4 # Copyright (C) 2006, 2007 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Functions used by the node daemon
24 @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
25 the L{UploadFile} function
44 from ganeti import errors
45 from ganeti import utils
46 from ganeti import ssh
47 from ganeti import hypervisor
48 from ganeti import constants
49 from ganeti import bdev
50 from ganeti import objects
51 from ganeti import ssconf
54 class RPCFail(Exception):
55 """Class denoting RPC failure.
57 Its argument is the error message.
61 def _Fail(msg, *args, **kwargs):
62 """Log an error and the raise an RPCFail exception.
64 This exception is then handled specially in the ganeti daemon and
65 turned into a 'failed' return type. As such, this function is a
66 useful shortcut for logging the error and returning it to the master
70 @param msg: the text of the exception
76 if "log" not in kwargs or kwargs["log"]: # if we should log this error
77 if "exc" in kwargs and kwargs["exc"]:
78 logging.exception(msg)
85 """Simple wrapper to return a SimpleStore.
87 @rtype: L{ssconf.SimpleStore}
88 @return: a SimpleStore instance
91 return ssconf.SimpleStore()
94 def _GetSshRunner(cluster_name):
95 """Simple wrapper to return an SshRunner.
97 @type cluster_name: str
98 @param cluster_name: the cluster name, which is needed
99 by the SshRunner constructor
100 @rtype: L{ssh.SshRunner}
101 @return: an SshRunner instance
104 return ssh.SshRunner(cluster_name)
107 def _Decompress(data):
108 """Unpacks data compressed by the RPC client.
110 @type data: list or tuple
111 @param data: Data sent by RPC client
113 @return: Decompressed data
116 assert isinstance(data, (list, tuple))
117 assert len(data) == 2
118 (encoding, content) = data
119 if encoding == constants.RPC_ENCODING_NONE:
121 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
122 return zlib.decompress(base64.b64decode(content))
124 raise AssertionError("Unknown data encoding")
127 def _CleanDirectory(path, exclude=None):
128 """Removes all regular files in a directory.
131 @param path: the directory to clean
133 @param exclude: list of files to be excluded, defaults
137 if not os.path.isdir(path):
142 # Normalize excluded paths
143 exclude = [os.path.normpath(i) for i in exclude]
145 for rel_name in utils.ListVisibleFiles(path):
146 full_name = os.path.normpath(os.path.join(path, rel_name))
147 if full_name in exclude:
149 if os.path.isfile(full_name) and not os.path.islink(full_name):
150 utils.RemoveFile(full_name)
153 def _BuildUploadFileList():
154 """Build the list of allowed upload files.
156 This is abstracted so that it's built only once at module import time.
159 allowed_files = set([
160 constants.CLUSTER_CONF_FILE,
162 constants.SSH_KNOWN_HOSTS_FILE,
163 constants.VNC_PASSWORD_FILE,
164 constants.RAPI_CERT_FILE,
165 constants.RAPI_USERS_FILE,
166 constants.HMAC_CLUSTER_KEY,
169 for hv_name in constants.HYPER_TYPES:
170 hv_class = hypervisor.GetHypervisorClass(hv_name)
171 allowed_files.update(hv_class.GetAncillaryFiles())
173 return frozenset(allowed_files)
176 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
180 """Removes job queue files and archived jobs.
186 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
187 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
191 """Returns master information.
193 This is an utility function to compute master information, either
194 for consumption here or from the node daemon.
197 @return: master_netdev, master_ip, master_name
198 @raise RPCFail: in case of errors
203 master_netdev = cfg.GetMasterNetdev()
204 master_ip = cfg.GetMasterIP()
205 master_node = cfg.GetMasterNode()
206 except errors.ConfigurationError, err:
207 _Fail("Cluster configuration incomplete: %s", err, exc=True)
208 return (master_netdev, master_ip, master_node)
211 def StartMaster(start_daemons, no_voting):
212 """Activate local node as master node.
214 The function will always try activate the IP address of the master
215 (unless someone else has it). It will also start the master daemons,
216 based on the start_daemons parameter.
218 @type start_daemons: boolean
219 @param start_daemons: whether to also start the master
220 daemons (ganeti-masterd and ganeti-rapi)
221 @type no_voting: boolean
222 @param no_voting: whether to start ganeti-masterd without a node vote
223 (if start_daemons is True), but still non-interactively
227 # GetMasterInfo will raise an exception if not able to return data
228 master_netdev, master_ip, _ = GetMasterInfo()
231 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
232 if utils.OwnIpAddress(master_ip):
233 # we already have the ip:
234 logging.debug("Master IP already configured, doing nothing")
236 msg = "Someone else has the master ip, not activating"
240 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
241 "dev", master_netdev, "label",
242 "%s:0" % master_netdev])
244 msg = "Can't activate master IP: %s" % result.output
248 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
249 "-s", master_ip, master_ip])
250 # we'll ignore the exit code of arping
252 # and now start the master and rapi daemons
255 'ganeti-masterd': [],
259 daemons_params['ganeti-masterd'].append('--no-voting')
260 daemons_params['ganeti-masterd'].append('--yes-do-it')
261 for daemon in daemons_params:
263 cmd.extend(daemons_params[daemon])
264 result = utils.RunCmd(cmd)
266 msg = "Can't start daemon %s: %s" % (daemon, result.output)
271 _Fail("; ".join(err_msgs))
274 def StopMaster(stop_daemons):
275 """Deactivate this node as master.
277 The function will always try to deactivate the IP address of the
278 master. It will also stop the master daemons depending on the
279 stop_daemons parameter.
281 @type stop_daemons: boolean
282 @param stop_daemons: whether to also stop the master daemons
283 (ganeti-masterd and ganeti-rapi)
287 # TODO: log and report back to the caller the error failures; we
288 # need to decide in which case we fail the RPC for this
290 # GetMasterInfo will raise an exception if not able to return data
291 master_netdev, master_ip, _ = GetMasterInfo()
293 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
294 "dev", master_netdev])
296 logging.error("Can't remove the master IP, error: %s", result.output)
297 # but otherwise ignore the failure
300 # stop/kill the rapi and the master daemon
301 for daemon in constants.RAPI, constants.MASTERD:
302 utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon)))
305 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
306 """Joins this node to the cluster.
308 This does the following:
309 - updates the hostkeys of the machine (rsa and dsa)
310 - adds the ssh private key to the user
311 - adds the ssh public key to the users' authorized_keys file
314 @param dsa: the DSA private key to write
316 @param dsapub: the DSA public key to write
318 @param rsa: the RSA private key to write
320 @param rsapub: the RSA public key to write
322 @param sshkey: the SSH private key to write
324 @param sshpub: the SSH public key to write
326 @return: the success of the operation
329 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
330 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
331 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
332 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
333 for name, content, mode in sshd_keys:
334 utils.WriteFile(name, data=content, mode=mode)
337 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
339 except errors.OpExecError, err:
340 _Fail("Error while processing user ssh files: %s", err, exc=True)
342 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
343 utils.WriteFile(name, data=content, mode=0600)
345 utils.AddAuthorizedKey(auth_keys, sshpub)
347 utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
351 """Cleans up and remove the current node.
353 This function cleans up and prepares the current node to be removed
356 If processing is successful, then it raises an
357 L{errors.QuitGanetiException} which is used as a special case to
358 shutdown the node daemon.
361 _CleanDirectory(constants.DATA_DIR)
365 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
367 f = open(pub_key, 'r')
369 utils.RemoveAuthorizedKey(auth_keys, f.read(8192))
373 utils.RemoveFile(priv_key)
374 utils.RemoveFile(pub_key)
375 except errors.OpExecError:
376 logging.exception("Error while processing ssh files")
378 # Raise a custom exception (handled in ganeti-noded)
379 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
382 def GetNodeInfo(vgname, hypervisor_type):
383 """Gives back a hash with different information about the node.
385 @type vgname: C{string}
386 @param vgname: the name of the volume group to ask for disk space information
387 @type hypervisor_type: C{str}
388 @param hypervisor_type: the name of the hypervisor to ask for
391 @return: dictionary with the following keys:
392 - vg_size is the size of the configured volume group in MiB
393 - vg_free is the free size of the volume group in MiB
394 - memory_dom0 is the memory allocated for domain0 in MiB
395 - memory_free is the currently available (free) ram in MiB
396 - memory_total is the total number of ram in MiB
400 vginfo = _GetVGInfo(vgname)
401 outputarray['vg_size'] = vginfo['vg_size']
402 outputarray['vg_free'] = vginfo['vg_free']
404 hyper = hypervisor.GetHypervisor(hypervisor_type)
405 hyp_info = hyper.GetNodeInfo()
406 if hyp_info is not None:
407 outputarray.update(hyp_info)
409 f = open("/proc/sys/kernel/random/boot_id", 'r')
411 outputarray["bootid"] = f.read(128).rstrip("\n")
418 def VerifyNode(what, cluster_name):
419 """Verify the status of the local node.
421 Based on the input L{what} parameter, various checks are done on the
424 If the I{filelist} key is present, this list of
425 files is checksummed and the file/checksum pairs are returned.
427 If the I{nodelist} key is present, we check that we have
428 connectivity via ssh with the target nodes (and check the hostname
431 If the I{node-net-test} key is present, we check that we have
432 connectivity to the given nodes via both primary IP and, if
433 applicable, secondary IPs.
436 @param what: a dictionary of things to check:
437 - filelist: list of files for which to compute checksums
438 - nodelist: list of nodes we should check ssh communication with
439 - node-net-test: list of nodes we should check node daemon port
441 - hypervisor: list with hypervisors to run the verify for
443 @return: a dictionary with the same keys as the input dict, and
444 values representing the result of the checks
449 if constants.NV_HYPERVISOR in what:
450 result[constants.NV_HYPERVISOR] = tmp = {}
451 for hv_name in what[constants.NV_HYPERVISOR]:
452 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
454 if constants.NV_FILELIST in what:
455 result[constants.NV_FILELIST] = utils.FingerprintFiles(
456 what[constants.NV_FILELIST])
458 if constants.NV_NODELIST in what:
459 result[constants.NV_NODELIST] = tmp = {}
460 random.shuffle(what[constants.NV_NODELIST])
461 for node in what[constants.NV_NODELIST]:
462 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
466 if constants.NV_NODENETTEST in what:
467 result[constants.NV_NODENETTEST] = tmp = {}
468 my_name = utils.HostInfo().name
469 my_pip = my_sip = None
470 for name, pip, sip in what[constants.NV_NODENETTEST]:
476 tmp[my_name] = ("Can't find my own primary/secondary IP"
479 port = utils.GetDaemonPort(constants.NODED)
480 for name, pip, sip in what[constants.NV_NODENETTEST]:
482 if not utils.TcpPing(pip, port, source=my_pip):
483 fail.append("primary")
485 if not utils.TcpPing(sip, port, source=my_sip):
486 fail.append("secondary")
488 tmp[name] = ("failure using the %s interface(s)" %
491 if constants.NV_LVLIST in what:
492 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
494 if constants.NV_INSTANCELIST in what:
495 result[constants.NV_INSTANCELIST] = GetInstanceList(
496 what[constants.NV_INSTANCELIST])
498 if constants.NV_VGLIST in what:
499 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
501 if constants.NV_VERSION in what:
502 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
503 constants.RELEASE_VERSION)
505 if constants.NV_HVINFO in what:
506 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
507 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
509 if constants.NV_DRBDLIST in what:
511 used_minors = bdev.DRBD8.GetUsedDevs().keys()
512 except errors.BlockDeviceError, err:
513 logging.warning("Can't get used minors list", exc_info=True)
514 used_minors = str(err)
515 result[constants.NV_DRBDLIST] = used_minors
520 def GetVolumeList(vg_name):
521 """Compute list of logical volumes and their size.
524 @param vg_name: the volume group whose LVs we should list
527 dictionary of all partions (key) with value being a tuple of
528 their size (in MiB), inactive and online status::
530 {'test1': ('20.06', True, True)}
532 in case of errors, a string is returned with the error
538 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
539 "--separator=%s" % sep,
540 "-olv_name,lv_size,lv_attr", vg_name])
542 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
544 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
545 for line in result.stdout.splitlines():
547 match = valid_line_re.match(line)
549 logging.error("Invalid line returned from lvs output: '%s'", line)
551 name, size, attr = match.groups()
552 inactive = attr[4] == '-'
553 online = attr[5] == 'o'
554 lvs[name] = (size, inactive, online)
559 def ListVolumeGroups():
560 """List the volume groups and their size.
563 @return: dictionary with keys volume name and values the
567 return utils.ListVolumeGroups()
571 """List all volumes on this node.
575 A list of dictionaries, each having four keys:
576 - name: the logical volume name,
577 - size: the size of the logical volume
578 - dev: the physical device on which the LV lives
579 - vg: the volume group to which it belongs
581 In case of errors, we return an empty list and log the
584 Note that since a logical volume can live on multiple physical
585 volumes, the resulting list might include a logical volume
589 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
591 "--options=lv_name,lv_size,devices,vg_name"])
593 _Fail("Failed to list logical volumes, lvs output: %s",
598 return dev.split('(')[0]
604 'name': line[0].strip(),
605 'size': line[1].strip(),
606 'dev': parse_dev(line[2].strip()),
607 'vg': line[3].strip(),
610 return [map_line(line.split('|')) for line in result.stdout.splitlines()
611 if line.count('|') >= 3]
614 def BridgesExist(bridges_list):
615 """Check if a list of bridges exist on the current node.
618 @return: C{True} if all of them exist, C{False} otherwise
622 for bridge in bridges_list:
623 if not utils.BridgeExists(bridge):
624 missing.append(bridge)
627 _Fail("Missing bridges %s", ", ".join(missing))
630 def GetInstanceList(hypervisor_list):
631 """Provides a list of instances.
633 @type hypervisor_list: list
634 @param hypervisor_list: the list of hypervisors to query information
637 @return: a list of all running instances on the current node
638 - instance1.example.com
639 - instance2.example.com
643 for hname in hypervisor_list:
645 names = hypervisor.GetHypervisor(hname).ListInstances()
646 results.extend(names)
647 except errors.HypervisorError, err:
648 _Fail("Error enumerating instances (hypervisor %s): %s",
649 hname, err, exc=True)
654 def GetInstanceInfo(instance, hname):
655 """Gives back the information about an instance as a dictionary.
657 @type instance: string
658 @param instance: the instance name
660 @param hname: the hypervisor type of the instance
663 @return: dictionary with the following keys:
664 - memory: memory size of instance (int)
665 - state: xen state of instance (string)
666 - time: cpu time of instance (float)
671 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
672 if iinfo is not None:
673 output['memory'] = iinfo[2]
674 output['state'] = iinfo[4]
675 output['time'] = iinfo[5]
680 def GetInstanceMigratable(instance):
681 """Gives whether an instance can be migrated.
683 @type instance: L{objects.Instance}
684 @param instance: object representing the instance to be checked.
687 @return: tuple of (result, description) where:
688 - result: whether the instance can be migrated or not
689 - description: a description of the issue, if relevant
692 hyper = hypervisor.GetHypervisor(instance.hypervisor)
693 iname = instance.name
694 if iname not in hyper.ListInstances():
695 _Fail("Instance %s is not running", iname)
697 for idx in range(len(instance.disks)):
698 link_name = _GetBlockDevSymlinkPath(iname, idx)
699 if not os.path.islink(link_name):
700 _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
703 def GetAllInstancesInfo(hypervisor_list):
704 """Gather data about all instances.
706 This is the equivalent of L{GetInstanceInfo}, except that it
707 computes data for all instances at once, thus being faster if one
708 needs data about more than one instance.
710 @type hypervisor_list: list
711 @param hypervisor_list: list of hypervisors to query for instance data
714 @return: dictionary of instance: data, with data having the following keys:
715 - memory: memory size of instance (int)
716 - state: xen state of instance (string)
717 - time: cpu time of instance (float)
718 - vcpus: the number of vcpus
723 for hname in hypervisor_list:
724 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
726 for name, _, memory, vcpus, state, times in iinfo:
734 # we only check static parameters, like memory and vcpus,
735 # and not state and time which can change between the
736 # invocations of the different hypervisors
737 for key in 'memory', 'vcpus':
738 if value[key] != output[name][key]:
739 _Fail("Instance %s is running twice"
740 " with different parameters", name)
746 def InstanceOsAdd(instance, reinstall):
747 """Add an OS to an instance.
749 @type instance: L{objects.Instance}
750 @param instance: Instance whose OS is to be installed
751 @type reinstall: boolean
752 @param reinstall: whether this is an instance reinstall
756 inst_os = OSFromDisk(instance.os)
758 create_env = OSEnvironment(instance, inst_os)
760 create_env['INSTANCE_REINSTALL'] = "1"
762 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
763 instance.name, int(time.time()))
765 result = utils.RunCmd([inst_os.create_script], env=create_env,
766 cwd=inst_os.path, output=logfile,)
768 logging.error("os create command '%s' returned error: %s, logfile: %s,"
769 " output: %s", result.cmd, result.fail_reason, logfile,
771 lines = [utils.SafeEncode(val)
772 for val in utils.TailFile(logfile, lines=20)]
773 _Fail("OS create script failed (%s), last lines in the"
774 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
777 def RunRenameInstance(instance, old_name):
778 """Run the OS rename script for an instance.
780 @type instance: L{objects.Instance}
781 @param instance: Instance whose OS is to be installed
782 @type old_name: string
783 @param old_name: previous instance name
785 @return: the success of the operation
788 inst_os = OSFromDisk(instance.os)
790 rename_env = OSEnvironment(instance, inst_os)
791 rename_env['OLD_INSTANCE_NAME'] = old_name
793 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
795 instance.name, int(time.time()))
797 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
798 cwd=inst_os.path, output=logfile)
801 logging.error("os create command '%s' returned error: %s output: %s",
802 result.cmd, result.fail_reason, result.output)
803 lines = [utils.SafeEncode(val)
804 for val in utils.TailFile(logfile, lines=20)]
805 _Fail("OS rename script failed (%s), last lines in the"
806 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
809 def _GetVGInfo(vg_name):
810 """Get information about the volume group.
813 @param vg_name: the volume group which we query
816 A dictionary with the following keys:
817 - C{vg_size} is the total size of the volume group in MiB
818 - C{vg_free} is the free size of the volume group in MiB
819 - C{pv_count} are the number of physical disks in that VG
821 If an error occurs during gathering of data, we return the same dict
822 with keys all set to None.
825 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
827 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
828 "--nosuffix", "--units=m", "--separator=:", vg_name])
831 logging.error("volume group %s not present", vg_name)
833 valarr = retval.stdout.strip().rstrip(':').split(':')
837 "vg_size": int(round(float(valarr[0]), 0)),
838 "vg_free": int(round(float(valarr[1]), 0)),
839 "pv_count": int(valarr[2]),
841 except ValueError, err:
842 logging.exception("Fail to parse vgs output: %s", err)
844 logging.error("vgs output has the wrong number of fields (expected"
845 " three): %s", str(valarr))
849 def _GetBlockDevSymlinkPath(instance_name, idx):
850 return os.path.join(constants.DISK_LINKS_DIR,
851 "%s:%d" % (instance_name, idx))
854 def _SymlinkBlockDev(instance_name, device_path, idx):
855 """Set up symlinks to a instance's block device.
857 This is an auxiliary function run when an instance is start (on the primary
858 node) or when an instance is migrated (on the target node).
861 @param instance_name: the name of the target instance
862 @param device_path: path of the physical block device, on the node
863 @param idx: the disk index
864 @return: absolute path to the disk's symlink
867 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
869 os.symlink(device_path, link_name)
871 if err.errno == errno.EEXIST:
872 if (not os.path.islink(link_name) or
873 os.readlink(link_name) != device_path):
875 os.symlink(device_path, link_name)
882 def _RemoveBlockDevLinks(instance_name, disks):
883 """Remove the block device symlinks belonging to the given instance.
886 for idx, _ in enumerate(disks):
887 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
888 if os.path.islink(link_name):
892 logging.exception("Can't remove symlink '%s'", link_name)
895 def _GatherAndLinkBlockDevs(instance):
896 """Set up an instance's block device(s).
898 This is run on the primary node at instance startup. The block
899 devices must be already assembled.
901 @type instance: L{objects.Instance}
902 @param instance: the instance whose disks we shoul assemble
904 @return: list of (disk_object, device_path)
908 for idx, disk in enumerate(instance.disks):
909 device = _RecursiveFindBD(disk)
911 raise errors.BlockDeviceError("Block device '%s' is not set up." %
915 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
917 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
920 block_devices.append((disk, link_name))
925 def StartInstance(instance):
926 """Start an instance.
928 @type instance: L{objects.Instance}
929 @param instance: the instance object
933 running_instances = GetInstanceList([instance.hypervisor])
935 if instance.name in running_instances:
936 logging.info("Instance %s already running, not starting", instance.name)
940 block_devices = _GatherAndLinkBlockDevs(instance)
941 hyper = hypervisor.GetHypervisor(instance.hypervisor)
942 hyper.StartInstance(instance, block_devices)
943 except errors.BlockDeviceError, err:
944 _Fail("Block device error: %s", err, exc=True)
945 except errors.HypervisorError, err:
946 _RemoveBlockDevLinks(instance.name, instance.disks)
947 _Fail("Hypervisor error: %s", err, exc=True)
950 def InstanceShutdown(instance):
951 """Shut an instance down.
953 @note: this functions uses polling with a hardcoded timeout.
955 @type instance: L{objects.Instance}
956 @param instance: the instance object
960 hv_name = instance.hypervisor
961 running_instances = GetInstanceList([hv_name])
962 iname = instance.name
964 if iname not in running_instances:
965 logging.info("Instance %s not running, doing nothing", iname)
968 hyper = hypervisor.GetHypervisor(hv_name)
970 hyper.StopInstance(instance)
971 except errors.HypervisorError, err:
972 _Fail("Failed to stop instance %s: %s", iname, err)
974 # test every 10secs for 2min
978 if instance.name not in GetInstanceList([hv_name]):
982 # the shutdown did not succeed
983 logging.error("Shutdown of '%s' unsuccessful, using destroy", iname)
986 hyper.StopInstance(instance, force=True)
987 except errors.HypervisorError, err:
988 _Fail("Failed to force stop instance %s: %s", iname, err)
991 if instance.name in GetInstanceList([hv_name]):
992 _Fail("Could not shutdown instance %s even by destroy", iname)
994 _RemoveBlockDevLinks(iname, instance.disks)
997 def InstanceReboot(instance, reboot_type):
998 """Reboot an instance.
1000 @type instance: L{objects.Instance}
1001 @param instance: the instance object to reboot
1002 @type reboot_type: str
1003 @param reboot_type: the type of reboot, one the following
1005 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1006 instance OS, do not recreate the VM
1007 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1008 restart the VM (at the hypervisor level)
1009 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1010 not accepted here, since that mode is handled differently, in
1011 cmdlib, and translates into full stop and start of the
1012 instance (instead of a call_instance_reboot RPC)
1016 running_instances = GetInstanceList([instance.hypervisor])
1018 if instance.name not in running_instances:
1019 _Fail("Cannot reboot instance %s that is not running", instance.name)
1021 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1022 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1024 hyper.RebootInstance(instance)
1025 except errors.HypervisorError, err:
1026 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1027 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1029 InstanceShutdown(instance)
1030 return StartInstance(instance)
1031 except errors.HypervisorError, err:
1032 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1034 _Fail("Invalid reboot_type received: %s", reboot_type)
1037 def MigrationInfo(instance):
1038 """Gather information about an instance to be migrated.
1040 @type instance: L{objects.Instance}
1041 @param instance: the instance definition
1044 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1046 info = hyper.MigrationInfo(instance)
1047 except errors.HypervisorError, err:
1048 _Fail("Failed to fetch migration information: %s", err, exc=True)
1052 def AcceptInstance(instance, info, target):
1053 """Prepare the node to accept an instance.
1055 @type instance: L{objects.Instance}
1056 @param instance: the instance definition
1057 @type info: string/data (opaque)
1058 @param info: migration information, from the source node
1059 @type target: string
1060 @param target: target host (usually ip), on this node
1063 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1065 hyper.AcceptInstance(instance, info, target)
1066 except errors.HypervisorError, err:
1067 _Fail("Failed to accept instance: %s", err, exc=True)
1070 def FinalizeMigration(instance, info, success):
1071 """Finalize any preparation to accept an instance.
1073 @type instance: L{objects.Instance}
1074 @param instance: the instance definition
1075 @type info: string/data (opaque)
1076 @param info: migration information, from the source node
1077 @type success: boolean
1078 @param success: whether the migration was a success or a failure
1081 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1083 hyper.FinalizeMigration(instance, info, success)
1084 except errors.HypervisorError, err:
1085 _Fail("Failed to finalize migration: %s", err, exc=True)
1088 def MigrateInstance(instance, target, live):
1089 """Migrates an instance to another node.
1091 @type instance: L{objects.Instance}
1092 @param instance: the instance definition
1093 @type target: string
1094 @param target: the target node name
1096 @param live: whether the migration should be done live or not (the
1097 interpretation of this parameter is left to the hypervisor)
1099 @return: a tuple of (success, msg) where:
1100 - succes is a boolean denoting the success/failure of the operation
1101 - msg is a string with details in case of failure
1104 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1107 hyper.MigrateInstance(instance.name, target, live)
1108 except errors.HypervisorError, err:
1109 _Fail("Failed to migrate instance: %s", err, exc=True)
1112 def BlockdevCreate(disk, size, owner, on_primary, info):
1113 """Creates a block device for an instance.
1115 @type disk: L{objects.Disk}
1116 @param disk: the object describing the disk we should create
1118 @param size: the size of the physical underlying device, in MiB
1120 @param owner: the name of the instance for which disk is created,
1121 used for device cache data
1122 @type on_primary: boolean
1123 @param on_primary: indicates if it is the primary node or not
1125 @param info: string that will be sent to the physical device
1126 creation, used for example to set (LVM) tags on LVs
1128 @return: the new unique_id of the device (this can sometime be
1129 computed only after creation), or None. On secondary nodes,
1130 it's not required to return anything.
1135 for child in disk.children:
1137 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1138 except errors.BlockDeviceError, err:
1139 _Fail("Can't assemble device %s: %s", child, err)
1140 if on_primary or disk.AssembleOnSecondary():
1141 # we need the children open in case the device itself has to
1145 except errors.BlockDeviceError, err:
1146 _Fail("Can't make child '%s' read-write: %s", child, err)
1150 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1151 except errors.BlockDeviceError, err:
1152 _Fail("Can't create block device: %s", err)
1154 if on_primary or disk.AssembleOnSecondary():
1157 except errors.BlockDeviceError, err:
1158 _Fail("Can't assemble device after creation, unusual event: %s", err)
1159 device.SetSyncSpeed(constants.SYNC_SPEED)
1160 if on_primary or disk.OpenOnSecondary():
1162 device.Open(force=True)
1163 except errors.BlockDeviceError, err:
1164 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1165 DevCacheManager.UpdateCache(device.dev_path, owner,
1166 on_primary, disk.iv_name)
1168 device.SetInfo(info)
1170 return device.unique_id
1173 def BlockdevRemove(disk):
1174 """Remove a block device.
1176 @note: This is intended to be called recursively.
1178 @type disk: L{objects.Disk}
1179 @param disk: the disk object we should remove
1181 @return: the success of the operation
1186 rdev = _RecursiveFindBD(disk)
1187 except errors.BlockDeviceError, err:
1188 # probably can't attach
1189 logging.info("Can't attach to device %s in remove", disk)
1191 if rdev is not None:
1192 r_path = rdev.dev_path
1195 except errors.BlockDeviceError, err:
1196 msgs.append(str(err))
1198 DevCacheManager.RemoveCache(r_path)
1201 for child in disk.children:
1203 BlockdevRemove(child)
1204 except RPCFail, err:
1205 msgs.append(str(err))
1208 _Fail("; ".join(msgs))
1211 def _RecursiveAssembleBD(disk, owner, as_primary):
1212 """Activate a block device for an instance.
1214 This is run on the primary and secondary nodes for an instance.
1216 @note: this function is called recursively.
1218 @type disk: L{objects.Disk}
1219 @param disk: the disk we try to assemble
1221 @param owner: the name of the instance which owns the disk
1222 @type as_primary: boolean
1223 @param as_primary: if we should make the block device
1226 @return: the assembled device or None (in case no device
1228 @raise errors.BlockDeviceError: in case there is an error
1229 during the activation of the children or the device
1235 mcn = disk.ChildrenNeeded()
1237 mcn = 0 # max number of Nones allowed
1239 mcn = len(disk.children) - mcn # max number of Nones
1240 for chld_disk in disk.children:
1242 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1243 except errors.BlockDeviceError, err:
1244 if children.count(None) >= mcn:
1247 logging.error("Error in child activation (but continuing): %s",
1249 children.append(cdev)
1251 if as_primary or disk.AssembleOnSecondary():
1252 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1253 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1255 if as_primary or disk.OpenOnSecondary():
1257 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1258 as_primary, disk.iv_name)
1265 def BlockdevAssemble(disk, owner, as_primary):
1266 """Activate a block device for an instance.
1268 This is a wrapper over _RecursiveAssembleBD.
1270 @rtype: str or boolean
1271 @return: a C{/dev/...} path for primary nodes, and
1272 C{True} for secondary nodes
1276 result = _RecursiveAssembleBD(disk, owner, as_primary)
1277 if isinstance(result, bdev.BlockDev):
1278 result = result.dev_path
1279 except errors.BlockDeviceError, err:
1280 _Fail("Error while assembling disk: %s", err, exc=True)
1285 def BlockdevShutdown(disk):
1286 """Shut down a block device.
1288 First, if the device is assembled (Attach() is successful), then
1289 the device is shutdown. Then the children of the device are
1292 This function is called recursively. Note that we don't cache the
1293 children or such, as oppossed to assemble, shutdown of different
1294 devices doesn't require that the upper device was active.
1296 @type disk: L{objects.Disk}
1297 @param disk: the description of the disk we should
1303 r_dev = _RecursiveFindBD(disk)
1304 if r_dev is not None:
1305 r_path = r_dev.dev_path
1308 DevCacheManager.RemoveCache(r_path)
1309 except errors.BlockDeviceError, err:
1310 msgs.append(str(err))
1313 for child in disk.children:
1315 BlockdevShutdown(child)
1316 except RPCFail, err:
1317 msgs.append(str(err))
1320 _Fail("; ".join(msgs))
1323 def BlockdevAddchildren(parent_cdev, new_cdevs):
1324 """Extend a mirrored block device.
1326 @type parent_cdev: L{objects.Disk}
1327 @param parent_cdev: the disk to which we should add children
1328 @type new_cdevs: list of L{objects.Disk}
1329 @param new_cdevs: the list of children which we should add
1333 parent_bdev = _RecursiveFindBD(parent_cdev)
1334 if parent_bdev is None:
1335 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1336 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1337 if new_bdevs.count(None) > 0:
1338 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1339 parent_bdev.AddChildren(new_bdevs)
1342 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1343 """Shrink a mirrored block device.
1345 @type parent_cdev: L{objects.Disk}
1346 @param parent_cdev: the disk from which we should remove children
1347 @type new_cdevs: list of L{objects.Disk}
1348 @param new_cdevs: the list of children which we should remove
1352 parent_bdev = _RecursiveFindBD(parent_cdev)
1353 if parent_bdev is None:
1354 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1356 for disk in new_cdevs:
1357 rpath = disk.StaticDevPath()
1359 bd = _RecursiveFindBD(disk)
1361 _Fail("Can't find device %s while removing children", disk)
1363 devs.append(bd.dev_path)
1366 parent_bdev.RemoveChildren(devs)
1369 def BlockdevGetmirrorstatus(disks):
1370 """Get the mirroring status of a list of devices.
1372 @type disks: list of L{objects.Disk}
1373 @param disks: the list of disks which we should query
1376 a list of (mirror_done, estimated_time) tuples, which
1377 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1378 @raise errors.BlockDeviceError: if any of the disks cannot be
1384 rbd = _RecursiveFindBD(dsk)
1386 _Fail("Can't find device %s", dsk)
1388 stats.append(rbd.CombinedSyncStatus())
1393 def _RecursiveFindBD(disk):
1394 """Check if a device is activated.
1396 If so, return information about the real device.
1398 @type disk: L{objects.Disk}
1399 @param disk: the disk object we need to find
1401 @return: None if the device can't be found,
1402 otherwise the device instance
1407 for chdisk in disk.children:
1408 children.append(_RecursiveFindBD(chdisk))
1410 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1413 def BlockdevFind(disk):
1414 """Check if a device is activated.
1416 If it is, return information about the real device.
1418 @type disk: L{objects.Disk}
1419 @param disk: the disk to find
1420 @rtype: None or objects.BlockDevStatus
1421 @return: None if the disk cannot be found, otherwise a the current
1426 rbd = _RecursiveFindBD(disk)
1427 except errors.BlockDeviceError, err:
1428 _Fail("Failed to find device: %s", err, exc=True)
1433 return rbd.GetSyncStatus()
1436 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1437 """Write a file to the filesystem.
1439 This allows the master to overwrite(!) a file. It will only perform
1440 the operation if the file belongs to a list of configuration files.
1442 @type file_name: str
1443 @param file_name: the target file name
1445 @param data: the new contents of the file
1447 @param mode: the mode to give the file (can be None)
1449 @param uid: the owner of the file (can be -1 for default)
1451 @param gid: the group of the file (can be -1 for default)
1453 @param atime: the atime to set on the file (can be None)
1455 @param mtime: the mtime to set on the file (can be None)
1459 if not os.path.isabs(file_name):
1460 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1462 if file_name not in _ALLOWED_UPLOAD_FILES:
1463 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1466 raw_data = _Decompress(data)
1468 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1469 atime=atime, mtime=mtime)
1472 def WriteSsconfFiles(values):
1473 """Update all ssconf files.
1475 Wrapper around the SimpleStore.WriteFiles.
1478 ssconf.SimpleStore().WriteFiles(values)
1481 def _ErrnoOrStr(err):
1482 """Format an EnvironmentError exception.
1484 If the L{err} argument has an errno attribute, it will be looked up
1485 and converted into a textual C{E...} description. Otherwise the
1486 string representation of the error will be returned.
1488 @type err: L{EnvironmentError}
1489 @param err: the exception to format
1492 if hasattr(err, 'errno'):
1493 detail = errno.errorcode[err.errno]
1499 def _OSOndiskAPIVersion(name, os_dir):
1500 """Compute and return the API version of a given OS.
1502 This function will try to read the API version of the OS given by
1503 the 'name' parameter and residing in the 'os_dir' directory.
1506 @param name: the OS name we should look for
1508 @param os_dir: the directory inwhich we should look for the OS
1510 @return: tuple (status, data) with status denoting the validity and
1511 data holding either the vaid versions or an error message
1514 api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
1517 st = os.stat(api_file)
1518 except EnvironmentError, err:
1519 return False, ("Required file 'ganeti_api_version' file not"
1520 " found under path %s: %s" % (os_dir, _ErrnoOrStr(err)))
1522 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1523 return False, ("File 'ganeti_api_version' file at %s is not"
1524 " a regular file" % os_dir)
1527 api_versions = utils.ReadFile(api_file).splitlines()
1528 except EnvironmentError, err:
1529 return False, ("Error while reading the API version file at %s: %s" %
1530 (api_file, _ErrnoOrStr(err)))
1533 api_versions = [int(version.strip()) for version in api_versions]
1534 except (TypeError, ValueError), err:
1535 return False, ("API version(s) can't be converted to integer: %s" %
1538 return True, api_versions
1541 def DiagnoseOS(top_dirs=None):
1542 """Compute the validity for all OSes.
1544 @type top_dirs: list
1545 @param top_dirs: the list of directories in which to
1546 search (if not given defaults to
1547 L{constants.OS_SEARCH_PATH})
1548 @rtype: list of L{objects.OS}
1549 @return: a list of tuples (name, path, status, diagnose)
1550 for all (potential) OSes under all search paths, where:
1551 - name is the (potential) OS name
1552 - path is the full path to the OS
1553 - status True/False is the validity of the OS
1554 - diagnose is the error message for an invalid OS, otherwise empty
1557 if top_dirs is None:
1558 top_dirs = constants.OS_SEARCH_PATH
1561 for dir_name in top_dirs:
1562 if os.path.isdir(dir_name):
1564 f_names = utils.ListVisibleFiles(dir_name)
1565 except EnvironmentError, err:
1566 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1568 for name in f_names:
1569 os_path = os.path.sep.join([dir_name, name])
1570 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1575 result.append((name, os_path, status, diagnose))
1580 def _TryOSFromDisk(name, base_dir=None):
1581 """Create an OS instance from disk.
1583 This function will return an OS instance if the given name is a
1586 @type base_dir: string
1587 @keyword base_dir: Base directory containing OS installations.
1588 Defaults to a search in all the OS_SEARCH_PATH dirs.
1590 @return: success and either the OS instance if we find a valid one,
1594 if base_dir is None:
1595 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1597 return False, "Directory for OS %s not found in search path" % name
1599 os_dir = os.path.sep.join([base_dir, name])
1601 status, api_versions = _OSOndiskAPIVersion(name, os_dir)
1604 return status, api_versions
1606 if not constants.OS_API_VERSIONS.intersection(api_versions):
1607 return False, ("API version mismatch for path '%s': found %s, want %s." %
1608 (os_dir, api_versions, constants.OS_API_VERSIONS))
1610 # OS Scripts dictionary, we will populate it with the actual script names
1611 os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
1613 for script in os_scripts:
1614 os_scripts[script] = os.path.sep.join([os_dir, script])
1617 st = os.stat(os_scripts[script])
1618 except EnvironmentError, err:
1619 return False, ("Script '%s' under path '%s' is missing (%s)" %
1620 (script, os_dir, _ErrnoOrStr(err)))
1622 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1623 return False, ("Script '%s' under path '%s' is not executable" %
1626 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1627 return False, ("Script '%s' under path '%s' is not a regular file" %
1630 os_obj = objects.OS(name=name, path=os_dir,
1631 create_script=os_scripts[constants.OS_SCRIPT_CREATE],
1632 export_script=os_scripts[constants.OS_SCRIPT_EXPORT],
1633 import_script=os_scripts[constants.OS_SCRIPT_IMPORT],
1634 rename_script=os_scripts[constants.OS_SCRIPT_RENAME],
1635 api_versions=api_versions)
1639 def OSFromDisk(name, base_dir=None):
1640 """Create an OS instance from disk.
1642 This function will return an OS instance if the given name is a
1643 valid OS name. Otherwise, it will raise an appropriate
1644 L{RPCFail} exception, detailing why this is not a valid OS.
1646 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1647 an exception but returns true/false status data.
1649 @type base_dir: string
1650 @keyword base_dir: Base directory containing OS installations.
1651 Defaults to a search in all the OS_SEARCH_PATH dirs.
1652 @rtype: L{objects.OS}
1653 @return: the OS instance if we find a valid one
1654 @raise RPCFail: if we don't find a valid OS
1657 status, payload = _TryOSFromDisk(name, base_dir)
1665 def OSEnvironment(instance, os, debug=0):
1666 """Calculate the environment for an os script.
1668 @type instance: L{objects.Instance}
1669 @param instance: target instance for the os script run
1670 @type os: L{objects.OS}
1671 @param os: operating system for which the environment is being built
1672 @type debug: integer
1673 @param debug: debug level (0 or 1, for OS Api 10)
1675 @return: dict of environment variables
1676 @raise errors.BlockDeviceError: if the block device
1681 api_version = max(constants.OS_API_VERSIONS.intersection(os.api_versions))
1682 result['OS_API_VERSION'] = '%d' % api_version
1683 result['INSTANCE_NAME'] = instance.name
1684 result['INSTANCE_OS'] = instance.os
1685 result['HYPERVISOR'] = instance.hypervisor
1686 result['DISK_COUNT'] = '%d' % len(instance.disks)
1687 result['NIC_COUNT'] = '%d' % len(instance.nics)
1688 result['DEBUG_LEVEL'] = '%d' % debug
1689 for idx, disk in enumerate(instance.disks):
1690 real_disk = _RecursiveFindBD(disk)
1691 if real_disk is None:
1692 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1695 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1696 result['DISK_%d_ACCESS' % idx] = disk.mode
1697 if constants.HV_DISK_TYPE in instance.hvparams:
1698 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1699 instance.hvparams[constants.HV_DISK_TYPE]
1700 if disk.dev_type in constants.LDS_BLOCK:
1701 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1702 elif disk.dev_type == constants.LD_FILE:
1703 result['DISK_%d_BACKEND_TYPE' % idx] = \
1704 'file:%s' % disk.physical_id[0]
1705 for idx, nic in enumerate(instance.nics):
1706 result['NIC_%d_MAC' % idx] = nic.mac
1708 result['NIC_%d_IP' % idx] = nic.ip
1709 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1710 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1711 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1712 if nic.nicparams[constants.NIC_LINK]:
1713 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1714 if constants.HV_NIC_TYPE in instance.hvparams:
1715 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1716 instance.hvparams[constants.HV_NIC_TYPE]
1718 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1719 for key, value in source.items():
1720 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1724 def BlockdevGrow(disk, amount):
1725 """Grow a stack of block devices.
1727 This function is called recursively, with the childrens being the
1728 first ones to resize.
1730 @type disk: L{objects.Disk}
1731 @param disk: the disk to be grown
1732 @rtype: (status, result)
1733 @return: a tuple with the status of the operation
1734 (True/False), and the errors message if status
1738 r_dev = _RecursiveFindBD(disk)
1740 _Fail("Cannot find block device %s", disk)
1744 except errors.BlockDeviceError, err:
1745 _Fail("Failed to grow block device: %s", err, exc=True)
1748 def BlockdevSnapshot(disk):
1749 """Create a snapshot copy of a block device.
1751 This function is called recursively, and the snapshot is actually created
1752 just for the leaf lvm backend device.
1754 @type disk: L{objects.Disk}
1755 @param disk: the disk to be snapshotted
1757 @return: snapshot disk path
1761 if len(disk.children) == 1:
1762 # only one child, let's recurse on it
1763 return BlockdevSnapshot(disk.children[0])
1765 # more than one child, choose one that matches
1766 for child in disk.children:
1767 if child.size == disk.size:
1768 # return implies breaking the loop
1769 return BlockdevSnapshot(child)
1770 elif disk.dev_type == constants.LD_LV:
1771 r_dev = _RecursiveFindBD(disk)
1772 if r_dev is not None:
1773 # let's stay on the safe side and ask for the full size, for now
1774 return r_dev.Snapshot(disk.size)
1776 _Fail("Cannot find block device %s", disk)
1778 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1779 disk.unique_id, disk.dev_type)
1782 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1783 """Export a block device snapshot to a remote node.
1785 @type disk: L{objects.Disk}
1786 @param disk: the description of the disk to export
1787 @type dest_node: str
1788 @param dest_node: the destination node to export to
1789 @type instance: L{objects.Instance}
1790 @param instance: the instance object to whom the disk belongs
1791 @type cluster_name: str
1792 @param cluster_name: the cluster name, needed for SSH hostalias
1794 @param idx: the index of the disk in the instance's disk list,
1795 used to export to the OS scripts environment
1799 inst_os = OSFromDisk(instance.os)
1800 export_env = OSEnvironment(instance, inst_os)
1802 export_script = inst_os.export_script
1804 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1805 instance.name, int(time.time()))
1806 if not os.path.exists(constants.LOG_OS_DIR):
1807 os.mkdir(constants.LOG_OS_DIR, 0750)
1808 real_disk = _RecursiveFindBD(disk)
1809 if real_disk is None:
1810 _Fail("Block device '%s' is not set up", disk)
1814 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1815 export_env['EXPORT_INDEX'] = str(idx)
1817 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1818 destfile = disk.physical_id[1]
1820 # the target command is built out of three individual commands,
1821 # which are joined by pipes; we check each individual command for
1823 expcmd = utils.BuildShellCmd("cd %s; %s 2>%s", inst_os.path,
1824 export_script, logfile)
1828 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1829 destdir, destdir, destfile)
1830 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1831 constants.GANETI_RUNAS,
1834 # all commands have been checked, so we're safe to combine them
1835 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1837 result = utils.RunCmd(command, env=export_env)
1840 _Fail("OS snapshot export command '%s' returned error: %s"
1841 " output: %s", command, result.fail_reason, result.output)
1844 def FinalizeExport(instance, snap_disks):
1845 """Write out the export configuration information.
1847 @type instance: L{objects.Instance}
1848 @param instance: the instance which we export, used for
1849 saving configuration
1850 @type snap_disks: list of L{objects.Disk}
1851 @param snap_disks: list of snapshot block devices, which
1852 will be used to get the actual name of the dump file
1857 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1858 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1860 config = objects.SerializableConfigParser()
1862 config.add_section(constants.INISECT_EXP)
1863 config.set(constants.INISECT_EXP, 'version', '0')
1864 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1865 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1866 config.set(constants.INISECT_EXP, 'os', instance.os)
1867 config.set(constants.INISECT_EXP, 'compression', 'gzip')
1869 config.add_section(constants.INISECT_INS)
1870 config.set(constants.INISECT_INS, 'name', instance.name)
1871 config.set(constants.INISECT_INS, 'memory', '%d' %
1872 instance.beparams[constants.BE_MEMORY])
1873 config.set(constants.INISECT_INS, 'vcpus', '%d' %
1874 instance.beparams[constants.BE_VCPUS])
1875 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1878 for nic_count, nic in enumerate(instance.nics):
1880 config.set(constants.INISECT_INS, 'nic%d_mac' %
1881 nic_count, '%s' % nic.mac)
1882 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
1883 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
1885 # TODO: redundant: on load can read nics until it doesn't exist
1886 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
1889 for disk_count, disk in enumerate(snap_disks):
1892 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
1893 ('%s' % disk.iv_name))
1894 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
1895 ('%s' % disk.physical_id[1]))
1896 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
1899 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
1901 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
1902 data=config.Dumps())
1903 shutil.rmtree(finaldestdir, True)
1904 shutil.move(destdir, finaldestdir)
1907 def ExportInfo(dest):
1908 """Get export configuration information.
1911 @param dest: directory containing the export
1913 @rtype: L{objects.SerializableConfigParser}
1914 @return: a serializable config file containing the
1918 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
1920 config = objects.SerializableConfigParser()
1923 if (not config.has_section(constants.INISECT_EXP) or
1924 not config.has_section(constants.INISECT_INS)):
1925 _Fail("Export info file doesn't have the required fields")
1927 return config.Dumps()
1930 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
1931 """Import an os image into an instance.
1933 @type instance: L{objects.Instance}
1934 @param instance: instance to import the disks into
1935 @type src_node: string
1936 @param src_node: source node for the disk images
1937 @type src_images: list of string
1938 @param src_images: absolute paths of the disk images
1939 @rtype: list of boolean
1940 @return: each boolean represent the success of importing the n-th disk
1943 inst_os = OSFromDisk(instance.os)
1944 import_env = OSEnvironment(instance, inst_os)
1945 import_script = inst_os.import_script
1947 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
1948 instance.name, int(time.time()))
1949 if not os.path.exists(constants.LOG_OS_DIR):
1950 os.mkdir(constants.LOG_OS_DIR, 0750)
1953 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
1954 import_script, logfile)
1957 for idx, image in enumerate(src_images):
1959 destcmd = utils.BuildShellCmd('cat %s', image)
1960 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
1961 constants.GANETI_RUNAS,
1963 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
1964 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
1965 import_env['IMPORT_INDEX'] = str(idx)
1966 result = utils.RunCmd(command, env=import_env)
1968 logging.error("Disk import command '%s' returned error: %s"
1969 " output: %s", command, result.fail_reason,
1971 final_result.append("error importing disk %d: %s, %s" %
1972 (idx, result.fail_reason, result.output[-100]))
1975 _Fail("; ".join(final_result), log=False)
1979 """Return a list of exports currently available on this machine.
1982 @return: list of the exports
1985 if os.path.isdir(constants.EXPORT_DIR):
1986 return utils.ListVisibleFiles(constants.EXPORT_DIR)
1988 _Fail("No exports directory")
1991 def RemoveExport(export):
1992 """Remove an existing export from the node.
1995 @param export: the name of the export to remove
1999 target = os.path.join(constants.EXPORT_DIR, export)
2002 shutil.rmtree(target)
2003 except EnvironmentError, err:
2004 _Fail("Error while removing the export: %s", err, exc=True)
2007 def BlockdevRename(devlist):
2008 """Rename a list of block devices.
2010 @type devlist: list of tuples
2011 @param devlist: list of tuples of the form (disk,
2012 new_logical_id, new_physical_id); disk is an
2013 L{objects.Disk} object describing the current disk,
2014 and new logical_id/physical_id is the name we
2017 @return: True if all renames succeeded, False otherwise
2022 for disk, unique_id in devlist:
2023 dev = _RecursiveFindBD(disk)
2025 msgs.append("Can't find device %s in rename" % str(disk))
2029 old_rpath = dev.dev_path
2030 dev.Rename(unique_id)
2031 new_rpath = dev.dev_path
2032 if old_rpath != new_rpath:
2033 DevCacheManager.RemoveCache(old_rpath)
2034 # FIXME: we should add the new cache information here, like:
2035 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2036 # but we don't have the owner here - maybe parse from existing
2037 # cache? for now, we only lose lvm data when we rename, which
2038 # is less critical than DRBD or MD
2039 except errors.BlockDeviceError, err:
2040 msgs.append("Can't rename device '%s' to '%s': %s" %
2041 (dev, unique_id, err))
2042 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2045 _Fail("; ".join(msgs))
2048 def _TransformFileStorageDir(file_storage_dir):
2049 """Checks whether given file_storage_dir is valid.
2051 Checks wheter the given file_storage_dir is within the cluster-wide
2052 default file_storage_dir stored in SimpleStore. Only paths under that
2053 directory are allowed.
2055 @type file_storage_dir: str
2056 @param file_storage_dir: the path to check
2058 @return: the normalized path if valid, None otherwise
2062 file_storage_dir = os.path.normpath(file_storage_dir)
2063 base_file_storage_dir = cfg.GetFileStorageDir()
2064 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2065 base_file_storage_dir):
2066 _Fail("File storage directory '%s' is not under base file"
2067 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2068 return file_storage_dir
2071 def CreateFileStorageDir(file_storage_dir):
2072 """Create file storage directory.
2074 @type file_storage_dir: str
2075 @param file_storage_dir: directory to create
2078 @return: tuple with first element a boolean indicating wheter dir
2079 creation was successful or not
2082 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2083 if os.path.exists(file_storage_dir):
2084 if not os.path.isdir(file_storage_dir):
2085 _Fail("Specified storage dir '%s' is not a directory",
2089 os.makedirs(file_storage_dir, 0750)
2090 except OSError, err:
2091 _Fail("Cannot create file storage directory '%s': %s",
2092 file_storage_dir, err, exc=True)
2095 def RemoveFileStorageDir(file_storage_dir):
2096 """Remove file storage directory.
2098 Remove it only if it's empty. If not log an error and return.
2100 @type file_storage_dir: str
2101 @param file_storage_dir: the directory we should cleanup
2102 @rtype: tuple (success,)
2103 @return: tuple of one element, C{success}, denoting
2104 whether the operation was successful
2107 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2108 if os.path.exists(file_storage_dir):
2109 if not os.path.isdir(file_storage_dir):
2110 _Fail("Specified Storage directory '%s' is not a directory",
2112 # deletes dir only if empty, otherwise we want to fail the rpc call
2114 os.rmdir(file_storage_dir)
2115 except OSError, err:
2116 _Fail("Cannot remove file storage directory '%s': %s",
2117 file_storage_dir, err)
2120 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2121 """Rename the file storage directory.
2123 @type old_file_storage_dir: str
2124 @param old_file_storage_dir: the current path
2125 @type new_file_storage_dir: str
2126 @param new_file_storage_dir: the name we should rename to
2127 @rtype: tuple (success,)
2128 @return: tuple of one element, C{success}, denoting
2129 whether the operation was successful
2132 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2133 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2134 if not os.path.exists(new_file_storage_dir):
2135 if os.path.isdir(old_file_storage_dir):
2137 os.rename(old_file_storage_dir, new_file_storage_dir)
2138 except OSError, err:
2139 _Fail("Cannot rename '%s' to '%s': %s",
2140 old_file_storage_dir, new_file_storage_dir, err)
2142 _Fail("Specified storage dir '%s' is not a directory",
2143 old_file_storage_dir)
2145 if os.path.exists(old_file_storage_dir):
2146 _Fail("Cannot rename '%s' to '%s': both locations exist",
2147 old_file_storage_dir, new_file_storage_dir)
2150 def _EnsureJobQueueFile(file_name):
2151 """Checks whether the given filename is in the queue directory.
2153 @type file_name: str
2154 @param file_name: the file name we should check
2156 @raises RPCFail: if the file is not valid
2159 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2160 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2163 _Fail("Passed job queue file '%s' does not belong to"
2164 " the queue directory '%s'", file_name, queue_dir)
2167 def JobQueueUpdate(file_name, content):
2168 """Updates a file in the queue directory.
2170 This is just a wrapper over L{utils.WriteFile}, with proper
2173 @type file_name: str
2174 @param file_name: the job file name
2176 @param content: the new job contents
2178 @return: the success of the operation
2181 _EnsureJobQueueFile(file_name)
2183 # Write and replace the file atomically
2184 utils.WriteFile(file_name, data=_Decompress(content))
2187 def JobQueueRename(old, new):
2188 """Renames a job queue file.
2190 This is just a wrapper over os.rename with proper checking.
2193 @param old: the old (actual) file name
2195 @param new: the desired file name
2197 @return: the success of the operation and payload
2200 _EnsureJobQueueFile(old)
2201 _EnsureJobQueueFile(new)
2203 utils.RenameFile(old, new, mkdir=True)
2206 def JobQueueSetDrainFlag(drain_flag):
2207 """Set the drain flag for the queue.
2209 This will set or unset the queue drain flag.
2211 @type drain_flag: boolean
2212 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2214 @return: always True, None
2215 @warning: the function always returns True
2219 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2221 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2224 def BlockdevClose(instance_name, disks):
2225 """Closes the given block devices.
2227 This means they will be switched to secondary mode (in case of
2230 @param instance_name: if the argument is not empty, the symlinks
2231 of this instance will be removed
2232 @type disks: list of L{objects.Disk}
2233 @param disks: the list of disks to be closed
2234 @rtype: tuple (success, message)
2235 @return: a tuple of success and message, where success
2236 indicates the succes of the operation, and message
2237 which will contain the error details in case we
2243 rd = _RecursiveFindBD(cf)
2245 _Fail("Can't find device %s", cf)
2252 except errors.BlockDeviceError, err:
2253 msg.append(str(err))
2255 _Fail("Can't make devices secondary: %s", ",".join(msg))
2258 _RemoveBlockDevLinks(instance_name, disks)
2261 def ValidateHVParams(hvname, hvparams):
2262 """Validates the given hypervisor parameters.
2264 @type hvname: string
2265 @param hvname: the hypervisor name
2266 @type hvparams: dict
2267 @param hvparams: the hypervisor parameters to be validated
2272 hv_type = hypervisor.GetHypervisor(hvname)
2273 hv_type.ValidateParameters(hvparams)
2274 except errors.HypervisorError, err:
2275 _Fail(str(err), log=False)
2279 """Demotes the current node from master candidate role.
2282 # try to ensure we're not the master by mistake
2283 master, myself = ssconf.GetMasterAndMyself()
2284 if master == myself:
2285 _Fail("ssconf status shows I'm the master node, will not demote")
2286 pid_file = utils.DaemonPidFileName(constants.MASTERD)
2287 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2288 _Fail("The master daemon is running, will not demote")
2290 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2291 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2292 except EnvironmentError, err:
2293 if err.errno != errno.ENOENT:
2294 _Fail("Error while backing up cluster file: %s", err, exc=True)
2295 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2298 def _FindDisks(nodes_ip, disks):
2299 """Sets the physical ID on disks and returns the block devices.
2302 # set the correct physical ID
2303 my_name = utils.HostInfo().name
2305 cf.SetPhysicalID(my_name, nodes_ip)
2310 rd = _RecursiveFindBD(cf)
2312 _Fail("Can't find device %s", cf)
2317 def DrbdDisconnectNet(nodes_ip, disks):
2318 """Disconnects the network on a list of drbd devices.
2321 bdevs = _FindDisks(nodes_ip, disks)
2327 except errors.BlockDeviceError, err:
2328 _Fail("Can't change network configuration to standalone mode: %s",
2332 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2333 """Attaches the network on a list of drbd devices.
2336 bdevs = _FindDisks(nodes_ip, disks)
2339 for idx, rd in enumerate(bdevs):
2341 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2342 except EnvironmentError, err:
2343 _Fail("Can't create symlink: %s", err)
2344 # reconnect disks, switch to new master configuration and if
2345 # needed primary mode
2348 rd.AttachNet(multimaster)
2349 except errors.BlockDeviceError, err:
2350 _Fail("Can't change network configuration: %s", err)
2351 # wait until the disks are connected; we need to retry the re-attach
2352 # if the device becomes standalone, as this might happen if the one
2353 # node disconnects and reconnects in a different mode before the
2354 # other node reconnects; in this case, one or both of the nodes will
2355 # decide it has wrong configuration and switch to standalone
2356 RECONNECT_TIMEOUT = 2 * 60
2357 sleep_time = 0.100 # start with 100 miliseconds
2358 timeout_limit = time.time() + RECONNECT_TIMEOUT
2359 while time.time() < timeout_limit:
2360 all_connected = True
2362 stats = rd.GetProcStatus()
2363 if not (stats.is_connected or stats.is_in_resync):
2364 all_connected = False
2365 if stats.is_standalone:
2366 # peer had different config info and this node became
2367 # standalone, even though this should not happen with the
2368 # new staged way of changing disk configs
2370 rd.AttachNet(multimaster)
2371 except errors.BlockDeviceError, err:
2372 _Fail("Can't change network configuration: %s", err)
2375 time.sleep(sleep_time)
2376 sleep_time = min(5, sleep_time * 1.5)
2377 if not all_connected:
2378 _Fail("Timeout in disk reconnecting")
2380 # change to primary mode
2384 except errors.BlockDeviceError, err:
2385 _Fail("Can't change to primary mode: %s", err)
2388 def DrbdWaitSync(nodes_ip, disks):
2389 """Wait until DRBDs have synchronized.
2392 bdevs = _FindDisks(nodes_ip, disks)
2397 stats = rd.GetProcStatus()
2398 if not (stats.is_connected or stats.is_in_resync):
2399 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2400 alldone = alldone and (not stats.is_in_resync)
2401 if stats.sync_percent is not None:
2402 min_resync = min(min_resync, stats.sync_percent)
2404 return (alldone, min_resync)
2407 def PowercycleNode(hypervisor_type):
2408 """Hard-powercycle the node.
2410 Because we need to return first, and schedule the powercycle in the
2411 background, we won't be able to report failures nicely.
2414 hyper = hypervisor.GetHypervisor(hypervisor_type)
2418 # if we can't fork, we'll pretend that we're in the child process
2421 return "Reboot scheduled in 5 seconds"
2423 hyper.PowercycleNode()
2426 class HooksRunner(object):
2429 This class is instantiated on the node side (ganeti-noded) and not
2433 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2435 def __init__(self, hooks_base_dir=None):
2436 """Constructor for hooks runner.
2438 @type hooks_base_dir: str or None
2439 @param hooks_base_dir: if not None, this overrides the
2440 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2443 if hooks_base_dir is None:
2444 hooks_base_dir = constants.HOOKS_BASE_DIR
2445 self._BASE_DIR = hooks_base_dir
2448 def ExecHook(script, env):
2449 """Exec one hook script.
2452 @param script: the full path to the script
2454 @param env: the environment with which to exec the script
2455 @rtype: tuple (success, message)
2456 @return: a tuple of success and message, where success
2457 indicates the succes of the operation, and message
2458 which will contain the error details in case we
2462 # exec the process using subprocess and log the output
2465 fdstdin = open("/dev/null", "r")
2466 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2467 stderr=subprocess.STDOUT, close_fds=True,
2468 shell=False, cwd="/", env=env)
2471 output = child.stdout.read(4096)
2472 child.stdout.close()
2473 except EnvironmentError, err:
2474 output += "Hook script error: %s" % str(err)
2478 result = child.wait()
2480 except EnvironmentError, err:
2481 if err.errno == errno.EINTR:
2485 # try not to leak fds
2486 for fd in (fdstdin, ):
2490 except EnvironmentError, err:
2491 # just log the error
2492 #logging.exception("Error while closing fd %s", fd)
2495 return result == 0, utils.SafeEncode(output.strip())
2497 def RunHooks(self, hpath, phase, env):
2498 """Run the scripts in the hooks directory.
2501 @param hpath: the path to the hooks directory which
2504 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2505 L{constants.HOOKS_PHASE_POST}
2507 @param env: dictionary with the environment for the hook
2509 @return: list of 3-element tuples:
2511 - script result, either L{constants.HKR_SUCCESS} or
2512 L{constants.HKR_FAIL}
2513 - output of the script
2515 @raise errors.ProgrammerError: for invalid input
2519 if phase == constants.HOOKS_PHASE_PRE:
2521 elif phase == constants.HOOKS_PHASE_POST:
2524 _Fail("Unknown hooks phase '%s'", phase)
2528 subdir = "%s-%s.d" % (hpath, suffix)
2529 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2531 dir_contents = utils.ListVisibleFiles(dir_name)
2533 # FIXME: must log output in case of failures
2536 # we use the standard python sort order,
2537 # so 00name is the recommended naming scheme
2539 for relname in dir_contents:
2540 fname = os.path.join(dir_name, relname)
2541 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2542 self.RE_MASK.match(relname) is not None):
2543 rrval = constants.HKR_SKIP
2546 result, output = self.ExecHook(fname, env)
2548 rrval = constants.HKR_FAIL
2550 rrval = constants.HKR_SUCCESS
2551 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2556 class IAllocatorRunner(object):
2557 """IAllocator runner.
2559 This class is instantiated on the node side (ganeti-noded) and not on
2563 def Run(self, name, idata):
2564 """Run an iallocator script.
2567 @param name: the iallocator script name
2569 @param idata: the allocator input data
2572 @return: two element tuple of:
2574 - either error message or stdout of allocator (for success)
2577 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2579 if alloc_script is None:
2580 _Fail("iallocator module '%s' not found in the search path", name)
2582 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2586 result = utils.RunCmd([alloc_script, fin_name])
2588 _Fail("iallocator module '%s' failed: %s, output '%s'",
2589 name, result.fail_reason, result.output)
2593 return result.stdout
2596 class DevCacheManager(object):
2597 """Simple class for managing a cache of block device information.
2600 _DEV_PREFIX = "/dev/"
2601 _ROOT_DIR = constants.BDEV_CACHE_DIR
2604 def _ConvertPath(cls, dev_path):
2605 """Converts a /dev/name path to the cache file name.
2607 This replaces slashes with underscores and strips the /dev
2608 prefix. It then returns the full path to the cache file.
2611 @param dev_path: the C{/dev/} path name
2613 @return: the converted path name
2616 if dev_path.startswith(cls._DEV_PREFIX):
2617 dev_path = dev_path[len(cls._DEV_PREFIX):]
2618 dev_path = dev_path.replace("/", "_")
2619 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2623 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2624 """Updates the cache information for a given device.
2627 @param dev_path: the pathname of the device
2629 @param owner: the owner (instance name) of the device
2630 @type on_primary: bool
2631 @param on_primary: whether this is the primary
2634 @param iv_name: the instance-visible name of the
2635 device, as in objects.Disk.iv_name
2640 if dev_path is None:
2641 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2643 fpath = cls._ConvertPath(dev_path)
2649 iv_name = "not_visible"
2650 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2652 utils.WriteFile(fpath, data=fdata)
2653 except EnvironmentError, err:
2654 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2657 def RemoveCache(cls, dev_path):
2658 """Remove data for a dev_path.
2660 This is just a wrapper over L{utils.RemoveFile} with a converted
2661 path name and logging.
2664 @param dev_path: the pathname of the device
2669 if dev_path is None:
2670 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2672 fpath = cls._ConvertPath(dev_path)
2674 utils.RemoveFile(fpath)
2675 except EnvironmentError, err:
2676 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)