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"""
39 from ganeti import errors
40 from ganeti import utils
41 from ganeti import ssh
42 from ganeti import hypervisor
43 from ganeti import constants
44 from ganeti import bdev
45 from ganeti import objects
46 from ganeti import ssconf
49 class RPCFail(Exception):
50 """Class denoting RPC failure.
52 Its argument is the error message.
56 def _Fail(msg, *args, **kwargs):
57 """Log an error and the raise an RPCFail exception.
59 This exception is then handled specially in the ganeti daemon and
60 turned into a 'failed' return type. As such, this function is a
61 useful shortcut for logging the error and returning it to the master
65 @param msg: the text of the exception
71 if "exc" in kwargs and kwargs["exc"]:
72 logging.exception(msg)
79 """Simple wrapper to return a SimpleStore.
81 @rtype: L{ssconf.SimpleStore}
82 @return: a SimpleStore instance
85 return ssconf.SimpleStore()
88 def _GetSshRunner(cluster_name):
89 """Simple wrapper to return an SshRunner.
91 @type cluster_name: str
92 @param cluster_name: the cluster name, which is needed
93 by the SshRunner constructor
94 @rtype: L{ssh.SshRunner}
95 @return: an SshRunner instance
98 return ssh.SshRunner(cluster_name)
101 def _Decompress(data):
102 """Unpacks data compressed by the RPC client.
104 @type data: list or tuple
105 @param data: Data sent by RPC client
107 @return: Decompressed data
110 assert isinstance(data, (list, tuple))
111 assert len(data) == 2
112 (encoding, content) = data
113 if encoding == constants.RPC_ENCODING_NONE:
115 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
116 return zlib.decompress(base64.b64decode(content))
118 raise AssertionError("Unknown data encoding")
121 def _CleanDirectory(path, exclude=None):
122 """Removes all regular files in a directory.
125 @param path: the directory to clean
127 @param exclude: list of files to be excluded, defaults
131 if not os.path.isdir(path):
136 # Normalize excluded paths
137 exclude = [os.path.normpath(i) for i in exclude]
139 for rel_name in utils.ListVisibleFiles(path):
140 full_name = os.path.normpath(os.path.join(path, rel_name))
141 if full_name in exclude:
143 if os.path.isfile(full_name) and not os.path.islink(full_name):
144 utils.RemoveFile(full_name)
148 """Removes job queue files and archived jobs.
153 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
154 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
158 """Returns master information.
160 This is an utility function to compute master information, either
161 for consumption here or from the node daemon.
164 @return: (master_netdev, master_ip, master_name) if we have a good
165 configuration, otherwise (None, None, None)
170 master_netdev = cfg.GetMasterNetdev()
171 master_ip = cfg.GetMasterIP()
172 master_node = cfg.GetMasterNode()
173 except errors.ConfigurationError, err:
174 logging.exception("Cluster configuration incomplete")
175 return (None, None, None)
176 return (master_netdev, master_ip, master_node)
179 def StartMaster(start_daemons):
180 """Activate local node as master node.
182 The function will always try activate the IP address of the master
183 (unless someone else has it). It will also start the master daemons,
184 based on the start_daemons parameter.
186 @type start_daemons: boolean
187 @param start_daemons: whther to also start the master
188 daemons (ganeti-masterd and ganeti-rapi)
193 master_netdev, master_ip, _ = GetMasterInfo()
194 if not master_netdev:
197 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
198 if utils.OwnIpAddress(master_ip):
199 # we already have the ip:
200 logging.debug("Already started")
202 logging.error("Someone else has the master ip, not activating")
205 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
206 "dev", master_netdev, "label",
207 "%s:0" % master_netdev])
209 logging.error("Can't activate master IP: %s", result.output)
212 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
213 "-s", master_ip, master_ip])
214 # we'll ignore the exit code of arping
216 # and now start the master and rapi daemons
218 for daemon in 'ganeti-masterd', 'ganeti-rapi':
219 result = utils.RunCmd([daemon])
221 logging.error("Can't start daemon %s: %s", daemon, result.output)
226 def StopMaster(stop_daemons):
227 """Deactivate this node as master.
229 The function will always try to deactivate the IP address of the
230 master. It will also stop the master daemons depending on the
231 stop_daemons parameter.
233 @type stop_daemons: boolean
234 @param stop_daemons: whether to also stop the master daemons
235 (ganeti-masterd and ganeti-rapi)
239 master_netdev, master_ip, _ = GetMasterInfo()
240 if not master_netdev:
243 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
244 "dev", master_netdev])
246 logging.error("Can't remove the master IP, error: %s", result.output)
247 # but otherwise ignore the failure
250 # stop/kill the rapi and the master daemon
251 for daemon in constants.RAPI_PID, constants.MASTERD_PID:
252 utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon)))
257 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
258 """Joins this node to the cluster.
260 This does the following:
261 - updates the hostkeys of the machine (rsa and dsa)
262 - adds the ssh private key to the user
263 - adds the ssh public key to the users' authorized_keys file
266 @param dsa: the DSA private key to write
268 @param dsapub: the DSA public key to write
270 @param rsa: the RSA private key to write
272 @param rsapub: the RSA public key to write
274 @param sshkey: the SSH private key to write
276 @param sshpub: the SSH public key to write
278 @return: the success of the operation
281 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
282 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
283 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
284 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
285 for name, content, mode in sshd_keys:
286 utils.WriteFile(name, data=content, mode=mode)
289 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
291 except errors.OpExecError, err:
292 _Fail("Error while processing user ssh files: %s", err, exc=True)
294 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
295 utils.WriteFile(name, data=content, mode=0600)
297 utils.AddAuthorizedKey(auth_keys, sshpub)
299 utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
301 return (True, "Node added successfully")
305 """Cleans up and remove the current node.
307 This function cleans up and prepares the current node to be removed
310 If processing is successful, then it raises an
311 L{errors.QuitGanetiException} which is used as a special case to
312 shutdown the node daemon.
315 _CleanDirectory(constants.DATA_DIR)
319 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
320 except errors.OpExecError:
321 logging.exception("Error while processing ssh files")
324 f = open(pub_key, 'r')
326 utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
330 utils.RemoveFile(priv_key)
331 utils.RemoveFile(pub_key)
333 # Return a reassuring string to the caller, and quit
334 raise errors.QuitGanetiException(False, 'Shutdown scheduled')
337 def GetNodeInfo(vgname, hypervisor_type):
338 """Gives back a hash with different informations about the node.
340 @type vgname: C{string}
341 @param vgname: the name of the volume group to ask for disk space information
342 @type hypervisor_type: C{str}
343 @param hypervisor_type: the name of the hypervisor to ask for
346 @return: dictionary with the following keys:
347 - vg_size is the size of the configured volume group in MiB
348 - vg_free is the free size of the volume group in MiB
349 - memory_dom0 is the memory allocated for domain0 in MiB
350 - memory_free is the currently available (free) ram in MiB
351 - memory_total is the total number of ram in MiB
355 vginfo = _GetVGInfo(vgname)
356 outputarray['vg_size'] = vginfo['vg_size']
357 outputarray['vg_free'] = vginfo['vg_free']
359 hyper = hypervisor.GetHypervisor(hypervisor_type)
360 hyp_info = hyper.GetNodeInfo()
361 if hyp_info is not None:
362 outputarray.update(hyp_info)
364 f = open("/proc/sys/kernel/random/boot_id", 'r')
366 outputarray["bootid"] = f.read(128).rstrip("\n")
373 def VerifyNode(what, cluster_name):
374 """Verify the status of the local node.
376 Based on the input L{what} parameter, various checks are done on the
379 If the I{filelist} key is present, this list of
380 files is checksummed and the file/checksum pairs are returned.
382 If the I{nodelist} key is present, we check that we have
383 connectivity via ssh with the target nodes (and check the hostname
386 If the I{node-net-test} key is present, we check that we have
387 connectivity to the given nodes via both primary IP and, if
388 applicable, secondary IPs.
391 @param what: a dictionary of things to check:
392 - filelist: list of files for which to compute checksums
393 - nodelist: list of nodes we should check ssh communication with
394 - node-net-test: list of nodes we should check node daemon port
396 - hypervisor: list with hypervisors to run the verify for
398 @return: a dictionary with the same keys as the input dict, and
399 values representing the result of the checks
404 if constants.NV_HYPERVISOR in what:
405 result[constants.NV_HYPERVISOR] = tmp = {}
406 for hv_name in what[constants.NV_HYPERVISOR]:
407 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
409 if constants.NV_FILELIST in what:
410 result[constants.NV_FILELIST] = utils.FingerprintFiles(
411 what[constants.NV_FILELIST])
413 if constants.NV_NODELIST in what:
414 result[constants.NV_NODELIST] = tmp = {}
415 random.shuffle(what[constants.NV_NODELIST])
416 for node in what[constants.NV_NODELIST]:
417 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
421 if constants.NV_NODENETTEST in what:
422 result[constants.NV_NODENETTEST] = tmp = {}
423 my_name = utils.HostInfo().name
424 my_pip = my_sip = None
425 for name, pip, sip in what[constants.NV_NODENETTEST]:
431 tmp[my_name] = ("Can't find my own primary/secondary IP"
434 port = utils.GetNodeDaemonPort()
435 for name, pip, sip in what[constants.NV_NODENETTEST]:
437 if not utils.TcpPing(pip, port, source=my_pip):
438 fail.append("primary")
440 if not utils.TcpPing(sip, port, source=my_sip):
441 fail.append("secondary")
443 tmp[name] = ("failure using the %s interface(s)" %
446 if constants.NV_LVLIST in what:
447 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
449 if constants.NV_INSTANCELIST in what:
450 result[constants.NV_INSTANCELIST] = GetInstanceList(
451 what[constants.NV_INSTANCELIST])
453 if constants.NV_VGLIST in what:
454 result[constants.NV_VGLIST] = ListVolumeGroups()
456 if constants.NV_VERSION in what:
457 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
458 constants.RELEASE_VERSION)
460 if constants.NV_HVINFO in what:
461 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
462 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
464 if constants.NV_DRBDLIST in what:
466 used_minors = bdev.DRBD8.GetUsedDevs().keys()
467 except errors.BlockDeviceError, err:
468 logging.warning("Can't get used minors list", exc_info=True)
469 used_minors = str(err)
470 result[constants.NV_DRBDLIST] = used_minors
475 def GetVolumeList(vg_name):
476 """Compute list of logical volumes and their size.
479 @param vg_name: the volume group whose LVs we should list
482 dictionary of all partions (key) with value being a tuple of
483 their size (in MiB), inactive and online status::
485 {'test1': ('20.06', True, True)}
487 in case of errors, a string is returned with the error
493 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
494 "--separator=%s" % sep,
495 "-olv_name,lv_size,lv_attr", vg_name])
497 logging.error("Failed to list logical volumes, lvs output: %s",
501 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
502 for line in result.stdout.splitlines():
504 match = valid_line_re.match(line)
506 logging.error("Invalid line returned from lvs output: '%s'", line)
508 name, size, attr = match.groups()
509 inactive = attr[4] == '-'
510 online = attr[5] == 'o'
511 lvs[name] = (size, inactive, online)
516 def ListVolumeGroups():
517 """List the volume groups and their size.
520 @return: dictionary with keys volume name and values the
524 return utils.ListVolumeGroups()
528 """List all volumes on this node.
532 A list of dictionaries, each having four keys:
533 - name: the logical volume name,
534 - size: the size of the logical volume
535 - dev: the physical device on which the LV lives
536 - vg: the volume group to which it belongs
538 In case of errors, we return an empty list and log the
541 Note that since a logical volume can live on multiple physical
542 volumes, the resulting list might include a logical volume
546 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
548 "--options=lv_name,lv_size,devices,vg_name"])
550 logging.error("Failed to list logical volumes, lvs output: %s",
556 return dev.split('(')[0]
562 'name': line[0].strip(),
563 'size': line[1].strip(),
564 'dev': parse_dev(line[2].strip()),
565 'vg': line[3].strip(),
568 return [map_line(line.split('|')) for line in result.stdout.splitlines()
569 if line.count('|') >= 3]
572 def BridgesExist(bridges_list):
573 """Check if a list of bridges exist on the current node.
576 @return: C{True} if all of them exist, C{False} otherwise
579 for bridge in bridges_list:
580 if not utils.BridgeExists(bridge):
586 def GetInstanceList(hypervisor_list):
587 """Provides a list of instances.
589 @type hypervisor_list: list
590 @param hypervisor_list: the list of hypervisors to query information
593 @return: a list of all running instances on the current node
594 - instance1.example.com
595 - instance2.example.com
599 for hname in hypervisor_list:
601 names = hypervisor.GetHypervisor(hname).ListInstances()
602 results.extend(names)
603 except errors.HypervisorError, err:
604 logging.exception("Error enumerating instances for hypevisor %s", hname)
610 def GetInstanceInfo(instance, hname):
611 """Gives back the informations about an instance as a dictionary.
613 @type instance: string
614 @param instance: the instance name
616 @param hname: the hypervisor type of the instance
619 @return: dictionary with the following keys:
620 - memory: memory size of instance (int)
621 - state: xen state of instance (string)
622 - time: cpu time of instance (float)
627 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
628 if iinfo is not None:
629 output['memory'] = iinfo[2]
630 output['state'] = iinfo[4]
631 output['time'] = iinfo[5]
636 def GetInstanceMigratable(instance):
637 """Gives whether an instance can be migrated.
639 @type instance: L{objects.Instance}
640 @param instance: object representing the instance to be checked.
643 @return: tuple of (result, description) where:
644 - result: whether the instance can be migrated or not
645 - description: a description of the issue, if relevant
648 hyper = hypervisor.GetHypervisor(instance.hypervisor)
649 if instance.name not in hyper.ListInstances():
650 return (False, 'not running')
652 for idx in range(len(instance.disks)):
653 link_name = _GetBlockDevSymlinkPath(instance.name, idx)
654 if not os.path.islink(link_name):
655 return (False, 'not restarted since ganeti 1.2.5')
660 def GetAllInstancesInfo(hypervisor_list):
661 """Gather data about all instances.
663 This is the equivalent of L{GetInstanceInfo}, except that it
664 computes data for all instances at once, thus being faster if one
665 needs data about more than one instance.
667 @type hypervisor_list: list
668 @param hypervisor_list: list of hypervisors to query for instance data
671 @return: dictionary of instance: data, with data having the following keys:
672 - memory: memory size of instance (int)
673 - state: xen state of instance (string)
674 - time: cpu time of instance (float)
675 - vcpus: the number of vcpus
680 for hname in hypervisor_list:
681 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
683 for name, inst_id, memory, vcpus, state, times in iinfo:
691 # we only check static parameters, like memory and vcpus,
692 # and not state and time which can change between the
693 # invocations of the different hypervisors
694 for key in 'memory', 'vcpus':
695 if value[key] != output[name][key]:
696 raise errors.HypervisorError("Instance %s is running twice"
697 " with different parameters" % name)
703 def InstanceOsAdd(instance, reinstall):
704 """Add an OS to an instance.
706 @type instance: L{objects.Instance}
707 @param instance: Instance whose OS is to be installed
708 @type reinstall: boolean
709 @param reinstall: whether this is an instance reinstall
711 @return: the success of the operation
715 inst_os = OSFromDisk(instance.os)
716 except errors.InvalidOS, err:
717 os_name, os_dir, os_err = err.args
719 return (False, "Can't find OS '%s': %s" % (os_name, os_err))
721 return (False, "Error parsing OS '%s' in directory %s: %s" %
722 (os_name, os_dir, os_err))
724 create_env = OSEnvironment(instance)
726 create_env['INSTANCE_REINSTALL'] = "1"
728 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
729 instance.name, int(time.time()))
731 result = utils.RunCmd([inst_os.create_script], env=create_env,
732 cwd=inst_os.path, output=logfile,)
734 logging.error("os create command '%s' returned error: %s, logfile: %s,"
735 " output: %s", result.cmd, result.fail_reason, logfile,
737 lines = [utils.SafeEncode(val)
738 for val in utils.TailFile(logfile, lines=20)]
739 return (False, "OS create script failed (%s), last lines in the"
740 " log file:\n%s" % (result.fail_reason, "\n".join(lines)))
742 return (True, "Successfully installed")
745 def RunRenameInstance(instance, old_name):
746 """Run the OS rename script for an instance.
748 @type instance: L{objects.Instance}
749 @param instance: Instance whose OS is to be installed
750 @type old_name: string
751 @param old_name: previous instance name
753 @return: the success of the operation
756 inst_os = OSFromDisk(instance.os)
758 rename_env = OSEnvironment(instance)
759 rename_env['OLD_INSTANCE_NAME'] = old_name
761 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
763 instance.name, int(time.time()))
765 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
766 cwd=inst_os.path, output=logfile)
769 logging.error("os create command '%s' returned error: %s output: %s",
770 result.cmd, result.fail_reason, result.output)
771 lines = [utils.SafeEncode(val)
772 for val in utils.TailFile(logfile, lines=20)]
773 return (False, "OS rename script failed (%s), last lines in the"
774 " log file:\n%s" % (result.fail_reason, "\n".join(lines)))
776 return (True, "Rename successful")
779 def _GetVGInfo(vg_name):
780 """Get informations about the volume group.
783 @param vg_name: the volume group which we query
786 A dictionary with the following keys:
787 - C{vg_size} is the total size of the volume group in MiB
788 - C{vg_free} is the free size of the volume group in MiB
789 - C{pv_count} are the number of physical disks in that VG
791 If an error occurs during gathering of data, we return the same dict
792 with keys all set to None.
795 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
797 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
798 "--nosuffix", "--units=m", "--separator=:", vg_name])
801 logging.error("volume group %s not present", vg_name)
803 valarr = retval.stdout.strip().rstrip(':').split(':')
807 "vg_size": int(round(float(valarr[0]), 0)),
808 "vg_free": int(round(float(valarr[1]), 0)),
809 "pv_count": int(valarr[2]),
811 except ValueError, err:
812 logging.exception("Fail to parse vgs output")
814 logging.error("vgs output has the wrong number of fields (expected"
815 " three): %s", str(valarr))
819 def _GetBlockDevSymlinkPath(instance_name, idx):
820 return os.path.join(constants.DISK_LINKS_DIR,
821 "%s:%d" % (instance_name, idx))
824 def _SymlinkBlockDev(instance_name, device_path, idx):
825 """Set up symlinks to a instance's block device.
827 This is an auxiliary function run when an instance is start (on the primary
828 node) or when an instance is migrated (on the target node).
831 @param instance_name: the name of the target instance
832 @param device_path: path of the physical block device, on the node
833 @param idx: the disk index
834 @return: absolute path to the disk's symlink
837 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
839 os.symlink(device_path, link_name)
841 if err.errno == errno.EEXIST:
842 if (not os.path.islink(link_name) or
843 os.readlink(link_name) != device_path):
845 os.symlink(device_path, link_name)
852 def _RemoveBlockDevLinks(instance_name, disks):
853 """Remove the block device symlinks belonging to the given instance.
856 for idx, disk in enumerate(disks):
857 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
858 if os.path.islink(link_name):
862 logging.exception("Can't remove symlink '%s'", link_name)
865 def _GatherAndLinkBlockDevs(instance):
866 """Set up an instance's block device(s).
868 This is run on the primary node at instance startup. The block
869 devices must be already assembled.
871 @type instance: L{objects.Instance}
872 @param instance: the instance whose disks we shoul assemble
874 @return: list of (disk_object, device_path)
878 for idx, disk in enumerate(instance.disks):
879 device = _RecursiveFindBD(disk)
881 raise errors.BlockDeviceError("Block device '%s' is not set up." %
885 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
887 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
890 block_devices.append((disk, link_name))
895 def StartInstance(instance):
896 """Start an instance.
898 @type instance: L{objects.Instance}
899 @param instance: the instance object
901 @return: whether the startup was successful or not
904 running_instances = GetInstanceList([instance.hypervisor])
906 if instance.name in running_instances:
907 return (True, "Already running")
910 block_devices = _GatherAndLinkBlockDevs(instance)
911 hyper = hypervisor.GetHypervisor(instance.hypervisor)
912 hyper.StartInstance(instance, block_devices)
913 except errors.BlockDeviceError, err:
914 _Fail("Block device error: %s", err, exc=True)
915 except errors.HypervisorError, err:
916 _RemoveBlockDevLinks(instance.name, instance.disks)
917 _Fail("Hypervisor error: %s", err, exc=True)
919 return (True, "Instance started successfully")
922 def InstanceShutdown(instance):
923 """Shut an instance down.
925 @note: this functions uses polling with a hardcoded timeout.
927 @type instance: L{objects.Instance}
928 @param instance: the instance object
930 @return: whether the startup was successful or not
933 hv_name = instance.hypervisor
934 running_instances = GetInstanceList([hv_name])
936 if instance.name not in running_instances:
937 return (True, "Instance already stopped")
939 hyper = hypervisor.GetHypervisor(hv_name)
941 hyper.StopInstance(instance)
942 except errors.HypervisorError, err:
943 _Fail("Failed to stop instance %s: %s", instance.name, err)
945 # test every 10secs for 2min
948 for dummy in range(11):
949 if instance.name not in GetInstanceList([hv_name]):
953 # the shutdown did not succeed
954 logging.error("Shutdown of '%s' unsuccessful, using destroy",
958 hyper.StopInstance(instance, force=True)
959 except errors.HypervisorError, err:
960 _Fail("Failed to force stop instance %s: %s", instance.name, err)
963 if instance.name in GetInstanceList([hv_name]):
964 _Fail("Could not shutdown instance %s even by destroy", instance.name)
966 _RemoveBlockDevLinks(instance.name, instance.disks)
968 return (True, "Instance has been shutdown successfully")
971 def InstanceReboot(instance, reboot_type):
972 """Reboot an instance.
974 @type instance: L{objects.Instance}
975 @param instance: the instance object to reboot
976 @type reboot_type: str
977 @param reboot_type: the type of reboot, one the following
979 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
980 instance OS, do not recreate the VM
981 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
982 restart the VM (at the hypervisor level)
983 - the other reboot type (L{constants.INSTANCE_REBOOT_HARD})
984 is not accepted here, since that mode is handled
987 @return: the success of the operation
990 running_instances = GetInstanceList([instance.hypervisor])
992 if instance.name not in running_instances:
993 _Fail("Cannot reboot instance %s that is not running", instance.name)
995 hyper = hypervisor.GetHypervisor(instance.hypervisor)
996 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
998 hyper.RebootInstance(instance)
999 except errors.HypervisorError, err:
1000 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1001 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1003 stop_result = InstanceShutdown(instance)
1004 if not stop_result[0]:
1006 return StartInstance(instance)
1007 except errors.HypervisorError, err:
1008 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1010 _Fail("Invalid reboot_type received: %s", reboot_type)
1012 return (True, "Reboot successful")
1015 def MigrationInfo(instance):
1016 """Gather information about an instance to be migrated.
1018 @type instance: L{objects.Instance}
1019 @param instance: the instance definition
1022 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1024 info = hyper.MigrationInfo(instance)
1025 except errors.HypervisorError, err:
1026 _Fail("Failed to fetch migration information: %s", err, exc=True)
1030 def AcceptInstance(instance, info, target):
1031 """Prepare the node to accept an instance.
1033 @type instance: L{objects.Instance}
1034 @param instance: the instance definition
1035 @type info: string/data (opaque)
1036 @param info: migration information, from the source node
1037 @type target: string
1038 @param target: target host (usually ip), on this node
1041 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1043 hyper.AcceptInstance(instance, info, target)
1044 except errors.HypervisorError, err:
1045 _Fail("Failed to accept instance: %s", err, exc=True)
1046 return (True, "Accept successfull")
1049 def FinalizeMigration(instance, info, success):
1050 """Finalize any preparation to accept an instance.
1052 @type instance: L{objects.Instance}
1053 @param instance: the instance definition
1054 @type info: string/data (opaque)
1055 @param info: migration information, from the source node
1056 @type success: boolean
1057 @param success: whether the migration was a success or a failure
1060 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1062 hyper.FinalizeMigration(instance, info, success)
1063 except errors.HypervisorError, err:
1064 _Fail("Failed to finalize migration: %s", err, exc=True)
1065 return (True, "Migration Finalized")
1068 def MigrateInstance(instance, target, live):
1069 """Migrates an instance to another node.
1071 @type instance: L{objects.Instance}
1072 @param instance: the instance definition
1073 @type target: string
1074 @param target: the target node name
1076 @param live: whether the migration should be done live or not (the
1077 interpretation of this parameter is left to the hypervisor)
1079 @return: a tuple of (success, msg) where:
1080 - succes is a boolean denoting the success/failure of the operation
1081 - msg is a string with details in case of failure
1084 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1087 hyper.MigrateInstance(instance.name, target, live)
1088 except errors.HypervisorError, err:
1089 _Fail("Failed to migrate instance: %s", err, exc=True)
1090 return (True, "Migration successfull")
1093 def BlockdevCreate(disk, size, owner, on_primary, info):
1094 """Creates a block device for an instance.
1096 @type disk: L{objects.Disk}
1097 @param disk: the object describing the disk we should create
1099 @param size: the size of the physical underlying device, in MiB
1101 @param owner: the name of the instance for which disk is created,
1102 used for device cache data
1103 @type on_primary: boolean
1104 @param on_primary: indicates if it is the primary node or not
1106 @param info: string that will be sent to the physical device
1107 creation, used for example to set (LVM) tags on LVs
1109 @return: the new unique_id of the device (this can sometime be
1110 computed only after creation), or None. On secondary nodes,
1111 it's not required to return anything.
1116 for child in disk.children:
1118 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1119 except errors.BlockDeviceError, err:
1120 _Fail("Can't assemble device %s: %s", child, err)
1121 if on_primary or disk.AssembleOnSecondary():
1122 # we need the children open in case the device itself has to
1126 except errors.BlockDeviceError, err:
1127 _Fail("Can't make child '%s' read-write: %s", child, err)
1131 device = bdev.Create(disk.dev_type, disk.physical_id, clist, size)
1132 except errors.BlockDeviceError, err:
1133 _Fail("Can't create block device: %s", err)
1135 if on_primary or disk.AssembleOnSecondary():
1138 except errors.BlockDeviceError, err:
1139 _Fail("Can't assemble device after creation, unusual event: %s", err)
1140 device.SetSyncSpeed(constants.SYNC_SPEED)
1141 if on_primary or disk.OpenOnSecondary():
1143 device.Open(force=True)
1144 except errors.BlockDeviceError, err:
1145 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1146 DevCacheManager.UpdateCache(device.dev_path, owner,
1147 on_primary, disk.iv_name)
1149 device.SetInfo(info)
1151 physical_id = device.unique_id
1152 return True, physical_id
1155 def BlockdevRemove(disk):
1156 """Remove a block device.
1158 @note: This is intended to be called recursively.
1160 @type disk: L{objects.Disk}
1161 @param disk: the disk object we should remove
1163 @return: the success of the operation
1169 rdev = _RecursiveFindBD(disk)
1170 except errors.BlockDeviceError, err:
1171 # probably can't attach
1172 logging.info("Can't attach to device %s in remove", disk)
1174 if rdev is not None:
1175 r_path = rdev.dev_path
1178 except errors.BlockDeviceError, err:
1179 msgs.append(str(err))
1182 DevCacheManager.RemoveCache(r_path)
1185 for child in disk.children:
1186 c_status, c_msg = BlockdevRemove(child)
1187 result = result and c_status
1188 if c_msg: # not an empty message
1191 return (result, "; ".join(msgs))
1194 def _RecursiveAssembleBD(disk, owner, as_primary):
1195 """Activate a block device for an instance.
1197 This is run on the primary and secondary nodes for an instance.
1199 @note: this function is called recursively.
1201 @type disk: L{objects.Disk}
1202 @param disk: the disk we try to assemble
1204 @param owner: the name of the instance which owns the disk
1205 @type as_primary: boolean
1206 @param as_primary: if we should make the block device
1209 @return: the assembled device or None (in case no device
1211 @raise errors.BlockDeviceError: in case there is an error
1212 during the activation of the children or the device
1218 mcn = disk.ChildrenNeeded()
1220 mcn = 0 # max number of Nones allowed
1222 mcn = len(disk.children) - mcn # max number of Nones
1223 for chld_disk in disk.children:
1225 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1226 except errors.BlockDeviceError, err:
1227 if children.count(None) >= mcn:
1230 logging.error("Error in child activation (but continuing): %s",
1232 children.append(cdev)
1234 if as_primary or disk.AssembleOnSecondary():
1235 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children)
1236 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1238 if as_primary or disk.OpenOnSecondary():
1240 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1241 as_primary, disk.iv_name)
1248 def BlockdevAssemble(disk, owner, as_primary):
1249 """Activate a block device for an instance.
1251 This is a wrapper over _RecursiveAssembleBD.
1253 @rtype: str or boolean
1254 @return: a C{/dev/...} path for primary nodes, and
1255 C{True} for secondary nodes
1259 result = "no error information"
1261 result = _RecursiveAssembleBD(disk, owner, as_primary)
1262 if isinstance(result, bdev.BlockDev):
1263 result = result.dev_path
1264 except errors.BlockDeviceError, err:
1265 result = "Error while assembling disk: %s" % str(err)
1267 return (status, result)
1270 def BlockdevShutdown(disk):
1271 """Shut down a block device.
1273 First, if the device is assembled (Attach() is successfull), then
1274 the device is shutdown. Then the children of the device are
1277 This function is called recursively. Note that we don't cache the
1278 children or such, as oppossed to assemble, shutdown of different
1279 devices doesn't require that the upper device was active.
1281 @type disk: L{objects.Disk}
1282 @param disk: the description of the disk we should
1285 @return: the success of the operation
1290 r_dev = _RecursiveFindBD(disk)
1291 if r_dev is not None:
1292 r_path = r_dev.dev_path
1295 DevCacheManager.RemoveCache(r_path)
1296 except errors.BlockDeviceError, err:
1297 msgs.append(str(err))
1301 for child in disk.children:
1302 c_status, c_msg = BlockdevShutdown(child)
1303 result = result and c_status
1304 if c_msg: # not an empty message
1307 return (result, "; ".join(msgs))
1310 def BlockdevAddchildren(parent_cdev, new_cdevs):
1311 """Extend a mirrored block device.
1313 @type parent_cdev: L{objects.Disk}
1314 @param parent_cdev: the disk to which we should add children
1315 @type new_cdevs: list of L{objects.Disk}
1316 @param new_cdevs: the list of children which we should add
1318 @return: the success of the operation
1321 parent_bdev = _RecursiveFindBD(parent_cdev)
1322 if parent_bdev is None:
1323 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1324 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1325 if new_bdevs.count(None) > 0:
1326 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1327 parent_bdev.AddChildren(new_bdevs)
1331 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1332 """Shrink a mirrored block device.
1334 @type parent_cdev: L{objects.Disk}
1335 @param parent_cdev: the disk from which we should remove children
1336 @type new_cdevs: list of L{objects.Disk}
1337 @param new_cdevs: the list of children which we should remove
1339 @return: the success of the operation
1342 parent_bdev = _RecursiveFindBD(parent_cdev)
1343 if parent_bdev is None:
1344 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1346 for disk in new_cdevs:
1347 rpath = disk.StaticDevPath()
1349 bd = _RecursiveFindBD(disk)
1351 _Fail("Can't find device %s while removing children", disk)
1353 devs.append(bd.dev_path)
1356 parent_bdev.RemoveChildren(devs)
1360 def BlockdevGetmirrorstatus(disks):
1361 """Get the mirroring status of a list of devices.
1363 @type disks: list of L{objects.Disk}
1364 @param disks: the list of disks which we should query
1367 a list of (mirror_done, estimated_time) tuples, which
1368 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1369 @raise errors.BlockDeviceError: if any of the disks cannot be
1375 rbd = _RecursiveFindBD(dsk)
1377 raise errors.BlockDeviceError("Can't find device %s" % str(dsk))
1378 stats.append(rbd.CombinedSyncStatus())
1382 def _RecursiveFindBD(disk):
1383 """Check if a device is activated.
1385 If so, return informations about the real device.
1387 @type disk: L{objects.Disk}
1388 @param disk: the disk object we need to find
1390 @return: None if the device can't be found,
1391 otherwise the device instance
1396 for chdisk in disk.children:
1397 children.append(_RecursiveFindBD(chdisk))
1399 return bdev.FindDevice(disk.dev_type, disk.physical_id, children)
1402 def BlockdevFind(disk):
1403 """Check if a device is activated.
1405 If it is, return informations about the real device.
1407 @type disk: L{objects.Disk}
1408 @param disk: the disk to find
1409 @rtype: None or tuple
1410 @return: None if the disk cannot be found, otherwise a
1411 tuple (device_path, major, minor, sync_percent,
1412 estimated_time, is_degraded)
1416 rbd = _RecursiveFindBD(disk)
1417 except errors.BlockDeviceError, err:
1418 _Fail("Failed to find device: %s", err, exc=True)
1421 return (True, (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus())
1424 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1425 """Write a file to the filesystem.
1427 This allows the master to overwrite(!) a file. It will only perform
1428 the operation if the file belongs to a list of configuration files.
1430 @type file_name: str
1431 @param file_name: the target file name
1433 @param data: the new contents of the file
1435 @param mode: the mode to give the file (can be None)
1437 @param uid: the owner of the file (can be -1 for default)
1439 @param gid: the group of the file (can be -1 for default)
1441 @param atime: the atime to set on the file (can be None)
1443 @param mtime: the mtime to set on the file (can be None)
1445 @return: the success of the operation; errors are logged
1446 in the node daemon log
1449 if not os.path.isabs(file_name):
1450 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1452 allowed_files = set([
1453 constants.CLUSTER_CONF_FILE,
1454 constants.ETC_HOSTS,
1455 constants.SSH_KNOWN_HOSTS_FILE,
1456 constants.VNC_PASSWORD_FILE,
1457 constants.RAPI_CERT_FILE,
1458 constants.RAPI_USERS_FILE,
1461 for hv_name in constants.HYPER_TYPES:
1462 hv_class = hypervisor.GetHypervisor(hv_name)
1463 allowed_files.update(hv_class.GetAncillaryFiles())
1465 if file_name not in allowed_files:
1466 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1469 raw_data = _Decompress(data)
1471 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1472 atime=atime, mtime=mtime)
1473 return (True, "success")
1476 def WriteSsconfFiles(values):
1477 """Update all ssconf files.
1479 Wrapper around the SimpleStore.WriteFiles.
1482 ssconf.SimpleStore().WriteFiles(values)
1485 def _ErrnoOrStr(err):
1486 """Format an EnvironmentError exception.
1488 If the L{err} argument has an errno attribute, it will be looked up
1489 and converted into a textual C{E...} description. Otherwise the
1490 string representation of the error will be returned.
1492 @type err: L{EnvironmentError}
1493 @param err: the exception to format
1496 if hasattr(err, 'errno'):
1497 detail = errno.errorcode[err.errno]
1503 def _OSOndiskVersion(name, os_dir):
1504 """Compute and return the API version of a given OS.
1506 This function will try to read the API version of the OS given by
1507 the 'name' parameter and residing in the 'os_dir' directory.
1510 @param name: the OS name we should look for
1512 @param os_dir: the directory inwhich we should look for the OS
1515 Either an integer denoting the version or None in the
1516 case when this is not a valid OS name.
1517 @raise errors.InvalidOS: if the OS cannot be found
1520 api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
1523 st = os.stat(api_file)
1524 except EnvironmentError, err:
1525 raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file not"
1526 " found (%s)" % _ErrnoOrStr(err))
1528 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1529 raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file is not"
1535 api_versions = f.readlines()
1538 except EnvironmentError, err:
1539 raise errors.InvalidOS(name, os_dir, "error while reading the"
1540 " API version (%s)" % _ErrnoOrStr(err))
1542 api_versions = [version.strip() for version in api_versions]
1544 api_versions = [int(version) for version in api_versions]
1545 except (TypeError, ValueError), err:
1546 raise errors.InvalidOS(name, os_dir,
1547 "API version is not integer (%s)" % str(err))
1552 def DiagnoseOS(top_dirs=None):
1553 """Compute the validity for all OSes.
1555 @type top_dirs: list
1556 @param top_dirs: the list of directories in which to
1557 search (if not given defaults to
1558 L{constants.OS_SEARCH_PATH})
1559 @rtype: list of L{objects.OS}
1560 @return: an OS object for each name in all the given
1564 if top_dirs is None:
1565 top_dirs = constants.OS_SEARCH_PATH
1568 for dir_name in top_dirs:
1569 if os.path.isdir(dir_name):
1571 f_names = utils.ListVisibleFiles(dir_name)
1572 except EnvironmentError, err:
1573 logging.exception("Can't list the OS directory %s", dir_name)
1575 for name in f_names:
1577 os_inst = OSFromDisk(name, base_dir=dir_name)
1578 result.append(os_inst)
1579 except errors.InvalidOS, err:
1580 result.append(objects.OS.FromInvalidOS(err))
1585 def OSFromDisk(name, base_dir=None):
1586 """Create an OS instance from disk.
1588 This function will return an OS instance if the given name is a
1589 valid OS name. Otherwise, it will raise an appropriate
1590 L{errors.InvalidOS} exception, detailing why this is not a valid OS.
1592 @type base_dir: string
1593 @keyword base_dir: Base directory containing OS installations.
1594 Defaults to a search in all the OS_SEARCH_PATH dirs.
1595 @rtype: L{objects.OS}
1596 @return: the OS instance if we find a valid one
1597 @raise errors.InvalidOS: if we don't find a valid OS
1600 if base_dir is None:
1601 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1603 raise errors.InvalidOS(name, None, "OS dir not found in search path")
1605 os_dir = os.path.sep.join([base_dir, name])
1607 api_versions = _OSOndiskVersion(name, os_dir)
1609 if constants.OS_API_VERSION not in api_versions:
1610 raise errors.InvalidOS(name, os_dir, "API version mismatch"
1611 " (found %s want %s)"
1612 % (api_versions, constants.OS_API_VERSION))
1614 # OS Scripts dictionary, we will populate it with the actual script names
1615 os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
1617 for script in os_scripts:
1618 os_scripts[script] = os.path.sep.join([os_dir, script])
1621 st = os.stat(os_scripts[script])
1622 except EnvironmentError, err:
1623 raise errors.InvalidOS(name, os_dir, "'%s' script missing (%s)" %
1624 (script, _ErrnoOrStr(err)))
1626 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1627 raise errors.InvalidOS(name, os_dir, "'%s' script not executable" %
1630 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1631 raise errors.InvalidOS(name, os_dir, "'%s' is not a regular file" %
1635 return objects.OS(name=name, path=os_dir, status=constants.OS_VALID_STATUS,
1636 create_script=os_scripts[constants.OS_SCRIPT_CREATE],
1637 export_script=os_scripts[constants.OS_SCRIPT_EXPORT],
1638 import_script=os_scripts[constants.OS_SCRIPT_IMPORT],
1639 rename_script=os_scripts[constants.OS_SCRIPT_RENAME],
1640 api_versions=api_versions)
1642 def OSEnvironment(instance, debug=0):
1643 """Calculate the environment for an os script.
1645 @type instance: L{objects.Instance}
1646 @param instance: target instance for the os script run
1647 @type debug: integer
1648 @param debug: debug level (0 or 1, for OS Api 10)
1650 @return: dict of environment variables
1651 @raise errors.BlockDeviceError: if the block device
1656 result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION
1657 result['INSTANCE_NAME'] = instance.name
1658 result['INSTANCE_OS'] = instance.os
1659 result['HYPERVISOR'] = instance.hypervisor
1660 result['DISK_COUNT'] = '%d' % len(instance.disks)
1661 result['NIC_COUNT'] = '%d' % len(instance.nics)
1662 result['DEBUG_LEVEL'] = '%d' % debug
1663 for idx, disk in enumerate(instance.disks):
1664 real_disk = _RecursiveFindBD(disk)
1665 if real_disk is None:
1666 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1669 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1670 result['DISK_%d_ACCESS' % idx] = disk.mode
1671 if constants.HV_DISK_TYPE in instance.hvparams:
1672 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1673 instance.hvparams[constants.HV_DISK_TYPE]
1674 if disk.dev_type in constants.LDS_BLOCK:
1675 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1676 elif disk.dev_type == constants.LD_FILE:
1677 result['DISK_%d_BACKEND_TYPE' % idx] = \
1678 'file:%s' % disk.physical_id[0]
1679 for idx, nic in enumerate(instance.nics):
1680 result['NIC_%d_MAC' % idx] = nic.mac
1682 result['NIC_%d_IP' % idx] = nic.ip
1683 result['NIC_%d_BRIDGE' % idx] = nic.bridge
1684 if constants.HV_NIC_TYPE in instance.hvparams:
1685 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1686 instance.hvparams[constants.HV_NIC_TYPE]
1690 def BlockdevGrow(disk, amount):
1691 """Grow a stack of block devices.
1693 This function is called recursively, with the childrens being the
1694 first ones to resize.
1696 @type disk: L{objects.Disk}
1697 @param disk: the disk to be grown
1698 @rtype: (status, result)
1699 @return: a tuple with the status of the operation
1700 (True/False), and the errors message if status
1704 r_dev = _RecursiveFindBD(disk)
1706 return False, "Cannot find block device %s" % (disk,)
1710 except errors.BlockDeviceError, err:
1711 _Fail("Failed to grow block device: %s", err, exc=True)
1716 def BlockdevSnapshot(disk):
1717 """Create a snapshot copy of a block device.
1719 This function is called recursively, and the snapshot is actually created
1720 just for the leaf lvm backend device.
1722 @type disk: L{objects.Disk}
1723 @param disk: the disk to be snapshotted
1725 @return: snapshot disk path
1729 if len(disk.children) == 1:
1730 # only one child, let's recurse on it
1731 return BlockdevSnapshot(disk.children[0])
1733 # more than one child, choose one that matches
1734 for child in disk.children:
1735 if child.size == disk.size:
1736 # return implies breaking the loop
1737 return BlockdevSnapshot(child)
1738 elif disk.dev_type == constants.LD_LV:
1739 r_dev = _RecursiveFindBD(disk)
1740 if r_dev is not None:
1741 # let's stay on the safe side and ask for the full size, for now
1742 return r_dev.Snapshot(disk.size)
1746 raise errors.ProgrammerError("Cannot snapshot non-lvm block device"
1747 " '%s' of type '%s'" %
1748 (disk.unique_id, disk.dev_type))
1751 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1752 """Export a block device snapshot to a remote node.
1754 @type disk: L{objects.Disk}
1755 @param disk: the description of the disk to export
1756 @type dest_node: str
1757 @param dest_node: the destination node to export to
1758 @type instance: L{objects.Instance}
1759 @param instance: the instance object to whom the disk belongs
1760 @type cluster_name: str
1761 @param cluster_name: the cluster name, needed for SSH hostalias
1763 @param idx: the index of the disk in the instance's disk list,
1764 used to export to the OS scripts environment
1766 @return: the success of the operation
1769 export_env = OSEnvironment(instance)
1771 inst_os = OSFromDisk(instance.os)
1772 export_script = inst_os.export_script
1774 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1775 instance.name, int(time.time()))
1776 if not os.path.exists(constants.LOG_OS_DIR):
1777 os.mkdir(constants.LOG_OS_DIR, 0750)
1778 real_disk = _RecursiveFindBD(disk)
1779 if real_disk is None:
1780 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1784 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1785 export_env['EXPORT_INDEX'] = str(idx)
1787 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1788 destfile = disk.physical_id[1]
1790 # the target command is built out of three individual commands,
1791 # which are joined by pipes; we check each individual command for
1793 expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
1794 export_script, logfile)
1798 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1799 destdir, destdir, destfile)
1800 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1801 constants.GANETI_RUNAS,
1804 # all commands have been checked, so we're safe to combine them
1805 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1807 result = utils.RunCmd(command, env=export_env)
1810 logging.error("os snapshot export command '%s' returned error: %s"
1811 " output: %s", command, result.fail_reason, result.output)
1817 def FinalizeExport(instance, snap_disks):
1818 """Write out the export configuration information.
1820 @type instance: L{objects.Instance}
1821 @param instance: the instance which we export, used for
1822 saving configuration
1823 @type snap_disks: list of L{objects.Disk}
1824 @param snap_disks: list of snapshot block devices, which
1825 will be used to get the actual name of the dump file
1828 @return: the success of the operation
1831 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1832 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1834 config = objects.SerializableConfigParser()
1836 config.add_section(constants.INISECT_EXP)
1837 config.set(constants.INISECT_EXP, 'version', '0')
1838 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1839 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1840 config.set(constants.INISECT_EXP, 'os', instance.os)
1841 config.set(constants.INISECT_EXP, 'compression', 'gzip')
1843 config.add_section(constants.INISECT_INS)
1844 config.set(constants.INISECT_INS, 'name', instance.name)
1845 config.set(constants.INISECT_INS, 'memory', '%d' %
1846 instance.beparams[constants.BE_MEMORY])
1847 config.set(constants.INISECT_INS, 'vcpus', '%d' %
1848 instance.beparams[constants.BE_VCPUS])
1849 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1852 for nic_count, nic in enumerate(instance.nics):
1854 config.set(constants.INISECT_INS, 'nic%d_mac' %
1855 nic_count, '%s' % nic.mac)
1856 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
1857 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
1859 # TODO: redundant: on load can read nics until it doesn't exist
1860 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
1863 for disk_count, disk in enumerate(snap_disks):
1866 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
1867 ('%s' % disk.iv_name))
1868 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
1869 ('%s' % disk.physical_id[1]))
1870 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
1873 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
1875 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
1876 data=config.Dumps())
1877 shutil.rmtree(finaldestdir, True)
1878 shutil.move(destdir, finaldestdir)
1883 def ExportInfo(dest):
1884 """Get export configuration information.
1887 @param dest: directory containing the export
1889 @rtype: L{objects.SerializableConfigParser}
1890 @return: a serializable config file containing the
1894 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
1896 config = objects.SerializableConfigParser()
1899 if (not config.has_section(constants.INISECT_EXP) or
1900 not config.has_section(constants.INISECT_INS)):
1906 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
1907 """Import an os image into an instance.
1909 @type instance: L{objects.Instance}
1910 @param instance: instance to import the disks into
1911 @type src_node: string
1912 @param src_node: source node for the disk images
1913 @type src_images: list of string
1914 @param src_images: absolute paths of the disk images
1915 @rtype: list of boolean
1916 @return: each boolean represent the success of importing the n-th disk
1919 import_env = OSEnvironment(instance)
1920 inst_os = OSFromDisk(instance.os)
1921 import_script = inst_os.import_script
1923 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
1924 instance.name, int(time.time()))
1925 if not os.path.exists(constants.LOG_OS_DIR):
1926 os.mkdir(constants.LOG_OS_DIR, 0750)
1929 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
1930 import_script, logfile)
1933 for idx, image in enumerate(src_images):
1935 destcmd = utils.BuildShellCmd('cat %s', image)
1936 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
1937 constants.GANETI_RUNAS,
1939 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
1940 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
1941 import_env['IMPORT_INDEX'] = str(idx)
1942 result = utils.RunCmd(command, env=import_env)
1944 logging.error("Disk import command '%s' returned error: %s"
1945 " output: %s", command, result.fail_reason,
1947 final_result.append(False)
1949 final_result.append(True)
1951 final_result.append(True)
1957 """Return a list of exports currently available on this machine.
1960 @return: list of the exports
1963 if os.path.isdir(constants.EXPORT_DIR):
1964 return utils.ListVisibleFiles(constants.EXPORT_DIR)
1969 def RemoveExport(export):
1970 """Remove an existing export from the node.
1973 @param export: the name of the export to remove
1975 @return: the success of the operation
1978 target = os.path.join(constants.EXPORT_DIR, export)
1980 shutil.rmtree(target)
1981 # TODO: catch some of the relevant exceptions and provide a pretty
1982 # error message if rmtree fails.
1987 def BlockdevRename(devlist):
1988 """Rename a list of block devices.
1990 @type devlist: list of tuples
1991 @param devlist: list of tuples of the form (disk,
1992 new_logical_id, new_physical_id); disk is an
1993 L{objects.Disk} object describing the current disk,
1994 and new logical_id/physical_id is the name we
1997 @return: True if all renames succeeded, False otherwise
2002 for disk, unique_id in devlist:
2003 dev = _RecursiveFindBD(disk)
2005 msgs.append("Can't find device %s in rename" % str(disk))
2009 old_rpath = dev.dev_path
2010 dev.Rename(unique_id)
2011 new_rpath = dev.dev_path
2012 if old_rpath != new_rpath:
2013 DevCacheManager.RemoveCache(old_rpath)
2014 # FIXME: we should add the new cache information here, like:
2015 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2016 # but we don't have the owner here - maybe parse from existing
2017 # cache? for now, we only lose lvm data when we rename, which
2018 # is less critical than DRBD or MD
2019 except errors.BlockDeviceError, err:
2020 msgs.append("Can't rename device '%s' to '%s': %s" %
2021 (dev, unique_id, err))
2022 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2024 return (result, "; ".join(msgs))
2027 def _TransformFileStorageDir(file_storage_dir):
2028 """Checks whether given file_storage_dir is valid.
2030 Checks wheter the given file_storage_dir is within the cluster-wide
2031 default file_storage_dir stored in SimpleStore. Only paths under that
2032 directory are allowed.
2034 @type file_storage_dir: str
2035 @param file_storage_dir: the path to check
2037 @return: the normalized path if valid, None otherwise
2041 file_storage_dir = os.path.normpath(file_storage_dir)
2042 base_file_storage_dir = cfg.GetFileStorageDir()
2043 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2044 base_file_storage_dir):
2045 logging.error("file storage directory '%s' is not under base file"
2046 " storage directory '%s'",
2047 file_storage_dir, base_file_storage_dir)
2049 return file_storage_dir
2052 def CreateFileStorageDir(file_storage_dir):
2053 """Create file storage directory.
2055 @type file_storage_dir: str
2056 @param file_storage_dir: directory to create
2059 @return: tuple with first element a boolean indicating wheter dir
2060 creation was successful or not
2063 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2065 if not file_storage_dir:
2068 if os.path.exists(file_storage_dir):
2069 if not os.path.isdir(file_storage_dir):
2070 logging.error("'%s' is not a directory", file_storage_dir)
2074 os.makedirs(file_storage_dir, 0750)
2075 except OSError, err:
2076 logging.error("Cannot create file storage directory '%s': %s",
2077 file_storage_dir, err)
2082 def RemoveFileStorageDir(file_storage_dir):
2083 """Remove file storage directory.
2085 Remove it only if it's empty. If not log an error and return.
2087 @type file_storage_dir: str
2088 @param file_storage_dir: the directory we should cleanup
2089 @rtype: tuple (success,)
2090 @return: tuple of one element, C{success}, denoting
2091 whether the operation was successfull
2094 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2096 if not file_storage_dir:
2099 if os.path.exists(file_storage_dir):
2100 if not os.path.isdir(file_storage_dir):
2101 logging.error("'%s' is not a directory", file_storage_dir)
2103 # deletes dir only if empty, otherwise we want to return False
2105 os.rmdir(file_storage_dir)
2106 except OSError, err:
2107 logging.exception("Cannot remove file storage directory '%s'",
2113 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2114 """Rename the file storage directory.
2116 @type old_file_storage_dir: str
2117 @param old_file_storage_dir: the current path
2118 @type new_file_storage_dir: str
2119 @param new_file_storage_dir: the name we should rename to
2120 @rtype: tuple (success,)
2121 @return: tuple of one element, C{success}, denoting
2122 whether the operation was successful
2125 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2126 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2128 if not old_file_storage_dir or not new_file_storage_dir:
2131 if not os.path.exists(new_file_storage_dir):
2132 if os.path.isdir(old_file_storage_dir):
2134 os.rename(old_file_storage_dir, new_file_storage_dir)
2135 except OSError, err:
2136 logging.exception("Cannot rename '%s' to '%s'",
2137 old_file_storage_dir, new_file_storage_dir)
2140 logging.error("'%s' is not a directory", old_file_storage_dir)
2143 if os.path.exists(old_file_storage_dir):
2144 logging.error("Cannot rename '%s' to '%s'. Both locations exist.",
2145 old_file_storage_dir, new_file_storage_dir)
2150 def _IsJobQueueFile(file_name):
2151 """Checks whether the given filename is in the queue directory.
2153 @type file_name: str
2154 @param file_name: the file name we should check
2156 @return: whether the file is under the queue directory
2159 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2160 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2163 logging.error("'%s' is not a file in the queue directory",
2169 def JobQueueUpdate(file_name, content):
2170 """Updates a file in the queue directory.
2172 This is just a wrapper over L{utils.WriteFile}, with proper
2175 @type file_name: str
2176 @param file_name: the job file name
2178 @param content: the new job contents
2180 @return: the success of the operation
2183 if not _IsJobQueueFile(file_name):
2186 # Write and replace the file atomically
2187 utils.WriteFile(file_name, data=_Decompress(content))
2192 def JobQueueRename(old, new):
2193 """Renames a job queue file.
2195 This is just a wrapper over os.rename with proper checking.
2198 @param old: the old (actual) file name
2200 @param new: the desired file name
2202 @return: the success of the operation
2205 if not (_IsJobQueueFile(old) and _IsJobQueueFile(new)):
2208 utils.RenameFile(old, new, mkdir=True)
2213 def JobQueueSetDrainFlag(drain_flag):
2214 """Set the drain flag for the queue.
2216 This will set or unset the queue drain flag.
2218 @type drain_flag: boolean
2219 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2221 @return: always True
2222 @warning: the function always returns True
2226 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2228 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2233 def BlockdevClose(instance_name, disks):
2234 """Closes the given block devices.
2236 This means they will be switched to secondary mode (in case of
2239 @param instance_name: if the argument is not empty, the symlinks
2240 of this instance will be removed
2241 @type disks: list of L{objects.Disk}
2242 @param disks: the list of disks to be closed
2243 @rtype: tuple (success, message)
2244 @return: a tuple of success and message, where success
2245 indicates the succes of the operation, and message
2246 which will contain the error details in case we
2252 rd = _RecursiveFindBD(cf)
2254 _Fail("Can't find device %s", cf)
2261 except errors.BlockDeviceError, err:
2262 msg.append(str(err))
2264 return (False, "Can't make devices secondary: %s" % ",".join(msg))
2267 _RemoveBlockDevLinks(instance_name, disks)
2268 return (True, "All devices secondary")
2271 def ValidateHVParams(hvname, hvparams):
2272 """Validates the given hypervisor parameters.
2274 @type hvname: string
2275 @param hvname: the hypervisor name
2276 @type hvparams: dict
2277 @param hvparams: the hypervisor parameters to be validated
2278 @rtype: tuple (success, message)
2279 @return: a tuple of success and message, where success
2280 indicates the succes of the operation, and message
2281 which will contain the error details in case we
2286 hv_type = hypervisor.GetHypervisor(hvname)
2287 hv_type.ValidateParameters(hvparams)
2288 return (True, "Validation passed")
2289 except errors.HypervisorError, err:
2290 return (False, str(err))
2294 """Demotes the current node from master candidate role.
2297 # try to ensure we're not the master by mistake
2298 master, myself = ssconf.GetMasterAndMyself()
2299 if master == myself:
2300 return (False, "ssconf status shows I'm the master node, will not demote")
2301 pid_file = utils.DaemonPidFileName(constants.MASTERD_PID)
2302 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2303 return (False, "The master daemon is running, will not demote")
2305 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2306 except EnvironmentError, err:
2307 if err.errno != errno.ENOENT:
2308 return (False, "Error while backing up cluster file: %s" % str(err))
2309 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2310 return (True, "Done")
2313 def _FindDisks(nodes_ip, disks):
2314 """Sets the physical ID on disks and returns the block devices.
2317 # set the correct physical ID
2318 my_name = utils.HostInfo().name
2320 cf.SetPhysicalID(my_name, nodes_ip)
2325 rd = _RecursiveFindBD(cf)
2327 return (False, "Can't find device %s" % cf)
2329 return (True, bdevs)
2332 def DrbdDisconnectNet(nodes_ip, disks):
2333 """Disconnects the network on a list of drbd devices.
2336 status, bdevs = _FindDisks(nodes_ip, disks)
2338 return status, bdevs
2344 except errors.BlockDeviceError, err:
2345 _Fail("Can't change network configuration to standalone mode: %s",
2347 return (True, "All disks are now disconnected")
2350 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2351 """Attaches the network on a list of drbd devices.
2354 status, bdevs = _FindDisks(nodes_ip, disks)
2356 return status, bdevs
2359 for idx, rd in enumerate(bdevs):
2361 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2362 except EnvironmentError, err:
2363 _Fail("Can't create symlink: %s", err)
2364 # reconnect disks, switch to new master configuration and if
2365 # needed primary mode
2368 rd.AttachNet(multimaster)
2369 except errors.BlockDeviceError, err:
2370 _Fail("Can't change network configuration: %s", err)
2371 # wait until the disks are connected; we need to retry the re-attach
2372 # if the device becomes standalone, as this might happen if the one
2373 # node disconnects and reconnects in a different mode before the
2374 # other node reconnects; in this case, one or both of the nodes will
2375 # decide it has wrong configuration and switch to standalone
2376 RECONNECT_TIMEOUT = 2 * 60
2377 sleep_time = 0.100 # start with 100 miliseconds
2378 timeout_limit = time.time() + RECONNECT_TIMEOUT
2379 while time.time() < timeout_limit:
2380 all_connected = True
2382 stats = rd.GetProcStatus()
2383 if not (stats.is_connected or stats.is_in_resync):
2384 all_connected = False
2385 if stats.is_standalone:
2386 # peer had different config info and this node became
2387 # standalone, even though this should not happen with the
2388 # new staged way of changing disk configs
2390 rd.ReAttachNet(multimaster)
2391 except errors.BlockDeviceError, err:
2392 _Fail("Can't change network configuration: %s", err)
2395 time.sleep(sleep_time)
2396 sleep_time = min(5, sleep_time * 1.5)
2397 if not all_connected:
2398 return (False, "Timeout in disk reconnecting")
2400 # change to primary mode
2404 except errors.BlockDeviceError, err:
2405 _Fail("Can't change to primary mode: %s", err)
2407 msg = "multi-master and primary"
2409 msg = "single-master"
2410 return (True, "Disks are now configured as %s" % msg)
2413 def DrbdWaitSync(nodes_ip, disks):
2414 """Wait until DRBDs have synchronized.
2417 status, bdevs = _FindDisks(nodes_ip, disks)
2419 return status, bdevs
2425 stats = rd.GetProcStatus()
2426 if not (stats.is_connected or stats.is_in_resync):
2429 alldone = alldone and (not stats.is_in_resync)
2430 if stats.sync_percent is not None:
2431 min_resync = min(min_resync, stats.sync_percent)
2432 return (not failure, (alldone, min_resync))
2435 def PowercycleNode(hypervisor_type):
2436 """Hard-powercycle the node.
2438 Because we need to return first, and schedule the powercycle in the
2439 background, we won't be able to report failures nicely.
2442 hyper = hypervisor.GetHypervisor(hypervisor_type)
2445 except OSError, err:
2446 # if we can't fork, we'll pretend that we're in the child process
2449 return (True, "Reboot scheduled in 5 seconds")
2451 hyper.PowercycleNode()
2454 class HooksRunner(object):
2457 This class is instantiated on the node side (ganeti-noded) and not
2461 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2463 def __init__(self, hooks_base_dir=None):
2464 """Constructor for hooks runner.
2466 @type hooks_base_dir: str or None
2467 @param hooks_base_dir: if not None, this overrides the
2468 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2471 if hooks_base_dir is None:
2472 hooks_base_dir = constants.HOOKS_BASE_DIR
2473 self._BASE_DIR = hooks_base_dir
2476 def ExecHook(script, env):
2477 """Exec one hook script.
2480 @param script: the full path to the script
2482 @param env: the environment with which to exec the script
2483 @rtype: tuple (success, message)
2484 @return: a tuple of success and message, where success
2485 indicates the succes of the operation, and message
2486 which will contain the error details in case we
2490 # exec the process using subprocess and log the output
2493 fdstdin = open("/dev/null", "r")
2494 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2495 stderr=subprocess.STDOUT, close_fds=True,
2496 shell=False, cwd="/", env=env)
2499 output = child.stdout.read(4096)
2500 child.stdout.close()
2501 except EnvironmentError, err:
2502 output += "Hook script error: %s" % str(err)
2506 result = child.wait()
2508 except EnvironmentError, err:
2509 if err.errno == errno.EINTR:
2513 # try not to leak fds
2514 for fd in (fdstdin, ):
2518 except EnvironmentError, err:
2519 # just log the error
2520 #logging.exception("Error while closing fd %s", fd)
2523 return result == 0, utils.SafeEncode(output.strip())
2525 def RunHooks(self, hpath, phase, env):
2526 """Run the scripts in the hooks directory.
2529 @param hpath: the path to the hooks directory which
2532 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2533 L{constants.HOOKS_PHASE_POST}
2535 @param env: dictionary with the environment for the hook
2537 @return: list of 3-element tuples:
2539 - script result, either L{constants.HKR_SUCCESS} or
2540 L{constants.HKR_FAIL}
2541 - output of the script
2543 @raise errors.ProgrammerError: for invalid input
2547 if phase == constants.HOOKS_PHASE_PRE:
2549 elif phase == constants.HOOKS_PHASE_POST:
2552 raise errors.ProgrammerError("Unknown hooks phase: '%s'" % phase)
2555 subdir = "%s-%s.d" % (hpath, suffix)
2556 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2558 dir_contents = utils.ListVisibleFiles(dir_name)
2559 except OSError, err:
2560 # FIXME: must log output in case of failures
2563 # we use the standard python sort order,
2564 # so 00name is the recommended naming scheme
2566 for relname in dir_contents:
2567 fname = os.path.join(dir_name, relname)
2568 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2569 self.RE_MASK.match(relname) is not None):
2570 rrval = constants.HKR_SKIP
2573 result, output = self.ExecHook(fname, env)
2575 rrval = constants.HKR_FAIL
2577 rrval = constants.HKR_SUCCESS
2578 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2583 class IAllocatorRunner(object):
2584 """IAllocator runner.
2586 This class is instantiated on the node side (ganeti-noded) and not on
2590 def Run(self, name, idata):
2591 """Run an iallocator script.
2594 @param name: the iallocator script name
2596 @param idata: the allocator input data
2599 @return: four element tuple of:
2600 - run status (one of the IARUN_ constants)
2603 - fail reason (as from L{utils.RunResult})
2606 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2608 if alloc_script is None:
2609 return (constants.IARUN_NOTFOUND, None, None, None)
2611 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2615 result = utils.RunCmd([alloc_script, fin_name])
2617 return (constants.IARUN_FAILURE, result.stdout, result.stderr,
2622 return (constants.IARUN_SUCCESS, result.stdout, result.stderr, None)
2625 class DevCacheManager(object):
2626 """Simple class for managing a cache of block device information.
2629 _DEV_PREFIX = "/dev/"
2630 _ROOT_DIR = constants.BDEV_CACHE_DIR
2633 def _ConvertPath(cls, dev_path):
2634 """Converts a /dev/name path to the cache file name.
2636 This replaces slashes with underscores and strips the /dev
2637 prefix. It then returns the full path to the cache file.
2640 @param dev_path: the C{/dev/} path name
2642 @return: the converted path name
2645 if dev_path.startswith(cls._DEV_PREFIX):
2646 dev_path = dev_path[len(cls._DEV_PREFIX):]
2647 dev_path = dev_path.replace("/", "_")
2648 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2652 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2653 """Updates the cache information for a given device.
2656 @param dev_path: the pathname of the device
2658 @param owner: the owner (instance name) of the device
2659 @type on_primary: bool
2660 @param on_primary: whether this is the primary
2663 @param iv_name: the instance-visible name of the
2664 device, as in objects.Disk.iv_name
2669 if dev_path is None:
2670 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2672 fpath = cls._ConvertPath(dev_path)
2678 iv_name = "not_visible"
2679 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2681 utils.WriteFile(fpath, data=fdata)
2682 except EnvironmentError, err:
2683 logging.exception("Can't update bdev cache for %s", dev_path)
2686 def RemoveCache(cls, dev_path):
2687 """Remove data for a dev_path.
2689 This is just a wrapper over L{utils.RemoveFile} with a converted
2690 path name and logging.
2693 @param dev_path: the pathname of the device
2698 if dev_path is None:
2699 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2701 fpath = cls._ConvertPath(dev_path)
2703 utils.RemoveFile(fpath)
2704 except EnvironmentError, err:
2705 logging.exception("Can't update bdev cache for %s", dev_path)