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
50 """Simple wrapper to return a SimpleStore.
52 @rtype: L{ssconf.SimpleStore}
53 @return: a SimpleStore instance
56 return ssconf.SimpleStore()
59 def _GetSshRunner(cluster_name):
60 """Simple wrapper to return an SshRunner.
62 @type cluster_name: str
63 @param cluster_name: the cluster name, which is needed
64 by the SshRunner constructor
65 @rtype: L{ssh.SshRunner}
66 @return: an SshRunner instance
69 return ssh.SshRunner(cluster_name)
72 def _Decompress(data):
73 """Unpacks data compressed by the RPC client.
75 @type data: list or tuple
76 @param data: Data sent by RPC client
78 @return: Decompressed data
81 assert isinstance(data, (list, tuple))
83 (encoding, content) = data
84 if encoding == constants.RPC_ENCODING_NONE:
86 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
87 return zlib.decompress(base64.b64decode(content))
89 raise AssertionError("Unknown data encoding")
92 def _CleanDirectory(path, exclude=None):
93 """Removes all regular files in a directory.
96 @param path: the directory to clean
98 @param exclude: list of files to be excluded, defaults
102 if not os.path.isdir(path):
107 # Normalize excluded paths
108 exclude = [os.path.normpath(i) for i in exclude]
110 for rel_name in utils.ListVisibleFiles(path):
111 full_name = os.path.normpath(os.path.join(path, rel_name))
112 if full_name in exclude:
114 if os.path.isfile(full_name) and not os.path.islink(full_name):
115 utils.RemoveFile(full_name)
119 """Removes job queue files and archived jobs.
124 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
125 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
129 """Returns master information.
131 This is an utility function to compute master information, either
132 for consumption here or from the node daemon.
135 @return: (master_netdev, master_ip, master_name) if we have a good
136 configuration, otherwise (None, None, None)
141 master_netdev = cfg.GetMasterNetdev()
142 master_ip = cfg.GetMasterIP()
143 master_node = cfg.GetMasterNode()
144 except errors.ConfigurationError, err:
145 logging.exception("Cluster configuration incomplete")
146 return (None, None, None)
147 return (master_netdev, master_ip, master_node)
150 def StartMaster(start_daemons):
151 """Activate local node as master node.
153 The function will always try activate the IP address of the master
154 (unless someone else has it). It will also start the master daemons,
155 based on the start_daemons parameter.
157 @type start_daemons: boolean
158 @param start_daemons: whther to also start the master
159 daemons (ganeti-masterd and ganeti-rapi)
164 master_netdev, master_ip, _ = GetMasterInfo()
165 if not master_netdev:
168 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
169 if utils.OwnIpAddress(master_ip):
170 # we already have the ip:
171 logging.debug("Already started")
173 logging.error("Someone else has the master ip, not activating")
176 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
177 "dev", master_netdev, "label",
178 "%s:0" % master_netdev])
180 logging.error("Can't activate master IP: %s", result.output)
183 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
184 "-s", master_ip, master_ip])
185 # we'll ignore the exit code of arping
187 # and now start the master and rapi daemons
189 for daemon in 'ganeti-masterd', 'ganeti-rapi':
190 result = utils.RunCmd([daemon])
192 logging.error("Can't start daemon %s: %s", daemon, result.output)
197 def StopMaster(stop_daemons):
198 """Deactivate this node as master.
200 The function will always try to deactivate the IP address of the
201 master. It will also stop the master daemons depending on the
202 stop_daemons parameter.
204 @type stop_daemons: boolean
205 @param stop_daemons: whether to also stop the master daemons
206 (ganeti-masterd and ganeti-rapi)
210 master_netdev, master_ip, _ = GetMasterInfo()
211 if not master_netdev:
214 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
215 "dev", master_netdev])
217 logging.error("Can't remove the master IP, error: %s", result.output)
218 # but otherwise ignore the failure
221 # stop/kill the rapi and the master daemon
222 for daemon in constants.RAPI_PID, constants.MASTERD_PID:
223 utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon)))
228 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
229 """Joins this node to the cluster.
231 This does the following:
232 - updates the hostkeys of the machine (rsa and dsa)
233 - adds the ssh private key to the user
234 - adds the ssh public key to the users' authorized_keys file
237 @param dsa: the DSA private key to write
239 @param dsapub: the DSA public key to write
241 @param rsa: the RSA private key to write
243 @param rsapub: the RSA public key to write
245 @param sshkey: the SSH private key to write
247 @param sshpub: the SSH public key to write
249 @return: the success of the operation
252 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
253 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
254 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
255 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
256 for name, content, mode in sshd_keys:
257 utils.WriteFile(name, data=content, mode=mode)
260 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
262 except errors.OpExecError, err:
263 msg = "Error while processing user ssh files"
264 logging.exception(msg)
265 return (False, "%s: %s" % (msg, err))
267 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
268 utils.WriteFile(name, data=content, mode=0600)
270 utils.AddAuthorizedKey(auth_keys, sshpub)
272 utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
274 return (True, "Node added successfully")
278 """Cleans up and remove the current node.
280 This function cleans up and prepares the current node to be removed
283 If processing is successful, then it raises an
284 L{errors.QuitGanetiException} which is used as a special case to
285 shutdown the node daemon.
288 _CleanDirectory(constants.DATA_DIR)
292 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
293 except errors.OpExecError:
294 logging.exception("Error while processing ssh files")
297 f = open(pub_key, 'r')
299 utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
303 utils.RemoveFile(priv_key)
304 utils.RemoveFile(pub_key)
306 # Return a reassuring string to the caller, and quit
307 raise errors.QuitGanetiException(False, 'Shutdown scheduled')
310 def GetNodeInfo(vgname, hypervisor_type):
311 """Gives back a hash with different informations about the node.
313 @type vgname: C{string}
314 @param vgname: the name of the volume group to ask for disk space information
315 @type hypervisor_type: C{str}
316 @param hypervisor_type: the name of the hypervisor to ask for
319 @return: dictionary with the following keys:
320 - vg_size is the size of the configured volume group in MiB
321 - vg_free is the free size of the volume group in MiB
322 - memory_dom0 is the memory allocated for domain0 in MiB
323 - memory_free is the currently available (free) ram in MiB
324 - memory_total is the total number of ram in MiB
328 vginfo = _GetVGInfo(vgname)
329 outputarray['vg_size'] = vginfo['vg_size']
330 outputarray['vg_free'] = vginfo['vg_free']
332 hyper = hypervisor.GetHypervisor(hypervisor_type)
333 hyp_info = hyper.GetNodeInfo()
334 if hyp_info is not None:
335 outputarray.update(hyp_info)
337 f = open("/proc/sys/kernel/random/boot_id", 'r')
339 outputarray["bootid"] = f.read(128).rstrip("\n")
346 def VerifyNode(what, cluster_name):
347 """Verify the status of the local node.
349 Based on the input L{what} parameter, various checks are done on the
352 If the I{filelist} key is present, this list of
353 files is checksummed and the file/checksum pairs are returned.
355 If the I{nodelist} key is present, we check that we have
356 connectivity via ssh with the target nodes (and check the hostname
359 If the I{node-net-test} key is present, we check that we have
360 connectivity to the given nodes via both primary IP and, if
361 applicable, secondary IPs.
364 @param what: a dictionary of things to check:
365 - filelist: list of files for which to compute checksums
366 - nodelist: list of nodes we should check ssh communication with
367 - node-net-test: list of nodes we should check node daemon port
369 - hypervisor: list with hypervisors to run the verify for
371 @return: a dictionary with the same keys as the input dict, and
372 values representing the result of the checks
377 if constants.NV_HYPERVISOR in what:
378 result[constants.NV_HYPERVISOR] = tmp = {}
379 for hv_name in what[constants.NV_HYPERVISOR]:
380 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
382 if constants.NV_FILELIST in what:
383 result[constants.NV_FILELIST] = utils.FingerprintFiles(
384 what[constants.NV_FILELIST])
386 if constants.NV_NODELIST in what:
387 result[constants.NV_NODELIST] = tmp = {}
388 random.shuffle(what[constants.NV_NODELIST])
389 for node in what[constants.NV_NODELIST]:
390 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
394 if constants.NV_NODENETTEST in what:
395 result[constants.NV_NODENETTEST] = tmp = {}
396 my_name = utils.HostInfo().name
397 my_pip = my_sip = None
398 for name, pip, sip in what[constants.NV_NODENETTEST]:
404 tmp[my_name] = ("Can't find my own primary/secondary IP"
407 port = utils.GetNodeDaemonPort()
408 for name, pip, sip in what[constants.NV_NODENETTEST]:
410 if not utils.TcpPing(pip, port, source=my_pip):
411 fail.append("primary")
413 if not utils.TcpPing(sip, port, source=my_sip):
414 fail.append("secondary")
416 tmp[name] = ("failure using the %s interface(s)" %
419 if constants.NV_LVLIST in what:
420 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
422 if constants.NV_INSTANCELIST in what:
423 result[constants.NV_INSTANCELIST] = GetInstanceList(
424 what[constants.NV_INSTANCELIST])
426 if constants.NV_VGLIST in what:
427 result[constants.NV_VGLIST] = ListVolumeGroups()
429 if constants.NV_VERSION in what:
430 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
431 constants.RELEASE_VERSION)
433 if constants.NV_HVINFO in what:
434 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
435 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
437 if constants.NV_DRBDLIST in what:
439 used_minors = bdev.DRBD8.GetUsedDevs().keys()
440 except errors.BlockDeviceError, err:
441 logging.warning("Can't get used minors list", exc_info=True)
442 used_minors = str(err)
443 result[constants.NV_DRBDLIST] = used_minors
448 def GetVolumeList(vg_name):
449 """Compute list of logical volumes and their size.
452 @param vg_name: the volume group whose LVs we should list
455 dictionary of all partions (key) with value being a tuple of
456 their size (in MiB), inactive and online status::
458 {'test1': ('20.06', True, True)}
460 in case of errors, a string is returned with the error
466 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
467 "--separator=%s" % sep,
468 "-olv_name,lv_size,lv_attr", vg_name])
470 logging.error("Failed to list logical volumes, lvs output: %s",
474 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
475 for line in result.stdout.splitlines():
477 match = valid_line_re.match(line)
479 logging.error("Invalid line returned from lvs output: '%s'", line)
481 name, size, attr = match.groups()
482 inactive = attr[4] == '-'
483 online = attr[5] == 'o'
484 lvs[name] = (size, inactive, online)
489 def ListVolumeGroups():
490 """List the volume groups and their size.
493 @return: dictionary with keys volume name and values the
497 return utils.ListVolumeGroups()
501 """List all volumes on this node.
505 A list of dictionaries, each having four keys:
506 - name: the logical volume name,
507 - size: the size of the logical volume
508 - dev: the physical device on which the LV lives
509 - vg: the volume group to which it belongs
511 In case of errors, we return an empty list and log the
514 Note that since a logical volume can live on multiple physical
515 volumes, the resulting list might include a logical volume
519 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
521 "--options=lv_name,lv_size,devices,vg_name"])
523 logging.error("Failed to list logical volumes, lvs output: %s",
529 return dev.split('(')[0]
535 'name': line[0].strip(),
536 'size': line[1].strip(),
537 'dev': parse_dev(line[2].strip()),
538 'vg': line[3].strip(),
541 return [map_line(line.split('|')) for line in result.stdout.splitlines()
542 if line.count('|') >= 3]
545 def BridgesExist(bridges_list):
546 """Check if a list of bridges exist on the current node.
549 @return: C{True} if all of them exist, C{False} otherwise
552 for bridge in bridges_list:
553 if not utils.BridgeExists(bridge):
559 def GetInstanceList(hypervisor_list):
560 """Provides a list of instances.
562 @type hypervisor_list: list
563 @param hypervisor_list: the list of hypervisors to query information
566 @return: a list of all running instances on the current node
567 - instance1.example.com
568 - instance2.example.com
572 for hname in hypervisor_list:
574 names = hypervisor.GetHypervisor(hname).ListInstances()
575 results.extend(names)
576 except errors.HypervisorError, err:
577 logging.exception("Error enumerating instances for hypevisor %s", hname)
583 def GetInstanceInfo(instance, hname):
584 """Gives back the informations about an instance as a dictionary.
586 @type instance: string
587 @param instance: the instance name
589 @param hname: the hypervisor type of the instance
592 @return: dictionary with the following keys:
593 - memory: memory size of instance (int)
594 - state: xen state of instance (string)
595 - time: cpu time of instance (float)
600 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
601 if iinfo is not None:
602 output['memory'] = iinfo[2]
603 output['state'] = iinfo[4]
604 output['time'] = iinfo[5]
609 def GetInstanceMigratable(instance):
610 """Gives whether an instance can be migrated.
612 @type instance: L{objects.Instance}
613 @param instance: object representing the instance to be checked.
616 @return: tuple of (result, description) where:
617 - result: whether the instance can be migrated or not
618 - description: a description of the issue, if relevant
621 hyper = hypervisor.GetHypervisor(instance.hypervisor)
622 if instance.name not in hyper.ListInstances():
623 return (False, 'not running')
625 for idx in range(len(instance.disks)):
626 link_name = _GetBlockDevSymlinkPath(instance.name, idx)
627 if not os.path.islink(link_name):
628 return (False, 'not restarted since ganeti 1.2.5')
633 def GetAllInstancesInfo(hypervisor_list):
634 """Gather data about all instances.
636 This is the equivalent of L{GetInstanceInfo}, except that it
637 computes data for all instances at once, thus being faster if one
638 needs data about more than one instance.
640 @type hypervisor_list: list
641 @param hypervisor_list: list of hypervisors to query for instance data
644 @return: dictionary of instance: data, with data having the following keys:
645 - memory: memory size of instance (int)
646 - state: xen state of instance (string)
647 - time: cpu time of instance (float)
648 - vcpus: the number of vcpus
653 for hname in hypervisor_list:
654 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
656 for name, inst_id, memory, vcpus, state, times in iinfo:
664 # we only check static parameters, like memory and vcpus,
665 # and not state and time which can change between the
666 # invocations of the different hypervisors
667 for key in 'memory', 'vcpus':
668 if value[key] != output[name][key]:
669 raise errors.HypervisorError("Instance %s is running twice"
670 " with different parameters" % name)
676 def InstanceOsAdd(instance, reinstall):
677 """Add an OS to an instance.
679 @type instance: L{objects.Instance}
680 @param instance: Instance whose OS is to be installed
681 @type reinstall: boolean
682 @param reinstall: whether this is an instance reinstall
684 @return: the success of the operation
688 inst_os = OSFromDisk(instance.os)
689 except errors.InvalidOS, err:
690 os_name, os_dir, os_err = err.args
692 return (False, "Can't find OS '%s': %s" % (os_name, os_err))
694 return (False, "Error parsing OS '%s' in directory %s: %s" %
695 (os_name, os_dir, os_err))
697 create_env = OSEnvironment(instance)
699 create_env['INSTANCE_REINSTALL'] = "1"
701 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
702 instance.name, int(time.time()))
704 result = utils.RunCmd([inst_os.create_script], env=create_env,
705 cwd=inst_os.path, output=logfile,)
707 logging.error("os create command '%s' returned error: %s, logfile: %s,"
708 " output: %s", result.cmd, result.fail_reason, logfile,
710 lines = [utils.SafeEncode(val)
711 for val in utils.TailFile(logfile, lines=20)]
712 return (False, "OS create script failed (%s), last lines in the"
713 " log file:\n%s" % (result.fail_reason, "\n".join(lines)))
715 return (True, "Successfully installed")
718 def RunRenameInstance(instance, old_name):
719 """Run the OS rename script for an instance.
721 @type instance: L{objects.Instance}
722 @param instance: Instance whose OS is to be installed
723 @type old_name: string
724 @param old_name: previous instance name
726 @return: the success of the operation
729 inst_os = OSFromDisk(instance.os)
731 rename_env = OSEnvironment(instance)
732 rename_env['OLD_INSTANCE_NAME'] = old_name
734 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
736 instance.name, int(time.time()))
738 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
739 cwd=inst_os.path, output=logfile)
742 logging.error("os create command '%s' returned error: %s output: %s",
743 result.cmd, result.fail_reason, result.output)
744 lines = [utils.SafeEncode(val)
745 for val in utils.TailFile(logfile, lines=20)]
746 return (False, "OS rename script failed (%s), last lines in the"
747 " log file:\n%s" % (result.fail_reason, "\n".join(lines)))
749 return (True, "Rename successful")
752 def _GetVGInfo(vg_name):
753 """Get informations about the volume group.
756 @param vg_name: the volume group which we query
759 A dictionary with the following keys:
760 - C{vg_size} is the total size of the volume group in MiB
761 - C{vg_free} is the free size of the volume group in MiB
762 - C{pv_count} are the number of physical disks in that VG
764 If an error occurs during gathering of data, we return the same dict
765 with keys all set to None.
768 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
770 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
771 "--nosuffix", "--units=m", "--separator=:", vg_name])
774 logging.error("volume group %s not present", vg_name)
776 valarr = retval.stdout.strip().rstrip(':').split(':')
780 "vg_size": int(round(float(valarr[0]), 0)),
781 "vg_free": int(round(float(valarr[1]), 0)),
782 "pv_count": int(valarr[2]),
784 except ValueError, err:
785 logging.exception("Fail to parse vgs output")
787 logging.error("vgs output has the wrong number of fields (expected"
788 " three): %s", str(valarr))
792 def _GetBlockDevSymlinkPath(instance_name, idx):
793 return os.path.join(constants.DISK_LINKS_DIR,
794 "%s:%d" % (instance_name, idx))
797 def _SymlinkBlockDev(instance_name, device_path, idx):
798 """Set up symlinks to a instance's block device.
800 This is an auxiliary function run when an instance is start (on the primary
801 node) or when an instance is migrated (on the target node).
804 @param instance_name: the name of the target instance
805 @param device_path: path of the physical block device, on the node
806 @param idx: the disk index
807 @return: absolute path to the disk's symlink
810 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
812 os.symlink(device_path, link_name)
814 if err.errno == errno.EEXIST:
815 if (not os.path.islink(link_name) or
816 os.readlink(link_name) != device_path):
818 os.symlink(device_path, link_name)
825 def _RemoveBlockDevLinks(instance_name, disks):
826 """Remove the block device symlinks belonging to the given instance.
829 for idx, disk in enumerate(disks):
830 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
831 if os.path.islink(link_name):
835 logging.exception("Can't remove symlink '%s'", link_name)
838 def _GatherAndLinkBlockDevs(instance):
839 """Set up an instance's block device(s).
841 This is run on the primary node at instance startup. The block
842 devices must be already assembled.
844 @type instance: L{objects.Instance}
845 @param instance: the instance whose disks we shoul assemble
847 @return: list of (disk_object, device_path)
851 for idx, disk in enumerate(instance.disks):
852 device = _RecursiveFindBD(disk)
854 raise errors.BlockDeviceError("Block device '%s' is not set up." %
858 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
860 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
863 block_devices.append((disk, link_name))
868 def StartInstance(instance):
869 """Start an instance.
871 @type instance: L{objects.Instance}
872 @param instance: the instance object
874 @return: whether the startup was successful or not
877 running_instances = GetInstanceList([instance.hypervisor])
879 if instance.name in running_instances:
880 return (True, "Already running")
883 block_devices = _GatherAndLinkBlockDevs(instance)
884 hyper = hypervisor.GetHypervisor(instance.hypervisor)
885 hyper.StartInstance(instance, block_devices)
886 except errors.BlockDeviceError, err:
887 logging.exception("Failed to start instance")
888 return (False, "Block device error: %s" % str(err))
889 except errors.HypervisorError, err:
890 logging.exception("Failed to start instance")
891 _RemoveBlockDevLinks(instance.name, instance.disks)
892 return (False, "Hypervisor error: %s" % str(err))
894 return (True, "Instance started successfully")
897 def InstanceShutdown(instance):
898 """Shut an instance down.
900 @note: this functions uses polling with a hardcoded timeout.
902 @type instance: L{objects.Instance}
903 @param instance: the instance object
905 @return: whether the startup was successful or not
908 hv_name = instance.hypervisor
909 running_instances = GetInstanceList([hv_name])
911 if instance.name not in running_instances:
912 return (True, "Instance already stopped")
914 hyper = hypervisor.GetHypervisor(hv_name)
916 hyper.StopInstance(instance)
917 except errors.HypervisorError, err:
918 msg = "Failed to stop instance %s: %s" % (instance.name, err)
922 # test every 10secs for 2min
925 for dummy in range(11):
926 if instance.name not in GetInstanceList([hv_name]):
930 # the shutdown did not succeed
931 logging.error("Shutdown of '%s' unsuccessful, using destroy",
935 hyper.StopInstance(instance, force=True)
936 except errors.HypervisorError, err:
937 msg = "Failed to force stop instance %s: %s" % (instance.name, err)
942 if instance.name in GetInstanceList([hv_name]):
943 msg = ("Could not shutdown instance %s even by destroy" %
948 _RemoveBlockDevLinks(instance.name, instance.disks)
950 return (True, "Instance has been shutdown successfully")
953 def InstanceReboot(instance, reboot_type):
954 """Reboot an instance.
956 @type instance: L{objects.Instance}
957 @param instance: the instance object to reboot
958 @type reboot_type: str
959 @param reboot_type: the type of reboot, one the following
961 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
962 instance OS, do not recreate the VM
963 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
964 restart the VM (at the hypervisor level)
965 - the other reboot type (L{constants.INSTANCE_REBOOT_HARD})
966 is not accepted here, since that mode is handled
969 @return: the success of the operation
972 running_instances = GetInstanceList([instance.hypervisor])
974 if instance.name not in running_instances:
975 msg = "Cannot reboot instance %s that is not running" % instance.name
979 hyper = hypervisor.GetHypervisor(instance.hypervisor)
980 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
982 hyper.RebootInstance(instance)
983 except errors.HypervisorError, err:
984 msg = "Failed to soft reboot instance %s: %s" % (instance.name, err)
987 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
989 stop_result = InstanceShutdown(instance)
990 if not stop_result[0]:
992 return StartInstance(instance)
993 except errors.HypervisorError, err:
994 msg = "Failed to hard reboot instance %s: %s" % (instance.name, err)
998 return (False, "Invalid reboot_type received: %s" % (reboot_type,))
1000 return (True, "Reboot successful")
1003 def MigrationInfo(instance):
1004 """Gather information about an instance to be migrated.
1006 @type instance: L{objects.Instance}
1007 @param instance: the instance definition
1010 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1012 info = hyper.MigrationInfo(instance)
1013 except errors.HypervisorError, err:
1014 msg = "Failed to fetch migration information"
1015 logging.exception(msg)
1016 return (False, '%s: %s' % (msg, err))
1020 def AcceptInstance(instance, info, target):
1021 """Prepare the node to accept an instance.
1023 @type instance: L{objects.Instance}
1024 @param instance: the instance definition
1025 @type info: string/data (opaque)
1026 @param info: migration information, from the source node
1027 @type target: string
1028 @param target: target host (usually ip), on this node
1031 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1033 hyper.AcceptInstance(instance, info, target)
1034 except errors.HypervisorError, err:
1035 msg = "Failed to accept instance"
1036 logging.exception(msg)
1037 return (False, '%s: %s' % (msg, err))
1038 return (True, "Accept successfull")
1041 def FinalizeMigration(instance, info, success):
1042 """Finalize any preparation to accept an instance.
1044 @type instance: L{objects.Instance}
1045 @param instance: the instance definition
1046 @type info: string/data (opaque)
1047 @param info: migration information, from the source node
1048 @type success: boolean
1049 @param success: whether the migration was a success or a failure
1052 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1054 hyper.FinalizeMigration(instance, info, success)
1055 except errors.HypervisorError, err:
1056 msg = "Failed to finalize migration"
1057 logging.exception(msg)
1058 return (False, '%s: %s' % (msg, err))
1059 return (True, "Migration Finalized")
1062 def MigrateInstance(instance, target, live):
1063 """Migrates an instance to another node.
1065 @type instance: L{objects.Instance}
1066 @param instance: the instance definition
1067 @type target: string
1068 @param target: the target node name
1070 @param live: whether the migration should be done live or not (the
1071 interpretation of this parameter is left to the hypervisor)
1073 @return: a tuple of (success, msg) where:
1074 - succes is a boolean denoting the success/failure of the operation
1075 - msg is a string with details in case of failure
1078 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1081 hyper.MigrateInstance(instance.name, target, live)
1082 except errors.HypervisorError, err:
1083 msg = "Failed to migrate instance"
1084 logging.exception(msg)
1085 return (False, "%s: %s" % (msg, err))
1086 return (True, "Migration successfull")
1089 def BlockdevCreate(disk, size, owner, on_primary, info):
1090 """Creates a block device for an instance.
1092 @type disk: L{objects.Disk}
1093 @param disk: the object describing the disk we should create
1095 @param size: the size of the physical underlying device, in MiB
1097 @param owner: the name of the instance for which disk is created,
1098 used for device cache data
1099 @type on_primary: boolean
1100 @param on_primary: indicates if it is the primary node or not
1102 @param info: string that will be sent to the physical device
1103 creation, used for example to set (LVM) tags on LVs
1105 @return: the new unique_id of the device (this can sometime be
1106 computed only after creation), or None. On secondary nodes,
1107 it's not required to return anything.
1112 for child in disk.children:
1114 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1115 except errors.BlockDeviceError, err:
1116 errmsg = "Can't assemble device %s: %s" % (child, err)
1117 logging.error(errmsg)
1118 return False, errmsg
1119 if on_primary or disk.AssembleOnSecondary():
1120 # we need the children open in case the device itself has to
1124 except errors.BlockDeviceError, err:
1125 errmsg = "Can't make child '%s' read-write: %s" % (child, err)
1126 logging.error(errmsg)
1127 return False, errmsg
1131 device = bdev.Create(disk.dev_type, disk.physical_id, clist, size)
1132 except errors.BlockDeviceError, err:
1133 return False, "Can't create block device: %s" % str(err)
1135 if on_primary or disk.AssembleOnSecondary():
1138 except errors.BlockDeviceError, err:
1139 errmsg = ("Can't assemble device after creation, very"
1140 " unusual event: %s" % str(err))
1141 logging.error(errmsg)
1142 return False, errmsg
1143 device.SetSyncSpeed(constants.SYNC_SPEED)
1144 if on_primary or disk.OpenOnSecondary():
1146 device.Open(force=True)
1147 except errors.BlockDeviceError, err:
1148 errmsg = ("Can't make device r/w after creation, very"
1149 " unusual event: %s" % str(err))
1150 logging.error(errmsg)
1151 return False, errmsg
1152 DevCacheManager.UpdateCache(device.dev_path, owner,
1153 on_primary, disk.iv_name)
1155 device.SetInfo(info)
1157 physical_id = device.unique_id
1158 return True, physical_id
1161 def BlockdevRemove(disk):
1162 """Remove a block device.
1164 @note: This is intended to be called recursively.
1166 @type disk: L{objects.Disk}
1167 @param disk: the disk object we should remove
1169 @return: the success of the operation
1175 rdev = _RecursiveFindBD(disk)
1176 except errors.BlockDeviceError, err:
1177 # probably can't attach
1178 logging.info("Can't attach to device %s in remove", disk)
1180 if rdev is not None:
1181 r_path = rdev.dev_path
1184 except errors.BlockDeviceError, err:
1185 msgs.append(str(err))
1188 DevCacheManager.RemoveCache(r_path)
1191 for child in disk.children:
1192 c_status, c_msg = BlockdevRemove(child)
1193 result = result and c_status
1194 if c_msg: # not an empty message
1197 return (result, "; ".join(msgs))
1200 def _RecursiveAssembleBD(disk, owner, as_primary):
1201 """Activate a block device for an instance.
1203 This is run on the primary and secondary nodes for an instance.
1205 @note: this function is called recursively.
1207 @type disk: L{objects.Disk}
1208 @param disk: the disk we try to assemble
1210 @param owner: the name of the instance which owns the disk
1211 @type as_primary: boolean
1212 @param as_primary: if we should make the block device
1215 @return: the assembled device or None (in case no device
1217 @raise errors.BlockDeviceError: in case there is an error
1218 during the activation of the children or the device
1224 mcn = disk.ChildrenNeeded()
1226 mcn = 0 # max number of Nones allowed
1228 mcn = len(disk.children) - mcn # max number of Nones
1229 for chld_disk in disk.children:
1231 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1232 except errors.BlockDeviceError, err:
1233 if children.count(None) >= mcn:
1236 logging.error("Error in child activation (but continuing): %s",
1238 children.append(cdev)
1240 if as_primary or disk.AssembleOnSecondary():
1241 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children)
1242 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1244 if as_primary or disk.OpenOnSecondary():
1246 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1247 as_primary, disk.iv_name)
1254 def BlockdevAssemble(disk, owner, as_primary):
1255 """Activate a block device for an instance.
1257 This is a wrapper over _RecursiveAssembleBD.
1259 @rtype: str or boolean
1260 @return: a C{/dev/...} path for primary nodes, and
1261 C{True} for secondary nodes
1265 result = "no error information"
1267 result = _RecursiveAssembleBD(disk, owner, as_primary)
1268 if isinstance(result, bdev.BlockDev):
1269 result = result.dev_path
1270 except errors.BlockDeviceError, err:
1271 result = "Error while assembling disk: %s" % str(err)
1273 return (status, result)
1276 def BlockdevShutdown(disk):
1277 """Shut down a block device.
1279 First, if the device is assembled (Attach() is successfull), then
1280 the device is shutdown. Then the children of the device are
1283 This function is called recursively. Note that we don't cache the
1284 children or such, as oppossed to assemble, shutdown of different
1285 devices doesn't require that the upper device was active.
1287 @type disk: L{objects.Disk}
1288 @param disk: the description of the disk we should
1291 @return: the success of the operation
1296 r_dev = _RecursiveFindBD(disk)
1297 if r_dev is not None:
1298 r_path = r_dev.dev_path
1301 DevCacheManager.RemoveCache(r_path)
1302 except errors.BlockDeviceError, err:
1303 msgs.append(str(err))
1307 for child in disk.children:
1308 c_status, c_msg = BlockdevShutdown(child)
1309 result = result and c_status
1310 if c_msg: # not an empty message
1313 return (result, "; ".join(msgs))
1316 def BlockdevAddchildren(parent_cdev, new_cdevs):
1317 """Extend a mirrored block device.
1319 @type parent_cdev: L{objects.Disk}
1320 @param parent_cdev: the disk to which we should add children
1321 @type new_cdevs: list of L{objects.Disk}
1322 @param new_cdevs: the list of children which we should add
1324 @return: the success of the operation
1327 parent_bdev = _RecursiveFindBD(parent_cdev)
1328 if parent_bdev is None:
1329 msg = "Can't find parent device %s" % str(parent_cdev)
1330 logging.error("BlockdevAddchildren: %s", msg)
1332 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1333 if new_bdevs.count(None) > 0:
1334 msg = "Can't find new device(s) to add: %s:%s" % (new_bdevs, new_cdevs)
1337 parent_bdev.AddChildren(new_bdevs)
1341 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1342 """Shrink a mirrored block device.
1344 @type parent_cdev: L{objects.Disk}
1345 @param parent_cdev: the disk from which we should remove children
1346 @type new_cdevs: list of L{objects.Disk}
1347 @param new_cdevs: the list of children which we should remove
1349 @return: the success of the operation
1352 parent_bdev = _RecursiveFindBD(parent_cdev)
1353 if parent_bdev is None:
1354 logging.error("Can't find parent in remove children: %s", parent_cdev)
1357 for disk in new_cdevs:
1358 rpath = disk.StaticDevPath()
1360 bd = _RecursiveFindBD(disk)
1362 logging.error("Can't find dynamic device %s while removing children",
1366 devs.append(bd.dev_path)
1369 parent_bdev.RemoveChildren(devs)
1373 def BlockdevGetmirrorstatus(disks):
1374 """Get the mirroring status of a list of devices.
1376 @type disks: list of L{objects.Disk}
1377 @param disks: the list of disks which we should query
1380 a list of (mirror_done, estimated_time) tuples, which
1381 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1382 @raise errors.BlockDeviceError: if any of the disks cannot be
1388 rbd = _RecursiveFindBD(dsk)
1390 raise errors.BlockDeviceError("Can't find device %s" % str(dsk))
1391 stats.append(rbd.CombinedSyncStatus())
1395 def _RecursiveFindBD(disk):
1396 """Check if a device is activated.
1398 If so, return informations about the real device.
1400 @type disk: L{objects.Disk}
1401 @param disk: the disk object we need to find
1403 @return: None if the device can't be found,
1404 otherwise the device instance
1409 for chdisk in disk.children:
1410 children.append(_RecursiveFindBD(chdisk))
1412 return bdev.FindDevice(disk.dev_type, disk.physical_id, children)
1415 def BlockdevFind(disk):
1416 """Check if a device is activated.
1418 If it is, return informations about the real device.
1420 @type disk: L{objects.Disk}
1421 @param disk: the disk to find
1422 @rtype: None or tuple
1423 @return: None if the disk cannot be found, otherwise a
1424 tuple (device_path, major, minor, sync_percent,
1425 estimated_time, is_degraded)
1429 rbd = _RecursiveFindBD(disk)
1430 except errors.BlockDeviceError, err:
1431 return (False, str(err))
1434 return (True, (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus())
1437 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1438 """Write a file to the filesystem.
1440 This allows the master to overwrite(!) a file. It will only perform
1441 the operation if the file belongs to a list of configuration files.
1443 @type file_name: str
1444 @param file_name: the target file name
1446 @param data: the new contents of the file
1448 @param mode: the mode to give the file (can be None)
1450 @param uid: the owner of the file (can be -1 for default)
1452 @param gid: the group of the file (can be -1 for default)
1454 @param atime: the atime to set on the file (can be None)
1456 @param mtime: the mtime to set on the file (can be None)
1458 @return: the success of the operation; errors are logged
1459 in the node daemon log
1462 if not os.path.isabs(file_name):
1463 err = "Filename passed to UploadFile is not absolute: '%s'" % file_name
1467 allowed_files = set([
1468 constants.CLUSTER_CONF_FILE,
1469 constants.ETC_HOSTS,
1470 constants.SSH_KNOWN_HOSTS_FILE,
1471 constants.VNC_PASSWORD_FILE,
1472 constants.RAPI_CERT_FILE,
1473 constants.RAPI_USERS_FILE,
1476 for hv_name in constants.HYPER_TYPES:
1477 hv_class = hypervisor.GetHypervisor(hv_name)
1478 allowed_files.update(hv_class.GetAncillaryFiles())
1480 if file_name not in allowed_files:
1481 err = "Filename passed to UploadFile not in allowed upload targets: '%s'" \
1486 raw_data = _Decompress(data)
1488 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1489 atime=atime, mtime=mtime)
1490 return (True, "success")
1493 def WriteSsconfFiles(values):
1494 """Update all ssconf files.
1496 Wrapper around the SimpleStore.WriteFiles.
1499 ssconf.SimpleStore().WriteFiles(values)
1502 def _ErrnoOrStr(err):
1503 """Format an EnvironmentError exception.
1505 If the L{err} argument has an errno attribute, it will be looked up
1506 and converted into a textual C{E...} description. Otherwise the
1507 string representation of the error will be returned.
1509 @type err: L{EnvironmentError}
1510 @param err: the exception to format
1513 if hasattr(err, 'errno'):
1514 detail = errno.errorcode[err.errno]
1520 def _OSOndiskVersion(name, os_dir):
1521 """Compute and return the API version of a given OS.
1523 This function will try to read the API version of the OS given by
1524 the 'name' parameter and residing in the 'os_dir' directory.
1527 @param name: the OS name we should look for
1529 @param os_dir: the directory inwhich we should look for the OS
1532 Either an integer denoting the version or None in the
1533 case when this is not a valid OS name.
1534 @raise errors.InvalidOS: if the OS cannot be found
1537 api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
1540 st = os.stat(api_file)
1541 except EnvironmentError, err:
1542 raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file not"
1543 " found (%s)" % _ErrnoOrStr(err))
1545 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1546 raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file is not"
1552 api_versions = f.readlines()
1555 except EnvironmentError, err:
1556 raise errors.InvalidOS(name, os_dir, "error while reading the"
1557 " API version (%s)" % _ErrnoOrStr(err))
1559 api_versions = [version.strip() for version in api_versions]
1561 api_versions = [int(version) for version in api_versions]
1562 except (TypeError, ValueError), err:
1563 raise errors.InvalidOS(name, os_dir,
1564 "API version is not integer (%s)" % str(err))
1569 def DiagnoseOS(top_dirs=None):
1570 """Compute the validity for all OSes.
1572 @type top_dirs: list
1573 @param top_dirs: the list of directories in which to
1574 search (if not given defaults to
1575 L{constants.OS_SEARCH_PATH})
1576 @rtype: list of L{objects.OS}
1577 @return: an OS object for each name in all the given
1581 if top_dirs is None:
1582 top_dirs = constants.OS_SEARCH_PATH
1585 for dir_name in top_dirs:
1586 if os.path.isdir(dir_name):
1588 f_names = utils.ListVisibleFiles(dir_name)
1589 except EnvironmentError, err:
1590 logging.exception("Can't list the OS directory %s", dir_name)
1592 for name in f_names:
1594 os_inst = OSFromDisk(name, base_dir=dir_name)
1595 result.append(os_inst)
1596 except errors.InvalidOS, err:
1597 result.append(objects.OS.FromInvalidOS(err))
1602 def OSFromDisk(name, base_dir=None):
1603 """Create an OS instance from disk.
1605 This function will return an OS instance if the given name is a
1606 valid OS name. Otherwise, it will raise an appropriate
1607 L{errors.InvalidOS} exception, detailing why this is not a valid OS.
1609 @type base_dir: string
1610 @keyword base_dir: Base directory containing OS installations.
1611 Defaults to a search in all the OS_SEARCH_PATH dirs.
1612 @rtype: L{objects.OS}
1613 @return: the OS instance if we find a valid one
1614 @raise errors.InvalidOS: if we don't find a valid OS
1617 if base_dir is None:
1618 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1620 raise errors.InvalidOS(name, None, "OS dir not found in search path")
1622 os_dir = os.path.sep.join([base_dir, name])
1624 api_versions = _OSOndiskVersion(name, os_dir)
1626 if constants.OS_API_VERSION not in api_versions:
1627 raise errors.InvalidOS(name, os_dir, "API version mismatch"
1628 " (found %s want %s)"
1629 % (api_versions, constants.OS_API_VERSION))
1631 # OS Scripts dictionary, we will populate it with the actual script names
1632 os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
1634 for script in os_scripts:
1635 os_scripts[script] = os.path.sep.join([os_dir, script])
1638 st = os.stat(os_scripts[script])
1639 except EnvironmentError, err:
1640 raise errors.InvalidOS(name, os_dir, "'%s' script missing (%s)" %
1641 (script, _ErrnoOrStr(err)))
1643 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1644 raise errors.InvalidOS(name, os_dir, "'%s' script not executable" %
1647 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1648 raise errors.InvalidOS(name, os_dir, "'%s' is not a regular file" %
1652 return objects.OS(name=name, path=os_dir, status=constants.OS_VALID_STATUS,
1653 create_script=os_scripts[constants.OS_SCRIPT_CREATE],
1654 export_script=os_scripts[constants.OS_SCRIPT_EXPORT],
1655 import_script=os_scripts[constants.OS_SCRIPT_IMPORT],
1656 rename_script=os_scripts[constants.OS_SCRIPT_RENAME],
1657 api_versions=api_versions)
1659 def OSEnvironment(instance, debug=0):
1660 """Calculate the environment for an os script.
1662 @type instance: L{objects.Instance}
1663 @param instance: target instance for the os script run
1664 @type debug: integer
1665 @param debug: debug level (0 or 1, for OS Api 10)
1667 @return: dict of environment variables
1668 @raise errors.BlockDeviceError: if the block device
1673 result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION
1674 result['INSTANCE_NAME'] = instance.name
1675 result['INSTANCE_OS'] = instance.os
1676 result['HYPERVISOR'] = instance.hypervisor
1677 result['DISK_COUNT'] = '%d' % len(instance.disks)
1678 result['NIC_COUNT'] = '%d' % len(instance.nics)
1679 result['DEBUG_LEVEL'] = '%d' % debug
1680 for idx, disk in enumerate(instance.disks):
1681 real_disk = _RecursiveFindBD(disk)
1682 if real_disk is None:
1683 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1686 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1687 result['DISK_%d_ACCESS' % idx] = disk.mode
1688 if constants.HV_DISK_TYPE in instance.hvparams:
1689 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1690 instance.hvparams[constants.HV_DISK_TYPE]
1691 if disk.dev_type in constants.LDS_BLOCK:
1692 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1693 elif disk.dev_type == constants.LD_FILE:
1694 result['DISK_%d_BACKEND_TYPE' % idx] = \
1695 'file:%s' % disk.physical_id[0]
1696 for idx, nic in enumerate(instance.nics):
1697 result['NIC_%d_MAC' % idx] = nic.mac
1699 result['NIC_%d_IP' % idx] = nic.ip
1700 result['NIC_%d_BRIDGE' % idx] = nic.bridge
1701 if constants.HV_NIC_TYPE in instance.hvparams:
1702 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1703 instance.hvparams[constants.HV_NIC_TYPE]
1707 def BlockdevGrow(disk, amount):
1708 """Grow a stack of block devices.
1710 This function is called recursively, with the childrens being the
1711 first ones to resize.
1713 @type disk: L{objects.Disk}
1714 @param disk: the disk to be grown
1715 @rtype: (status, result)
1716 @return: a tuple with the status of the operation
1717 (True/False), and the errors message if status
1721 r_dev = _RecursiveFindBD(disk)
1723 return False, "Cannot find block device %s" % (disk,)
1727 except errors.BlockDeviceError, err:
1728 return False, str(err)
1733 def BlockdevSnapshot(disk):
1734 """Create a snapshot copy of a block device.
1736 This function is called recursively, and the snapshot is actually created
1737 just for the leaf lvm backend device.
1739 @type disk: L{objects.Disk}
1740 @param disk: the disk to be snapshotted
1742 @return: snapshot disk path
1746 if len(disk.children) == 1:
1747 # only one child, let's recurse on it
1748 return BlockdevSnapshot(disk.children[0])
1750 # more than one child, choose one that matches
1751 for child in disk.children:
1752 if child.size == disk.size:
1753 # return implies breaking the loop
1754 return BlockdevSnapshot(child)
1755 elif disk.dev_type == constants.LD_LV:
1756 r_dev = _RecursiveFindBD(disk)
1757 if r_dev is not None:
1758 # let's stay on the safe side and ask for the full size, for now
1759 return r_dev.Snapshot(disk.size)
1763 raise errors.ProgrammerError("Cannot snapshot non-lvm block device"
1764 " '%s' of type '%s'" %
1765 (disk.unique_id, disk.dev_type))
1768 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1769 """Export a block device snapshot to a remote node.
1771 @type disk: L{objects.Disk}
1772 @param disk: the description of the disk to export
1773 @type dest_node: str
1774 @param dest_node: the destination node to export to
1775 @type instance: L{objects.Instance}
1776 @param instance: the instance object to whom the disk belongs
1777 @type cluster_name: str
1778 @param cluster_name: the cluster name, needed for SSH hostalias
1780 @param idx: the index of the disk in the instance's disk list,
1781 used to export to the OS scripts environment
1783 @return: the success of the operation
1786 export_env = OSEnvironment(instance)
1788 inst_os = OSFromDisk(instance.os)
1789 export_script = inst_os.export_script
1791 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1792 instance.name, int(time.time()))
1793 if not os.path.exists(constants.LOG_OS_DIR):
1794 os.mkdir(constants.LOG_OS_DIR, 0750)
1795 real_disk = _RecursiveFindBD(disk)
1796 if real_disk is None:
1797 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1801 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1802 export_env['EXPORT_INDEX'] = str(idx)
1804 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1805 destfile = disk.physical_id[1]
1807 # the target command is built out of three individual commands,
1808 # which are joined by pipes; we check each individual command for
1810 expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
1811 export_script, logfile)
1815 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1816 destdir, destdir, destfile)
1817 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1818 constants.GANETI_RUNAS,
1821 # all commands have been checked, so we're safe to combine them
1822 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1824 result = utils.RunCmd(command, env=export_env)
1827 logging.error("os snapshot export command '%s' returned error: %s"
1828 " output: %s", command, result.fail_reason, result.output)
1834 def FinalizeExport(instance, snap_disks):
1835 """Write out the export configuration information.
1837 @type instance: L{objects.Instance}
1838 @param instance: the instance which we export, used for
1839 saving configuration
1840 @type snap_disks: list of L{objects.Disk}
1841 @param snap_disks: list of snapshot block devices, which
1842 will be used to get the actual name of the dump file
1845 @return: the success of the operation
1848 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1849 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1851 config = objects.SerializableConfigParser()
1853 config.add_section(constants.INISECT_EXP)
1854 config.set(constants.INISECT_EXP, 'version', '0')
1855 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1856 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1857 config.set(constants.INISECT_EXP, 'os', instance.os)
1858 config.set(constants.INISECT_EXP, 'compression', 'gzip')
1860 config.add_section(constants.INISECT_INS)
1861 config.set(constants.INISECT_INS, 'name', instance.name)
1862 config.set(constants.INISECT_INS, 'memory', '%d' %
1863 instance.beparams[constants.BE_MEMORY])
1864 config.set(constants.INISECT_INS, 'vcpus', '%d' %
1865 instance.beparams[constants.BE_VCPUS])
1866 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1869 for nic_count, nic in enumerate(instance.nics):
1871 config.set(constants.INISECT_INS, 'nic%d_mac' %
1872 nic_count, '%s' % nic.mac)
1873 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
1874 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
1876 # TODO: redundant: on load can read nics until it doesn't exist
1877 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
1880 for disk_count, disk in enumerate(snap_disks):
1883 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
1884 ('%s' % disk.iv_name))
1885 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
1886 ('%s' % disk.physical_id[1]))
1887 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
1890 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
1892 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
1893 data=config.Dumps())
1894 shutil.rmtree(finaldestdir, True)
1895 shutil.move(destdir, finaldestdir)
1900 def ExportInfo(dest):
1901 """Get export configuration information.
1904 @param dest: directory containing the export
1906 @rtype: L{objects.SerializableConfigParser}
1907 @return: a serializable config file containing the
1911 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
1913 config = objects.SerializableConfigParser()
1916 if (not config.has_section(constants.INISECT_EXP) or
1917 not config.has_section(constants.INISECT_INS)):
1923 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
1924 """Import an os image into an instance.
1926 @type instance: L{objects.Instance}
1927 @param instance: instance to import the disks into
1928 @type src_node: string
1929 @param src_node: source node for the disk images
1930 @type src_images: list of string
1931 @param src_images: absolute paths of the disk images
1932 @rtype: list of boolean
1933 @return: each boolean represent the success of importing the n-th disk
1936 import_env = OSEnvironment(instance)
1937 inst_os = OSFromDisk(instance.os)
1938 import_script = inst_os.import_script
1940 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
1941 instance.name, int(time.time()))
1942 if not os.path.exists(constants.LOG_OS_DIR):
1943 os.mkdir(constants.LOG_OS_DIR, 0750)
1946 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
1947 import_script, logfile)
1950 for idx, image in enumerate(src_images):
1952 destcmd = utils.BuildShellCmd('cat %s', image)
1953 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
1954 constants.GANETI_RUNAS,
1956 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
1957 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
1958 import_env['IMPORT_INDEX'] = str(idx)
1959 result = utils.RunCmd(command, env=import_env)
1961 logging.error("Disk import command '%s' returned error: %s"
1962 " output: %s", command, result.fail_reason,
1964 final_result.append(False)
1966 final_result.append(True)
1968 final_result.append(True)
1974 """Return a list of exports currently available on this machine.
1977 @return: list of the exports
1980 if os.path.isdir(constants.EXPORT_DIR):
1981 return utils.ListVisibleFiles(constants.EXPORT_DIR)
1986 def RemoveExport(export):
1987 """Remove an existing export from the node.
1990 @param export: the name of the export to remove
1992 @return: the success of the operation
1995 target = os.path.join(constants.EXPORT_DIR, export)
1997 shutil.rmtree(target)
1998 # TODO: catch some of the relevant exceptions and provide a pretty
1999 # error message if rmtree fails.
2004 def BlockdevRename(devlist):
2005 """Rename a list of block devices.
2007 @type devlist: list of tuples
2008 @param devlist: list of tuples of the form (disk,
2009 new_logical_id, new_physical_id); disk is an
2010 L{objects.Disk} object describing the current disk,
2011 and new logical_id/physical_id is the name we
2014 @return: True if all renames succeeded, False otherwise
2019 for disk, unique_id in devlist:
2020 dev = _RecursiveFindBD(disk)
2022 msgs.append("Can't find device %s in rename" % str(disk))
2026 old_rpath = dev.dev_path
2027 dev.Rename(unique_id)
2028 new_rpath = dev.dev_path
2029 if old_rpath != new_rpath:
2030 DevCacheManager.RemoveCache(old_rpath)
2031 # FIXME: we should add the new cache information here, like:
2032 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2033 # but we don't have the owner here - maybe parse from existing
2034 # cache? for now, we only lose lvm data when we rename, which
2035 # is less critical than DRBD or MD
2036 except errors.BlockDeviceError, err:
2037 msgs.append("Can't rename device '%s' to '%s': %s" %
2038 (dev, unique_id, err))
2039 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2041 return (result, "; ".join(msgs))
2044 def _TransformFileStorageDir(file_storage_dir):
2045 """Checks whether given file_storage_dir is valid.
2047 Checks wheter the given file_storage_dir is within the cluster-wide
2048 default file_storage_dir stored in SimpleStore. Only paths under that
2049 directory are allowed.
2051 @type file_storage_dir: str
2052 @param file_storage_dir: the path to check
2054 @return: the normalized path if valid, None otherwise
2058 file_storage_dir = os.path.normpath(file_storage_dir)
2059 base_file_storage_dir = cfg.GetFileStorageDir()
2060 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2061 base_file_storage_dir):
2062 logging.error("file storage directory '%s' is not under base file"
2063 " storage directory '%s'",
2064 file_storage_dir, base_file_storage_dir)
2066 return file_storage_dir
2069 def CreateFileStorageDir(file_storage_dir):
2070 """Create file storage directory.
2072 @type file_storage_dir: str
2073 @param file_storage_dir: directory to create
2076 @return: tuple with first element a boolean indicating wheter dir
2077 creation was successful or not
2080 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2082 if not file_storage_dir:
2085 if os.path.exists(file_storage_dir):
2086 if not os.path.isdir(file_storage_dir):
2087 logging.error("'%s' is not a directory", file_storage_dir)
2091 os.makedirs(file_storage_dir, 0750)
2092 except OSError, err:
2093 logging.error("Cannot create file storage directory '%s': %s",
2094 file_storage_dir, err)
2099 def RemoveFileStorageDir(file_storage_dir):
2100 """Remove file storage directory.
2102 Remove it only if it's empty. If not log an error and return.
2104 @type file_storage_dir: str
2105 @param file_storage_dir: the directory we should cleanup
2106 @rtype: tuple (success,)
2107 @return: tuple of one element, C{success}, denoting
2108 whether the operation was successfull
2111 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2113 if not file_storage_dir:
2116 if os.path.exists(file_storage_dir):
2117 if not os.path.isdir(file_storage_dir):
2118 logging.error("'%s' is not a directory", file_storage_dir)
2120 # deletes dir only if empty, otherwise we want to return False
2122 os.rmdir(file_storage_dir)
2123 except OSError, err:
2124 logging.exception("Cannot remove file storage directory '%s'",
2130 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2131 """Rename the file storage directory.
2133 @type old_file_storage_dir: str
2134 @param old_file_storage_dir: the current path
2135 @type new_file_storage_dir: str
2136 @param new_file_storage_dir: the name we should rename to
2137 @rtype: tuple (success,)
2138 @return: tuple of one element, C{success}, denoting
2139 whether the operation was successful
2142 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2143 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2145 if not old_file_storage_dir or not new_file_storage_dir:
2148 if not os.path.exists(new_file_storage_dir):
2149 if os.path.isdir(old_file_storage_dir):
2151 os.rename(old_file_storage_dir, new_file_storage_dir)
2152 except OSError, err:
2153 logging.exception("Cannot rename '%s' to '%s'",
2154 old_file_storage_dir, new_file_storage_dir)
2157 logging.error("'%s' is not a directory", old_file_storage_dir)
2160 if os.path.exists(old_file_storage_dir):
2161 logging.error("Cannot rename '%s' to '%s'. Both locations exist.",
2162 old_file_storage_dir, new_file_storage_dir)
2167 def _IsJobQueueFile(file_name):
2168 """Checks whether the given filename is in the queue directory.
2170 @type file_name: str
2171 @param file_name: the file name we should check
2173 @return: whether the file is under the queue directory
2176 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2177 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2180 logging.error("'%s' is not a file in the queue directory",
2186 def JobQueueUpdate(file_name, content):
2187 """Updates a file in the queue directory.
2189 This is just a wrapper over L{utils.WriteFile}, with proper
2192 @type file_name: str
2193 @param file_name: the job file name
2195 @param content: the new job contents
2197 @return: the success of the operation
2200 if not _IsJobQueueFile(file_name):
2203 # Write and replace the file atomically
2204 utils.WriteFile(file_name, data=_Decompress(content))
2209 def JobQueueRename(old, new):
2210 """Renames a job queue file.
2212 This is just a wrapper over os.rename with proper checking.
2215 @param old: the old (actual) file name
2217 @param new: the desired file name
2219 @return: the success of the operation
2222 if not (_IsJobQueueFile(old) and _IsJobQueueFile(new)):
2225 utils.RenameFile(old, new, mkdir=True)
2230 def JobQueueSetDrainFlag(drain_flag):
2231 """Set the drain flag for the queue.
2233 This will set or unset the queue drain flag.
2235 @type drain_flag: boolean
2236 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2238 @return: always True
2239 @warning: the function always returns True
2243 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2245 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2250 def BlockdevClose(instance_name, disks):
2251 """Closes the given block devices.
2253 This means they will be switched to secondary mode (in case of
2256 @param instance_name: if the argument is not empty, the symlinks
2257 of this instance will be removed
2258 @type disks: list of L{objects.Disk}
2259 @param disks: the list of disks to be closed
2260 @rtype: tuple (success, message)
2261 @return: a tuple of success and message, where success
2262 indicates the succes of the operation, and message
2263 which will contain the error details in case we
2269 rd = _RecursiveFindBD(cf)
2271 return (False, "Can't find device %s" % cf)
2278 except errors.BlockDeviceError, err:
2279 msg.append(str(err))
2281 return (False, "Can't make devices secondary: %s" % ",".join(msg))
2284 _RemoveBlockDevLinks(instance_name, disks)
2285 return (True, "All devices secondary")
2288 def ValidateHVParams(hvname, hvparams):
2289 """Validates the given hypervisor parameters.
2291 @type hvname: string
2292 @param hvname: the hypervisor name
2293 @type hvparams: dict
2294 @param hvparams: the hypervisor parameters to be validated
2295 @rtype: tuple (success, message)
2296 @return: a tuple of success and message, where success
2297 indicates the succes of the operation, and message
2298 which will contain the error details in case we
2303 hv_type = hypervisor.GetHypervisor(hvname)
2304 hv_type.ValidateParameters(hvparams)
2305 return (True, "Validation passed")
2306 except errors.HypervisorError, err:
2307 return (False, str(err))
2311 """Demotes the current node from master candidate role.
2314 # try to ensure we're not the master by mistake
2315 master, myself = ssconf.GetMasterAndMyself()
2316 if master == myself:
2317 return (False, "ssconf status shows I'm the master node, will not demote")
2318 pid_file = utils.DaemonPidFileName(constants.MASTERD_PID)
2319 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2320 return (False, "The master daemon is running, will not demote")
2322 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2323 except EnvironmentError, err:
2324 if err.errno != errno.ENOENT:
2325 return (False, "Error while backing up cluster file: %s" % str(err))
2326 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2327 return (True, "Done")
2330 def _FindDisks(nodes_ip, disks):
2331 """Sets the physical ID on disks and returns the block devices.
2334 # set the correct physical ID
2335 my_name = utils.HostInfo().name
2337 cf.SetPhysicalID(my_name, nodes_ip)
2342 rd = _RecursiveFindBD(cf)
2344 return (False, "Can't find device %s" % cf)
2346 return (True, bdevs)
2349 def DrbdDisconnectNet(nodes_ip, disks):
2350 """Disconnects the network on a list of drbd devices.
2353 status, bdevs = _FindDisks(nodes_ip, disks)
2355 return status, bdevs
2361 except errors.BlockDeviceError, err:
2362 logging.exception("Failed to go into standalone mode")
2363 return (False, "Can't change network configuration: %s" % str(err))
2364 return (True, "All disks are now disconnected")
2367 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2368 """Attaches the network on a list of drbd devices.
2371 status, bdevs = _FindDisks(nodes_ip, disks)
2373 return status, bdevs
2376 for idx, rd in enumerate(bdevs):
2378 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2379 except EnvironmentError, err:
2380 return (False, "Can't create symlink: %s" % str(err))
2381 # reconnect disks, switch to new master configuration and if
2382 # needed primary mode
2385 rd.AttachNet(multimaster)
2386 except errors.BlockDeviceError, err:
2387 return (False, "Can't change network configuration: %s" % str(err))
2388 # wait until the disks are connected; we need to retry the re-attach
2389 # if the device becomes standalone, as this might happen if the one
2390 # node disconnects and reconnects in a different mode before the
2391 # other node reconnects; in this case, one or both of the nodes will
2392 # decide it has wrong configuration and switch to standalone
2393 RECONNECT_TIMEOUT = 2 * 60
2394 sleep_time = 0.100 # start with 100 miliseconds
2395 timeout_limit = time.time() + RECONNECT_TIMEOUT
2396 while time.time() < timeout_limit:
2397 all_connected = True
2399 stats = rd.GetProcStatus()
2400 if not (stats.is_connected or stats.is_in_resync):
2401 all_connected = False
2402 if stats.is_standalone:
2403 # peer had different config info and this node became
2404 # standalone, even though this should not happen with the
2405 # new staged way of changing disk configs
2407 rd.ReAttachNet(multimaster)
2408 except errors.BlockDeviceError, err:
2409 return (False, "Can't change network configuration: %s" % str(err))
2412 time.sleep(sleep_time)
2413 sleep_time = min(5, sleep_time * 1.5)
2414 if not all_connected:
2415 return (False, "Timeout in disk reconnecting")
2417 # change to primary mode
2421 except errors.BlockDeviceError, err:
2422 return (False, "Can't change to primary mode: %s" % str(err))
2424 msg = "multi-master and primary"
2426 msg = "single-master"
2427 return (True, "Disks are now configured as %s" % msg)
2430 def DrbdWaitSync(nodes_ip, disks):
2431 """Wait until DRBDs have synchronized.
2434 status, bdevs = _FindDisks(nodes_ip, disks)
2436 return status, bdevs
2442 stats = rd.GetProcStatus()
2443 if not (stats.is_connected or stats.is_in_resync):
2446 alldone = alldone and (not stats.is_in_resync)
2447 if stats.sync_percent is not None:
2448 min_resync = min(min_resync, stats.sync_percent)
2449 return (not failure, (alldone, min_resync))
2452 def PowercycleNode(hypervisor_type):
2453 """Hard-powercycle the node.
2455 Because we need to return first, and schedule the powercycle in the
2456 background, we won't be able to report failures nicely.
2459 hyper = hypervisor.GetHypervisor(hypervisor_type)
2462 except OSError, err:
2463 # if we can't fork, we'll pretend that we're in the child process
2466 return (True, "Reboot scheduled in 5 seconds")
2468 hyper.PowercycleNode()
2471 class HooksRunner(object):
2474 This class is instantiated on the node side (ganeti-noded) and not
2478 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2480 def __init__(self, hooks_base_dir=None):
2481 """Constructor for hooks runner.
2483 @type hooks_base_dir: str or None
2484 @param hooks_base_dir: if not None, this overrides the
2485 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2488 if hooks_base_dir is None:
2489 hooks_base_dir = constants.HOOKS_BASE_DIR
2490 self._BASE_DIR = hooks_base_dir
2493 def ExecHook(script, env):
2494 """Exec one hook script.
2497 @param script: the full path to the script
2499 @param env: the environment with which to exec the script
2500 @rtype: tuple (success, message)
2501 @return: a tuple of success and message, where success
2502 indicates the succes of the operation, and message
2503 which will contain the error details in case we
2507 # exec the process using subprocess and log the output
2510 fdstdin = open("/dev/null", "r")
2511 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2512 stderr=subprocess.STDOUT, close_fds=True,
2513 shell=False, cwd="/", env=env)
2516 output = child.stdout.read(4096)
2517 child.stdout.close()
2518 except EnvironmentError, err:
2519 output += "Hook script error: %s" % str(err)
2523 result = child.wait()
2525 except EnvironmentError, err:
2526 if err.errno == errno.EINTR:
2530 # try not to leak fds
2531 for fd in (fdstdin, ):
2535 except EnvironmentError, err:
2536 # just log the error
2537 #logging.exception("Error while closing fd %s", fd)
2540 return result == 0, utils.SafeEncode(output.strip())
2542 def RunHooks(self, hpath, phase, env):
2543 """Run the scripts in the hooks directory.
2546 @param hpath: the path to the hooks directory which
2549 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2550 L{constants.HOOKS_PHASE_POST}
2552 @param env: dictionary with the environment for the hook
2554 @return: list of 3-element tuples:
2556 - script result, either L{constants.HKR_SUCCESS} or
2557 L{constants.HKR_FAIL}
2558 - output of the script
2560 @raise errors.ProgrammerError: for invalid input
2564 if phase == constants.HOOKS_PHASE_PRE:
2566 elif phase == constants.HOOKS_PHASE_POST:
2569 raise errors.ProgrammerError("Unknown hooks phase: '%s'" % phase)
2572 subdir = "%s-%s.d" % (hpath, suffix)
2573 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2575 dir_contents = utils.ListVisibleFiles(dir_name)
2576 except OSError, err:
2577 # FIXME: must log output in case of failures
2580 # we use the standard python sort order,
2581 # so 00name is the recommended naming scheme
2583 for relname in dir_contents:
2584 fname = os.path.join(dir_name, relname)
2585 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2586 self.RE_MASK.match(relname) is not None):
2587 rrval = constants.HKR_SKIP
2590 result, output = self.ExecHook(fname, env)
2592 rrval = constants.HKR_FAIL
2594 rrval = constants.HKR_SUCCESS
2595 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2600 class IAllocatorRunner(object):
2601 """IAllocator runner.
2603 This class is instantiated on the node side (ganeti-noded) and not on
2607 def Run(self, name, idata):
2608 """Run an iallocator script.
2611 @param name: the iallocator script name
2613 @param idata: the allocator input data
2616 @return: four element tuple of:
2617 - run status (one of the IARUN_ constants)
2620 - fail reason (as from L{utils.RunResult})
2623 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2625 if alloc_script is None:
2626 return (constants.IARUN_NOTFOUND, None, None, None)
2628 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2632 result = utils.RunCmd([alloc_script, fin_name])
2634 return (constants.IARUN_FAILURE, result.stdout, result.stderr,
2639 return (constants.IARUN_SUCCESS, result.stdout, result.stderr, None)
2642 class DevCacheManager(object):
2643 """Simple class for managing a cache of block device information.
2646 _DEV_PREFIX = "/dev/"
2647 _ROOT_DIR = constants.BDEV_CACHE_DIR
2650 def _ConvertPath(cls, dev_path):
2651 """Converts a /dev/name path to the cache file name.
2653 This replaces slashes with underscores and strips the /dev
2654 prefix. It then returns the full path to the cache file.
2657 @param dev_path: the C{/dev/} path name
2659 @return: the converted path name
2662 if dev_path.startswith(cls._DEV_PREFIX):
2663 dev_path = dev_path[len(cls._DEV_PREFIX):]
2664 dev_path = dev_path.replace("/", "_")
2665 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2669 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2670 """Updates the cache information for a given device.
2673 @param dev_path: the pathname of the device
2675 @param owner: the owner (instance name) of the device
2676 @type on_primary: bool
2677 @param on_primary: whether this is the primary
2680 @param iv_name: the instance-visible name of the
2681 device, as in objects.Disk.iv_name
2686 if dev_path is None:
2687 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2689 fpath = cls._ConvertPath(dev_path)
2695 iv_name = "not_visible"
2696 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2698 utils.WriteFile(fpath, data=fdata)
2699 except EnvironmentError, err:
2700 logging.exception("Can't update bdev cache for %s", dev_path)
2703 def RemoveCache(cls, dev_path):
2704 """Remove data for a dev_path.
2706 This is just a wrapper over L{utils.RemoveFile} with a converted
2707 path name and logging.
2710 @param dev_path: the pathname of the device
2715 if dev_path is None:
2716 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2718 fpath = cls._ConvertPath(dev_path)
2720 utils.RemoveFile(fpath)
2721 except EnvironmentError, err:
2722 logging.exception("Can't update bdev cache for %s", dev_path)