInfo}, except that it
753 computes data for all instances at once, thus being faster if one
754 needs data about more than one instance.
756 @type hypervisor_list: list
757 @param hypervisor_list: list of hypervisors to query for instance data
760 @return: dictionary of instance: data, with data having the following keys:
761 - memory: memory size of instance (int)
762 - state: xen state of instance (string)
763 - time: cpu time of instance (float)
764 - vcpus: the number of vcpus
769 for hname in hypervisor_list:
770 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
772 for name, _, memory, vcpus, state, times in iinfo:
780 # we only check static parameters, like memory and vcpus,
781 # and not state and time which can change between the
782 # invocations of the different hypervisors
783 for key in 'memory', 'vcpus':
784 if value[key] != output[name][key]:
785 _Fail("Instance %s is running twice"
786 " with different parameters", name)
792 def InstanceOsAdd(instance, reinstall):
793 """Add an OS to an instance.
795 @type instance: L{objects.Instance}
796 @param instance: Instance whose OS is to be installed
797 @type reinstall: boolean
798 @param reinstall: whether this is an instance reinstall
802 inst_os = OSFromDisk(instance.os)
804 create_env = OSEnvironment(instance, inst_os)
806 create_env['INSTANCE_REINSTALL'] = "1"
808 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
809 instance.name, int(time.time()))
811 result = utils.RunCmd([inst_os.create_script], env=create_env,
812 cwd=inst_os.path, output=logfile,)
814 logging.error("os create command '%s' returned error: %s, logfile: %s,"
815 " output: %s", result.cmd, result.fail_reason, logfile,
817 lines = [utils.SafeEncode(val)
818 for val in utils.TailFile(logfile, lines=20)]
819 _Fail("OS create script failed (%s), last lines in the"
820 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
823 def RunRenameInstance(instance, old_name):
824 """Run the OS rename script for an instance.
826 @type instance: L{objects.Instance}
827 @param instance: Instance whose OS is to be installed
828 @type old_name: string
829 @param old_name: previous instance name
831 @return: the success of the operation
834 inst_os = OSFromDisk(instance.os)
836 rename_env = OSEnvironment(instance, inst_os)
837 rename_env['OLD_INSTANCE_NAME'] = old_name
839 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
841 instance.name, int(time.time()))
843 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
844 cwd=inst_os.path, output=logfile)
847 logging.error("os create command '%s' returned error: %s output: %s",
848 result.cmd, result.fail_reason, result.output)
849 lines = [utils.SafeEncode(val)
850 for val in utils.TailFile(logfile, lines=20)]
851 _Fail("OS rename script failed (%s), last lines in the"
852 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
855 def _GetVGInfo(vg_name):
856 """Get information about the volume group.
859 @param vg_name: the volume group which we query
862 A dictionary with the following keys:
863 - C{vg_size} is the total size of the volume group in MiB
864 - C{vg_free} is the free size of the volume group in MiB
865 - C{pv_count} are the number of physical disks in that VG
867 If an error occurs during gathering of data, we return the same dict
868 with keys all set to None.
871 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
873 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
874 "--nosuffix", "--units=m", "--separator=:", vg_name])
877 logging.error("volume group %s not present", vg_name)
879 valarr = retval.stdout.strip().rstrip(':').split(':')
883 "vg_size": int(round(float(valarr[0]), 0)),
884 "vg_free": int(round(float(valarr[1]), 0)),
885 "pv_count": int(valarr[2]),
887 except ValueError, err:
888 logging.exception("Fail to parse vgs output: %s", err)
890 logging.error("vgs output has the wrong number of fields (expected"
891 " three): %s", str(valarr))
895 def _GetBlockDevSymlinkPath(instance_name, idx):
896 return os.path.join(constants.DISK_LINKS_DIR,
897 "%s:%d" % (instance_name, idx))
900 def _SymlinkBlockDev(instance_name, device_path, idx):
901 """Set up symlinks to a instance's block device.
903 This is an auxiliary function run when an instance is start (on the primary
904 node) or when an instance is migrated (on the target node).
907 @param instance_name: the name of the target instance
908 @param device_path: path of the physical block device, on the node
909 @param idx: the disk index
910 @return: absolute path to the disk's symlink
913 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
915 os.symlink(device_path, link_name)
917 if err.errno == errno.EEXIST:
918 if (not os.path.islink(link_name) or
919 os.readlink(link_name) != device_path):
921 os.symlink(device_path, link_name)
928 def _RemoveBlockDevLinks(instance_name, disks):
929 """Remove the block device symlinks belonging to the given instance.
932 for idx, _ in enumerate(disks):
933 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
934 if os.path.islink(link_name):
938 logging.exception("Can't remove symlink '%s'", link_name)
941 def _GatherAndLinkBlockDevs(instance):
942 """Set up an instance's block device(s).
944 This is run on the primary node at instance startup. The block
945 devices must be already assembled.
947 @type instance: L{objects.Instance}
948 @param instance: the instance whose disks we shoul assemble
950 @return: list of (disk_object, device_path)
954 for idx, disk in enumerate(instance.disks):
955 device = _RecursiveFindBD(disk)
957 raise errors.BlockDeviceError("Block device '%s' is not set up." %
961 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
963 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
966 block_devices.append((disk, link_name))
971 def StartInstance(instance):
972 """Start an instance.
974 @type instance: L{objects.Instance}
975 @param instance: the instance object
979 running_instances = GetInstanceList([instance.hypervisor])
981 if instance.name in running_instances:
982 logging.info("Instance %s already running, not starting", instance.name)
986 block_devices = _GatherAndLinkBlockDevs(instance)
987 hyper = hypervisor.GetHypervisor(instance.hypervisor)
988 hyper.StartInstance(instance, block_devices)
989 except errors.BlockDeviceError, err:
990 _Fail("Block device error: %s", err, exc=True)
991 except errors.HypervisorError, err:
992 _RemoveBlockDevLinks(instance.name, instance.disks)
993 _Fail("Hypervisor error: %s", err, exc=True)
996 def InstanceShutdown(instance, timeout):
997 """Shut an instance down.
999 @note: this functions uses polling with a hardcoded timeout.
1001 @type instance: L{objects.Instance}
1002 @param instance: the instance object
1003 @type timeout: integer
1004 @param timeout: maximum timeout for soft shutdown
1008 hv_name = instance.hypervisor
1009 hyper = hypervisor.GetHypervisor(hv_name)
1010 iname = instance.name
1012 if instance.name not in hyper.ListInstances():
1013 logging.info("Instance %s not running, doing nothing", iname)
1018 self.tried_once = False
1021 if iname not in hyper.ListInstances():
1025 hyper.StopInstance(instance, retry=self.tried_once)
1026 except errors.HypervisorError, err:
1027 if iname not in hyper.ListInstances():
1028 # if the instance is no longer existing, consider this a
1029 # success and go to cleanup
1032 _Fail("Failed to stop instance %s: %s", iname, err)
1034 self.tried_once = True
1036 raise utils.RetryAgain()
1039 utils.Retry(_TryShutdown(), 5, timeout)
1040 except utils.RetryTimeout:
1041 # the shutdown did not succeed
1042 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1045 hyper.StopInstance(instance, force=True)
1046 except errors.HypervisorError, err:
1047 if iname in hyper.ListInstances():
1048 # only raise an error if the instance still exists, otherwise
1049 # the error could simply be "instance ... unknown"!
1050 _Fail("Failed to force stop instance %s: %s", iname, err)
1054 if iname in hyper.ListInstances():
1055 _Fail("Could not shutdown instance %s even by destroy", iname)
1057 _RemoveBlockDevLinks(iname, instance.disks)
1060 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1061 """Reboot an instance.
1063 @type instance: L{objects.Instance}
1064 @param instance: the instance object to reboot
1065 @type reboot_type: str
1066 @param reboot_type: the type of reboot, one the following
1068 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1069 instance OS, do not recreate the VM
1070 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1071 restart the VM (at the hypervisor level)
1072 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1073 not accepted here, since that mode is handled differently, in
1074 cmdlib, and translates into full stop and start of the
1075 instance (instead of a call_instance_reboot RPC)
1076 @type shutdown_timeout: integer
1077 @param shutdown_timeout: maximum timeout for soft shutdown
1081 running_instances = GetInstanceList([instance.hypervisor])
1083 if instance.name not in running_instances:
1084 _Fail("Cannot reboot instance %s that is not running", instance.name)
1086 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1087 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1089 hyper.RebootInstance(instance)
1090 except errors.HypervisorError, err:
1091 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1092 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1094 InstanceShutdown(instance, shutdown_timeout)
1095 return StartInstance(instance)
1096 except errors.HypervisorError, err:
1097 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1099 _Fail("Invalid reboot_type received: %s", reboot_type)
1102 def MigrationInfo(instance):
1103 """Gather information about an instance to be migrated.
1105 @type instance: L{objects.Instance}
1106 @param instance: the instance definition
1109 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1111 info = hyper.MigrationInfo(instance)
1112 except errors.HypervisorError, err:
1113 _Fail("Failed to fetch migration information: %s", err, exc=True)
1117 def AcceptInstance(instance, info, target):
1118 """Prepare the node to accept an instance.
1120 @type instance: L{objects.Instance}
1121 @param instance: the instance definition
1122 @type info: string/data (opaque)
1123 @param info: migration information, from the source node
1124 @type target: string
1125 @param target: target host (usually ip), on this node
1128 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1130 hyper.AcceptInstance(instance, info, target)
1131 except errors.HypervisorError, err:
1132 _Fail("Failed to accept instance: %s", err, exc=True)
1135 def FinalizeMigration(instance, info, success):
1136 """Finalize any preparation to accept an instance.
1138 @type instance: L{objects.Instance}
1139 @param instance: the instance definition
1140 @type info: string/data (opaque)
1141 @param info: migration information, from the source node
1142 @type success: boolean
1143 @param success: whether the migration was a success or a failure
1146 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1148 hyper.FinalizeMigration(instance, info, success)
1149 except errors.HypervisorError, err:
1150 _Fail("Failed to finalize migration: %s", err, exc=True)
1153 def MigrateInstance(instance, target, live):
1154 """Migrates an instance to another node.
1156 @type instance: L{objects.Instance}
1157 @param instance: the instance definition
1158 @type target: string
1159 @param target: the target node name
1161 @param live: whether the migration should be done live or not (the
1162 interpretation of this parameter is left to the hypervisor)
1164 @return: a tuple of (success, msg) where:
1165 - succes is a boolean denoting the success/failure of the operation
1166 - msg is a string with details in case of failure
1169 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1172 hyper.MigrateInstance(instance, target, live)
1173 except errors.HypervisorError, err:
1174 _Fail("Failed to migrate instance: %s", err, exc=True)
1177 def BlockdevCreate(disk, size, owner, on_primary, info):
1178 """Creates a block device for an instance.
1180 @type disk: L{objects.Disk}
1181 @param disk: the object describing the disk we should create
1183 @param size: the size of the physical underlying device, in MiB
1185 @param owner: the name of the instance for which disk is created,
1186 used for device cache data
1187 @type on_primary: boolean
1188 @param on_primary: indicates if it is the primary node or not
1190 @param info: string that will be sent to the physical device
1191 creation, used for example to set (LVM) tags on LVs
1193 @return: the new unique_id of the device (this can sometime be
1194 computed only after creation), or None. On secondary nodes,
1195 it's not required to return anything.
1198 # TODO: remove the obsolete 'size' argument
1199 # pylint: disable-msg=W0613
1202 for child in disk.children:
1204 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1205 except errors.BlockDeviceError, err:
1206 _Fail("Can't assemble device %s: %s", child, err)
1207 if on_primary or disk.AssembleOnSecondary():
1208 # we need the children open in case the device itself has to
1211 # pylint: disable-msg=E1103
1213 except errors.BlockDeviceError, err:
1214 _Fail("Can't make child '%s' read-write: %s", child, err)
1218 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1219 except errors.BlockDeviceError, err:
1220 _Fail("Can't create block device: %s", err)
1222 if on_primary or disk.AssembleOnSecondary():
1225 except errors.BlockDeviceError, err:
1226 _Fail("Can't assemble device after creation, unusual event: %s", err)
1227 device.SetSyncSpeed(constants.SYNC_SPEED)
1228 if on_primary or disk.OpenOnSecondary():
1230 device.Open(force=True)
1231 except errors.BlockDeviceError, err:
1232 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1233 DevCacheManager.UpdateCache(device.dev_path, owner,
1234 on_primary, disk.iv_name)
1236 device.SetInfo(info)
1238 return device.unique_id
1241 def BlockdevRemove(disk):
1242 """Remove a block device.
1244 @note: This is intended to be called recursively.
1246 @type disk: L{objects.Disk}
1247 @param disk: the disk object we should remove
1249 @return: the success of the operation
1254 rdev = _RecursiveFindBD(disk)
1255 except errors.BlockDeviceError, err:
1256 # probably can't attach
1257 logging.info("Can't attach to device %s in remove", disk)
1259 if rdev is not None:
1260 r_path = rdev.dev_path
1263 except errors.BlockDeviceError, err:
1264 msgs.append(str(err))
1266 DevCacheManager.RemoveCache(r_path)
1269 for child in disk.children:
1271 BlockdevRemove(child)
1272 except RPCFail, err:
1273 msgs.append(str(err))
1276 _Fail("; ".join(msgs))
1279 def _RecursiveAssembleBD(disk, owner, as_primary):
1280 """Activate a block device for an instance.
1282 This is run on the primary and secondary nodes for an instance.
1284 @note: this function is called recursively.
1286 @type disk: L{objects.Disk}
1287 @param disk: the disk we try to assemble
1289 @param owner: the name of the instance which owns the disk
1290 @type as_primary: boolean
1291 @param as_primary: if we should make the block device
1294 @return: the assembled device or None (in case no device
1296 @raise errors.BlockDeviceError: in case there is an error
1297 during the activation of the children or the device
1303 mcn = disk.ChildrenNeeded()
1305 mcn = 0 # max number of Nones allowed
1307 mcn = len(disk.children) - mcn # max number of Nones
1308 for chld_disk in disk.children:
1310 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1311 except errors.BlockDeviceError, err:
1312 if children.count(None) >= mcn:
1315 logging.error("Error in child activation (but continuing): %s",
1317 children.append(cdev)
1319 if as_primary or disk.AssembleOnSecondary():
1320 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1321 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1323 if as_primary or disk.OpenOnSecondary():
1325 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1326 as_primary, disk.iv_name)
1333 def BlockdevAssemble(disk, owner, as_primary):
1334 """Activate a block device for an instance.
1336 This is a wrapper over _RecursiveAssembleBD.
1338 @rtype: str or boolean
1339 @return: a C{/dev/...} path for primary nodes, and
1340 C{True} for secondary nodes
1344 result = _RecursiveAssembleBD(disk, owner, as_primary)
1345 if isinstance(result, bdev.BlockDev):
1346 # pylint: disable-msg=E1103
1347 result = result.dev_path
1348 except errors.BlockDeviceError, err:
1349 _Fail("Error while assembling disk: %s", err, exc=True)
1354 def BlockdevShutdown(disk):
1355 """Shut down a block device.
1357 First, if the device is assembled (Attach() is successful), then
1358 the device is shutdown. Then the children of the device are
1361 This function is called recursively. Note that we don't cache the
1362 children or such, as oppossed to assemble, shutdown of different
1363 devices doesn't require that the upper device was active.
1365 @type disk: L{objects.Disk}
1366 @param disk: the description of the disk we should
1372 r_dev = _RecursiveFindBD(disk)
1373 if r_dev is not None:
1374 r_path = r_dev.dev_path
1377 DevCacheManager.RemoveCache(r_path)
1378 except errors.BlockDeviceError, err:
1379 msgs.append(str(err))
1382 for child in disk.children:
1384 BlockdevShutdown(child)
1385 except RPCFail, err:
1386 msgs.append(str(err))
1389 _Fail("; ".join(msgs))
1392 def BlockdevAddchildren(parent_cdev, new_cdevs):
1393 """Extend a mirrored block device.
1395 @type parent_cdev: L{objects.Disk}
1396 @param parent_cdev: the disk to which we should add children
1397 @type new_cdevs: list of L{objects.Disk}
1398 @param new_cdevs: the list of children which we should add
1402 parent_bdev = _RecursiveFindBD(parent_cdev)
1403 if parent_bdev is None:
1404 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1405 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1406 if new_bdevs.count(None) > 0:
1407 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1408 parent_bdev.AddChildren(new_bdevs)
1411 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1412 """Shrink a mirrored block device.
1414 @type parent_cdev: L{objects.Disk}
1415 @param parent_cdev: the disk from which we should remove children
1416 @type new_cdevs: list of L{objects.Disk}
1417 @param new_cdevs: the list of children which we should remove
1421 parent_bdev = _RecursiveFindBD(parent_cdev)
1422 if parent_bdev is None:
1423 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1425 for disk in new_cdevs:
1426 rpath = disk.StaticDevPath()
1428 bd = _RecursiveFindBD(disk)
1430 _Fail("Can't find device %s while removing children", disk)
1432 devs.append(bd.dev_path)
1435 parent_bdev.RemoveChildren(devs)
1438 def BlockdevGetmirrorstatus(disks):
1439 """Get the mirroring status of a list of devices.
1441 @type disks: list of L{objects.Disk}
1442 @param disks: the list of disks which we should query
1445 a list of (mirror_done, estimated_time) tuples, which
1446 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1447 @raise errors.BlockDeviceError: if any of the disks cannot be
1453 rbd = _RecursiveFindBD(dsk)
1455 _Fail("Can't find device %s", dsk)
1457 stats.append(rbd.CombinedSyncStatus())
1462 def _RecursiveFindBD(disk):
1463 """Check if a device is activated.
1465 If so, return information about the real device.
1467 @type disk: L{objects.Disk}
1468 @param disk: the disk object we need to find
1470 @return: None if the device can't be found,
1471 otherwise the device instance
1476 for chdisk in disk.children:
1477 children.append(_RecursiveFindBD(chdisk))
1479 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1482 def BlockdevFind(disk):
1483 """Check if a device is activated.
1485 If it is, return information about the real device.
1487 @type disk: L{objects.Disk}
1488 @param disk: the disk to find
1489 @rtype: None or objects.BlockDevStatus
1490 @return: None if the disk cannot be found, otherwise a the current
1495 rbd = _RecursiveFindBD(disk)
1496 except errors.BlockDeviceError, err:
1497 _Fail("Failed to find device: %s", err, exc=True)
1502 return rbd.GetSyncStatus()
1505 def BlockdevGetsize(disks):
1506 """Computes the size of the given disks.
1508 If a disk is not found, returns None instead.
1510 @type disks: list of L{objects.Disk}
1511 @param disks: the list of disk to compute the size for
1513 @return: list with elements None if the disk cannot be found,
1520 rbd = _RecursiveFindBD(cf)
1521 except errors.BlockDeviceError:
1527 result.append(rbd.GetActualSize())
1531 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1532 """Export a block device to a remote node.
1534 @type disk: L{objects.Disk}
1535 @param disk: the description of the disk to export
1536 @type dest_node: str
1537 @param dest_node: the destination node to export to
1538 @type dest_path: str
1539 @param dest_path: the destination path on the target node
1540 @type cluster_name: str
1541 @param cluster_name: the cluster name, needed for SSH hostalias
1545 real_disk = _RecursiveFindBD(disk)
1546 if real_disk is None:
1547 _Fail("Block device '%s' is not set up", disk)
1551 # the block size on the read dd is 1MiB to match our units
1552 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1553 "dd if=%s bs=1048576 count=%s",
1554 real_disk.dev_path, str(disk.size))
1556 # we set here a smaller block size as, due to ssh buffering, more
1557 # than 64-128k will mostly ignored; we use nocreat to fail if the
1558 # device is not already there or we pass a wrong path; we use
1559 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1560 # to not buffer too much memory; this means that at best, we flush
1561 # every 64k, which will not be very fast
1562 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1563 " oflag=dsync", dest_path)
1565 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1566 constants.GANETI_RUNAS,
1569 # all commands have been checked, so we're safe to combine them
1570 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1572 result = utils.RunCmd(["bash", "-c", command])
1575 _Fail("Disk copy command '%s' returned error: %s"
1576 " output: %s", command, result.fail_reason, result.output)
1579 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1580 """Write a file to the filesystem.
1582 This allows the master to overwrite(!) a file. It will only perform
1583 the operation if the file belongs to a list of configuration files.
1585 @type file_name: str
1586 @param file_name: the target file name
1588 @param data: the new contents of the file
1590 @param mode: the mode to give the file (can be None)
1592 @param uid: the owner of the file (can be -1 for default)
1594 @param gid: the group of the file (can be -1 for default)
1596 @param atime: the atime to set on the file (can be None)
1598 @param mtime: the mtime to set on the file (can be None)
1602 if not os.path.isabs(file_name):
1603 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1605 if file_name not in _ALLOWED_UPLOAD_FILES:
1606 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1609 raw_data = _Decompress(data)
1611 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1612 atime=atime, mtime=mtime)
1615 def WriteSsconfFiles(values):
1616 """Update all ssconf files.
1618 Wrapper around the SimpleStore.WriteFiles.
1621 ssconf.SimpleStore().WriteFiles(values)
1624 def _ErrnoOrStr(err):
1625 """Format an EnvironmentError exception.
1627 If the L{err} argument has an errno attribute, it will be looked up
1628 and converted into a textual C{E...} description. Otherwise the
1629 string representation of the error will be returned.
1631 @type err: L{EnvironmentError}
1632 @param err: the exception to format
1635 if hasattr(err, 'errno'):
1636 detail = errno.errorcode[err.errno]
1642 def _OSOndiskAPIVersion(name, os_dir):
1643 """Compute and return the API version of a given OS.
1645 This function will try to read the API version of the OS given by
1646 the 'name' parameter and residing in the 'os_dir' directory.
1649 @param name: the OS name we should look for
1651 @param os_dir: the directory inwhich we should look for the OS
1653 @return: tuple (status, data) with status denoting the validity and
1654 data holding either the vaid versions or an error message
1657 api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
1660 st = os.stat(api_file)
1661 except EnvironmentError, err:
1662 return False, ("Required file '%s' not found under path %s: %s" %
1663 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1665 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1666 return False, ("File '%s' in %s is not a regular file" %
1667 (constants.OS_API_FILE, os_dir))
1670 api_versions = utils.ReadFile(api_file).splitlines()
1671 except EnvironmentError, err:
1672 return False, ("Error while reading the API version file at %s: %s" %
1673 (api_file, _ErrnoOrStr(err)))
1676 api_versions = [int(version.strip()) for version in api_versions]
1677 except (TypeError, ValueError), err:
1678 return False, ("API version(s) can't be converted to integer: %s" %
1681 return True, api_versions
1684 def DiagnoseOS(top_dirs=None):
1685 """Compute the validity for all OSes.
1687 @type top_dirs: list
1688 @param top_dirs: the list of directories in which to
1689 search (if not given defaults to
1690 L{constants.OS_SEARCH_PATH})
1691 @rtype: list of L{objects.OS}
1692 @return: a list of tuples (name, path, status, diagnose, variants)
1693 for all (potential) OSes under all search paths, where:
1694 - name is the (potential) OS name
1695 - path is the full path to the OS
1696 - status True/False is the validity of the OS
1697 - diagnose is the error message for an invalid OS, otherwise empty
1698 - variants is a list of supported OS variants, if any
1701 if top_dirs is None:
1702 top_dirs = constants.OS_SEARCH_PATH
1705 for dir_name in top_dirs:
1706 if os.path.isdir(dir_name):
1708 f_names = utils.ListVisibleFiles(dir_name)
1709 except EnvironmentError, err:
1710 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1712 for name in f_names:
1713 os_path = os.path.sep.join([dir_name, name])
1714 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1717 variants = os_inst.supported_variants
1721 result.append((name, os_path, status, diagnose, variants))
1726 def _TryOSFromDisk(name, base_dir=None):
1727 """Create an OS instance from disk.
1729 This function will return an OS instance if the given name is a
1732 @type base_dir: string
1733 @keyword base_dir: Base directory containing OS installations.
1734 Defaults to a search in all the OS_SEARCH_PATH dirs.
1736 @return: success and either the OS instance if we find a valid one,
1740 if base_dir is None:
1741 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1743 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1746 return False, "Directory for OS %s not found in search path" % name
1748 status, api_versions = _OSOndiskAPIVersion(name, os_dir)
1751 return status, api_versions
1753 if not constants.OS_API_VERSIONS.intersection(api_versions):
1754 return False, ("API version mismatch for path '%s': found %s, want %s." %
1755 (os_dir, api_versions, constants.OS_API_VERSIONS))
1757 # OS Files dictionary, we will populate it with the absolute path names
1758 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1760 if max(api_versions) >= constants.OS_API_V15:
1761 os_files[constants.OS_VARIANTS_FILE] = ''
1763 for filename in os_files:
1764 os_files[filename] = os.path.sep.join([os_dir, filename])
1767 st = os.stat(os_files[filename])
1768 except EnvironmentError, err:
1769 return False, ("File '%s' under path '%s' is missing (%s)" %
1770 (filename, os_dir, _ErrnoOrStr(err)))
1772 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1773 return False, ("File '%s' under path '%s' is not a regular file" %
1776 if filename in constants.OS_SCRIPTS:
1777 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1778 return False, ("File '%s' under path '%s' is not executable" %
1782 if constants.OS_VARIANTS_FILE in os_files:
1783 variants_file = os_files[constants.OS_VARIANTS_FILE]
1785 variants = utils.ReadFile(variants_file).splitlines()
1786 except EnvironmentError, err:
1787 return False, ("Error while reading the OS variants file at %s: %s" %
1788 (variants_file, _ErrnoOrStr(err)))
1790 return False, ("No supported os variant found")
1792 os_obj = objects.OS(name=name, path=os_dir,
1793 create_script=os_files[constants.OS_SCRIPT_CREATE],
1794 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1795 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1796 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1797 supported_variants=variants,
1798 api_versions=api_versions)
1802 def OSFromDisk(name, base_dir=None):
1803 """Create an OS instance from disk.
1805 This function will return an OS instance if the given name is a
1806 valid OS name. Otherwise, it will raise an appropriate
1807 L{RPCFail} exception, detailing why this is not a valid OS.
1809 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1810 an exception but returns true/false status data.
1812 @type base_dir: string
1813 @keyword base_dir: Base directory containing OS installations.
1814 Defaults to a search in all the OS_SEARCH_PATH dirs.
1815 @rtype: L{objects.OS}
1816 @return: the OS instance if we find a valid one
1817 @raise RPCFail: if we don't find a valid OS
1820 name_only = name.split("+", 1)[0]
1821 status, payload = _TryOSFromDisk(name_only, base_dir)
1829 def OSEnvironment(instance, inst_os, debug=0):
1830 """Calculate the environment for an os script.
1832 @type instance: L{objects.Instance}
1833 @param instance: target instance for the os script run
1834 @type inst_os: L{objects.OS}
1835 @param inst_os: operating system for which the environment is being built
1836 @type debug: integer
1837 @param debug: debug level (0 or 1, for OS Api 10)
1839 @return: dict of environment variables
1840 @raise errors.BlockDeviceError: if the block device
1846 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1847 result['OS_API_VERSION'] = '%d' % api_version
1848 result['INSTANCE_NAME'] = instance.name
1849 result['INSTANCE_OS'] = instance.os
1850 result['HYPERVISOR'] = instance.hypervisor
1851 result['DISK_COUNT'] = '%d' % len(instance.disks)
1852 result['NIC_COUNT'] = '%d' % len(instance.nics)
1853 result['DEBUG_LEVEL'] = '%d' % debug
1854 if api_version >= constants.OS_API_V15:
1856 variant = instance.os.split('+', 1)[1]
1858 variant = inst_os.supported_variants[0]
1859 result['OS_VARIANT'] = variant
1860 for idx, disk in enumerate(instance.disks):
1861 real_disk = _RecursiveFindBD(disk)
1862 if real_disk is None:
1863 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1866 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1867 result['DISK_%d_ACCESS' % idx] = disk.mode
1868 if constants.HV_DISK_TYPE in instance.hvparams:
1869 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1870 instance.hvparams[constants.HV_DISK_TYPE]
1871 if disk.dev_type in constants.LDS_BLOCK:
1872 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1873 elif disk.dev_type == constants.LD_FILE:
1874 result['DISK_%d_BACKEND_TYPE' % idx] = \
1875 'file:%s' % disk.physical_id[0]
1876 for idx, nic in enumerate(instance.nics):
1877 result['NIC_%d_MAC' % idx] = nic.mac
1879 result['NIC_%d_IP' % idx] = nic.ip
1880 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1881 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1882 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1883 if nic.nicparams[constants.NIC_LINK]:
1884 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1885 if constants.HV_NIC_TYPE in instance.hvparams:
1886 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1887 instance.hvparams[constants.HV_NIC_TYPE]
1889 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1890 for key, value in source.items():
1891 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1895 def BlockdevGrow(disk, amount):
1896 """Grow a stack of block devices.
1898 This function is called recursively, with the childrens being the
1899 first ones to resize.
1901 @type disk: L{objects.Disk}
1902 @param disk: the disk to be grown
1903 @rtype: (status, result)
1904 @return: a tuple with the status of the operation
1905 (True/False), and the errors message if status
1909 r_dev = _RecursiveFindBD(disk)
1911 _Fail("Cannot find block device %s", disk)
1915 except errors.BlockDeviceError, err:
1916 _Fail("Failed to grow block device: %s", err, exc=True)
1919 def BlockdevSnapshot(disk):
1920 """Create a snapshot copy of a block device.
1922 This function is called recursively, and the snapshot is actually created
1923 just for the leaf lvm backend device.
1925 @type disk: L{objects.Disk}
1926 @param disk: the disk to be snapshotted
1928 @return: snapshot disk path
1932 if len(disk.children) == 1:
1933 # only one child, let's recurse on it
1934 return BlockdevSnapshot(disk.children[0])
1936 # more than one child, choose one that matches
1937 for child in disk.children:
1938 if child.size == disk.size:
1939 # return implies breaking the loop
1940 return BlockdevSnapshot(child)
1941 elif disk.dev_type == constants.LD_LV:
1942 r_dev = _RecursiveFindBD(disk)
1943 if r_dev is not None:
1944 # let's stay on the safe side and ask for the full size, for now
1945 return r_dev.Snapshot(disk.size)
1947 _Fail("Cannot find block device %s", disk)
1949 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1950 disk.unique_id, disk.dev_type)
1953 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1954 """Export a block device snapshot to a remote node.
1956 @type disk: L{objects.Disk}
1957 @param disk: the description of the disk to export
1958 @type dest_node: str
1959 @param dest_node: the destination node to export to
1960 @type instance: L{objects.Instance}
1961 @param instance: the instance object to whom the disk belongs
1962 @type cluster_name: str
1963 @param cluster_name: the cluster name, needed for SSH hostalias
1965 @param idx: the index of the disk in the instance's disk list,
1966 used to export to the OS scripts environment
1970 inst_os = OSFromDisk(instance.os)
1971 export_env = OSEnvironment(instance, inst_os)
1973 export_script = inst_os.export_script
1975 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1976 instance.name, int(time.time()))
1977 if not os.path.exists(constants.LOG_OS_DIR):
1978 os.mkdir(constants.LOG_OS_DIR, 0750)
1979 real_disk = _RecursiveFindBD(disk)
1980 if real_disk is None:
1981 _Fail("Block device '%s' is not set up", disk)
1985 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1986 export_env['EXPORT_INDEX'] = str(idx)
1988 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1989 destfile = disk.physical_id[1]
1991 # the target command is built out of three individual commands,
1992 # which are joined by pipes; we check each individual command for
1994 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
1995 inst_os.path, export_script, logfile)
1999 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
2000 destdir, destdir, destfile)
2001 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2002 constants.GANETI_RUNAS,
2005 # all commands have been checked, so we're safe to combine them
2006 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
2008 result = utils.RunCmd(["bash", "-c", command], env=export_env)
2011 _Fail("OS snapshot export command '%s' returned error: %s"
2012 " output: %s", command, result.fail_reason, result.output)
2015 def FinalizeExport(instance, snap_disks):
2016 """Write out the export configuration information.
2018 @type instance: L{objects.Instance}
2019 @param instance: the instance which we export, used for
2020 saving configuration
2021 @type snap_disks: list of L{objects.Disk}
2022 @param snap_disks: list of snapshot block devices, which
2023 will be used to get the actual name of the dump file
2028 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
2029 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
2031 config = objects.SerializableConfigParser()
2033 config.add_section(constants.INISECT_EXP)
2034 config.set(constants.INISECT_EXP, 'version', '0')
2035 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2036 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2037 config.set(constants.INISECT_EXP, 'os', instance.os)
2038 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2040 config.add_section(constants.INISECT_INS)
2041 config.set(constants.INISECT_INS, 'name', instance.name)
2042 config.set(constants.INISECT_INS, 'memory', '%d' %
2043 instance.beparams[constants.BE_MEMORY])
2044 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2045 instance.beparams[constants.BE_VCPUS])
2046 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2049 for nic_count, nic in enumerate(instance.nics):
2051 config.set(constants.INISECT_INS, 'nic%d_mac' %
2052 nic_count, '%s' % nic.mac)
2053 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2054 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
2056 # TODO: redundant: on load can read nics until it doesn't exist
2057 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2060 for disk_count, disk in enumerate(snap_disks):
2063 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2064 ('%s' % disk.iv_name))
2065 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2066 ('%s' % disk.physical_id[1]))
2067 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2070 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2072 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
2073 data=config.Dumps())
2074 shutil.rmtree(finaldestdir, True)
2075 shutil.move(destdir, finaldestdir)
2078 def ExportInfo(dest):
2079 """Get export configuration information.
2082 @param dest: directory containing the export
2084 @rtype: L{objects.SerializableConfigParser}
2085 @return: a serializable config file containing the
2089 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
2091 config = objects.SerializableConfigParser()
2094 if (not config.has_section(constants.INISECT_EXP) or
2095 not config.has_section(constants.INISECT_INS)):
2096 _Fail("Export info file doesn't have the required fields")
2098 return config.Dumps()
2101 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
2102 """Import an os image into an instance.
2104 @type instance: L{objects.Instance}
2105 @param instance: instance to import the disks into
2106 @type src_node: string
2107 @param src_node: source node for the disk images
2108 @type src_images: list of string
2109 @param src_images: absolute paths of the disk images
2110 @rtype: list of boolean
2111 @return: each boolean represent the success of importing the n-th disk
2114 inst_os = OSFromDisk(instance.os)
2115 import_env = OSEnvironment(instance, inst_os)
2116 import_script = inst_os.import_script
2118 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2119 instance.name, int(time.time()))
2120 if not os.path.exists(constants.LOG_OS_DIR):
2121 os.mkdir(constants.LOG_OS_DIR, 0750)
2124 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2125 import_script, logfile)
2128 for idx, image in enumerate(src_images):
2130 destcmd = utils.BuildShellCmd('cat %s', image)
2131 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2132 constants.GANETI_RUNAS,
2134 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2135 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2136 import_env['IMPORT_INDEX'] = str(idx)
2137 result = utils.RunCmd(command, env=import_env)
2139 logging.error("Disk import command '%s' returned error: %s"
2140 " output: %s", command, result.fail_reason,
2142 final_result.append("error importing disk %d: %s, %s" %
2143 (idx, result.fail_reason, result.output[-100]))
2146 _Fail("; ".join(final_result), log=False)
2150 """Return a list of exports currently available on this machine.
2153 @return: list of the exports
2156 if os.path.isdir(constants.EXPORT_DIR):
2157 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2159 _Fail("No exports directory")
2162 def RemoveExport(export):
2163 """Remove an existing export from the node.
2166 @param export: the name of the export to remove
2170 target = os.path.join(constants.EXPORT_DIR, export)
2173 shutil.rmtree(target)
2174 except EnvironmentError, err:
2175 _Fail("Error while removing the export: %s", err, exc=True)
2178 def BlockdevRename(devlist):
2179 """Rename a list of block devices.
2181 @type devlist: list of tuples
2182 @param devlist: list of tuples of the form (disk,
2183 new_logical_id, new_physical_id); disk is an
2184 L{objects.Disk} object describing the current disk,
2185 and new logical_id/physical_id is the name we
2188 @return: True if all renames succeeded, False otherwise
2193 for disk, unique_id in devlist:
2194 dev = _RecursiveFindBD(disk)
2196 msgs.append("Can't find device %s in rename" % str(disk))
2200 old_rpath = dev.dev_path
2201 dev.Rename(unique_id)
2202 new_rpath = dev.dev_path
2203 if old_rpath != new_rpath:
2204 DevCacheManager.RemoveCache(old_rpath)
2205 # FIXME: we should add the new cache information here, like:
2206 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2207 # but we don't have the owner here - maybe parse from existing
2208 # cache? for now, we only lose lvm data when we rename, which
2209 # is less critical than DRBD or MD
2210 except errors.BlockDeviceError, err:
2211 msgs.append("Can't rename device '%s' to '%s': %s" %
2212 (dev, unique_id, err))
2213 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2216 _Fail("; ".join(msgs))
2219 def _TransformFileStorageDir(file_storage_dir):
2220 """Checks whether given file_storage_dir is valid.
2222 Checks wheter the given file_storage_dir is within the cluster-wide
2223 default file_storage_dir stored in SimpleStore. Only paths under that
2224 directory are allowed.
2226 @type file_storage_dir: str
2227 @param file_storage_dir: the path to check
2229 @return: the normalized path if valid, None otherwise
2233 file_storage_dir = os.path.normpath(file_storage_dir)
2234 base_file_storage_dir = cfg.GetFileStorageDir()
2235 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2236 base_file_storage_dir):
2237 _Fail("File storage directory '%s' is not under base file"
2238 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2239 return file_storage_dir
2242 def CreateFileStorageDir(file_storage_dir):
2243 """Create file storage directory.
2245 @type file_storage_dir: str
2246 @param file_storage_dir: directory to create
2249 @return: tuple with first element a boolean indicating wheter dir
2250 creation was successful or not
2253 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2254 if os.path.exists(file_storage_dir):
2255 if not os.path.isdir(file_storage_dir):
2256 _Fail("Specified storage dir '%s' is not a directory",
2260 os.makedirs(file_storage_dir, 0750)
2261 except OSError, err:
2262 _Fail("Cannot create file storage directory '%s': %s",
2263 file_storage_dir, err, exc=True)
2266 def RemoveFileStorageDir(file_storage_dir):
2267 """Remove file storage directory.
2269 Remove it only if it's empty. If not log an error and return.
2271 @type file_storage_dir: str
2272 @param file_storage_dir: the directory we should cleanup
2273 @rtype: tuple (success,)
2274 @return: tuple of one element, C{success}, denoting
2275 whether the operation was successful
2278 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2279 if os.path.exists(file_storage_dir):
2280 if not os.path.isdir(file_storage_dir):
2281 _Fail("Specified Storage directory '%s' is not a directory",
2283 # deletes dir only if empty, otherwise we want to fail the rpc call
2285 os.rmdir(file_storage_dir)
2286 except OSError, err:
2287 _Fail("Cannot remove file storage directory '%s': %s",
2288 file_storage_dir, err)
2291 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2292 """Rename the file storage directory.
2294 @type old_file_storage_dir: str
2295 @param old_file_storage_dir: the current path
2296 @type new_file_storage_dir: str
2297 @param new_file_storage_dir: the name we should rename to
2298 @rtype: tuple (success,)
2299 @return: tuple of one element, C{success}, denoting
2300 whether the operation was successful
2303 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2304 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2305 if not os.path.exists(new_file_storage_dir):
2306 if os.path.isdir(old_file_storage_dir):
2308 os.rename(old_file_storage_dir, new_file_storage_dir)
2309 except OSError, err:
2310 _Fail("Cannot rename '%s' to '%s': %s",
2311 old_file_storage_dir, new_file_storage_dir, err)
2313 _Fail("Specified storage dir '%s' is not a directory",
2314 old_file_storage_dir)
2316 if os.path.exists(old_file_storage_dir):
2317 _Fail("Cannot rename '%s' to '%s': both locations exist",
2318 old_file_storage_dir, new_file_storage_dir)
2321 def _EnsureJobQueueFile(file_name):
2322 """Checks whether the given filename is in the queue directory.
2324 @type file_name: str
2325 @param file_name: the file name we should check
2327 @raises RPCFail: if the file is not valid
2330 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2331 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2334 _Fail("Passed job queue file '%s' does not belong to"
2335 " the queue directory '%s'", file_name, queue_dir)
2338 def JobQueueUpdate(file_name, content):
2339 """Updates a file in the queue directory.
2341 This is just a wrapper over L{utils.WriteFile}, with proper
2344 @type file_name: str
2345 @param file_name: the job file name
2347 @param content: the new job contents
2349 @return: the success of the operation
2352 _EnsureJobQueueFile(file_name)
2354 # Write and replace the file atomically
2355 utils.WriteFile(file_name, data=_Decompress(content))
2358 def JobQueueRename(old, new):
2359 """Renames a job queue file.
2361 This is just a wrapper over os.rename with proper checking.
2364 @param old: the old (actual) file name
2366 @param new: the desired file name
2368 @return: the success of the operation and payload
2371 _EnsureJobQueueFile(old)
2372 _EnsureJobQueueFile(new)
2374 utils.RenameFile(old, new, mkdir=True)
2377 def JobQueueSetDrainFlag(drain_flag):
2378 """Set the drain flag for the queue.
2380 This will set or unset the queue drain flag.
2382 @type drain_flag: boolean
2383 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2385 @return: always True, None
2386 @warning: the function always returns True
2390 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2392 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2395 def BlockdevClose(instance_name, disks):
2396 """Closes the given block devices.
2398 This means they will be switched to secondary mode (in case of
2401 @param instance_name: if the argument is not empty, the symlinks
2402 of this instance will be removed
2403 @type disks: list of L{objects.Disk}
2404 @param disks: the list of disks to be closed
2405 @rtype: tuple (success, message)
2406 @return: a tuple of success and message, where success
2407 indicates the succes of the operation, and message
2408 which will contain the error details in case we
2414 rd = _RecursiveFindBD(cf)
2416 _Fail("Can't find device %s", cf)
2423 except errors.BlockDeviceError, err:
2424 msg.append(str(err))
2426 _Fail("Can't make devices secondary: %s", ",".join(msg))
2429 _RemoveBlockDevLinks(instance_name, disks)
2432 def ValidateHVParams(hvname, hvparams):
2433 """Validates the given hypervisor parameters.
2435 @type hvname: string
2436 @param hvname: the hypervisor name
2437 @type hvparams: dict
2438 @param hvparams: the hypervisor parameters to be validated
2443 hv_type = hypervisor.GetHypervisor(hvname)
2444 hv_type.ValidateParameters(hvparams)
2445 except errors.HypervisorError, err:
2446 _Fail(str(err), log=False)
2450 """Demotes the current node from master candidate role.
2453 # try to ensure we're not the master by mistake
2454 master, myself = ssconf.GetMasterAndMyself()
2455 if master == myself:
2456 _Fail("ssconf status shows I'm the master node, will not demote")
2458 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2459 if not result.failed:
2460 _Fail("The master daemon is running, will not demote")
2463 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2464 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2465 except EnvironmentError, err:
2466 if err.errno != errno.ENOENT:
2467 _Fail("Error while backing up cluster file: %s", err, exc=True)
2469 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2472 def _FindDisks(nodes_ip, disks):
2473 """Sets the physical ID on disks and returns the block devices.
2476 # set the correct physical ID
2477 my_name = utils.HostInfo().name
2479 cf.SetPhysicalID(my_name, nodes_ip)
2484 rd = _RecursiveFindBD(cf)
2486 _Fail("Can't find device %s", cf)
2491 def DrbdDisconnectNet(nodes_ip, disks):
2492 """Disconnects the network on a list of drbd devices.
2495 bdevs = _FindDisks(nodes_ip, disks)
2501 except errors.BlockDeviceError, err:
2502 _Fail("Can't change network configuration to standalone mode: %s",
2506 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2507 """Attaches the network on a list of drbd devices.
2510 bdevs = _FindDisks(nodes_ip, disks)
2513 for idx, rd in enumerate(bdevs):
2515 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2516 except EnvironmentError, err:
2517 _Fail("Can't create symlink: %s", err)
2518 # reconnect disks, switch to new master configuration and if
2519 # needed primary mode
2522 rd.AttachNet(multimaster)
2523 except errors.BlockDeviceError, err:
2524 _Fail("Can't change network configuration: %s", err)
2526 # wait until the disks are connected; we need to retry the re-attach
2527 # if the device becomes standalone, as this might happen if the one
2528 # node disconnects and reconnects in a different mode before the
2529 # other node reconnects; in this case, one or both of the nodes will
2530 # decide it has wrong configuration and switch to standalone
2533 all_connected = True
2536 stats = rd.GetProcStatus()
2538 all_connected = (all_connected and
2539 (stats.is_connected or stats.is_in_resync))
2541 if stats.is_standalone:
2542 # peer had different config info and this node became
2543 # standalone, even though this should not happen with the
2544 # new staged way of changing disk configs
2546 rd.AttachNet(multimaster)
2547 except errors.BlockDeviceError, err:
2548 _Fail("Can't change network configuration: %s", err)
2550 if not all_connected:
2551 raise utils.RetryAgain()
2554 # Start with a delay of 100 miliseconds and go up to 5 seconds
2555 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2556 except utils.RetryTimeout:
2557 _Fail("Timeout in disk reconnecting")
2560 # change to primary mode
2564 except errors.BlockDeviceError, err:
2565 _Fail("Can't change to primary mode: %s", err)
2568 def DrbdWaitSync(nodes_ip, disks):
2569 """Wait until DRBDs have synchronized.
2573 stats = rd.GetProcStatus()
2574 if not (stats.is_connected or stats.is_in_resync):
2575 raise utils.RetryAgain()
2578 bdevs = _FindDisks(nodes_ip, disks)
2584 # poll each second for 15 seconds
2585 stats = utils.Retry(_helper, 1, 15, args=[rd])
2586 except utils.RetryTimeout:
2587 stats = rd.GetProcStatus()
2589 if not (stats.is_connected or stats.is_in_resync):
2590 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2591 alldone = alldone and (not stats.is_in_resync)
2592 if stats.sync_percent is not None:
2593 min_resync = min(min_resync, stats.sync_percent)
2595 return (alldone, min_resync)
2598 def PowercycleNode(hypervisor_type):
2599 """Hard-powercycle the node.
2601 Because we need to return first, and schedule the powercycle in the
2602 background, we won't be able to report failures nicely.
2605 hyper = hypervisor.GetHypervisor(hypervisor_type)
2609 # if we can't fork, we'll pretend that we're in the child process
2612 return "Reboot scheduled in 5 seconds"
2614 hyper.PowercycleNode()
2617 class HooksRunner(object):
2620 This class is instantiated on the node side (ganeti-noded) and not
2624 def __init__(self, hooks_base_dir=None):
2625 """Constructor for hooks runner.
2627 @type hooks_base_dir: str or None
2628 @param hooks_base_dir: if not None, this overrides the
2629 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2632 if hooks_base_dir is None:
2633 hooks_base_dir = constants.HOOKS_BASE_DIR
2634 # yeah, _BASE_DIR is not valid for attributes, we use it like a
2636 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
2639 def ExecHook(script, env):
2640 """Exec one hook script.
2643 @param script: the full path to the script
2645 @param env: the environment with which to exec the script
2646 @rtype: tuple (success, message)
2647 @return: a tuple of success and message, where success
2648 indicates the succes of the operation, and message
2649 which will contain the error details in case we
2653 # exec the process using subprocess and log the output
2656 fdstdin = open("/dev/null", "r")
2657 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2658 stderr=subprocess.STDOUT, close_fds=True,
2659 shell=False, cwd="/", env=env)
2662 output = child.stdout.read(4096)
2663 child.stdout.close()
2664 except EnvironmentError, err:
2665 output += "Hook script error: %s" % str(err)
2669 result = child.wait()
2671 except EnvironmentError, err:
2672 if err.errno == errno.EINTR:
2676 # try not to leak fds
2677 for fd in (fdstdin, ):
2681 except EnvironmentError, err:
2682 # just log the error
2683 #logging.exception("Error while closing fd %s", fd)
2686 return result == 0, utils.SafeEncode(output.strip())
2688 def RunHooks(self, hpath, phase, env):
2689 """Run the scripts in the hooks directory.
2692 @param hpath: the path to the hooks directory which
2695 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2696 L{constants.HOOKS_PHASE_POST}
2698 @param env: dictionary with the environment for the hook
2700 @return: list of 3-element tuples:
2702 - script result, either L{constants.HKR_SUCCESS} or
2703 L{constants.HKR_FAIL}
2704 - output of the script
2706 @raise errors.ProgrammerError: for invalid input
2710 if phase == constants.HOOKS_PHASE_PRE:
2712 elif phase == constants.HOOKS_PHASE_POST:
2715 _Fail("Unknown hooks phase '%s'", phase)
2719 subdir = "%s-%s.d" % (hpath, suffix)
2720 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2722 dir_contents = utils.ListVisibleFiles(dir_name)
2724 # FIXME: must log output in case of failures
2727 # we use the standard python sort order,
2728 # so 00name is the recommended naming scheme
2730 for relname in dir_contents:
2731 fname = os.path.join(dir_name, relname)
2732 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2733 constants.EXT_PLUGIN_MASK.match(relname) is not None):
2734 rrval = constants.HKR_SKIP
2737 result, output = self.ExecHook(fname, env)
2739 rrval = constants.HKR_FAIL
2741 rrval = constants.HKR_SUCCESS
2742 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2747 class IAllocatorRunner(object):
2748 """IAllocator runner.
2750 This class is instantiated on the node side (ganeti-noded) and not on
2755 def Run(name, idata):
2756 """Run an iallocator script.
2759 @param name: the iallocator script name
2761 @param idata: the allocator input data
2764 @return: two element tuple of:
2766 - either error message or stdout of allocator (for success)
2769 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2771 if alloc_script is None:
2772 _Fail("iallocator module '%s' not found in the search path", name)
2774 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2778 result = utils.RunCmd([alloc_script, fin_name])
2780 _Fail("iallocator module '%s' failed: %s, output '%s'",
2781 name, result.fail_reason, result.output)
2785 return result.stdout
2788 class DevCacheManager(object):
2789 """Simple class for managing a cache of block device information.
2792 _DEV_PREFIX = "/dev/"
2793 _ROOT_DIR = constants.BDEV_CACHE_DIR
2796 def _ConvertPath(cls, dev_path):
2797 """Converts a /dev/name path to the cache file name.
2799 This replaces slashes with underscores and strips the /dev
2800 prefix. It then returns the full path to the cache file.
2803 @param dev_path: the C{/dev/} path name
2805 @return: the converted path name
2808 if dev_path.startswith(cls._DEV_PREFIX):
2809 dev_path = dev_path[len(cls._DEV_PREFIX):]
2810 dev_path = dev_path.replace("/", "_")
2811 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2815 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2816 """Updates the cache information for a given device.
2819 @param dev_path: the pathname of the device
2821 @param owner: the owner (instance name) of the device
2822 @type on_primary: bool
2823 @param on_primary: whether this is the primary
2826 @param iv_name: the instance-visible name of the
2827 device, as in objects.Disk.iv_name
2832 if dev_path is None:
2833 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2835 fpath = cls._ConvertPath(dev_path)
2841 iv_name = "not_visible"
2842 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2844 utils.WriteFile(fpath, data=fdata)
2845 except EnvironmentError, err:
2846 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2849 def RemoveCache(cls, dev_path):
2850 """Remove data for a dev_path.
2852 This is just a wrapper over L{utils.RemoveFile} with a converted
2853 path name and logging.
2856 @param dev_path: the pathname of the device
2861 if dev_path is None:
2862 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2864 fpath = cls._ConvertPath(dev_path)
2866 utils.RemoveFile(fpath)
2867 except EnvironmentError, err:
2868 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)