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, debug):
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
800 @param debug: debug level, passed to the OS scripts
804 inst_os = OSFromDisk(instance.os)
806 create_env = OSEnvironment(instance, inst_os, debug)
808 create_env['INSTANCE_REINSTALL'] = "1"
810 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
811 instance.name, int(time.time()))
813 result = utils.RunCmd([inst_os.create_script], env=create_env,
814 cwd=inst_os.path, output=logfile,)
816 logging.error("os create command '%s' returned error: %s, logfile: %s,"
817 " output: %s", result.cmd, result.fail_reason, logfile,
819 lines = [utils.SafeEncode(val)
820 for val in utils.TailFile(logfile, lines=20)]
821 _Fail("OS create script failed (%s), last lines in the"
822 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
825 def RunRenameInstance(instance, old_name, debug):
826 """Run the OS rename script for an instance.
828 @type instance: L{objects.Instance}
829 @param instance: Instance whose OS is to be installed
830 @type old_name: string
831 @param old_name: previous instance name
833 @param debug: debug level, passed to the OS scripts
835 @return: the success of the operation
838 inst_os = OSFromDisk(instance.os)
840 rename_env = OSEnvironment(instance, inst_os, debug)
841 rename_env['OLD_INSTANCE_NAME'] = old_name
843 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
845 instance.name, int(time.time()))
847 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
848 cwd=inst_os.path, output=logfile)
851 logging.error("os create command '%s' returned error: %s output: %s",
852 result.cmd, result.fail_reason, result.output)
853 lines = [utils.SafeEncode(val)
854 for val in utils.TailFile(logfile, lines=20)]
855 _Fail("OS rename script failed (%s), last lines in the"
856 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
859 def _GetVGInfo(vg_name):
860 """Get information about the volume group.
863 @param vg_name: the volume group which we query
866 A dictionary with the following keys:
867 - C{vg_size} is the total size of the volume group in MiB
868 - C{vg_free} is the free size of the volume group in MiB
869 - C{pv_count} are the number of physical disks in that VG
871 If an error occurs during gathering of data, we return the same dict
872 with keys all set to None.
875 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
877 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
878 "--nosuffix", "--units=m", "--separator=:", vg_name])
881 logging.error("volume group %s not present", vg_name)
883 valarr = retval.stdout.strip().rstrip(':').split(':')
887 "vg_size": int(round(float(valarr[0]), 0)),
888 "vg_free": int(round(float(valarr[1]), 0)),
889 "pv_count": int(valarr[2]),
891 except (TypeError, ValueError), err:
892 logging.exception("Fail to parse vgs output: %s", err)
894 logging.error("vgs output has the wrong number of fields (expected"
895 " three): %s", str(valarr))
899 def _GetBlockDevSymlinkPath(instance_name, idx):
900 return os.path.join(constants.DISK_LINKS_DIR,
901 "%s:%d" % (instance_name, idx))
904 def _SymlinkBlockDev(instance_name, device_path, idx):
905 """Set up symlinks to a instance's block device.
907 This is an auxiliary function run when an instance is start (on the primary
908 node) or when an instance is migrated (on the target node).
911 @param instance_name: the name of the target instance
912 @param device_path: path of the physical block device, on the node
913 @param idx: the disk index
914 @return: absolute path to the disk's symlink
917 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
919 os.symlink(device_path, link_name)
921 if err.errno == errno.EEXIST:
922 if (not os.path.islink(link_name) or
923 os.readlink(link_name) != device_path):
925 os.symlink(device_path, link_name)
932 def _RemoveBlockDevLinks(instance_name, disks):
933 """Remove the block device symlinks belonging to the given instance.
936 for idx, _ in enumerate(disks):
937 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
938 if os.path.islink(link_name):
942 logging.exception("Can't remove symlink '%s'", link_name)
945 def _GatherAndLinkBlockDevs(instance):
946 """Set up an instance's block device(s).
948 This is run on the primary node at instance startup. The block
949 devices must be already assembled.
951 @type instance: L{objects.Instance}
952 @param instance: the instance whose disks we shoul assemble
954 @return: list of (disk_object, device_path)
958 for idx, disk in enumerate(instance.disks):
959 device = _RecursiveFindBD(disk)
961 raise errors.BlockDeviceError("Block device '%s' is not set up." %
965 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
967 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
970 block_devices.append((disk, link_name))
975 def StartInstance(instance):
976 """Start an instance.
978 @type instance: L{objects.Instance}
979 @param instance: the instance object
983 running_instances = GetInstanceList([instance.hypervisor])
985 if instance.name in running_instances:
986 logging.info("Instance %s already running, not starting", instance.name)
990 block_devices = _GatherAndLinkBlockDevs(instance)
991 hyper = hypervisor.GetHypervisor(instance.hypervisor)
992 hyper.StartInstance(instance, block_devices)
993 except errors.BlockDeviceError, err:
994 _Fail("Block device error: %s", err, exc=True)
995 except errors.HypervisorError, err:
996 _RemoveBlockDevLinks(instance.name, instance.disks)
997 _Fail("Hypervisor error: %s", err, exc=True)
1000 def InstanceShutdown(instance, timeout):
1001 """Shut an instance down.
1003 @note: this functions uses polling with a hardcoded timeout.
1005 @type instance: L{objects.Instance}
1006 @param instance: the instance object
1007 @type timeout: integer
1008 @param timeout: maximum timeout for soft shutdown
1012 hv_name = instance.hypervisor
1013 hyper = hypervisor.GetHypervisor(hv_name)
1014 iname = instance.name
1016 if instance.name not in hyper.ListInstances():
1017 logging.info("Instance %s not running, doing nothing", iname)
1022 self.tried_once = False
1025 if iname not in hyper.ListInstances():
1029 hyper.StopInstance(instance, retry=self.tried_once)
1030 except errors.HypervisorError, err:
1031 if iname not in hyper.ListInstances():
1032 # if the instance is no longer existing, consider this a
1033 # success and go to cleanup
1036 _Fail("Failed to stop instance %s: %s", iname, err)
1038 self.tried_once = True
1040 raise utils.RetryAgain()
1043 utils.Retry(_TryShutdown(), 5, timeout)
1044 except utils.RetryTimeout:
1045 # the shutdown did not succeed
1046 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1049 hyper.StopInstance(instance, force=True)
1050 except errors.HypervisorError, err:
1051 if iname in hyper.ListInstances():
1052 # only raise an error if the instance still exists, otherwise
1053 # the error could simply be "instance ... unknown"!
1054 _Fail("Failed to force stop instance %s: %s", iname, err)
1058 if iname in hyper.ListInstances():
1059 _Fail("Could not shutdown instance %s even by destroy", iname)
1061 _RemoveBlockDevLinks(iname, instance.disks)
1064 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1065 """Reboot an instance.
1067 @type instance: L{objects.Instance}
1068 @param instance: the instance object to reboot
1069 @type reboot_type: str
1070 @param reboot_type: the type of reboot, one the following
1072 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1073 instance OS, do not recreate the VM
1074 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1075 restart the VM (at the hypervisor level)
1076 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1077 not accepted here, since that mode is handled differently, in
1078 cmdlib, and translates into full stop and start of the
1079 instance (instead of a call_instance_reboot RPC)
1080 @type shutdown_timeout: integer
1081 @param shutdown_timeout: maximum timeout for soft shutdown
1085 running_instances = GetInstanceList([instance.hypervisor])
1087 if instance.name not in running_instances:
1088 _Fail("Cannot reboot instance %s that is not running", instance.name)
1090 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1091 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1093 hyper.RebootInstance(instance)
1094 except errors.HypervisorError, err:
1095 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1096 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1098 InstanceShutdown(instance, shutdown_timeout)
1099 return StartInstance(instance)
1100 except errors.HypervisorError, err:
1101 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1103 _Fail("Invalid reboot_type received: %s", reboot_type)
1106 def MigrationInfo(instance):
1107 """Gather information about an instance to be migrated.
1109 @type instance: L{objects.Instance}
1110 @param instance: the instance definition
1113 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1115 info = hyper.MigrationInfo(instance)
1116 except errors.HypervisorError, err:
1117 _Fail("Failed to fetch migration information: %s", err, exc=True)
1121 def AcceptInstance(instance, info, target):
1122 """Prepare the node to accept an instance.
1124 @type instance: L{objects.Instance}
1125 @param instance: the instance definition
1126 @type info: string/data (opaque)
1127 @param info: migration information, from the source node
1128 @type target: string
1129 @param target: target host (usually ip), on this node
1132 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1134 hyper.AcceptInstance(instance, info, target)
1135 except errors.HypervisorError, err:
1136 _Fail("Failed to accept instance: %s", err, exc=True)
1139 def FinalizeMigration(instance, info, success):
1140 """Finalize any preparation to accept an instance.
1142 @type instance: L{objects.Instance}
1143 @param instance: the instance definition
1144 @type info: string/data (opaque)
1145 @param info: migration information, from the source node
1146 @type success: boolean
1147 @param success: whether the migration was a success or a failure
1150 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1152 hyper.FinalizeMigration(instance, info, success)
1153 except errors.HypervisorError, err:
1154 _Fail("Failed to finalize migration: %s", err, exc=True)
1157 def MigrateInstance(instance, target, live):
1158 """Migrates an instance to another node.
1160 @type instance: L{objects.Instance}
1161 @param instance: the instance definition
1162 @type target: string
1163 @param target: the target node name
1165 @param live: whether the migration should be done live or not (the
1166 interpretation of this parameter is left to the hypervisor)
1168 @return: a tuple of (success, msg) where:
1169 - succes is a boolean denoting the success/failure of the operation
1170 - msg is a string with details in case of failure
1173 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1176 hyper.MigrateInstance(instance, target, live)
1177 except errors.HypervisorError, err:
1178 _Fail("Failed to migrate instance: %s", err, exc=True)
1181 def BlockdevCreate(disk, size, owner, on_primary, info):
1182 """Creates a block device for an instance.
1184 @type disk: L{objects.Disk}
1185 @param disk: the object describing the disk we should create
1187 @param size: the size of the physical underlying device, in MiB
1189 @param owner: the name of the instance for which disk is created,
1190 used for device cache data
1191 @type on_primary: boolean
1192 @param on_primary: indicates if it is the primary node or not
1194 @param info: string that will be sent to the physical device
1195 creation, used for example to set (LVM) tags on LVs
1197 @return: the new unique_id of the device (this can sometime be
1198 computed only after creation), or None. On secondary nodes,
1199 it's not required to return anything.
1202 # TODO: remove the obsolete 'size' argument
1203 # pylint: disable-msg=W0613
1206 for child in disk.children:
1208 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1209 except errors.BlockDeviceError, err:
1210 _Fail("Can't assemble device %s: %s", child, err)
1211 if on_primary or disk.AssembleOnSecondary():
1212 # we need the children open in case the device itself has to
1215 # pylint: disable-msg=E1103
1217 except errors.BlockDeviceError, err:
1218 _Fail("Can't make child '%s' read-write: %s", child, err)
1222 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1223 except errors.BlockDeviceError, err:
1224 _Fail("Can't create block device: %s", err)
1226 if on_primary or disk.AssembleOnSecondary():
1229 except errors.BlockDeviceError, err:
1230 _Fail("Can't assemble device after creation, unusual event: %s", err)
1231 device.SetSyncSpeed(constants.SYNC_SPEED)
1232 if on_primary or disk.OpenOnSecondary():
1234 device.Open(force=True)
1235 except errors.BlockDeviceError, err:
1236 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1237 DevCacheManager.UpdateCache(device.dev_path, owner,
1238 on_primary, disk.iv_name)
1240 device.SetInfo(info)
1242 return device.unique_id
1245 def BlockdevRemove(disk):
1246 """Remove a block device.
1248 @note: This is intended to be called recursively.
1250 @type disk: L{objects.Disk}
1251 @param disk: the disk object we should remove
1253 @return: the success of the operation
1258 rdev = _RecursiveFindBD(disk)
1259 except errors.BlockDeviceError, err:
1260 # probably can't attach
1261 logging.info("Can't attach to device %s in remove", disk)
1263 if rdev is not None:
1264 r_path = rdev.dev_path
1267 except errors.BlockDeviceError, err:
1268 msgs.append(str(err))
1270 DevCacheManager.RemoveCache(r_path)
1273 for child in disk.children:
1275 BlockdevRemove(child)
1276 except RPCFail, err:
1277 msgs.append(str(err))
1280 _Fail("; ".join(msgs))
1283 def _RecursiveAssembleBD(disk, owner, as_primary):
1284 """Activate a block device for an instance.
1286 This is run on the primary and secondary nodes for an instance.
1288 @note: this function is called recursively.
1290 @type disk: L{objects.Disk}
1291 @param disk: the disk we try to assemble
1293 @param owner: the name of the instance which owns the disk
1294 @type as_primary: boolean
1295 @param as_primary: if we should make the block device
1298 @return: the assembled device or None (in case no device
1300 @raise errors.BlockDeviceError: in case there is an error
1301 during the activation of the children or the device
1307 mcn = disk.ChildrenNeeded()
1309 mcn = 0 # max number of Nones allowed
1311 mcn = len(disk.children) - mcn # max number of Nones
1312 for chld_disk in disk.children:
1314 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1315 except errors.BlockDeviceError, err:
1316 if children.count(None) >= mcn:
1319 logging.error("Error in child activation (but continuing): %s",
1321 children.append(cdev)
1323 if as_primary or disk.AssembleOnSecondary():
1324 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1325 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1327 if as_primary or disk.OpenOnSecondary():
1329 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1330 as_primary, disk.iv_name)
1337 def BlockdevAssemble(disk, owner, as_primary):
1338 """Activate a block device for an instance.
1340 This is a wrapper over _RecursiveAssembleBD.
1342 @rtype: str or boolean
1343 @return: a C{/dev/...} path for primary nodes, and
1344 C{True} for secondary nodes
1348 result = _RecursiveAssembleBD(disk, owner, as_primary)
1349 if isinstance(result, bdev.BlockDev):
1350 # pylint: disable-msg=E1103
1351 result = result.dev_path
1352 except errors.BlockDeviceError, err:
1353 _Fail("Error while assembling disk: %s", err, exc=True)
1358 def BlockdevShutdown(disk):
1359 """Shut down a block device.
1361 First, if the device is assembled (Attach() is successful), then
1362 the device is shutdown. Then the children of the device are
1365 This function is called recursively. Note that we don't cache the
1366 children or such, as oppossed to assemble, shutdown of different
1367 devices doesn't require that the upper device was active.
1369 @type disk: L{objects.Disk}
1370 @param disk: the description of the disk we should
1376 r_dev = _RecursiveFindBD(disk)
1377 if r_dev is not None:
1378 r_path = r_dev.dev_path
1381 DevCacheManager.RemoveCache(r_path)
1382 except errors.BlockDeviceError, err:
1383 msgs.append(str(err))
1386 for child in disk.children:
1388 BlockdevShutdown(child)
1389 except RPCFail, err:
1390 msgs.append(str(err))
1393 _Fail("; ".join(msgs))
1396 def BlockdevAddchildren(parent_cdev, new_cdevs):
1397 """Extend a mirrored block device.
1399 @type parent_cdev: L{objects.Disk}
1400 @param parent_cdev: the disk to which we should add children
1401 @type new_cdevs: list of L{objects.Disk}
1402 @param new_cdevs: the list of children which we should add
1406 parent_bdev = _RecursiveFindBD(parent_cdev)
1407 if parent_bdev is None:
1408 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1409 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1410 if new_bdevs.count(None) > 0:
1411 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1412 parent_bdev.AddChildren(new_bdevs)
1415 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1416 """Shrink a mirrored block device.
1418 @type parent_cdev: L{objects.Disk}
1419 @param parent_cdev: the disk from which we should remove children
1420 @type new_cdevs: list of L{objects.Disk}
1421 @param new_cdevs: the list of children which we should remove
1425 parent_bdev = _RecursiveFindBD(parent_cdev)
1426 if parent_bdev is None:
1427 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1429 for disk in new_cdevs:
1430 rpath = disk.StaticDevPath()
1432 bd = _RecursiveFindBD(disk)
1434 _Fail("Can't find device %s while removing children", disk)
1436 devs.append(bd.dev_path)
1439 parent_bdev.RemoveChildren(devs)
1442 def BlockdevGetmirrorstatus(disks):
1443 """Get the mirroring status of a list of devices.
1445 @type disks: list of L{objects.Disk}
1446 @param disks: the list of disks which we should query
1449 a list of (mirror_done, estimated_time) tuples, which
1450 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1451 @raise errors.BlockDeviceError: if any of the disks cannot be
1457 rbd = _RecursiveFindBD(dsk)
1459 _Fail("Can't find device %s", dsk)
1461 stats.append(rbd.CombinedSyncStatus())
1466 def _RecursiveFindBD(disk):
1467 """Check if a device is activated.
1469 If so, return information about the real device.
1471 @type disk: L{objects.Disk}
1472 @param disk: the disk object we need to find
1474 @return: None if the device can't be found,
1475 otherwise the device instance
1480 for chdisk in disk.children:
1481 children.append(_RecursiveFindBD(chdisk))
1483 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1486 def BlockdevFind(disk):
1487 """Check if a device is activated.
1489 If it is, return information about the real device.
1491 @type disk: L{objects.Disk}
1492 @param disk: the disk to find
1493 @rtype: None or objects.BlockDevStatus
1494 @return: None if the disk cannot be found, otherwise a the current
1499 rbd = _RecursiveFindBD(disk)
1500 except errors.BlockDeviceError, err:
1501 _Fail("Failed to find device: %s", err, exc=True)
1506 return rbd.GetSyncStatus()
1509 def BlockdevGetsize(disks):
1510 """Computes the size of the given disks.
1512 If a disk is not found, returns None instead.
1514 @type disks: list of L{objects.Disk}
1515 @param disks: the list of disk to compute the size for
1517 @return: list with elements None if the disk cannot be found,
1524 rbd = _RecursiveFindBD(cf)
1525 except errors.BlockDeviceError:
1531 result.append(rbd.GetActualSize())
1535 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1536 """Export a block device to a remote node.
1538 @type disk: L{objects.Disk}
1539 @param disk: the description of the disk to export
1540 @type dest_node: str
1541 @param dest_node: the destination node to export to
1542 @type dest_path: str
1543 @param dest_path: the destination path on the target node
1544 @type cluster_name: str
1545 @param cluster_name: the cluster name, needed for SSH hostalias
1549 real_disk = _RecursiveFindBD(disk)
1550 if real_disk is None:
1551 _Fail("Block device '%s' is not set up", disk)
1555 # the block size on the read dd is 1MiB to match our units
1556 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1557 "dd if=%s bs=1048576 count=%s",
1558 real_disk.dev_path, str(disk.size))
1560 # we set here a smaller block size as, due to ssh buffering, more
1561 # than 64-128k will mostly ignored; we use nocreat to fail if the
1562 # device is not already there or we pass a wrong path; we use
1563 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1564 # to not buffer too much memory; this means that at best, we flush
1565 # every 64k, which will not be very fast
1566 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1567 " oflag=dsync", dest_path)
1569 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1570 constants.GANETI_RUNAS,
1573 # all commands have been checked, so we're safe to combine them
1574 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1576 result = utils.RunCmd(["bash", "-c", command])
1579 _Fail("Disk copy command '%s' returned error: %s"
1580 " output: %s", command, result.fail_reason, result.output)
1583 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1584 """Write a file to the filesystem.
1586 This allows the master to overwrite(!) a file. It will only perform
1587 the operation if the file belongs to a list of configuration files.
1589 @type file_name: str
1590 @param file_name: the target file name
1592 @param data: the new contents of the file
1594 @param mode: the mode to give the file (can be None)
1596 @param uid: the owner of the file (can be -1 for default)
1598 @param gid: the group of the file (can be -1 for default)
1600 @param atime: the atime to set on the file (can be None)
1602 @param mtime: the mtime to set on the file (can be None)
1606 if not os.path.isabs(file_name):
1607 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1609 if file_name not in _ALLOWED_UPLOAD_FILES:
1610 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1613 raw_data = _Decompress(data)
1615 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1616 atime=atime, mtime=mtime)
1619 def WriteSsconfFiles(values):
1620 """Update all ssconf files.
1622 Wrapper around the SimpleStore.WriteFiles.
1625 ssconf.SimpleStore().WriteFiles(values)
1628 def _ErrnoOrStr(err):
1629 """Format an EnvironmentError exception.
1631 If the L{err} argument has an errno attribute, it will be looked up
1632 and converted into a textual C{E...} description. Otherwise the
1633 string representation of the error will be returned.
1635 @type err: L{EnvironmentError}
1636 @param err: the exception to format
1639 if hasattr(err, 'errno'):
1640 detail = errno.errorcode[err.errno]
1646 def _OSOndiskAPIVersion(os_dir):
1647 """Compute and return the API version of a given OS.
1649 This function will try to read the API version of the OS residing in
1650 the 'os_dir' directory.
1653 @param os_dir: the directory in which we should look for the OS
1655 @return: tuple (status, data) with status denoting the validity and
1656 data holding either the vaid versions or an error message
1659 api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
1662 st = os.stat(api_file)
1663 except EnvironmentError, err:
1664 return False, ("Required file '%s' not found under path %s: %s" %
1665 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1667 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1668 return False, ("File '%s' in %s is not a regular file" %
1669 (constants.OS_API_FILE, os_dir))
1672 api_versions = utils.ReadFile(api_file).splitlines()
1673 except EnvironmentError, err:
1674 return False, ("Error while reading the API version file at %s: %s" %
1675 (api_file, _ErrnoOrStr(err)))
1678 api_versions = [int(version.strip()) for version in api_versions]
1679 except (TypeError, ValueError), err:
1680 return False, ("API version(s) can't be converted to integer: %s" %
1683 return True, api_versions
1686 def DiagnoseOS(top_dirs=None):
1687 """Compute the validity for all OSes.
1689 @type top_dirs: list
1690 @param top_dirs: the list of directories in which to
1691 search (if not given defaults to
1692 L{constants.OS_SEARCH_PATH})
1693 @rtype: list of L{objects.OS}
1694 @return: a list of tuples (name, path, status, diagnose, variants)
1695 for all (potential) OSes under all search paths, where:
1696 - name is the (potential) OS name
1697 - path is the full path to the OS
1698 - status True/False is the validity of the OS
1699 - diagnose is the error message for an invalid OS, otherwise empty
1700 - variants is a list of supported OS variants, if any
1703 if top_dirs is None:
1704 top_dirs = constants.OS_SEARCH_PATH
1707 for dir_name in top_dirs:
1708 if os.path.isdir(dir_name):
1710 f_names = utils.ListVisibleFiles(dir_name)
1711 except EnvironmentError, err:
1712 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1714 for name in f_names:
1715 os_path = os.path.sep.join([dir_name, name])
1716 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1719 variants = os_inst.supported_variants
1723 result.append((name, os_path, status, diagnose, variants))
1728 def _TryOSFromDisk(name, base_dir=None):
1729 """Create an OS instance from disk.
1731 This function will return an OS instance if the given name is a
1734 @type base_dir: string
1735 @keyword base_dir: Base directory containing OS installations.
1736 Defaults to a search in all the OS_SEARCH_PATH dirs.
1738 @return: success and either the OS instance if we find a valid one,
1742 if base_dir is None:
1743 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1745 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1748 return False, "Directory for OS %s not found in search path" % name
1750 status, api_versions = _OSOndiskAPIVersion(os_dir)
1753 return status, api_versions
1755 if not constants.OS_API_VERSIONS.intersection(api_versions):
1756 return False, ("API version mismatch for path '%s': found %s, want %s." %
1757 (os_dir, api_versions, constants.OS_API_VERSIONS))
1759 # OS Files dictionary, we will populate it with the absolute path names
1760 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1762 if max(api_versions) >= constants.OS_API_V15:
1763 os_files[constants.OS_VARIANTS_FILE] = ''
1765 for filename in os_files:
1766 os_files[filename] = os.path.sep.join([os_dir, filename])
1769 st = os.stat(os_files[filename])
1770 except EnvironmentError, err:
1771 return False, ("File '%s' under path '%s' is missing (%s)" %
1772 (filename, os_dir, _ErrnoOrStr(err)))
1774 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1775 return False, ("File '%s' under path '%s' is not a regular file" %
1778 if filename in constants.OS_SCRIPTS:
1779 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1780 return False, ("File '%s' under path '%s' is not executable" %
1784 if constants.OS_VARIANTS_FILE in os_files:
1785 variants_file = os_files[constants.OS_VARIANTS_FILE]
1787 variants = utils.ReadFile(variants_file).splitlines()
1788 except EnvironmentError, err:
1789 return False, ("Error while reading the OS variants file at %s: %s" %
1790 (variants_file, _ErrnoOrStr(err)))
1792 return False, ("No supported os variant found")
1794 os_obj = objects.OS(name=name, path=os_dir,
1795 create_script=os_files[constants.OS_SCRIPT_CREATE],
1796 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1797 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1798 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1799 supported_variants=variants,
1800 api_versions=api_versions)
1804 def OSFromDisk(name, base_dir=None):
1805 """Create an OS instance from disk.
1807 This function will return an OS instance if the given name is a
1808 valid OS name. Otherwise, it will raise an appropriate
1809 L{RPCFail} exception, detailing why this is not a valid OS.
1811 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1812 an exception but returns true/false status data.
1814 @type base_dir: string
1815 @keyword base_dir: Base directory containing OS installations.
1816 Defaults to a search in all the OS_SEARCH_PATH dirs.
1817 @rtype: L{objects.OS}
1818 @return: the OS instance if we find a valid one
1819 @raise RPCFail: if we don't find a valid OS
1822 name_only = name.split("+", 1)[0]
1823 status, payload = _TryOSFromDisk(name_only, base_dir)
1831 def OSEnvironment(instance, inst_os, debug=0):
1832 """Calculate the environment for an os script.
1834 @type instance: L{objects.Instance}
1835 @param instance: target instance for the os script run
1836 @type inst_os: L{objects.OS}
1837 @param inst_os: operating system for which the environment is being built
1838 @type debug: integer
1839 @param debug: debug level (0 or 1, for OS Api 10)
1841 @return: dict of environment variables
1842 @raise errors.BlockDeviceError: if the block device
1848 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1849 result['OS_API_VERSION'] = '%d' % api_version
1850 result['INSTANCE_NAME'] = instance.name
1851 result['INSTANCE_OS'] = instance.os
1852 result['HYPERVISOR'] = instance.hypervisor
1853 result['DISK_COUNT'] = '%d' % len(instance.disks)
1854 result['NIC_COUNT'] = '%d' % len(instance.nics)
1855 result['DEBUG_LEVEL'] = '%d' % debug
1856 if api_version >= constants.OS_API_V15:
1858 variant = instance.os.split('+', 1)[1]
1860 variant = inst_os.supported_variants[0]
1861 result['OS_VARIANT'] = variant
1862 for idx, disk in enumerate(instance.disks):
1863 real_disk = _RecursiveFindBD(disk)
1864 if real_disk is None:
1865 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1868 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1869 result['DISK_%d_ACCESS' % idx] = disk.mode
1870 if constants.HV_DISK_TYPE in instance.hvparams:
1871 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1872 instance.hvparams[constants.HV_DISK_TYPE]
1873 if disk.dev_type in constants.LDS_BLOCK:
1874 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1875 elif disk.dev_type == constants.LD_FILE:
1876 result['DISK_%d_BACKEND_TYPE' % idx] = \
1877 'file:%s' % disk.physical_id[0]
1878 for idx, nic in enumerate(instance.nics):
1879 result['NIC_%d_MAC' % idx] = nic.mac
1881 result['NIC_%d_IP' % idx] = nic.ip
1882 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1883 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1884 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1885 if nic.nicparams[constants.NIC_LINK]:
1886 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1887 if constants.HV_NIC_TYPE in instance.hvparams:
1888 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1889 instance.hvparams[constants.HV_NIC_TYPE]
1891 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1892 for key, value in source.items():
1893 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1897 def BlockdevGrow(disk, amount):
1898 """Grow a stack of block devices.
1900 This function is called recursively, with the childrens being the
1901 first ones to resize.
1903 @type disk: L{objects.Disk}
1904 @param disk: the disk to be grown
1905 @rtype: (status, result)
1906 @return: a tuple with the status of the operation
1907 (True/False), and the errors message if status
1911 r_dev = _RecursiveFindBD(disk)
1913 _Fail("Cannot find block device %s", disk)
1917 except errors.BlockDeviceError, err:
1918 _Fail("Failed to grow block device: %s", err, exc=True)
1921 def BlockdevSnapshot(disk):
1922 """Create a snapshot copy of a block device.
1924 This function is called recursively, and the snapshot is actually created
1925 just for the leaf lvm backend device.
1927 @type disk: L{objects.Disk}
1928 @param disk: the disk to be snapshotted
1930 @return: snapshot disk path
1933 if disk.dev_type == constants.LD_DRBD8:
1934 if not disk.children:
1935 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
1937 return BlockdevSnapshot(disk.children[0])
1938 elif disk.dev_type == constants.LD_LV:
1939 r_dev = _RecursiveFindBD(disk)
1940 if r_dev is not None:
1941 # FIXME: choose a saner value for the snapshot size
1942 # let's stay on the safe side and ask for the full size, for now
1943 return r_dev.Snapshot(disk.size)
1945 _Fail("Cannot find block device %s", disk)
1947 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1948 disk.unique_id, disk.dev_type)
1951 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx, debug):
1952 """Export a block device snapshot to a remote node.
1954 @type disk: L{objects.Disk}
1955 @param disk: the description of the disk to export
1956 @type dest_node: str
1957 @param dest_node: the destination node to export to
1958 @type instance: L{objects.Instance}
1959 @param instance: the instance object to whom the disk belongs
1960 @type cluster_name: str
1961 @param cluster_name: the cluster name, needed for SSH hostalias
1963 @param idx: the index of the disk in the instance's disk list,
1964 used to export to the OS scripts environment
1965 @type debug: integer
1966 @param debug: debug level, passed to the OS scripts
1970 inst_os = OSFromDisk(instance.os)
1971 export_env = OSEnvironment(instance, inst_os, debug)
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, debug):
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 @type debug: integer
2111 @param debug: debug level, passed to the OS scripts
2112 @rtype: list of boolean
2113 @return: each boolean represent the success of importing the n-th disk
2116 inst_os = OSFromDisk(instance.os)
2117 import_env = OSEnvironment(instance, inst_os, debug)
2118 import_script = inst_os.import_script
2120 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2121 instance.name, int(time.time()))
2122 if not os.path.exists(constants.LOG_OS_DIR):
2123 os.mkdir(constants.LOG_OS_DIR, 0750)
2126 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2127 import_script, logfile)
2130 for idx, image in enumerate(src_images):
2132 destcmd = utils.BuildShellCmd('cat %s', image)
2133 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2134 constants.GANETI_RUNAS,
2136 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2137 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2138 import_env['IMPORT_INDEX'] = str(idx)
2139 result = utils.RunCmd(command, env=import_env)
2141 logging.error("Disk import command '%s' returned error: %s"
2142 " output: %s", command, result.fail_reason,
2144 final_result.append("error importing disk %d: %s, %s" %
2145 (idx, result.fail_reason, result.output[-100]))
2148 _Fail("; ".join(final_result), log=False)
2152 """Return a list of exports currently available on this machine.
2155 @return: list of the exports
2158 if os.path.isdir(constants.EXPORT_DIR):
2159 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2161 _Fail("No exports directory")
2164 def RemoveExport(export):
2165 """Remove an existing export from the node.
2168 @param export: the name of the export to remove
2172 target = os.path.join(constants.EXPORT_DIR, export)
2175 shutil.rmtree(target)
2176 except EnvironmentError, err:
2177 _Fail("Error while removing the export: %s", err, exc=True)
2180 def BlockdevRename(devlist):
2181 """Rename a list of block devices.
2183 @type devlist: list of tuples
2184 @param devlist: list of tuples of the form (disk,
2185 new_logical_id, new_physical_id); disk is an
2186 L{objects.Disk} object describing the current disk,
2187 and new logical_id/physical_id is the name we
2190 @return: True if all renames succeeded, False otherwise
2195 for disk, unique_id in devlist:
2196 dev = _RecursiveFindBD(disk)
2198 msgs.append("Can't find device %s in rename" % str(disk))
2202 old_rpath = dev.dev_path
2203 dev.Rename(unique_id)
2204 new_rpath = dev.dev_path
2205 if old_rpath != new_rpath:
2206 DevCacheManager.RemoveCache(old_rpath)
2207 # FIXME: we should add the new cache information here, like:
2208 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2209 # but we don't have the owner here - maybe parse from existing
2210 # cache? for now, we only lose lvm data when we rename, which
2211 # is less critical than DRBD or MD
2212 except errors.BlockDeviceError, err:
2213 msgs.append("Can't rename device '%s' to '%s': %s" %
2214 (dev, unique_id, err))
2215 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2218 _Fail("; ".join(msgs))
2221 def _TransformFileStorageDir(file_storage_dir):
2222 """Checks whether given file_storage_dir is valid.
2224 Checks wheter the given file_storage_dir is within the cluster-wide
2225 default file_storage_dir stored in SimpleStore. Only paths under that
2226 directory are allowed.
2228 @type file_storage_dir: str
2229 @param file_storage_dir: the path to check
2231 @return: the normalized path if valid, None otherwise
2235 file_storage_dir = os.path.normpath(file_storage_dir)
2236 base_file_storage_dir = cfg.GetFileStorageDir()
2237 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2238 base_file_storage_dir):
2239 _Fail("File storage directory '%s' is not under base file"
2240 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2241 return file_storage_dir
2244 def CreateFileStorageDir(file_storage_dir):
2245 """Create file storage directory.
2247 @type file_storage_dir: str
2248 @param file_storage_dir: directory to create
2251 @return: tuple with first element a boolean indicating wheter dir
2252 creation was successful or not
2255 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2256 if os.path.exists(file_storage_dir):
2257 if not os.path.isdir(file_storage_dir):
2258 _Fail("Specified storage dir '%s' is not a directory",
2262 os.makedirs(file_storage_dir, 0750)
2263 except OSError, err:
2264 _Fail("Cannot create file storage directory '%s': %s",
2265 file_storage_dir, err, exc=True)
2268 def RemoveFileStorageDir(file_storage_dir):
2269 """Remove file storage directory.
2271 Remove it only if it's empty. If not log an error and return.
2273 @type file_storage_dir: str
2274 @param file_storage_dir: the directory we should cleanup
2275 @rtype: tuple (success,)
2276 @return: tuple of one element, C{success}, denoting
2277 whether the operation was successful
2280 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2281 if os.path.exists(file_storage_dir):
2282 if not os.path.isdir(file_storage_dir):
2283 _Fail("Specified Storage directory '%s' is not a directory",
2285 # deletes dir only if empty, otherwise we want to fail the rpc call
2287 os.rmdir(file_storage_dir)
2288 except OSError, err:
2289 _Fail("Cannot remove file storage directory '%s': %s",
2290 file_storage_dir, err)
2293 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2294 """Rename the file storage directory.
2296 @type old_file_storage_dir: str
2297 @param old_file_storage_dir: the current path
2298 @type new_file_storage_dir: str
2299 @param new_file_storage_dir: the name we should rename to
2300 @rtype: tuple (success,)
2301 @return: tuple of one element, C{success}, denoting
2302 whether the operation was successful
2305 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2306 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2307 if not os.path.exists(new_file_storage_dir):
2308 if os.path.isdir(old_file_storage_dir):
2310 os.rename(old_file_storage_dir, new_file_storage_dir)
2311 except OSError, err:
2312 _Fail("Cannot rename '%s' to '%s': %s",
2313 old_file_storage_dir, new_file_storage_dir, err)
2315 _Fail("Specified storage dir '%s' is not a directory",
2316 old_file_storage_dir)
2318 if os.path.exists(old_file_storage_dir):
2319 _Fail("Cannot rename '%s' to '%s': both locations exist",
2320 old_file_storage_dir, new_file_storage_dir)
2323 def _EnsureJobQueueFile(file_name):
2324 """Checks whether the given filename is in the queue directory.
2326 @type file_name: str
2327 @param file_name: the file name we should check
2329 @raises RPCFail: if the file is not valid
2332 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2333 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2336 _Fail("Passed job queue file '%s' does not belong to"
2337 " the queue directory '%s'", file_name, queue_dir)
2340 def JobQueueUpdate(file_name, content):
2341 """Updates a file in the queue directory.
2343 This is just a wrapper over L{utils.WriteFile}, with proper
2346 @type file_name: str
2347 @param file_name: the job file name
2349 @param content: the new job contents
2351 @return: the success of the operation
2354 _EnsureJobQueueFile(file_name)
2356 # Write and replace the file atomically
2357 utils.WriteFile(file_name, data=_Decompress(content))
2360 def JobQueueRename(old, new):
2361 """Renames a job queue file.
2363 This is just a wrapper over os.rename with proper checking.
2366 @param old: the old (actual) file name
2368 @param new: the desired file name
2370 @return: the success of the operation and payload
2373 _EnsureJobQueueFile(old)
2374 _EnsureJobQueueFile(new)
2376 utils.RenameFile(old, new, mkdir=True)
2379 def JobQueueSetDrainFlag(drain_flag):
2380 """Set the drain flag for the queue.
2382 This will set or unset the queue drain flag.
2384 @type drain_flag: boolean
2385 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2387 @return: always True, None
2388 @warning: the function always returns True
2392 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2394 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2397 def BlockdevClose(instance_name, disks):
2398 """Closes the given block devices.
2400 This means they will be switched to secondary mode (in case of
2403 @param instance_name: if the argument is not empty, the symlinks
2404 of this instance will be removed
2405 @type disks: list of L{objects.Disk}
2406 @param disks: the list of disks to be closed
2407 @rtype: tuple (success, message)
2408 @return: a tuple of success and message, where success
2409 indicates the succes of the operation, and message
2410 which will contain the error details in case we
2416 rd = _RecursiveFindBD(cf)
2418 _Fail("Can't find device %s", cf)
2425 except errors.BlockDeviceError, err:
2426 msg.append(str(err))
2428 _Fail("Can't make devices secondary: %s", ",".join(msg))
2431 _RemoveBlockDevLinks(instance_name, disks)
2434 def ValidateHVParams(hvname, hvparams):
2435 """Validates the given hypervisor parameters.
2437 @type hvname: string
2438 @param hvname: the hypervisor name
2439 @type hvparams: dict
2440 @param hvparams: the hypervisor parameters to be validated
2445 hv_type = hypervisor.GetHypervisor(hvname)
2446 hv_type.ValidateParameters(hvparams)
2447 except errors.HypervisorError, err:
2448 _Fail(str(err), log=False)
2452 """Demotes the current node from master candidate role.
2455 # try to ensure we're not the master by mistake
2456 master, myself = ssconf.GetMasterAndMyself()
2457 if master == myself:
2458 _Fail("ssconf status shows I'm the master node, will not demote")
2460 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2461 if not result.failed:
2462 _Fail("The master daemon is running, will not demote")
2465 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2466 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2467 except EnvironmentError, err:
2468 if err.errno != errno.ENOENT:
2469 _Fail("Error while backing up cluster file: %s", err, exc=True)
2471 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2474 def _FindDisks(nodes_ip, disks):
2475 """Sets the physical ID on disks and returns the block devices.
2478 # set the correct physical ID
2479 my_name = utils.HostInfo().name
2481 cf.SetPhysicalID(my_name, nodes_ip)
2486 rd = _RecursiveFindBD(cf)
2488 _Fail("Can't find device %s", cf)
2493 def DrbdDisconnectNet(nodes_ip, disks):
2494 """Disconnects the network on a list of drbd devices.
2497 bdevs = _FindDisks(nodes_ip, disks)
2503 except errors.BlockDeviceError, err:
2504 _Fail("Can't change network configuration to standalone mode: %s",
2508 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2509 """Attaches the network on a list of drbd devices.
2512 bdevs = _FindDisks(nodes_ip, disks)
2515 for idx, rd in enumerate(bdevs):
2517 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2518 except EnvironmentError, err:
2519 _Fail("Can't create symlink: %s", err)
2520 # reconnect disks, switch to new master configuration and if
2521 # needed primary mode
2524 rd.AttachNet(multimaster)
2525 except errors.BlockDeviceError, err:
2526 _Fail("Can't change network configuration: %s", err)
2528 # wait until the disks are connected; we need to retry the re-attach
2529 # if the device becomes standalone, as this might happen if the one
2530 # node disconnects and reconnects in a different mode before the
2531 # other node reconnects; in this case, one or both of the nodes will
2532 # decide it has wrong configuration and switch to standalone
2535 all_connected = True
2538 stats = rd.GetProcStatus()
2540 all_connected = (all_connected and
2541 (stats.is_connected or stats.is_in_resync))
2543 if stats.is_standalone:
2544 # peer had different config info and this node became
2545 # standalone, even though this should not happen with the
2546 # new staged way of changing disk configs
2548 rd.AttachNet(multimaster)
2549 except errors.BlockDeviceError, err:
2550 _Fail("Can't change network configuration: %s", err)
2552 if not all_connected:
2553 raise utils.RetryAgain()
2556 # Start with a delay of 100 miliseconds and go up to 5 seconds
2557 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2558 except utils.RetryTimeout:
2559 _Fail("Timeout in disk reconnecting")
2562 # change to primary mode
2566 except errors.BlockDeviceError, err:
2567 _Fail("Can't change to primary mode: %s", err)
2570 def DrbdWaitSync(nodes_ip, disks):
2571 """Wait until DRBDs have synchronized.
2575 stats = rd.GetProcStatus()
2576 if not (stats.is_connected or stats.is_in_resync):
2577 raise utils.RetryAgain()
2580 bdevs = _FindDisks(nodes_ip, disks)
2586 # poll each second for 15 seconds
2587 stats = utils.Retry(_helper, 1, 15, args=[rd])
2588 except utils.RetryTimeout:
2589 stats = rd.GetProcStatus()
2591 if not (stats.is_connected or stats.is_in_resync):
2592 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2593 alldone = alldone and (not stats.is_in_resync)
2594 if stats.sync_percent is not None:
2595 min_resync = min(min_resync, stats.sync_percent)
2597 return (alldone, min_resync)
2600 def PowercycleNode(hypervisor_type):
2601 """Hard-powercycle the node.
2603 Because we need to return first, and schedule the powercycle in the
2604 background, we won't be able to report failures nicely.
2607 hyper = hypervisor.GetHypervisor(hypervisor_type)
2611 # if we can't fork, we'll pretend that we're in the child process
2614 return "Reboot scheduled in 5 seconds"
2616 hyper.PowercycleNode()
2619 class HooksRunner(object):
2622 This class is instantiated on the node side (ganeti-noded) and not
2626 def __init__(self, hooks_base_dir=None):
2627 """Constructor for hooks runner.
2629 @type hooks_base_dir: str or None
2630 @param hooks_base_dir: if not None, this overrides the
2631 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2634 if hooks_base_dir is None:
2635 hooks_base_dir = constants.HOOKS_BASE_DIR
2636 # yeah, _BASE_DIR is not valid for attributes, we use it like a
2638 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
2640 def RunHooks(self, hpath, phase, env):
2641 """Run the scripts in the hooks directory.
2644 @param hpath: the path to the hooks directory which
2647 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2648 L{constants.HOOKS_PHASE_POST}
2650 @param env: dictionary with the environment for the hook
2652 @return: list of 3-element tuples:
2654 - script result, either L{constants.HKR_SUCCESS} or
2655 L{constants.HKR_FAIL}
2656 - output of the script
2658 @raise errors.ProgrammerError: for invalid input
2662 if phase == constants.HOOKS_PHASE_PRE:
2664 elif phase == constants.HOOKS_PHASE_POST:
2667 _Fail("Unknown hooks phase '%s'", phase)
2671 subdir = "%s-%s.d" % (hpath, suffix)
2672 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2674 dir_contents = utils.ListVisibleFiles(dir_name)
2676 # FIXME: must log output in case of failures
2679 # we use the standard python sort order,
2680 # so 00name is the recommended naming scheme
2682 for relname in dir_contents:
2683 fname = os.path.join(dir_name, relname)
2684 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2685 constants.EXT_PLUGIN_MASK.match(relname) is not None):
2686 rrval = constants.HKR_SKIP
2690 result = utils.RunCmd([fname], env=env, reset_env=True)
2691 except (OpExecError, EnvironmentError), err:
2692 rrval = constants.HKR_FAIL
2693 output = "Hook script error: %s" % str(err)
2696 rrval = constants.HKR_FAIL
2698 rrval = constants.HKR_SUCCESS
2699 output = utils.SafeEncode(result.output.strip())
2700 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2705 class IAllocatorRunner(object):
2706 """IAllocator runner.
2708 This class is instantiated on the node side (ganeti-noded) and not on
2713 def Run(name, idata):
2714 """Run an iallocator script.
2717 @param name: the iallocator script name
2719 @param idata: the allocator input data
2722 @return: two element tuple of:
2724 - either error message or stdout of allocator (for success)
2727 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2729 if alloc_script is None:
2730 _Fail("iallocator module '%s' not found in the search path", name)
2732 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2736 result = utils.RunCmd([alloc_script, fin_name])
2738 _Fail("iallocator module '%s' failed: %s, output '%s'",
2739 name, result.fail_reason, result.output)
2743 return result.stdout
2746 class DevCacheManager(object):
2747 """Simple class for managing a cache of block device information.
2750 _DEV_PREFIX = "/dev/"
2751 _ROOT_DIR = constants.BDEV_CACHE_DIR
2754 def _ConvertPath(cls, dev_path):
2755 """Converts a /dev/name path to the cache file name.
2757 This replaces slashes with underscores and strips the /dev
2758 prefix. It then returns the full path to the cache file.
2761 @param dev_path: the C{/dev/} path name
2763 @return: the converted path name
2766 if dev_path.startswith(cls._DEV_PREFIX):
2767 dev_path = dev_path[len(cls._DEV_PREFIX):]
2768 dev_path = dev_path.replace("/", "_")
2769 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2773 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2774 """Updates the cache information for a given device.
2777 @param dev_path: the pathname of the device
2779 @param owner: the owner (instance name) of the device
2780 @type on_primary: bool
2781 @param on_primary: whether this is the primary
2784 @param iv_name: the instance-visible name of the
2785 device, as in objects.Disk.iv_name
2790 if dev_path is None:
2791 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2793 fpath = cls._ConvertPath(dev_path)
2799 iv_name = "not_visible"
2800 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2802 utils.WriteFile(fpath, data=fdata)
2803 except EnvironmentError, err:
2804 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2807 def RemoveCache(cls, dev_path):
2808 """Remove data for a dev_path.
2810 This is just a wrapper over L{utils.RemoveFile} with a converted
2811 path name and logging.
2814 @param dev_path: the pathname of the device
2819 if dev_path is None:
2820 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2822 fpath = cls._ConvertPath(dev_path)
2824 utils.RemoveFile(fpath)
2825 except EnvironmentError, err:
2826 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)