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)
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.
1200 for child in disk.children:
1202 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1203 except errors.BlockDeviceError, err:
1204 _Fail("Can't assemble device %s: %s", child, err)
1205 if on_primary or disk.AssembleOnSecondary():
1206 # we need the children open in case the device itself has to
1210 except errors.BlockDeviceError, err:
1211 _Fail("Can't make child '%s' read-write: %s", child, err)
1215 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1216 except errors.BlockDeviceError, err:
1217 _Fail("Can't create block device: %s", err)
1219 if on_primary or disk.AssembleOnSecondary():
1222 except errors.BlockDeviceError, err:
1223 _Fail("Can't assemble device after creation, unusual event: %s", err)
1224 device.SetSyncSpeed(constants.SYNC_SPEED)
1225 if on_primary or disk.OpenOnSecondary():
1227 device.Open(force=True)
1228 except errors.BlockDeviceError, err:
1229 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1230 DevCacheManager.UpdateCache(device.dev_path, owner,
1231 on_primary, disk.iv_name)
1233 device.SetInfo(info)
1235 return device.unique_id
1238 def BlockdevRemove(disk):
1239 """Remove a block device.
1241 @note: This is intended to be called recursively.
1243 @type disk: L{objects.Disk}
1244 @param disk: the disk object we should remove
1246 @return: the success of the operation
1251 rdev = _RecursiveFindBD(disk)
1252 except errors.BlockDeviceError, err:
1253 # probably can't attach
1254 logging.info("Can't attach to device %s in remove", disk)
1256 if rdev is not None:
1257 r_path = rdev.dev_path
1260 except errors.BlockDeviceError, err:
1261 msgs.append(str(err))
1263 DevCacheManager.RemoveCache(r_path)
1266 for child in disk.children:
1268 BlockdevRemove(child)
1269 except RPCFail, err:
1270 msgs.append(str(err))
1273 _Fail("; ".join(msgs))
1276 def _RecursiveAssembleBD(disk, owner, as_primary):
1277 """Activate a block device for an instance.
1279 This is run on the primary and secondary nodes for an instance.
1281 @note: this function is called recursively.
1283 @type disk: L{objects.Disk}
1284 @param disk: the disk we try to assemble
1286 @param owner: the name of the instance which owns the disk
1287 @type as_primary: boolean
1288 @param as_primary: if we should make the block device
1291 @return: the assembled device or None (in case no device
1293 @raise errors.BlockDeviceError: in case there is an error
1294 during the activation of the children or the device
1300 mcn = disk.ChildrenNeeded()
1302 mcn = 0 # max number of Nones allowed
1304 mcn = len(disk.children) - mcn # max number of Nones
1305 for chld_disk in disk.children:
1307 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1308 except errors.BlockDeviceError, err:
1309 if children.count(None) >= mcn:
1312 logging.error("Error in child activation (but continuing): %s",
1314 children.append(cdev)
1316 if as_primary or disk.AssembleOnSecondary():
1317 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1318 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1320 if as_primary or disk.OpenOnSecondary():
1322 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1323 as_primary, disk.iv_name)
1330 def BlockdevAssemble(disk, owner, as_primary):
1331 """Activate a block device for an instance.
1333 This is a wrapper over _RecursiveAssembleBD.
1335 @rtype: str or boolean
1336 @return: a C{/dev/...} path for primary nodes, and
1337 C{True} for secondary nodes
1341 result = _RecursiveAssembleBD(disk, owner, as_primary)
1342 if isinstance(result, bdev.BlockDev):
1343 result = result.dev_path
1344 except errors.BlockDeviceError, err:
1345 _Fail("Error while assembling disk: %s", err, exc=True)
1350 def BlockdevShutdown(disk):
1351 """Shut down a block device.
1353 First, if the device is assembled (Attach() is successful), then
1354 the device is shutdown. Then the children of the device are
1357 This function is called recursively. Note that we don't cache the
1358 children or such, as oppossed to assemble, shutdown of different
1359 devices doesn't require that the upper device was active.
1361 @type disk: L{objects.Disk}
1362 @param disk: the description of the disk we should
1368 r_dev = _RecursiveFindBD(disk)
1369 if r_dev is not None:
1370 r_path = r_dev.dev_path
1373 DevCacheManager.RemoveCache(r_path)
1374 except errors.BlockDeviceError, err:
1375 msgs.append(str(err))
1378 for child in disk.children:
1380 BlockdevShutdown(child)
1381 except RPCFail, err:
1382 msgs.append(str(err))
1385 _Fail("; ".join(msgs))
1388 def BlockdevAddchildren(parent_cdev, new_cdevs):
1389 """Extend a mirrored block device.
1391 @type parent_cdev: L{objects.Disk}
1392 @param parent_cdev: the disk to which we should add children
1393 @type new_cdevs: list of L{objects.Disk}
1394 @param new_cdevs: the list of children which we should add
1398 parent_bdev = _RecursiveFindBD(parent_cdev)
1399 if parent_bdev is None:
1400 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1401 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1402 if new_bdevs.count(None) > 0:
1403 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1404 parent_bdev.AddChildren(new_bdevs)
1407 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1408 """Shrink a mirrored block device.
1410 @type parent_cdev: L{objects.Disk}
1411 @param parent_cdev: the disk from which we should remove children
1412 @type new_cdevs: list of L{objects.Disk}
1413 @param new_cdevs: the list of children which we should remove
1417 parent_bdev = _RecursiveFindBD(parent_cdev)
1418 if parent_bdev is None:
1419 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1421 for disk in new_cdevs:
1422 rpath = disk.StaticDevPath()
1424 bd = _RecursiveFindBD(disk)
1426 _Fail("Can't find device %s while removing children", disk)
1428 devs.append(bd.dev_path)
1431 parent_bdev.RemoveChildren(devs)
1434 def BlockdevGetmirrorstatus(disks):
1435 """Get the mirroring status of a list of devices.
1437 @type disks: list of L{objects.Disk}
1438 @param disks: the list of disks which we should query
1441 a list of (mirror_done, estimated_time) tuples, which
1442 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1443 @raise errors.BlockDeviceError: if any of the disks cannot be
1449 rbd = _RecursiveFindBD(dsk)
1451 _Fail("Can't find device %s", dsk)
1453 stats.append(rbd.CombinedSyncStatus())
1458 def _RecursiveFindBD(disk):
1459 """Check if a device is activated.
1461 If so, return information about the real device.
1463 @type disk: L{objects.Disk}
1464 @param disk: the disk object we need to find
1466 @return: None if the device can't be found,
1467 otherwise the device instance
1472 for chdisk in disk.children:
1473 children.append(_RecursiveFindBD(chdisk))
1475 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1478 def BlockdevFind(disk):
1479 """Check if a device is activated.
1481 If it is, return information about the real device.
1483 @type disk: L{objects.Disk}
1484 @param disk: the disk to find
1485 @rtype: None or objects.BlockDevStatus
1486 @return: None if the disk cannot be found, otherwise a the current
1491 rbd = _RecursiveFindBD(disk)
1492 except errors.BlockDeviceError, err:
1493 _Fail("Failed to find device: %s", err, exc=True)
1498 return rbd.GetSyncStatus()
1501 def BlockdevGetsize(disks):
1502 """Computes the size of the given disks.
1504 If a disk is not found, returns None instead.
1506 @type disks: list of L{objects.Disk}
1507 @param disks: the list of disk to compute the size for
1509 @return: list with elements None if the disk cannot be found,
1516 rbd = _RecursiveFindBD(cf)
1517 except errors.BlockDeviceError, err:
1523 result.append(rbd.GetActualSize())
1527 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1528 """Export a block device to a remote node.
1530 @type disk: L{objects.Disk}
1531 @param disk: the description of the disk to export
1532 @type dest_node: str
1533 @param dest_node: the destination node to export to
1534 @type dest_path: str
1535 @param dest_path: the destination path on the target node
1536 @type cluster_name: str
1537 @param cluster_name: the cluster name, needed for SSH hostalias
1541 real_disk = _RecursiveFindBD(disk)
1542 if real_disk is None:
1543 _Fail("Block device '%s' is not set up", disk)
1547 # the block size on the read dd is 1MiB to match our units
1548 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1549 "dd if=%s bs=1048576 count=%s",
1550 real_disk.dev_path, str(disk.size))
1552 # we set here a smaller block size as, due to ssh buffering, more
1553 # than 64-128k will mostly ignored; we use nocreat to fail if the
1554 # device is not already there or we pass a wrong path; we use
1555 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1556 # to not buffer too much memory; this means that at best, we flush
1557 # every 64k, which will not be very fast
1558 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1559 " oflag=dsync", dest_path)
1561 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1562 constants.GANETI_RUNAS,
1565 # all commands have been checked, so we're safe to combine them
1566 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1568 result = utils.RunCmd(["bash", "-c", command])
1571 _Fail("Disk copy command '%s' returned error: %s"
1572 " output: %s", command, result.fail_reason, result.output)
1575 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1576 """Write a file to the filesystem.
1578 This allows the master to overwrite(!) a file. It will only perform
1579 the operation if the file belongs to a list of configuration files.
1581 @type file_name: str
1582 @param file_name: the target file name
1584 @param data: the new contents of the file
1586 @param mode: the mode to give the file (can be None)
1588 @param uid: the owner of the file (can be -1 for default)
1590 @param gid: the group of the file (can be -1 for default)
1592 @param atime: the atime to set on the file (can be None)
1594 @param mtime: the mtime to set on the file (can be None)
1598 if not os.path.isabs(file_name):
1599 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1601 if file_name not in _ALLOWED_UPLOAD_FILES:
1602 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1605 raw_data = _Decompress(data)
1607 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1608 atime=atime, mtime=mtime)
1611 def WriteSsconfFiles(values):
1612 """Update all ssconf files.
1614 Wrapper around the SimpleStore.WriteFiles.
1617 ssconf.SimpleStore().WriteFiles(values)
1620 def _ErrnoOrStr(err):
1621 """Format an EnvironmentError exception.
1623 If the L{err} argument has an errno attribute, it will be looked up
1624 and converted into a textual C{E...} description. Otherwise the
1625 string representation of the error will be returned.
1627 @type err: L{EnvironmentError}
1628 @param err: the exception to format
1631 if hasattr(err, 'errno'):
1632 detail = errno.errorcode[err.errno]
1638 def _OSOndiskAPIVersion(name, os_dir):
1639 """Compute and return the API version of a given OS.
1641 This function will try to read the API version of the OS given by
1642 the 'name' parameter and residing in the 'os_dir' directory.
1645 @param name: the OS name we should look for
1647 @param os_dir: the directory inwhich we should look for the OS
1649 @return: tuple (status, data) with status denoting the validity and
1650 data holding either the vaid versions or an error message
1653 api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
1656 st = os.stat(api_file)
1657 except EnvironmentError, err:
1658 return False, ("Required file '%s' not found under path %s: %s" %
1659 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1661 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1662 return False, ("File '%s' in %s is not a regular file" %
1663 (constants.OS_API_FILE, os_dir))
1666 api_versions = utils.ReadFile(api_file).splitlines()
1667 except EnvironmentError, err:
1668 return False, ("Error while reading the API version file at %s: %s" %
1669 (api_file, _ErrnoOrStr(err)))
1672 api_versions = [int(version.strip()) for version in api_versions]
1673 except (TypeError, ValueError), err:
1674 return False, ("API version(s) can't be converted to integer: %s" %
1677 return True, api_versions
1680 def DiagnoseOS(top_dirs=None):
1681 """Compute the validity for all OSes.
1683 @type top_dirs: list
1684 @param top_dirs: the list of directories in which to
1685 search (if not given defaults to
1686 L{constants.OS_SEARCH_PATH})
1687 @rtype: list of L{objects.OS}
1688 @return: a list of tuples (name, path, status, diagnose, variants)
1689 for all (potential) OSes under all search paths, where:
1690 - name is the (potential) OS name
1691 - path is the full path to the OS
1692 - status True/False is the validity of the OS
1693 - diagnose is the error message for an invalid OS, otherwise empty
1694 - variants is a list of supported OS variants, if any
1697 if top_dirs is None:
1698 top_dirs = constants.OS_SEARCH_PATH
1701 for dir_name in top_dirs:
1702 if os.path.isdir(dir_name):
1704 f_names = utils.ListVisibleFiles(dir_name)
1705 except EnvironmentError, err:
1706 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1708 for name in f_names:
1709 os_path = os.path.sep.join([dir_name, name])
1710 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1713 variants = os_inst.supported_variants
1717 result.append((name, os_path, status, diagnose, variants))
1722 def _TryOSFromDisk(name, base_dir=None):
1723 """Create an OS instance from disk.
1725 This function will return an OS instance if the given name is a
1728 @type base_dir: string
1729 @keyword base_dir: Base directory containing OS installations.
1730 Defaults to a search in all the OS_SEARCH_PATH dirs.
1732 @return: success and either the OS instance if we find a valid one,
1736 if base_dir is None:
1737 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1739 return False, "Directory for OS %s not found in search path" % name
1741 os_dir = os.path.sep.join([base_dir, name])
1743 status, api_versions = _OSOndiskAPIVersion(name, os_dir)
1746 return status, api_versions
1748 if not constants.OS_API_VERSIONS.intersection(api_versions):
1749 return False, ("API version mismatch for path '%s': found %s, want %s." %
1750 (os_dir, api_versions, constants.OS_API_VERSIONS))
1752 # OS Files dictionary, we will populate it with the absolute path names
1753 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1755 if max(api_versions) >= constants.OS_API_V15:
1756 os_files[constants.OS_VARIANTS_FILE] = ''
1758 for filename in os_files:
1759 os_files[filename] = os.path.sep.join([os_dir, filename])
1762 st = os.stat(os_files[filename])
1763 except EnvironmentError, err:
1764 return False, ("File '%s' under path '%s' is missing (%s)" %
1765 (filename, os_dir, _ErrnoOrStr(err)))
1767 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1768 return False, ("File '%s' under path '%s' is not a regular file" %
1771 if filename in constants.OS_SCRIPTS:
1772 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1773 return False, ("File '%s' under path '%s' is not executable" %
1777 if constants.OS_VARIANTS_FILE in os_files:
1778 variants_file = os_files[constants.OS_VARIANTS_FILE]
1780 variants = utils.ReadFile(variants_file).splitlines()
1781 except EnvironmentError, err:
1782 return False, ("Error while reading the OS variants file at %s: %s" %
1783 (variants_file, _ErrnoOrStr(err)))
1785 return False, ("No supported os variant found")
1787 os_obj = objects.OS(name=name, path=os_dir,
1788 create_script=os_files[constants.OS_SCRIPT_CREATE],
1789 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1790 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1791 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1792 supported_variants=variants,
1793 api_versions=api_versions)
1797 def OSFromDisk(name, base_dir=None):
1798 """Create an OS instance from disk.
1800 This function will return an OS instance if the given name is a
1801 valid OS name. Otherwise, it will raise an appropriate
1802 L{RPCFail} exception, detailing why this is not a valid OS.
1804 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1805 an exception but returns true/false status data.
1807 @type base_dir: string
1808 @keyword base_dir: Base directory containing OS installations.
1809 Defaults to a search in all the OS_SEARCH_PATH dirs.
1810 @rtype: L{objects.OS}
1811 @return: the OS instance if we find a valid one
1812 @raise RPCFail: if we don't find a valid OS
1815 name_only = name.split("+", 1)[0]
1816 status, payload = _TryOSFromDisk(name_only, base_dir)
1824 def OSEnvironment(instance, inst_os, debug=0):
1825 """Calculate the environment for an os script.
1827 @type instance: L{objects.Instance}
1828 @param instance: target instance for the os script run
1829 @type inst_os: L{objects.OS}
1830 @param inst_os: operating system for which the environment is being built
1831 @type debug: integer
1832 @param debug: debug level (0 or 1, for OS Api 10)
1834 @return: dict of environment variables
1835 @raise errors.BlockDeviceError: if the block device
1841 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1842 result['OS_API_VERSION'] = '%d' % api_version
1843 result['INSTANCE_NAME'] = instance.name
1844 result['INSTANCE_OS'] = instance.os
1845 result['HYPERVISOR'] = instance.hypervisor
1846 result['DISK_COUNT'] = '%d' % len(instance.disks)
1847 result['NIC_COUNT'] = '%d' % len(instance.nics)
1848 result['DEBUG_LEVEL'] = '%d' % debug
1849 if api_version >= constants.OS_API_V15:
1851 variant = instance.os.split('+', 1)[1]
1853 variant = inst_os.supported_variants[0]
1854 result['OS_VARIANT'] = variant
1855 for idx, disk in enumerate(instance.disks):
1856 real_disk = _RecursiveFindBD(disk)
1857 if real_disk is None:
1858 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1861 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1862 result['DISK_%d_ACCESS' % idx] = disk.mode
1863 if constants.HV_DISK_TYPE in instance.hvparams:
1864 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1865 instance.hvparams[constants.HV_DISK_TYPE]
1866 if disk.dev_type in constants.LDS_BLOCK:
1867 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1868 elif disk.dev_type == constants.LD_FILE:
1869 result['DISK_%d_BACKEND_TYPE' % idx] = \
1870 'file:%s' % disk.physical_id[0]
1871 for idx, nic in enumerate(instance.nics):
1872 result['NIC_%d_MAC' % idx] = nic.mac
1874 result['NIC_%d_IP' % idx] = nic.ip
1875 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1876 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1877 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1878 if nic.nicparams[constants.NIC_LINK]:
1879 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1880 if constants.HV_NIC_TYPE in instance.hvparams:
1881 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1882 instance.hvparams[constants.HV_NIC_TYPE]
1884 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1885 for key, value in source.items():
1886 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1890 def BlockdevGrow(disk, amount):
1891 """Grow a stack of block devices.
1893 This function is called recursively, with the childrens being the
1894 first ones to resize.
1896 @type disk: L{objects.Disk}
1897 @param disk: the disk to be grown
1898 @rtype: (status, result)
1899 @return: a tuple with the status of the operation
1900 (True/False), and the errors message if status
1904 r_dev = _RecursiveFindBD(disk)
1906 _Fail("Cannot find block device %s", disk)
1910 except errors.BlockDeviceError, err:
1911 _Fail("Failed to grow block device: %s", err, exc=True)
1914 def BlockdevSnapshot(disk):
1915 """Create a snapshot copy of a block device.
1917 This function is called recursively, and the snapshot is actually created
1918 just for the leaf lvm backend device.
1920 @type disk: L{objects.Disk}
1921 @param disk: the disk to be snapshotted
1923 @return: snapshot disk path
1927 if len(disk.children) == 1:
1928 # only one child, let's recurse on it
1929 return BlockdevSnapshot(disk.children[0])
1931 # more than one child, choose one that matches
1932 for child in disk.children:
1933 if child.size == disk.size:
1934 # return implies breaking the loop
1935 return BlockdevSnapshot(child)
1936 elif disk.dev_type == constants.LD_LV:
1937 r_dev = _RecursiveFindBD(disk)
1938 if r_dev is not None:
1939 # let's stay on the safe side and ask for the full size, for now
1940 return r_dev.Snapshot(disk.size)
1942 _Fail("Cannot find block device %s", disk)
1944 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1945 disk.unique_id, disk.dev_type)
1948 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1949 """Export a block device snapshot to a remote node.
1951 @type disk: L{objects.Disk}
1952 @param disk: the description of the disk to export
1953 @type dest_node: str
1954 @param dest_node: the destination node to export to
1955 @type instance: L{objects.Instance}
1956 @param instance: the instance object to whom the disk belongs
1957 @type cluster_name: str
1958 @param cluster_name: the cluster name, needed for SSH hostalias
1960 @param idx: the index of the disk in the instance's disk list,
1961 used to export to the OS scripts environment
1965 inst_os = OSFromDisk(instance.os)
1966 export_env = OSEnvironment(instance, inst_os)
1968 export_script = inst_os.export_script
1970 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1971 instance.name, int(time.time()))
1972 if not os.path.exists(constants.LOG_OS_DIR):
1973 os.mkdir(constants.LOG_OS_DIR, 0750)
1974 real_disk = _RecursiveFindBD(disk)
1975 if real_disk is None:
1976 _Fail("Block device '%s' is not set up", disk)
1980 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1981 export_env['EXPORT_INDEX'] = str(idx)
1983 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1984 destfile = disk.physical_id[1]
1986 # the target command is built out of three individual commands,
1987 # which are joined by pipes; we check each individual command for
1989 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
1990 inst_os.path, export_script, logfile)
1994 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1995 destdir, destdir, destfile)
1996 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1997 constants.GANETI_RUNAS,
2000 # all commands have been checked, so we're safe to combine them
2001 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
2003 result = utils.RunCmd(["bash", "-c", command], env=export_env)
2006 _Fail("OS snapshot export command '%s' returned error: %s"
2007 " output: %s", command, result.fail_reason, result.output)
2010 def FinalizeExport(instance, snap_disks):
2011 """Write out the export configuration information.
2013 @type instance: L{objects.Instance}
2014 @param instance: the instance which we export, used for
2015 saving configuration
2016 @type snap_disks: list of L{objects.Disk}
2017 @param snap_disks: list of snapshot block devices, which
2018 will be used to get the actual name of the dump file
2023 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
2024 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
2026 config = objects.SerializableConfigParser()
2028 config.add_section(constants.INISECT_EXP)
2029 config.set(constants.INISECT_EXP, 'version', '0')
2030 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2031 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2032 config.set(constants.INISECT_EXP, 'os', instance.os)
2033 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2035 config.add_section(constants.INISECT_INS)
2036 config.set(constants.INISECT_INS, 'name', instance.name)
2037 config.set(constants.INISECT_INS, 'memory', '%d' %
2038 instance.beparams[constants.BE_MEMORY])
2039 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2040 instance.beparams[constants.BE_VCPUS])
2041 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2044 for nic_count, nic in enumerate(instance.nics):
2046 config.set(constants.INISECT_INS, 'nic%d_mac' %
2047 nic_count, '%s' % nic.mac)
2048 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2049 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
2051 # TODO: redundant: on load can read nics until it doesn't exist
2052 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2055 for disk_count, disk in enumerate(snap_disks):
2058 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2059 ('%s' % disk.iv_name))
2060 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2061 ('%s' % disk.physical_id[1]))
2062 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2065 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2067 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
2068 data=config.Dumps())
2069 shutil.rmtree(finaldestdir, True)
2070 shutil.move(destdir, finaldestdir)
2073 def ExportInfo(dest):
2074 """Get export configuration information.
2077 @param dest: directory containing the export
2079 @rtype: L{objects.SerializableConfigParser}
2080 @return: a serializable config file containing the
2084 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
2086 config = objects.SerializableConfigParser()
2089 if (not config.has_section(constants.INISECT_EXP) or
2090 not config.has_section(constants.INISECT_INS)):
2091 _Fail("Export info file doesn't have the required fields")
2093 return config.Dumps()
2096 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
2097 """Import an os image into an instance.
2099 @type instance: L{objects.Instance}
2100 @param instance: instance to import the disks into
2101 @type src_node: string
2102 @param src_node: source node for the disk images
2103 @type src_images: list of string
2104 @param src_images: absolute paths of the disk images
2105 @rtype: list of boolean
2106 @return: each boolean represent the success of importing the n-th disk
2109 inst_os = OSFromDisk(instance.os)
2110 import_env = OSEnvironment(instance, inst_os)
2111 import_script = inst_os.import_script
2113 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2114 instance.name, int(time.time()))
2115 if not os.path.exists(constants.LOG_OS_DIR):
2116 os.mkdir(constants.LOG_OS_DIR, 0750)
2119 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2120 import_script, logfile)
2123 for idx, image in enumerate(src_images):
2125 destcmd = utils.BuildShellCmd('cat %s', image)
2126 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2127 constants.GANETI_RUNAS,
2129 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2130 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2131 import_env['IMPORT_INDEX'] = str(idx)
2132 result = utils.RunCmd(command, env=import_env)
2134 logging.error("Disk import command '%s' returned error: %s"
2135 " output: %s", command, result.fail_reason,
2137 final_result.append("error importing disk %d: %s, %s" %
2138 (idx, result.fail_reason, result.output[-100]))
2141 _Fail("; ".join(final_result), log=False)
2145 """Return a list of exports currently available on this machine.
2148 @return: list of the exports
2151 if os.path.isdir(constants.EXPORT_DIR):
2152 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2154 _Fail("No exports directory")
2157 def RemoveExport(export):
2158 """Remove an existing export from the node.
2161 @param export: the name of the export to remove
2165 target = os.path.join(constants.EXPORT_DIR, export)
2168 shutil.rmtree(target)
2169 except EnvironmentError, err:
2170 _Fail("Error while removing the export: %s", err, exc=True)
2173 def BlockdevRename(devlist):
2174 """Rename a list of block devices.
2176 @type devlist: list of tuples
2177 @param devlist: list of tuples of the form (disk,
2178 new_logical_id, new_physical_id); disk is an
2179 L{objects.Disk} object describing the current disk,
2180 and new logical_id/physical_id is the name we
2183 @return: True if all renames succeeded, False otherwise
2188 for disk, unique_id in devlist:
2189 dev = _RecursiveFindBD(disk)
2191 msgs.append("Can't find device %s in rename" % str(disk))
2195 old_rpath = dev.dev_path
2196 dev.Rename(unique_id)
2197 new_rpath = dev.dev_path
2198 if old_rpath != new_rpath:
2199 DevCacheManager.RemoveCache(old_rpath)
2200 # FIXME: we should add the new cache information here, like:
2201 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2202 # but we don't have the owner here - maybe parse from existing
2203 # cache? for now, we only lose lvm data when we rename, which
2204 # is less critical than DRBD or MD
2205 except errors.BlockDeviceError, err:
2206 msgs.append("Can't rename device '%s' to '%s': %s" %
2207 (dev, unique_id, err))
2208 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2211 _Fail("; ".join(msgs))
2214 def _TransformFileStorageDir(file_storage_dir):
2215 """Checks whether given file_storage_dir is valid.
2217 Checks wheter the given file_storage_dir is within the cluster-wide
2218 default file_storage_dir stored in SimpleStore. Only paths under that
2219 directory are allowed.
2221 @type file_storage_dir: str
2222 @param file_storage_dir: the path to check
2224 @return: the normalized path if valid, None otherwise
2228 file_storage_dir = os.path.normpath(file_storage_dir)
2229 base_file_storage_dir = cfg.GetFileStorageDir()
2230 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2231 base_file_storage_dir):
2232 _Fail("File storage directory '%s' is not under base file"
2233 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2234 return file_storage_dir
2237 def CreateFileStorageDir(file_storage_dir):
2238 """Create file storage directory.
2240 @type file_storage_dir: str
2241 @param file_storage_dir: directory to create
2244 @return: tuple with first element a boolean indicating wheter dir
2245 creation was successful or not
2248 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2249 if os.path.exists(file_storage_dir):
2250 if not os.path.isdir(file_storage_dir):
2251 _Fail("Specified storage dir '%s' is not a directory",
2255 os.makedirs(file_storage_dir, 0750)
2256 except OSError, err:
2257 _Fail("Cannot create file storage directory '%s': %s",
2258 file_storage_dir, err, exc=True)
2261 def RemoveFileStorageDir(file_storage_dir):
2262 """Remove file storage directory.
2264 Remove it only if it's empty. If not log an error and return.
2266 @type file_storage_dir: str
2267 @param file_storage_dir: the directory we should cleanup
2268 @rtype: tuple (success,)
2269 @return: tuple of one element, C{success}, denoting
2270 whether the operation was successful
2273 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2274 if os.path.exists(file_storage_dir):
2275 if not os.path.isdir(file_storage_dir):
2276 _Fail("Specified Storage directory '%s' is not a directory",
2278 # deletes dir only if empty, otherwise we want to fail the rpc call
2280 os.rmdir(file_storage_dir)
2281 except OSError, err:
2282 _Fail("Cannot remove file storage directory '%s': %s",
2283 file_storage_dir, err)
2286 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2287 """Rename the file storage directory.
2289 @type old_file_storage_dir: str
2290 @param old_file_storage_dir: the current path
2291 @type new_file_storage_dir: str
2292 @param new_file_storage_dir: the name we should rename to
2293 @rtype: tuple (success,)
2294 @return: tuple of one element, C{success}, denoting
2295 whether the operation was successful
2298 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2299 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2300 if not os.path.exists(new_file_storage_dir):
2301 if os.path.isdir(old_file_storage_dir):
2303 os.rename(old_file_storage_dir, new_file_storage_dir)
2304 except OSError, err:
2305 _Fail("Cannot rename '%s' to '%s': %s",
2306 old_file_storage_dir, new_file_storage_dir, err)
2308 _Fail("Specified storage dir '%s' is not a directory",
2309 old_file_storage_dir)
2311 if os.path.exists(old_file_storage_dir):
2312 _Fail("Cannot rename '%s' to '%s': both locations exist",
2313 old_file_storage_dir, new_file_storage_dir)
2316 def _EnsureJobQueueFile(file_name):
2317 """Checks whether the given filename is in the queue directory.
2319 @type file_name: str
2320 @param file_name: the file name we should check
2322 @raises RPCFail: if the file is not valid
2325 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2326 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2329 _Fail("Passed job queue file '%s' does not belong to"
2330 " the queue directory '%s'", file_name, queue_dir)
2333 def JobQueueUpdate(file_name, content):
2334 """Updates a file in the queue directory.
2336 This is just a wrapper over L{utils.WriteFile}, with proper
2339 @type file_name: str
2340 @param file_name: the job file name
2342 @param content: the new job contents
2344 @return: the success of the operation
2347 _EnsureJobQueueFile(file_name)
2349 # Write and replace the file atomically
2350 utils.WriteFile(file_name, data=_Decompress(content))
2353 def JobQueueRename(old, new):
2354 """Renames a job queue file.
2356 This is just a wrapper over os.rename with proper checking.
2359 @param old: the old (actual) file name
2361 @param new: the desired file name
2363 @return: the success of the operation and payload
2366 _EnsureJobQueueFile(old)
2367 _EnsureJobQueueFile(new)
2369 utils.RenameFile(old, new, mkdir=True)
2372 def JobQueueSetDrainFlag(drain_flag):
2373 """Set the drain flag for the queue.
2375 This will set or unset the queue drain flag.
2377 @type drain_flag: boolean
2378 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2380 @return: always True, None
2381 @warning: the function always returns True
2385 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2387 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2390 def BlockdevClose(instance_name, disks):
2391 """Closes the given block devices.
2393 This means they will be switched to secondary mode (in case of
2396 @param instance_name: if the argument is not empty, the symlinks
2397 of this instance will be removed
2398 @type disks: list of L{objects.Disk}
2399 @param disks: the list of disks to be closed
2400 @rtype: tuple (success, message)
2401 @return: a tuple of success and message, where success
2402 indicates the succes of the operation, and message
2403 which will contain the error details in case we
2409 rd = _RecursiveFindBD(cf)
2411 _Fail("Can't find device %s", cf)
2418 except errors.BlockDeviceError, err:
2419 msg.append(str(err))
2421 _Fail("Can't make devices secondary: %s", ",".join(msg))
2424 _RemoveBlockDevLinks(instance_name, disks)
2427 def ValidateHVParams(hvname, hvparams):
2428 """Validates the given hypervisor parameters.
2430 @type hvname: string
2431 @param hvname: the hypervisor name
2432 @type hvparams: dict
2433 @param hvparams: the hypervisor parameters to be validated
2438 hv_type = hypervisor.GetHypervisor(hvname)
2439 hv_type.ValidateParameters(hvparams)
2440 except errors.HypervisorError, err:
2441 _Fail(str(err), log=False)
2445 """Demotes the current node from master candidate role.
2448 # try to ensure we're not the master by mistake
2449 master, myself = ssconf.GetMasterAndMyself()
2450 if master == myself:
2451 _Fail("ssconf status shows I'm the master node, will not demote")
2453 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2454 if not result.failed:
2455 _Fail("The master daemon is running, will not demote")
2458 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2459 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2460 except EnvironmentError, err:
2461 if err.errno != errno.ENOENT:
2462 _Fail("Error while backing up cluster file: %s", err, exc=True)
2464 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2467 def _FindDisks(nodes_ip, disks):
2468 """Sets the physical ID on disks and returns the block devices.
2471 # set the correct physical ID
2472 my_name = utils.HostInfo().name
2474 cf.SetPhysicalID(my_name, nodes_ip)
2479 rd = _RecursiveFindBD(cf)
2481 _Fail("Can't find device %s", cf)
2486 def DrbdDisconnectNet(nodes_ip, disks):
2487 """Disconnects the network on a list of drbd devices.
2490 bdevs = _FindDisks(nodes_ip, disks)
2496 except errors.BlockDeviceError, err:
2497 _Fail("Can't change network configuration to standalone mode: %s",
2501 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2502 """Attaches the network on a list of drbd devices.
2505 bdevs = _FindDisks(nodes_ip, disks)
2508 for idx, rd in enumerate(bdevs):
2510 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2511 except EnvironmentError, err:
2512 _Fail("Can't create symlink: %s", err)
2513 # reconnect disks, switch to new master configuration and if
2514 # needed primary mode
2517 rd.AttachNet(multimaster)
2518 except errors.BlockDeviceError, err:
2519 _Fail("Can't change network configuration: %s", err)
2521 # wait until the disks are connected; we need to retry the re-attach
2522 # if the device becomes standalone, as this might happen if the one
2523 # node disconnects and reconnects in a different mode before the
2524 # other node reconnects; in this case, one or both of the nodes will
2525 # decide it has wrong configuration and switch to standalone
2528 all_connected = True
2531 stats = rd.GetProcStatus()
2533 all_connected = (all_connected and
2534 (stats.is_connected or stats.is_in_resync))
2536 if stats.is_standalone:
2537 # peer had different config info and this node became
2538 # standalone, even though this should not happen with the
2539 # new staged way of changing disk configs
2541 rd.AttachNet(multimaster)
2542 except errors.BlockDeviceError, err:
2543 _Fail("Can't change network configuration: %s", err)
2545 if not all_connected:
2546 raise utils.RetryAgain()
2549 # Start with a delay of 100 miliseconds and go up to 5 seconds
2550 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2551 except utils.RetryTimeout:
2552 _Fail("Timeout in disk reconnecting")
2555 # change to primary mode
2559 except errors.BlockDeviceError, err:
2560 _Fail("Can't change to primary mode: %s", err)
2563 def DrbdWaitSync(nodes_ip, disks):
2564 """Wait until DRBDs have synchronized.
2568 stats = rd.GetProcStatus()
2569 if not (stats.is_connected or stats.is_in_resync):
2570 raise utils.RetryAgain()
2573 bdevs = _FindDisks(nodes_ip, disks)
2579 # poll each second for 15 seconds
2580 stats = utils.Retry(_helper, 1, 15, args=[rd])
2581 except utils.RetryTimeout:
2582 stats = rd.GetProcStatus()
2584 if not (stats.is_connected or stats.is_in_resync):
2585 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2586 alldone = alldone and (not stats.is_in_resync)
2587 if stats.sync_percent is not None:
2588 min_resync = min(min_resync, stats.sync_percent)
2590 return (alldone, min_resync)
2593 def PowercycleNode(hypervisor_type):
2594 """Hard-powercycle the node.
2596 Because we need to return first, and schedule the powercycle in the
2597 background, we won't be able to report failures nicely.
2600 hyper = hypervisor.GetHypervisor(hypervisor_type)
2604 # if we can't fork, we'll pretend that we're in the child process
2607 return "Reboot scheduled in 5 seconds"
2609 hyper.PowercycleNode()
2612 class HooksRunner(object):
2615 This class is instantiated on the node side (ganeti-noded) and not
2619 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2621 def __init__(self, hooks_base_dir=None):
2622 """Constructor for hooks runner.
2624 @type hooks_base_dir: str or None
2625 @param hooks_base_dir: if not None, this overrides the
2626 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2629 if hooks_base_dir is None:
2630 hooks_base_dir = constants.HOOKS_BASE_DIR
2631 self._BASE_DIR = hooks_base_dir
2634 def ExecHook(script, env):
2635 """Exec one hook script.
2638 @param script: the full path to the script
2640 @param env: the environment with which to exec the script
2641 @rtype: tuple (success, message)
2642 @return: a tuple of success and message, where success
2643 indicates the succes of the operation, and message
2644 which will contain the error details in case we
2648 # exec the process using subprocess and log the output
2651 fdstdin = open("/dev/null", "r")
2652 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2653 stderr=subprocess.STDOUT, close_fds=True,
2654 shell=False, cwd="/", env=env)
2657 output = child.stdout.read(4096)
2658 child.stdout.close()
2659 except EnvironmentError, err:
2660 output += "Hook script error: %s" % str(err)
2664 result = child.wait()
2666 except EnvironmentError, err:
2667 if err.errno == errno.EINTR:
2671 # try not to leak fds
2672 for fd in (fdstdin, ):
2676 except EnvironmentError, err:
2677 # just log the error
2678 #logging.exception("Error while closing fd %s", fd)
2681 return result == 0, utils.SafeEncode(output.strip())
2683 def RunHooks(self, hpath, phase, env):
2684 """Run the scripts in the hooks directory.
2687 @param hpath: the path to the hooks directory which
2690 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2691 L{constants.HOOKS_PHASE_POST}
2693 @param env: dictionary with the environment for the hook
2695 @return: list of 3-element tuples:
2697 - script result, either L{constants.HKR_SUCCESS} or
2698 L{constants.HKR_FAIL}
2699 - output of the script
2701 @raise errors.ProgrammerError: for invalid input
2705 if phase == constants.HOOKS_PHASE_PRE:
2707 elif phase == constants.HOOKS_PHASE_POST:
2710 _Fail("Unknown hooks phase '%s'", phase)
2714 subdir = "%s-%s.d" % (hpath, suffix)
2715 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2717 dir_contents = utils.ListVisibleFiles(dir_name)
2719 # FIXME: must log output in case of failures
2722 # we use the standard python sort order,
2723 # so 00name is the recommended naming scheme
2725 for relname in dir_contents:
2726 fname = os.path.join(dir_name, relname)
2727 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2728 self.RE_MASK.match(relname) is not None):
2729 rrval = constants.HKR_SKIP
2732 result, output = self.ExecHook(fname, env)
2734 rrval = constants.HKR_FAIL
2736 rrval = constants.HKR_SUCCESS
2737 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2742 class IAllocatorRunner(object):
2743 """IAllocator runner.
2745 This class is instantiated on the node side (ganeti-noded) and not on
2749 def Run(self, name, idata):
2750 """Run an iallocator script.
2753 @param name: the iallocator script name
2755 @param idata: the allocator input data
2758 @return: two element tuple of:
2760 - either error message or stdout of allocator (for success)
2763 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2765 if alloc_script is None:
2766 _Fail("iallocator module '%s' not found in the search path", name)
2768 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2772 result = utils.RunCmd([alloc_script, fin_name])
2774 _Fail("iallocator module '%s' failed: %s, output '%s'",
2775 name, result.fail_reason, result.output)
2779 return result.stdout
2782 class DevCacheManager(object):
2783 """Simple class for managing a cache of block device information.
2786 _DEV_PREFIX = "/dev/"
2787 _ROOT_DIR = constants.BDEV_CACHE_DIR
2790 def _ConvertPath(cls, dev_path):
2791 """Converts a /dev/name path to the cache file name.
2793 This replaces slashes with underscores and strips the /dev
2794 prefix. It then returns the full path to the cache file.
2797 @param dev_path: the C{/dev/} path name
2799 @return: the converted path name
2802 if dev_path.startswith(cls._DEV_PREFIX):
2803 dev_path = dev_path[len(cls._DEV_PREFIX):]
2804 dev_path = dev_path.replace("/", "_")
2805 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2809 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2810 """Updates the cache information for a given device.
2813 @param dev_path: the pathname of the device
2815 @param owner: the owner (instance name) of the device
2816 @type on_primary: bool
2817 @param on_primary: whether this is the primary
2820 @param iv_name: the instance-visible name of the
2821 device, as in objects.Disk.iv_name
2826 if dev_path is None:
2827 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2829 fpath = cls._ConvertPath(dev_path)
2835 iv_name = "not_visible"
2836 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2838 utils.WriteFile(fpath, data=fdata)
2839 except EnvironmentError, err:
2840 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2843 def RemoveCache(cls, dev_path):
2844 """Remove data for a dev_path.
2846 This is just a wrapper over L{utils.RemoveFile} with a converted
2847 path name and logging.
2850 @param dev_path: the pathname of the device
2855 if dev_path is None:
2856 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2858 fpath = cls._ConvertPath(dev_path)
2860 utils.RemoveFile(fpath)
2861 except EnvironmentError, err:
2862 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)