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 logging.exception("Error while processing user ssh files")
266 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
267 utils.WriteFile(name, data=content, mode=0600)
269 utils.AddAuthorizedKey(auth_keys, sshpub)
271 utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
277 """Cleans up and remove the current node.
279 This function cleans up and prepares the current node to be removed
282 If processing is successful, then it raises an
283 L{errors.QuitGanetiException} which is used as a special case to
284 shutdown the node daemon.
287 _CleanDirectory(constants.DATA_DIR)
291 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
292 except errors.OpExecError:
293 logging.exception("Error while processing ssh files")
296 f = open(pub_key, 'r')
298 utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
302 utils.RemoveFile(priv_key)
303 utils.RemoveFile(pub_key)
305 # Return a reassuring string to the caller, and quit
306 raise errors.QuitGanetiException(False, 'Shutdown scheduled')
309 def GetNodeInfo(vgname, hypervisor_type):
310 """Gives back a hash with different informations about the node.
312 @type vgname: C{string}
313 @param vgname: the name of the volume group to ask for disk space information
314 @type hypervisor_type: C{str}
315 @param hypervisor_type: the name of the hypervisor to ask for
318 @return: dictionary with the following keys:
319 - vg_size is the size of the configured volume group in MiB
320 - vg_free is the free size of the volume group in MiB
321 - memory_dom0 is the memory allocated for domain0 in MiB
322 - memory_free is the currently available (free) ram in MiB
323 - memory_total is the total number of ram in MiB
327 vginfo = _GetVGInfo(vgname)
328 outputarray['vg_size'] = vginfo['vg_size']
329 outputarray['vg_free'] = vginfo['vg_free']
331 hyper = hypervisor.GetHypervisor(hypervisor_type)
332 hyp_info = hyper.GetNodeInfo()
333 if hyp_info is not None:
334 outputarray.update(hyp_info)
336 f = open("/proc/sys/kernel/random/boot_id", 'r')
338 outputarray["bootid"] = f.read(128).rstrip("\n")
345 def VerifyNode(what, cluster_name):
346 """Verify the status of the local node.
348 Based on the input L{what} parameter, various checks are done on the
351 If the I{filelist} key is present, this list of
352 files is checksummed and the file/checksum pairs are returned.
354 If the I{nodelist} key is present, we check that we have
355 connectivity via ssh with the target nodes (and check the hostname
358 If the I{node-net-test} key is present, we check that we have
359 connectivity to the given nodes via both primary IP and, if
360 applicable, secondary IPs.
363 @param what: a dictionary of things to check:
364 - filelist: list of files for which to compute checksums
365 - nodelist: list of nodes we should check ssh communication with
366 - node-net-test: list of nodes we should check node daemon port
368 - hypervisor: list with hypervisors to run the verify for
370 @return: a dictionary with the same keys as the input dict, and
371 values representing the result of the checks
376 if constants.NV_HYPERVISOR in what:
377 result[constants.NV_HYPERVISOR] = tmp = {}
378 for hv_name in what[constants.NV_HYPERVISOR]:
379 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
381 if constants.NV_FILELIST in what:
382 result[constants.NV_FILELIST] = utils.FingerprintFiles(
383 what[constants.NV_FILELIST])
385 if constants.NV_NODELIST in what:
386 result[constants.NV_NODELIST] = tmp = {}
387 random.shuffle(what[constants.NV_NODELIST])
388 for node in what[constants.NV_NODELIST]:
389 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
393 if constants.NV_NODENETTEST in what:
394 result[constants.NV_NODENETTEST] = tmp = {}
395 my_name = utils.HostInfo().name
396 my_pip = my_sip = None
397 for name, pip, sip in what[constants.NV_NODENETTEST]:
403 tmp[my_name] = ("Can't find my own primary/secondary IP"
406 port = utils.GetNodeDaemonPort()
407 for name, pip, sip in what[constants.NV_NODENETTEST]:
409 if not utils.TcpPing(pip, port, source=my_pip):
410 fail.append("primary")
412 if not utils.TcpPing(sip, port, source=my_sip):
413 fail.append("secondary")
415 tmp[name] = ("failure using the %s interface(s)" %
418 if constants.NV_LVLIST in what:
419 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
421 if constants.NV_INSTANCELIST in what:
422 result[constants.NV_INSTANCELIST] = GetInstanceList(
423 what[constants.NV_INSTANCELIST])
425 if constants.NV_VGLIST in what:
426 result[constants.NV_VGLIST] = ListVolumeGroups()
428 if constants.NV_VERSION in what:
429 result[constants.NV_VERSION] = constants.PROTOCOL_VERSION
431 if constants.NV_HVINFO in what:
432 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
433 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
438 def GetVolumeList(vg_name):
439 """Compute list of logical volumes and their size.
442 @param vg_name: the volume group whose LVs we should list
445 dictionary of all partions (key) with value being a tuple of
446 their size (in MiB), inactive and online status::
448 {'test1': ('20.06', True, True)}
450 in case of errors, a string is returned with the error
456 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
457 "--separator=%s" % sep,
458 "-olv_name,lv_size,lv_attr", vg_name])
460 logging.error("Failed to list logical volumes, lvs output: %s",
464 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
465 for line in result.stdout.splitlines():
467 match = valid_line_re.match(line)
469 logging.error("Invalid line returned from lvs output: '%s'", line)
471 name, size, attr = match.groups()
472 inactive = attr[4] == '-'
473 online = attr[5] == 'o'
474 lvs[name] = (size, inactive, online)
479 def ListVolumeGroups():
480 """List the volume groups and their size.
483 @return: dictionary with keys volume name and values the
487 return utils.ListVolumeGroups()
491 """List all volumes on this node.
495 A list of dictionaries, each having four keys:
496 - name: the logical volume name,
497 - size: the size of the logical volume
498 - dev: the physical device on which the LV lives
499 - vg: the volume group to which it belongs
501 In case of errors, we return an empty list and log the
504 Note that since a logical volume can live on multiple physical
505 volumes, the resulting list might include a logical volume
509 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
511 "--options=lv_name,lv_size,devices,vg_name"])
513 logging.error("Failed to list logical volumes, lvs output: %s",
519 return dev.split('(')[0]
525 'name': line[0].strip(),
526 'size': line[1].strip(),
527 'dev': parse_dev(line[2].strip()),
528 'vg': line[3].strip(),
531 return [map_line(line.split('|')) for line in result.stdout.splitlines()
532 if line.count('|') >= 3]
535 def BridgesExist(bridges_list):
536 """Check if a list of bridges exist on the current node.
539 @return: C{True} if all of them exist, C{False} otherwise
542 for bridge in bridges_list:
543 if not utils.BridgeExists(bridge):
549 def GetInstanceList(hypervisor_list):
550 """Provides a list of instances.
552 @type hypervisor_list: list
553 @param hypervisor_list: the list of hypervisors to query information
556 @return: a list of all running instances on the current node
557 - instance1.example.com
558 - instance2.example.com
562 for hname in hypervisor_list:
564 names = hypervisor.GetHypervisor(hname).ListInstances()
565 results.extend(names)
566 except errors.HypervisorError, err:
567 logging.exception("Error enumerating instances for hypevisor %s", hname)
568 # FIXME: should we somehow not propagate this to the master?
574 def GetInstanceInfo(instance, hname):
575 """Gives back the informations about an instance as a dictionary.
577 @type instance: string
578 @param instance: the instance name
580 @param hname: the hypervisor type of the instance
583 @return: dictionary with the following keys:
584 - memory: memory size of instance (int)
585 - state: xen state of instance (string)
586 - time: cpu time of instance (float)
591 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
592 if iinfo is not None:
593 output['memory'] = iinfo[2]
594 output['state'] = iinfo[4]
595 output['time'] = iinfo[5]
600 def GetInstanceMigratable(instance):
601 """Gives whether an instance can be migrated.
603 @type instance: L{objects.Instance}
604 @param instance: object representing the instance to be checked.
607 @return: tuple of (result, description) where:
608 - result: whether the instance can be migrated or not
609 - description: a description of the issue, if relevant
612 hyper = hypervisor.GetHypervisor(instance.hypervisor)
613 if instance.name not in hyper.ListInstances():
614 return (False, 'not running')
616 for idx in range(len(instance.disks)):
617 link_name = _GetBlockDevSymlinkPath(instance.name, idx)
618 if not os.path.islink(link_name):
619 return (False, 'not restarted since ganeti 1.2.5')
624 def GetAllInstancesInfo(hypervisor_list):
625 """Gather data about all instances.
627 This is the equivalent of L{GetInstanceInfo}, except that it
628 computes data for all instances at once, thus being faster if one
629 needs data about more than one instance.
631 @type hypervisor_list: list
632 @param hypervisor_list: list of hypervisors to query for instance data
635 @return: dictionary of instance: data, with data having the following keys:
636 - memory: memory size of instance (int)
637 - state: xen state of instance (string)
638 - time: cpu time of instance (float)
639 - vcpus: the number of vcpus
644 for hname in hypervisor_list:
645 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
647 for name, inst_id, memory, vcpus, state, times in iinfo:
654 if name in output and output[name] != value:
655 raise errors.HypervisorError("Instance %s running duplicate"
656 " with different parameters" % name)
662 def AddOSToInstance(instance):
663 """Add an OS to an instance.
665 @type instance: L{objects.Instance}
666 @param instance: Instance whose OS is to be installed
668 @return: the success of the operation
671 inst_os = OSFromDisk(instance.os)
673 create_env = OSEnvironment(instance)
675 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
676 instance.name, int(time.time()))
678 result = utils.RunCmd([inst_os.create_script], env=create_env,
679 cwd=inst_os.path, output=logfile,)
681 logging.error("os create command '%s' returned error: %s, logfile: %s,"
682 " output: %s", result.cmd, result.fail_reason, logfile,
689 def RunRenameInstance(instance, old_name):
690 """Run the OS rename script for an instance.
692 @type instance: L{objects.Instance}
693 @param instance: Instance whose OS is to be installed
694 @type old_name: string
695 @param old_name: previous instance name
697 @return: the success of the operation
700 inst_os = OSFromDisk(instance.os)
702 rename_env = OSEnvironment(instance)
703 rename_env['OLD_INSTANCE_NAME'] = old_name
705 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
707 instance.name, int(time.time()))
709 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
710 cwd=inst_os.path, output=logfile)
713 logging.error("os create command '%s' returned error: %s output: %s",
714 result.cmd, result.fail_reason, result.output)
720 def _GetVGInfo(vg_name):
721 """Get informations about the volume group.
724 @param vg_name: the volume group which we query
727 A dictionary with the following keys:
728 - C{vg_size} is the total size of the volume group in MiB
729 - C{vg_free} is the free size of the volume group in MiB
730 - C{pv_count} are the number of physical disks in that VG
732 If an error occurs during gathering of data, we return the same dict
733 with keys all set to None.
736 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
738 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
739 "--nosuffix", "--units=m", "--separator=:", vg_name])
742 logging.error("volume group %s not present", vg_name)
744 valarr = retval.stdout.strip().rstrip(':').split(':')
748 "vg_size": int(round(float(valarr[0]), 0)),
749 "vg_free": int(round(float(valarr[1]), 0)),
750 "pv_count": int(valarr[2]),
752 except ValueError, err:
753 logging.exception("Fail to parse vgs output")
755 logging.error("vgs output has the wrong number of fields (expected"
756 " three): %s", str(valarr))
760 def _GetBlockDevSymlinkPath(instance_name, idx):
761 return os.path.join(constants.DISK_LINKS_DIR,
762 "%s:%d" % (instance_name, idx))
765 def _SymlinkBlockDev(instance_name, device_path, idx):
766 """Set up symlinks to a instance's block device.
768 This is an auxiliary function run when an instance is start (on the primary
769 node) or when an instance is migrated (on the target node).
772 @param instance_name: the name of the target instance
773 @param device_path: path of the physical block device, on the node
774 @param idx: the disk index
775 @return: absolute path to the disk's symlink
778 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
780 os.symlink(device_path, link_name)
782 if err.errno == errno.EEXIST:
783 if (not os.path.islink(link_name) or
784 os.readlink(link_name) != device_path):
786 os.symlink(device_path, link_name)
793 def _RemoveBlockDevLinks(instance_name, disks):
794 """Remove the block device symlinks belonging to the given instance.
797 for idx, disk in enumerate(disks):
798 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
799 if os.path.islink(link_name):
803 logging.exception("Can't remove symlink '%s'", link_name)
806 def _GatherAndLinkBlockDevs(instance):
807 """Set up an instance's block device(s).
809 This is run on the primary node at instance startup. The block
810 devices must be already assembled.
812 @type instance: L{objects.Instance}
813 @param instance: the instance whose disks we shoul assemble
815 @return: list of (disk_object, device_path)
819 for idx, disk in enumerate(instance.disks):
820 device = _RecursiveFindBD(disk)
822 raise errors.BlockDeviceError("Block device '%s' is not set up." %
826 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
828 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
831 block_devices.append((disk, link_name))
836 def StartInstance(instance, extra_args):
837 """Start an instance.
839 @type instance: L{objects.Instance}
840 @param instance: the instance object
842 @return: whether the startup was successful or not
845 running_instances = GetInstanceList([instance.hypervisor])
847 if instance.name in running_instances:
848 return (True, "Already running")
851 block_devices = _GatherAndLinkBlockDevs(instance)
852 hyper = hypervisor.GetHypervisor(instance.hypervisor)
853 hyper.StartInstance(instance, block_devices, extra_args)
854 except errors.BlockDeviceError, err:
855 logging.exception("Failed to start instance")
856 return (False, "Block device error: %s" % str(err))
857 except errors.HypervisorError, err:
858 logging.exception("Failed to start instance")
859 _RemoveBlockDevLinks(instance.name, instance.disks)
860 return (False, "Hypervisor error: %s" % str(err))
862 return (True, "Instance started successfully")
865 def ShutdownInstance(instance):
866 """Shut an instance down.
868 @note: this functions uses polling with a hardcoded timeout.
870 @type instance: L{objects.Instance}
871 @param instance: the instance object
873 @return: whether the startup was successful or not
876 hv_name = instance.hypervisor
877 running_instances = GetInstanceList([hv_name])
879 if instance.name not in running_instances:
882 hyper = hypervisor.GetHypervisor(hv_name)
884 hyper.StopInstance(instance)
885 except errors.HypervisorError, err:
886 logging.error("Failed to stop instance")
889 # test every 10secs for 2min
892 for dummy in range(11):
893 if instance.name not in GetInstanceList([hv_name]):
897 # the shutdown did not succeed
898 logging.error("shutdown of '%s' unsuccessful, using destroy", instance)
901 hyper.StopInstance(instance, force=True)
902 except errors.HypervisorError, err:
903 logging.exception("Failed to stop instance")
907 if instance.name in GetInstanceList([hv_name]):
908 logging.error("could not shutdown instance '%s' even by destroy",
912 _RemoveBlockDevLinks(instance.name, instance.disks)
917 def RebootInstance(instance, reboot_type, extra_args):
918 """Reboot an instance.
920 @type instance: L{objects.Instance}
921 @param instance: the instance object to reboot
922 @type reboot_type: str
923 @param reboot_type: the type of reboot, one the following
925 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
926 instance OS, do not recreate the VM
927 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
928 restart the VM (at the hypervisor level)
929 - the other reboot type (L{constants.INSTANCE_REBOOT_HARD})
930 is not accepted here, since that mode is handled
933 @return: the success of the operation
936 running_instances = GetInstanceList([instance.hypervisor])
938 if instance.name not in running_instances:
939 logging.error("Cannot reboot instance that is not running")
942 hyper = hypervisor.GetHypervisor(instance.hypervisor)
943 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
945 hyper.RebootInstance(instance)
946 except errors.HypervisorError, err:
947 logging.exception("Failed to soft reboot instance")
949 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
951 ShutdownInstance(instance)
952 StartInstance(instance, extra_args)
953 except errors.HypervisorError, err:
954 logging.exception("Failed to hard reboot instance")
957 raise errors.ParameterError("reboot_type invalid")
962 def MigrateInstance(instance, target, live):
963 """Migrates an instance to another node.
965 @type instance: L{objects.Instance}
966 @param instance: the instance definition
968 @param target: the target node name
970 @param live: whether the migration should be done live or not (the
971 interpretation of this parameter is left to the hypervisor)
973 @return: a tuple of (success, msg) where:
974 - succes is a boolean denoting the success/failure of the operation
975 - msg is a string with details in case of failure
978 hyper = hypervisor.GetHypervisor(instance.hypervisor)
981 hyper.MigrateInstance(instance.name, target, live)
982 except errors.HypervisorError, err:
983 msg = "Failed to migrate instance"
984 logging.exception(msg)
985 return (False, "%s: %s" % (msg, err))
986 return (True, "Migration successfull")
989 def CreateBlockDevice(disk, size, owner, on_primary, info):
990 """Creates a block device for an instance.
992 @type disk: L{objects.Disk}
993 @param disk: the object describing the disk we should create
995 @param size: the size of the physical underlying device, in MiB
997 @param owner: the name of the instance for which disk is created,
998 used for device cache data
999 @type on_primary: boolean
1000 @param on_primary: indicates if it is the primary node or not
1002 @param info: string that will be sent to the physical device
1003 creation, used for example to set (LVM) tags on LVs
1005 @return: the new unique_id of the device (this can sometime be
1006 computed only after creation), or None. On secondary nodes,
1007 it's not required to return anything.
1012 for child in disk.children:
1013 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1014 if on_primary or disk.AssembleOnSecondary():
1015 # we need the children open in case the device itself has to
1021 device = bdev.Create(disk.dev_type, disk.physical_id, clist, size)
1022 except errors.GenericError, err:
1023 return False, "Can't create block device: %s" % str(err)
1025 if on_primary or disk.AssembleOnSecondary():
1026 if not device.Assemble():
1027 errorstring = "Can't assemble device after creation, very unusual event"
1028 logging.error(errorstring)
1029 return False, errorstring
1030 device.SetSyncSpeed(constants.SYNC_SPEED)
1031 if on_primary or disk.OpenOnSecondary():
1032 device.Open(force=True)
1033 DevCacheManager.UpdateCache(device.dev_path, owner,
1034 on_primary, disk.iv_name)
1036 device.SetInfo(info)
1038 physical_id = device.unique_id
1039 return True, physical_id
1042 def RemoveBlockDevice(disk):
1043 """Remove a block device.
1045 @note: This is intended to be called recursively.
1047 @type disk: L{objects.Disk}
1048 @param disk: the disk object we should remove
1050 @return: the success of the operation
1054 rdev = _RecursiveFindBD(disk)
1055 except errors.BlockDeviceError, err:
1056 # probably can't attach
1057 logging.info("Can't attach to device %s in remove", disk)
1059 if rdev is not None:
1060 r_path = rdev.dev_path
1061 result = rdev.Remove()
1063 DevCacheManager.RemoveCache(r_path)
1067 for child in disk.children:
1068 result = result and RemoveBlockDevice(child)
1072 def _RecursiveAssembleBD(disk, owner, as_primary):
1073 """Activate a block device for an instance.
1075 This is run on the primary and secondary nodes for an instance.
1077 @note: this function is called recursively.
1079 @type disk: L{objects.Disk}
1080 @param disk: the disk we try to assemble
1082 @param owner: the name of the instance which owns the disk
1083 @type as_primary: boolean
1084 @param as_primary: if we should make the block device
1087 @return: the assembled device or None (in case no device
1089 @raise errors.BlockDeviceError: in case there is an error
1090 during the activation of the children or the device
1096 mcn = disk.ChildrenNeeded()
1098 mcn = 0 # max number of Nones allowed
1100 mcn = len(disk.children) - mcn # max number of Nones
1101 for chld_disk in disk.children:
1103 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1104 except errors.BlockDeviceError, err:
1105 if children.count(None) >= mcn:
1108 logging.debug("Error in child activation: %s", str(err))
1109 children.append(cdev)
1111 if as_primary or disk.AssembleOnSecondary():
1112 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children)
1113 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1115 if as_primary or disk.OpenOnSecondary():
1117 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1118 as_primary, disk.iv_name)
1125 def AssembleBlockDevice(disk, owner, as_primary):
1126 """Activate a block device for an instance.
1128 This is a wrapper over _RecursiveAssembleBD.
1130 @rtype: str or boolean
1131 @return: a C{/dev/...} path for primary nodes, and
1132 C{True} for secondary nodes
1135 result = _RecursiveAssembleBD(disk, owner, as_primary)
1136 if isinstance(result, bdev.BlockDev):
1137 result = result.dev_path
1141 def ShutdownBlockDevice(disk):
1142 """Shut down a block device.
1144 First, if the device is assembled (Attach() is successfull), then
1145 the device is shutdown. Then the children of the device are
1148 This function is called recursively. Note that we don't cache the
1149 children or such, as oppossed to assemble, shutdown of different
1150 devices doesn't require that the upper device was active.
1152 @type disk: L{objects.Disk}
1153 @param disk: the description of the disk we should
1156 @return: the success of the operation
1159 r_dev = _RecursiveFindBD(disk)
1160 if r_dev is not None:
1161 r_path = r_dev.dev_path
1162 result = r_dev.Shutdown()
1164 DevCacheManager.RemoveCache(r_path)
1168 for child in disk.children:
1169 result = result and ShutdownBlockDevice(child)
1173 def MirrorAddChildren(parent_cdev, new_cdevs):
1174 """Extend a mirrored block device.
1176 @type parent_cdev: L{objects.Disk}
1177 @param parent_cdev: the disk to which we should add children
1178 @type new_cdevs: list of L{objects.Disk}
1179 @param new_cdevs: the list of children which we should add
1181 @return: the success of the operation
1184 parent_bdev = _RecursiveFindBD(parent_cdev)
1185 if parent_bdev is None:
1186 logging.error("Can't find parent device")
1188 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1189 if new_bdevs.count(None) > 0:
1190 logging.error("Can't find new device(s) to add: %s:%s",
1191 new_bdevs, new_cdevs)
1193 parent_bdev.AddChildren(new_bdevs)
1197 def MirrorRemoveChildren(parent_cdev, new_cdevs):
1198 """Shrink a mirrored block device.
1200 @type parent_cdev: L{objects.Disk}
1201 @param parent_cdev: the disk from which we should remove children
1202 @type new_cdevs: list of L{objects.Disk}
1203 @param new_cdevs: the list of children which we should remove
1205 @return: the success of the operation
1208 parent_bdev = _RecursiveFindBD(parent_cdev)
1209 if parent_bdev is None:
1210 logging.error("Can't find parent in remove children: %s", parent_cdev)
1213 for disk in new_cdevs:
1214 rpath = disk.StaticDevPath()
1216 bd = _RecursiveFindBD(disk)
1218 logging.error("Can't find dynamic device %s while removing children",
1222 devs.append(bd.dev_path)
1225 parent_bdev.RemoveChildren(devs)
1229 def GetMirrorStatus(disks):
1230 """Get the mirroring status of a list of devices.
1232 @type disks: list of L{objects.Disk}
1233 @param disks: the list of disks which we should query
1236 a list of (mirror_done, estimated_time) tuples, which
1237 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1238 @raise errors.BlockDeviceError: if any of the disks cannot be
1244 rbd = _RecursiveFindBD(dsk)
1246 raise errors.BlockDeviceError("Can't find device %s" % str(dsk))
1247 stats.append(rbd.CombinedSyncStatus())
1251 def _RecursiveFindBD(disk):
1252 """Check if a device is activated.
1254 If so, return informations about the real device.
1256 @type disk: L{objects.Disk}
1257 @param disk: the disk object we need to find
1259 @return: None if the device can't be found,
1260 otherwise the device instance
1265 for chdisk in disk.children:
1266 children.append(_RecursiveFindBD(chdisk))
1268 return bdev.FindDevice(disk.dev_type, disk.physical_id, children)
1271 def FindBlockDevice(disk):
1272 """Check if a device is activated.
1274 If it is, return informations about the real device.
1276 @type disk: L{objects.Disk}
1277 @param disk: the disk to find
1278 @rtype: None or tuple
1279 @return: None if the disk cannot be found, otherwise a
1280 tuple (device_path, major, minor, sync_percent,
1281 estimated_time, is_degraded)
1284 rbd = _RecursiveFindBD(disk)
1287 return (rbd.dev_path, rbd.major, rbd.minor) + rbd.GetSyncStatus()
1290 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1291 """Write a file to the filesystem.
1293 This allows the master to overwrite(!) a file. It will only perform
1294 the operation if the file belongs to a list of configuration files.
1296 @type file_name: str
1297 @param file_name: the target file name
1299 @param data: the new contents of the file
1301 @param mode: the mode to give the file (can be None)
1303 @param uid: the owner of the file (can be -1 for default)
1305 @param gid: the group of the file (can be -1 for default)
1307 @param atime: the atime to set on the file (can be None)
1309 @param mtime: the mtime to set on the file (can be None)
1311 @return: the success of the operation; errors are logged
1312 in the node daemon log
1315 if not os.path.isabs(file_name):
1316 logging.error("Filename passed to UploadFile is not absolute: '%s'",
1321 constants.CLUSTER_CONF_FILE,
1322 constants.ETC_HOSTS,
1323 constants.SSH_KNOWN_HOSTS_FILE,
1324 constants.VNC_PASSWORD_FILE,
1327 if file_name not in allowed_files:
1328 logging.error("Filename passed to UploadFile not in allowed"
1329 " upload targets: '%s'", file_name)
1332 raw_data = _Decompress(data)
1334 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1335 atime=atime, mtime=mtime)
1339 def WriteSsconfFiles(values):
1340 """Update all ssconf files.
1342 Wrapper around the SimpleStore.WriteFiles.
1345 ssconf.SimpleStore().WriteFiles(values)
1348 def _ErrnoOrStr(err):
1349 """Format an EnvironmentError exception.
1351 If the L{err} argument has an errno attribute, it will be looked up
1352 and converted into a textual C{E...} description. Otherwise the
1353 string representation of the error will be returned.
1355 @type err: L{EnvironmentError}
1356 @param err: the exception to format
1359 if hasattr(err, 'errno'):
1360 detail = errno.errorcode[err.errno]
1366 def _OSOndiskVersion(name, os_dir):
1367 """Compute and return the API version of a given OS.
1369 This function will try to read the API version of the OS given by
1370 the 'name' parameter and residing in the 'os_dir' directory.
1373 @param name: the OS name we should look for
1375 @param os_dir: the directory inwhich we should look for the OS
1378 Either an integer denoting the version or None in the
1379 case when this is not a valid OS name.
1380 @raise errors.InvalidOS: if the OS cannot be found
1383 api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
1386 st = os.stat(api_file)
1387 except EnvironmentError, err:
1388 raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file not"
1389 " found (%s)" % _ErrnoOrStr(err))
1391 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1392 raise errors.InvalidOS(name, os_dir, "'ganeti_api_version' file is not"
1398 api_versions = f.readlines()
1401 except EnvironmentError, err:
1402 raise errors.InvalidOS(name, os_dir, "error while reading the"
1403 " API version (%s)" % _ErrnoOrStr(err))
1405 api_versions = [version.strip() for version in api_versions]
1407 api_versions = [int(version) for version in api_versions]
1408 except (TypeError, ValueError), err:
1409 raise errors.InvalidOS(name, os_dir,
1410 "API version is not integer (%s)" % str(err))
1415 def DiagnoseOS(top_dirs=None):
1416 """Compute the validity for all OSes.
1418 @type top_dirs: list
1419 @param top_dirs: the list of directories in which to
1420 search (if not given defaults to
1421 L{constants.OS_SEARCH_PATH})
1422 @rtype: list of L{objects.OS}
1423 @return: an OS object for each name in all the given
1427 if top_dirs is None:
1428 top_dirs = constants.OS_SEARCH_PATH
1431 for dir_name in top_dirs:
1432 if os.path.isdir(dir_name):
1434 f_names = utils.ListVisibleFiles(dir_name)
1435 except EnvironmentError, err:
1436 logging.exception("Can't list the OS directory %s", dir_name)
1438 for name in f_names:
1440 os_inst = OSFromDisk(name, base_dir=dir_name)
1441 result.append(os_inst)
1442 except errors.InvalidOS, err:
1443 result.append(objects.OS.FromInvalidOS(err))
1448 def OSFromDisk(name, base_dir=None):
1449 """Create an OS instance from disk.
1451 This function will return an OS instance if the given name is a
1452 valid OS name. Otherwise, it will raise an appropriate
1453 L{errors.InvalidOS} exception, detailing why this is not a valid OS.
1455 @type base_dir: string
1456 @keyword base_dir: Base directory containing OS installations.
1457 Defaults to a search in all the OS_SEARCH_PATH dirs.
1458 @rtype: L{objects.OS}
1459 @return: the OS instance if we find a valid one
1460 @raise errors.InvalidOS: if we don't find a valid OS
1463 if base_dir is None:
1464 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1466 raise errors.InvalidOS(name, None, "OS dir not found in search path")
1468 os_dir = os.path.sep.join([base_dir, name])
1470 api_versions = _OSOndiskVersion(name, os_dir)
1472 if constants.OS_API_VERSION not in api_versions:
1473 raise errors.InvalidOS(name, os_dir, "API version mismatch"
1474 " (found %s want %s)"
1475 % (api_versions, constants.OS_API_VERSION))
1477 # OS Scripts dictionary, we will populate it with the actual script names
1478 os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
1480 for script in os_scripts:
1481 os_scripts[script] = os.path.sep.join([os_dir, script])
1484 st = os.stat(os_scripts[script])
1485 except EnvironmentError, err:
1486 raise errors.InvalidOS(name, os_dir, "'%s' script missing (%s)" %
1487 (script, _ErrnoOrStr(err)))
1489 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1490 raise errors.InvalidOS(name, os_dir, "'%s' script not executable" %
1493 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1494 raise errors.InvalidOS(name, os_dir, "'%s' is not a regular file" %
1498 return objects.OS(name=name, path=os_dir, status=constants.OS_VALID_STATUS,
1499 create_script=os_scripts[constants.OS_SCRIPT_CREATE],
1500 export_script=os_scripts[constants.OS_SCRIPT_EXPORT],
1501 import_script=os_scripts[constants.OS_SCRIPT_IMPORT],
1502 rename_script=os_scripts[constants.OS_SCRIPT_RENAME],
1503 api_versions=api_versions)
1505 def OSEnvironment(instance, debug=0):
1506 """Calculate the environment for an os script.
1508 @type instance: L{objects.Instance}
1509 @param instance: target instance for the os script run
1510 @type debug: integer
1511 @param debug: debug level (0 or 1, for OS Api 10)
1513 @return: dict of environment variables
1514 @raise errors.BlockDeviceError: if the block device
1519 result['OS_API_VERSION'] = '%d' % constants.OS_API_VERSION
1520 result['INSTANCE_NAME'] = instance.name
1521 result['HYPERVISOR'] = instance.hypervisor
1522 result['DISK_COUNT'] = '%d' % len(instance.disks)
1523 result['NIC_COUNT'] = '%d' % len(instance.nics)
1524 result['DEBUG_LEVEL'] = '%d' % debug
1525 for idx, disk in enumerate(instance.disks):
1526 real_disk = _RecursiveFindBD(disk)
1527 if real_disk is None:
1528 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1531 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1532 # FIXME: When disks will have read-only mode, populate this
1533 result['DISK_%d_ACCESS' % idx] = 'W'
1534 if constants.HV_DISK_TYPE in instance.hvparams:
1535 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1536 instance.hvparams[constants.HV_DISK_TYPE]
1537 if disk.dev_type in constants.LDS_BLOCK:
1538 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1539 elif disk.dev_type == constants.LD_FILE:
1540 result['DISK_%d_BACKEND_TYPE' % idx] = \
1541 'file:%s' % disk.physical_id[0]
1542 for idx, nic in enumerate(instance.nics):
1543 result['NIC_%d_MAC' % idx] = nic.mac
1545 result['NIC_%d_IP' % idx] = nic.ip
1546 result['NIC_%d_BRIDGE' % idx] = nic.bridge
1547 if constants.HV_NIC_TYPE in instance.hvparams:
1548 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1549 instance.hvparams[constants.HV_NIC_TYPE]
1553 def GrowBlockDevice(disk, amount):
1554 """Grow a stack of block devices.
1556 This function is called recursively, with the childrens being the
1557 first ones to resize.
1559 @type disk: L{objects.Disk}
1560 @param disk: the disk to be grown
1561 @rtype: (status, result)
1562 @return: a tuple with the status of the operation
1563 (True/False), and the errors message if status
1567 r_dev = _RecursiveFindBD(disk)
1569 return False, "Cannot find block device %s" % (disk,)
1573 except errors.BlockDeviceError, err:
1574 return False, str(err)
1579 def SnapshotBlockDevice(disk):
1580 """Create a snapshot copy of a block device.
1582 This function is called recursively, and the snapshot is actually created
1583 just for the leaf lvm backend device.
1585 @type disk: L{objects.Disk}
1586 @param disk: the disk to be snapshotted
1588 @return: snapshot disk path
1592 if len(disk.children) == 1:
1593 # only one child, let's recurse on it
1594 return SnapshotBlockDevice(disk.children[0])
1596 # more than one child, choose one that matches
1597 for child in disk.children:
1598 if child.size == disk.size:
1599 # return implies breaking the loop
1600 return SnapshotBlockDevice(child)
1601 elif disk.dev_type == constants.LD_LV:
1602 r_dev = _RecursiveFindBD(disk)
1603 if r_dev is not None:
1604 # let's stay on the safe side and ask for the full size, for now
1605 return r_dev.Snapshot(disk.size)
1609 raise errors.ProgrammerError("Cannot snapshot non-lvm block device"
1610 " '%s' of type '%s'" %
1611 (disk.unique_id, disk.dev_type))
1614 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1615 """Export a block device snapshot to a remote node.
1617 @type disk: L{objects.Disk}
1618 @param disk: the description of the disk to export
1619 @type dest_node: str
1620 @param dest_node: the destination node to export to
1621 @type instance: L{objects.Instance}
1622 @param instance: the instance object to whom the disk belongs
1623 @type cluster_name: str
1624 @param cluster_name: the cluster name, needed for SSH hostalias
1626 @param idx: the index of the disk in the instance's disk list,
1627 used to export to the OS scripts environment
1629 @return: the success of the operation
1632 export_env = OSEnvironment(instance)
1634 inst_os = OSFromDisk(instance.os)
1635 export_script = inst_os.export_script
1637 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1638 instance.name, int(time.time()))
1639 if not os.path.exists(constants.LOG_OS_DIR):
1640 os.mkdir(constants.LOG_OS_DIR, 0750)
1641 real_disk = _RecursiveFindBD(disk)
1642 if real_disk is None:
1643 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1647 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1648 export_env['EXPORT_INDEX'] = str(idx)
1650 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1651 destfile = disk.physical_id[1]
1653 # the target command is built out of three individual commands,
1654 # which are joined by pipes; we check each individual command for
1656 expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
1657 export_script, logfile)
1661 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1662 destdir, destdir, destfile)
1663 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1664 constants.GANETI_RUNAS,
1667 # all commands have been checked, so we're safe to combine them
1668 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1670 result = utils.RunCmd(command, env=export_env)
1673 logging.error("os snapshot export command '%s' returned error: %s"
1674 " output: %s", command, result.fail_reason, result.output)
1680 def FinalizeExport(instance, snap_disks):
1681 """Write out the export configuration information.
1683 @type instance: L{objects.Instance}
1684 @param instance: the instance which we export, used for
1685 saving configuration
1686 @type snap_disks: list of L{objects.Disk}
1687 @param snap_disks: list of snapshot block devices, which
1688 will be used to get the actual name of the dump file
1691 @return: the success of the operation
1694 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1695 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1697 config = objects.SerializableConfigParser()
1699 config.add_section(constants.INISECT_EXP)
1700 config.set(constants.INISECT_EXP, 'version', '0')
1701 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1702 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1703 config.set(constants.INISECT_EXP, 'os', instance.os)
1704 config.set(constants.INISECT_EXP, 'compression', 'gzip')
1706 config.add_section(constants.INISECT_INS)
1707 config.set(constants.INISECT_INS, 'name', instance.name)
1708 config.set(constants.INISECT_INS, 'memory', '%d' %
1709 instance.beparams[constants.BE_MEMORY])
1710 config.set(constants.INISECT_INS, 'vcpus', '%d' %
1711 instance.beparams[constants.BE_VCPUS])
1712 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1715 for nic_count, nic in enumerate(instance.nics):
1717 config.set(constants.INISECT_INS, 'nic%d_mac' %
1718 nic_count, '%s' % nic.mac)
1719 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
1720 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
1722 # TODO: redundant: on load can read nics until it doesn't exist
1723 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
1726 for disk_count, disk in enumerate(snap_disks):
1729 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
1730 ('%s' % disk.iv_name))
1731 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
1732 ('%s' % disk.physical_id[1]))
1733 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
1736 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
1738 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
1739 data=config.Dumps())
1740 shutil.rmtree(finaldestdir, True)
1741 shutil.move(destdir, finaldestdir)
1746 def ExportInfo(dest):
1747 """Get export configuration information.
1750 @param dest: directory containing the export
1752 @rtype: L{objects.SerializableConfigParser}
1753 @return: a serializable config file containing the
1757 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
1759 config = objects.SerializableConfigParser()
1762 if (not config.has_section(constants.INISECT_EXP) or
1763 not config.has_section(constants.INISECT_INS)):
1769 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
1770 """Import an os image into an instance.
1772 @type instance: L{objects.Instance}
1773 @param instance: instance to import the disks into
1774 @type src_node: string
1775 @param src_node: source node for the disk images
1776 @type src_images: list of string
1777 @param src_images: absolute paths of the disk images
1778 @rtype: list of boolean
1779 @return: each boolean represent the success of importing the n-th disk
1782 import_env = OSEnvironment(instance)
1783 inst_os = OSFromDisk(instance.os)
1784 import_script = inst_os.import_script
1786 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
1787 instance.name, int(time.time()))
1788 if not os.path.exists(constants.LOG_OS_DIR):
1789 os.mkdir(constants.LOG_OS_DIR, 0750)
1792 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
1793 import_script, logfile)
1796 for idx, image in enumerate(src_images):
1798 destcmd = utils.BuildShellCmd('cat %s', image)
1799 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
1800 constants.GANETI_RUNAS,
1802 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
1803 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
1804 import_env['IMPORT_INDEX'] = str(idx)
1805 result = utils.RunCmd(command, env=import_env)
1807 logging.error("Disk import command '%s' returned error: %s"
1808 " output: %s", command, result.fail_reason,
1810 final_result.append(False)
1812 final_result.append(True)
1814 final_result.append(True)
1820 """Return a list of exports currently available on this machine.
1823 @return: list of the exports
1826 if os.path.isdir(constants.EXPORT_DIR):
1827 return utils.ListVisibleFiles(constants.EXPORT_DIR)
1832 def RemoveExport(export):
1833 """Remove an existing export from the node.
1836 @param export: the name of the export to remove
1838 @return: the success of the operation
1841 target = os.path.join(constants.EXPORT_DIR, export)
1843 shutil.rmtree(target)
1844 # TODO: catch some of the relevant exceptions and provide a pretty
1845 # error message if rmtree fails.
1850 def RenameBlockDevices(devlist):
1851 """Rename a list of block devices.
1853 @type devlist: list of tuples
1854 @param devlist: list of tuples of the form (disk,
1855 new_logical_id, new_physical_id); disk is an
1856 L{objects.Disk} object describing the current disk,
1857 and new logical_id/physical_id is the name we
1860 @return: True if all renames succeeded, False otherwise
1864 for disk, unique_id in devlist:
1865 dev = _RecursiveFindBD(disk)
1870 old_rpath = dev.dev_path
1871 dev.Rename(unique_id)
1872 new_rpath = dev.dev_path
1873 if old_rpath != new_rpath:
1874 DevCacheManager.RemoveCache(old_rpath)
1875 # FIXME: we should add the new cache information here, like:
1876 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
1877 # but we don't have the owner here - maybe parse from existing
1878 # cache? for now, we only lose lvm data when we rename, which
1879 # is less critical than DRBD or MD
1880 except errors.BlockDeviceError, err:
1881 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
1886 def _TransformFileStorageDir(file_storage_dir):
1887 """Checks whether given file_storage_dir is valid.
1889 Checks wheter the given file_storage_dir is within the cluster-wide
1890 default file_storage_dir stored in SimpleStore. Only paths under that
1891 directory are allowed.
1893 @type file_storage_dir: str
1894 @param file_storage_dir: the path to check
1896 @return: the normalized path if valid, None otherwise
1900 file_storage_dir = os.path.normpath(file_storage_dir)
1901 base_file_storage_dir = cfg.GetFileStorageDir()
1902 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
1903 base_file_storage_dir):
1904 logging.error("file storage directory '%s' is not under base file"
1905 " storage directory '%s'",
1906 file_storage_dir, base_file_storage_dir)
1908 return file_storage_dir
1911 def CreateFileStorageDir(file_storage_dir):
1912 """Create file storage directory.
1914 @type file_storage_dir: str
1915 @param file_storage_dir: directory to create
1918 @return: tuple with first element a boolean indicating wheter dir
1919 creation was successful or not
1922 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
1924 if not file_storage_dir:
1927 if os.path.exists(file_storage_dir):
1928 if not os.path.isdir(file_storage_dir):
1929 logging.error("'%s' is not a directory", file_storage_dir)
1933 os.makedirs(file_storage_dir, 0750)
1934 except OSError, err:
1935 logging.error("Cannot create file storage directory '%s': %s",
1936 file_storage_dir, err)
1941 def RemoveFileStorageDir(file_storage_dir):
1942 """Remove file storage directory.
1944 Remove it only if it's empty. If not log an error and return.
1946 @type file_storage_dir: str
1947 @param file_storage_dir: the directory we should cleanup
1948 @rtype: tuple (success,)
1949 @return: tuple of one element, C{success}, denoting
1950 whether the operation was successfull
1953 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
1955 if not file_storage_dir:
1958 if os.path.exists(file_storage_dir):
1959 if not os.path.isdir(file_storage_dir):
1960 logging.error("'%s' is not a directory", file_storage_dir)
1962 # deletes dir only if empty, otherwise we want to return False
1964 os.rmdir(file_storage_dir)
1965 except OSError, err:
1966 logging.exception("Cannot remove file storage directory '%s'",
1972 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
1973 """Rename the file storage directory.
1975 @type old_file_storage_dir: str
1976 @param old_file_storage_dir: the current path
1977 @type new_file_storage_dir: str
1978 @param new_file_storage_dir: the name we should rename to
1979 @rtype: tuple (success,)
1980 @return: tuple of one element, C{success}, denoting
1981 whether the operation was successful
1984 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
1985 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
1987 if not old_file_storage_dir or not new_file_storage_dir:
1990 if not os.path.exists(new_file_storage_dir):
1991 if os.path.isdir(old_file_storage_dir):
1993 os.rename(old_file_storage_dir, new_file_storage_dir)
1994 except OSError, err:
1995 logging.exception("Cannot rename '%s' to '%s'",
1996 old_file_storage_dir, new_file_storage_dir)
1999 logging.error("'%s' is not a directory", old_file_storage_dir)
2002 if os.path.exists(old_file_storage_dir):
2003 logging.error("Cannot rename '%s' to '%s'. Both locations exist.",
2004 old_file_storage_dir, new_file_storage_dir)
2009 def _IsJobQueueFile(file_name):
2010 """Checks whether the given filename is in the queue directory.
2012 @type file_name: str
2013 @param file_name: the file name we should check
2015 @return: whether the file is under the queue directory
2018 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2019 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2022 logging.error("'%s' is not a file in the queue directory",
2028 def JobQueueUpdate(file_name, content):
2029 """Updates a file in the queue directory.
2031 This is just a wrapper over L{utils.WriteFile}, with proper
2034 @type file_name: str
2035 @param file_name: the job file name
2037 @param content: the new job contents
2039 @return: the success of the operation
2042 if not _IsJobQueueFile(file_name):
2045 # Write and replace the file atomically
2046 utils.WriteFile(file_name, data=_Decompress(content))
2051 def JobQueueRename(old, new):
2052 """Renames a job queue file.
2054 This is just a wrapper over os.rename with proper checking.
2057 @param old: the old (actual) file name
2059 @param new: the desired file name
2061 @return: the success of the operation
2064 if not (_IsJobQueueFile(old) and _IsJobQueueFile(new)):
2067 utils.RenameFile(old, new, mkdir=True)
2072 def JobQueueSetDrainFlag(drain_flag):
2073 """Set the drain flag for the queue.
2075 This will set or unset the queue drain flag.
2077 @type drain_flag: boolean
2078 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2080 @return: always True
2081 @warning: the function always returns True
2085 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2087 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2092 def CloseBlockDevices(instance_name, disks):
2093 """Closes the given block devices.
2095 This means they will be switched to secondary mode (in case of
2098 @param instance_name: if the argument is not empty, the symlinks
2099 of this instance will be removed
2100 @type disks: list of L{objects.Disk}
2101 @param disks: the list of disks to be closed
2102 @rtype: tuple (success, message)
2103 @return: a tuple of success and message, where success
2104 indicates the succes of the operation, and message
2105 which will contain the error details in case we
2111 rd = _RecursiveFindBD(cf)
2113 return (False, "Can't find device %s" % cf)
2120 except errors.BlockDeviceError, err:
2121 msg.append(str(err))
2123 return (False, "Can't make devices secondary: %s" % ",".join(msg))
2126 _RemoveBlockDevLinks(instance_name, disks)
2127 return (True, "All devices secondary")
2130 def ValidateHVParams(hvname, hvparams):
2131 """Validates the given hypervisor parameters.
2133 @type hvname: string
2134 @param hvname: the hypervisor name
2135 @type hvparams: dict
2136 @param hvparams: the hypervisor parameters to be validated
2137 @rtype: tuple (success, message)
2138 @return: a tuple of success and message, where success
2139 indicates the succes of the operation, and message
2140 which will contain the error details in case we
2145 hv_type = hypervisor.GetHypervisor(hvname)
2146 hv_type.ValidateParameters(hvparams)
2147 return (True, "Validation passed")
2148 except errors.HypervisorError, err:
2149 return (False, str(err))
2153 """Demotes the current node from master candidate role.
2156 # try to ensure we're not the master by mistake
2157 master, myself = ssconf.GetMasterAndMyself()
2158 if master == myself:
2159 return (False, "ssconf status shows I'm the master node, will not demote")
2160 pid_file = utils.DaemonPidFileName(constants.MASTERD_PID)
2161 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2162 return (False, "The master daemon is running, will not demote")
2164 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2165 except EnvironmentError, err:
2166 if err.errno != errno.ENOENT:
2167 return (False, "Error while backing up cluster file: %s" % str(err))
2168 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2169 return (True, "Done")
2172 def _FindDisks(nodes_ip, disks):
2173 """Sets the physical ID on disks and returns the block devices.
2176 # set the correct physical ID
2177 my_name = utils.HostInfo().name
2179 cf.SetPhysicalID(my_name, nodes_ip)
2184 rd = _RecursiveFindBD(cf)
2186 return (False, "Can't find device %s" % cf)
2188 return (True, bdevs)
2191 def DrbdDisconnectNet(nodes_ip, disks):
2192 """Disconnects the network on a list of drbd devices.
2195 status, bdevs = _FindDisks(nodes_ip, disks)
2197 return status, bdevs
2203 except errors.BlockDeviceError, err:
2204 logging.exception("Failed to go into standalone mode")
2205 return (False, "Can't change network configuration: %s" % str(err))
2206 return (True, "All disks are now disconnected")
2209 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2210 """Attaches the network on a list of drbd devices.
2213 status, bdevs = _FindDisks(nodes_ip, disks)
2215 return status, bdevs
2218 for idx, rd in enumerate(bdevs):
2220 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2221 except EnvironmentError, err:
2222 return (False, "Can't create symlink: %s" % str(err))
2223 # reconnect disks, switch to new master configuration and if
2224 # needed primary mode
2227 rd.AttachNet(multimaster)
2228 except errors.BlockDeviceError, err:
2229 return (False, "Can't change network configuration: %s" % str(err))
2230 # wait until the disks are connected; we need to retry the re-attach
2231 # if the device becomes standalone, as this might happen if the one
2232 # node disconnects and reconnects in a different mode before the
2233 # other node reconnects; in this case, one or both of the nodes will
2234 # decide it has wrong configuration and switch to standalone
2235 RECONNECT_TIMEOUT = 2 * 60
2236 sleep_time = 0.100 # start with 100 miliseconds
2237 timeout_limit = time.time() + RECONNECT_TIMEOUT
2238 while time.time() < timeout_limit:
2239 all_connected = True
2241 stats = rd.GetProcStatus()
2242 if not (stats.is_connected or stats.is_in_resync):
2243 all_connected = False
2244 if stats.is_standalone:
2245 # peer had different config info and this node became
2246 # standalone, even though this should not happen with the
2247 # new staged way of changing disk configs
2249 rd.ReAttachNet(multimaster)
2250 except errors.BlockDeviceError, err:
2251 return (False, "Can't change network configuration: %s" % str(err))
2254 time.sleep(sleep_time)
2255 sleep_time = min(5, sleep_time * 1.5)
2256 if not all_connected:
2257 return (False, "Timeout in disk reconnecting")
2259 # change to primary mode
2263 msg = "multi-master and primary"
2265 msg = "single-master"
2266 return (True, "Disks are now configured as %s" % msg)
2269 def DrbdWaitSync(nodes_ip, disks):
2270 """Wait until DRBDs have synchronized.
2273 status, bdevs = _FindDisks(nodes_ip, disks)
2275 return status, bdevs
2281 stats = rd.GetProcStatus()
2282 if not (stats.is_connected or stats.is_in_resync):
2285 alldone = alldone and (not stats.is_in_resync)
2286 if stats.sync_percent is not None:
2287 min_resync = min(min_resync, stats.sync_percent)
2288 return (not failure, (alldone, min_resync))
2291 class HooksRunner(object):
2294 This class is instantiated on the node side (ganeti-noded) and not
2298 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2300 def __init__(self, hooks_base_dir=None):
2301 """Constructor for hooks runner.
2303 @type hooks_base_dir: str or None
2304 @param hooks_base_dir: if not None, this overrides the
2305 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2308 if hooks_base_dir is None:
2309 hooks_base_dir = constants.HOOKS_BASE_DIR
2310 self._BASE_DIR = hooks_base_dir
2313 def ExecHook(script, env):
2314 """Exec one hook script.
2317 @param script: the full path to the script
2319 @param env: the environment with which to exec the script
2320 @rtype: tuple (success, message)
2321 @return: a tuple of success and message, where success
2322 indicates the succes of the operation, and message
2323 which will contain the error details in case we
2327 # exec the process using subprocess and log the output
2330 fdstdin = open("/dev/null", "r")
2331 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2332 stderr=subprocess.STDOUT, close_fds=True,
2333 shell=False, cwd="/", env=env)
2336 output = child.stdout.read(4096)
2337 child.stdout.close()
2338 except EnvironmentError, err:
2339 output += "Hook script error: %s" % str(err)
2343 result = child.wait()
2345 except EnvironmentError, err:
2346 if err.errno == errno.EINTR:
2350 # try not to leak fds
2351 for fd in (fdstdin, ):
2355 except EnvironmentError, err:
2356 # just log the error
2357 #logging.exception("Error while closing fd %s", fd)
2360 return result == 0, output
2362 def RunHooks(self, hpath, phase, env):
2363 """Run the scripts in the hooks directory.
2366 @param hpath: the path to the hooks directory which
2369 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2370 L{constants.HOOKS_PHASE_POST}
2372 @param env: dictionary with the environment for the hook
2374 @return: list of 3-element tuples:
2376 - script result, either L{constants.HKR_SUCCESS} or
2377 L{constants.HKR_FAIL}
2378 - output of the script
2380 @raise errors.ProgrammerError: for invalid input
2384 if phase == constants.HOOKS_PHASE_PRE:
2386 elif phase == constants.HOOKS_PHASE_POST:
2389 raise errors.ProgrammerError("Unknown hooks phase: '%s'" % phase)
2392 subdir = "%s-%s.d" % (hpath, suffix)
2393 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2395 dir_contents = utils.ListVisibleFiles(dir_name)
2396 except OSError, err:
2397 # FIXME: must log output in case of failures
2400 # we use the standard python sort order,
2401 # so 00name is the recommended naming scheme
2403 for relname in dir_contents:
2404 fname = os.path.join(dir_name, relname)
2405 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2406 self.RE_MASK.match(relname) is not None):
2407 rrval = constants.HKR_SKIP
2410 result, output = self.ExecHook(fname, env)
2412 rrval = constants.HKR_FAIL
2414 rrval = constants.HKR_SUCCESS
2415 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2420 class IAllocatorRunner(object):
2421 """IAllocator runner.
2423 This class is instantiated on the node side (ganeti-noded) and not on
2427 def Run(self, name, idata):
2428 """Run an iallocator script.
2431 @param name: the iallocator script name
2433 @param idata: the allocator input data
2436 @return: four element tuple of:
2437 - run status (one of the IARUN_ constants)
2440 - fail reason (as from L{utils.RunResult})
2443 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2445 if alloc_script is None:
2446 return (constants.IARUN_NOTFOUND, None, None, None)
2448 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2452 result = utils.RunCmd([alloc_script, fin_name])
2454 return (constants.IARUN_FAILURE, result.stdout, result.stderr,
2459 return (constants.IARUN_SUCCESS, result.stdout, result.stderr, None)
2462 class DevCacheManager(object):
2463 """Simple class for managing a cache of block device information.
2466 _DEV_PREFIX = "/dev/"
2467 _ROOT_DIR = constants.BDEV_CACHE_DIR
2470 def _ConvertPath(cls, dev_path):
2471 """Converts a /dev/name path to the cache file name.
2473 This replaces slashes with underscores and strips the /dev
2474 prefix. It then returns the full path to the cache file.
2477 @param dev_path: the C{/dev/} path name
2479 @return: the converted path name
2482 if dev_path.startswith(cls._DEV_PREFIX):
2483 dev_path = dev_path[len(cls._DEV_PREFIX):]
2484 dev_path = dev_path.replace("/", "_")
2485 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2489 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2490 """Updates the cache information for a given device.
2493 @param dev_path: the pathname of the device
2495 @param owner: the owner (instance name) of the device
2496 @type on_primary: bool
2497 @param on_primary: whether this is the primary
2500 @param iv_name: the instance-visible name of the
2501 device, as in objects.Disk.iv_name
2506 if dev_path is None:
2507 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2509 fpath = cls._ConvertPath(dev_path)
2515 iv_name = "not_visible"
2516 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2518 utils.WriteFile(fpath, data=fdata)
2519 except EnvironmentError, err:
2520 logging.exception("Can't update bdev cache for %s", dev_path)
2523 def RemoveCache(cls, dev_path):
2524 """Remove data for a dev_path.
2526 This is just a wrapper over L{utils.RemoveFile} with a converted
2527 path name and logging.
2530 @param dev_path: the pathname of the device
2535 if dev_path is None:
2536 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2538 fpath = cls._ConvertPath(dev_path)
2540 utils.RemoveFile(fpath)
2541 except EnvironmentError, err:
2542 logging.exception("Can't update bdev cache for %s", dev_path)