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 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
57 class RPCFail(Exception):
58 """Class denoting RPC failure.
60 Its argument is the error message.
65 def _Fail(msg, *args, **kwargs):
66 """Log an error and the raise an RPCFail exception.
68 This exception is then handled specially in the ganeti daemon and
69 turned into a 'failed' return type. As such, this function is a
70 useful shortcut for logging the error and returning it to the master
74 @param msg: the text of the exception
80 if "log" not in kwargs or kwargs["log"]: # if we should log this error
81 if "exc" in kwargs and kwargs["exc"]:
82 logging.exception(msg)
89 """Simple wrapper to return a SimpleStore.
91 @rtype: L{ssconf.SimpleStore}
92 @return: a SimpleStore instance
95 return ssconf.SimpleStore()
98 def _GetSshRunner(cluster_name):
99 """Simple wrapper to return an SshRunner.
101 @type cluster_name: str
102 @param cluster_name: the cluster name, which is needed
103 by the SshRunner constructor
104 @rtype: L{ssh.SshRunner}
105 @return: an SshRunner instance
108 return ssh.SshRunner(cluster_name)
111 def _Decompress(data):
112 """Unpacks data compressed by the RPC client.
114 @type data: list or tuple
115 @param data: Data sent by RPC client
117 @return: Decompressed data
120 assert isinstance(data, (list, tuple))
121 assert len(data) == 2
122 (encoding, content) = data
123 if encoding == constants.RPC_ENCODING_NONE:
125 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
126 return zlib.decompress(base64.b64decode(content))
128 raise AssertionError("Unknown data encoding")
131 def _CleanDirectory(path, exclude=None):
132 """Removes all regular files in a directory.
135 @param path: the directory to clean
137 @param exclude: list of files to be excluded, defaults
141 if not os.path.isdir(path):
146 # Normalize excluded paths
147 exclude = [os.path.normpath(i) for i in exclude]
149 for rel_name in utils.ListVisibleFiles(path):
150 full_name = os.path.normpath(os.path.join(path, rel_name))
151 if full_name in exclude:
153 if os.path.isfile(full_name) and not os.path.islink(full_name):
154 utils.RemoveFile(full_name)
157 def _BuildUploadFileList():
158 """Build the list of allowed upload files.
160 This is abstracted so that it's built only once at module import time.
163 allowed_files = set([
164 constants.CLUSTER_CONF_FILE,
166 constants.SSH_KNOWN_HOSTS_FILE,
167 constants.VNC_PASSWORD_FILE,
168 constants.RAPI_CERT_FILE,
169 constants.RAPI_USERS_FILE,
170 constants.HMAC_CLUSTER_KEY,
173 for hv_name in constants.HYPER_TYPES:
174 hv_class = hypervisor.GetHypervisorClass(hv_name)
175 allowed_files.update(hv_class.GetAncillaryFiles())
177 return frozenset(allowed_files)
180 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
184 """Removes job queue files and archived jobs.
190 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
191 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
195 """Returns master information.
197 This is an utility function to compute master information, either
198 for consumption here or from the node daemon.
201 @return: master_netdev, master_ip, master_name
202 @raise RPCFail: in case of errors
207 master_netdev = cfg.GetMasterNetdev()
208 master_ip = cfg.GetMasterIP()
209 master_node = cfg.GetMasterNode()
210 except errors.ConfigurationError, err:
211 _Fail("Cluster configuration incomplete: %s", err, exc=True)
212 return (master_netdev, master_ip, master_node)
215 def StartMaster(start_daemons, no_voting):
216 """Activate local node as master node.
218 The function will always try activate the IP address of the master
219 (unless someone else has it). It will also start the master daemons,
220 based on the start_daemons parameter.
222 @type start_daemons: boolean
223 @param start_daemons: whether to also start the master
224 daemons (ganeti-masterd and ganeti-rapi)
225 @type no_voting: boolean
226 @param no_voting: whether to start ganeti-masterd without a node vote
227 (if start_daemons is True), but still non-interactively
231 # GetMasterInfo will raise an exception if not able to return data
232 master_netdev, master_ip, _ = GetMasterInfo()
235 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
236 if utils.OwnIpAddress(master_ip):
237 # we already have the ip:
238 logging.debug("Master IP already configured, doing nothing")
240 msg = "Someone else has the master ip, not activating"
244 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
245 "dev", master_netdev, "label",
246 "%s:0" % master_netdev])
248 msg = "Can't activate master IP: %s" % result.output
252 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
253 "-s", master_ip, master_ip])
254 # we'll ignore the exit code of arping
256 # and now start the master and rapi daemons
259 'ganeti-masterd': [],
263 daemons_params['ganeti-masterd'].append('--no-voting')
264 daemons_params['ganeti-masterd'].append('--yes-do-it')
265 for daemon in daemons_params:
267 cmd.extend(daemons_params[daemon])
268 result = utils.RunCmd(cmd)
270 msg = "Can't start daemon %s: %s" % (daemon, result.output)
275 _Fail("; ".join(err_msgs))
278 def StopMaster(stop_daemons):
279 """Deactivate this node as master.
281 The function will always try to deactivate the IP address of the
282 master. It will also stop the master daemons depending on the
283 stop_daemons parameter.
285 @type stop_daemons: boolean
286 @param stop_daemons: whether to also stop the master daemons
287 (ganeti-masterd and ganeti-rapi)
291 # TODO: log and report back to the caller the error failures; we
292 # need to decide in which case we fail the RPC for this
294 # GetMasterInfo will raise an exception if not able to return data
295 master_netdev, master_ip, _ = GetMasterInfo()
297 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
298 "dev", master_netdev])
300 logging.error("Can't remove the master IP, error: %s", result.output)
301 # but otherwise ignore the failure
304 # stop/kill the rapi and the master daemon
305 for daemon in constants.RAPI, constants.MASTERD:
306 utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon)))
309 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
310 """Joins this node to the cluster.
312 This does the following:
313 - updates the hostkeys of the machine (rsa and dsa)
314 - adds the ssh private key to the user
315 - adds the ssh public key to the users' authorized_keys file
318 @param dsa: the DSA private key to write
320 @param dsapub: the DSA public key to write
322 @param rsa: the RSA private key to write
324 @param rsapub: the RSA public key to write
326 @param sshkey: the SSH private key to write
328 @param sshpub: the SSH public key to write
330 @return: the success of the operation
333 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
334 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
335 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
336 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
337 for name, content, mode in sshd_keys:
338 utils.WriteFile(name, data=content, mode=mode)
341 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
343 except errors.OpExecError, err:
344 _Fail("Error while processing user ssh files: %s", err, exc=True)
346 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
347 utils.WriteFile(name, data=content, mode=0600)
349 utils.AddAuthorizedKey(auth_keys, sshpub)
351 utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
355 """Cleans up and remove the current node.
357 This function cleans up and prepares the current node to be removed
360 If processing is successful, then it raises an
361 L{errors.QuitGanetiException} which is used as a special case to
362 shutdown the node daemon.
365 _CleanDirectory(constants.DATA_DIR)
369 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
371 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
373 utils.RemoveFile(priv_key)
374 utils.RemoveFile(pub_key)
375 except errors.OpExecError:
376 logging.exception("Error while processing ssh files")
379 utils.RemoveFile(constants.HMAC_CLUSTER_KEY)
380 utils.RemoveFile(constants.RAPI_CERT_FILE)
381 utils.RemoveFile(constants.SSL_CERT_FILE)
383 logging.exception("Error while removing cluster secrets")
385 confd_pid = utils.ReadPidFile(utils.DaemonPidFileName(constants.CONFD))
388 utils.KillProcess(confd_pid, timeout=2)
390 # Raise a custom exception (handled in ganeti-noded)
391 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
394 def GetNodeInfo(vgname, hypervisor_type):
395 """Gives back a hash with different information about the node.
397 @type vgname: C{string}
398 @param vgname: the name of the volume group to ask for disk space information
399 @type hypervisor_type: C{str}
400 @param hypervisor_type: the name of the hypervisor to ask for
403 @return: dictionary with the following keys:
404 - vg_size is the size of the configured volume group in MiB
405 - vg_free is the free size of the volume group in MiB
406 - memory_dom0 is the memory allocated for domain0 in MiB
407 - memory_free is the currently available (free) ram in MiB
408 - memory_total is the total number of ram in MiB
412 vginfo = _GetVGInfo(vgname)
413 outputarray['vg_size'] = vginfo['vg_size']
414 outputarray['vg_free'] = vginfo['vg_free']
416 hyper = hypervisor.GetHypervisor(hypervisor_type)
417 hyp_info = hyper.GetNodeInfo()
418 if hyp_info is not None:
419 outputarray.update(hyp_info)
421 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
426 def VerifyNode(what, cluster_name):
427 """Verify the status of the local node.
429 Based on the input L{what} parameter, various checks are done on the
432 If the I{filelist} key is present, this list of
433 files is checksummed and the file/checksum pairs are returned.
435 If the I{nodelist} key is present, we check that we have
436 connectivity via ssh with the target nodes (and check the hostname
439 If the I{node-net-test} key is present, we check that we have
440 connectivity to the given nodes via both primary IP and, if
441 applicable, secondary IPs.
444 @param what: a dictionary of things to check:
445 - filelist: list of files for which to compute checksums
446 - nodelist: list of nodes we should check ssh communication with
447 - node-net-test: list of nodes we should check node daemon port
449 - hypervisor: list with hypervisors to run the verify for
451 @return: a dictionary with the same keys as the input dict, and
452 values representing the result of the checks
457 if constants.NV_HYPERVISOR in what:
458 result[constants.NV_HYPERVISOR] = tmp = {}
459 for hv_name in what[constants.NV_HYPERVISOR]:
460 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
462 if constants.NV_FILELIST in what:
463 result[constants.NV_FILELIST] = utils.FingerprintFiles(
464 what[constants.NV_FILELIST])
466 if constants.NV_NODELIST in what:
467 result[constants.NV_NODELIST] = tmp = {}
468 random.shuffle(what[constants.NV_NODELIST])
469 for node in what[constants.NV_NODELIST]:
470 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
474 if constants.NV_NODENETTEST in what:
475 result[constants.NV_NODENETTEST] = tmp = {}
476 my_name = utils.HostInfo().name
477 my_pip = my_sip = None
478 for name, pip, sip in what[constants.NV_NODENETTEST]:
484 tmp[my_name] = ("Can't find my own primary/secondary IP"
487 port = utils.GetDaemonPort(constants.NODED)
488 for name, pip, sip in what[constants.NV_NODENETTEST]:
490 if not utils.TcpPing(pip, port, source=my_pip):
491 fail.append("primary")
493 if not utils.TcpPing(sip, port, source=my_sip):
494 fail.append("secondary")
496 tmp[name] = ("failure using the %s interface(s)" %
499 if constants.NV_LVLIST in what:
500 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
502 if constants.NV_INSTANCELIST in what:
503 result[constants.NV_INSTANCELIST] = GetInstanceList(
504 what[constants.NV_INSTANCELIST])
506 if constants.NV_VGLIST in what:
507 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
509 if constants.NV_VERSION in what:
510 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
511 constants.RELEASE_VERSION)
513 if constants.NV_HVINFO in what:
514 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
515 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
517 if constants.NV_DRBDLIST in what:
519 used_minors = bdev.DRBD8.GetUsedDevs().keys()
520 except errors.BlockDeviceError, err:
521 logging.warning("Can't get used minors list", exc_info=True)
522 used_minors = str(err)
523 result[constants.NV_DRBDLIST] = used_minors
528 def GetVolumeList(vg_name):
529 """Compute list of logical volumes and their size.
532 @param vg_name: the volume group whose LVs we should list
535 dictionary of all partions (key) with value being a tuple of
536 their size (in MiB), inactive and online status::
538 {'test1': ('20.06', True, True)}
540 in case of errors, a string is returned with the error
546 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
547 "--separator=%s" % sep,
548 "-olv_name,lv_size,lv_attr", vg_name])
550 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
552 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
553 for line in result.stdout.splitlines():
555 match = valid_line_re.match(line)
557 logging.error("Invalid line returned from lvs output: '%s'", line)
559 name, size, attr = match.groups()
560 inactive = attr[4] == '-'
561 online = attr[5] == 'o'
562 virtual = attr[0] == 'v'
564 # we don't want to report such volumes as existing, since they
565 # don't really hold data
567 lvs[name] = (size, inactive, online)
572 def ListVolumeGroups():
573 """List the volume groups and their size.
576 @return: dictionary with keys volume name and values the
580 return utils.ListVolumeGroups()
584 """List all volumes on this node.
588 A list of dictionaries, each having four keys:
589 - name: the logical volume name,
590 - size: the size of the logical volume
591 - dev: the physical device on which the LV lives
592 - vg: the volume group to which it belongs
594 In case of errors, we return an empty list and log the
597 Note that since a logical volume can live on multiple physical
598 volumes, the resulting list might include a logical volume
602 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
604 "--options=lv_name,lv_size,devices,vg_name"])
606 _Fail("Failed to list logical volumes, lvs output: %s",
611 return dev.split('(')[0]
617 'name': line[0].strip(),
618 'size': line[1].strip(),
619 'dev': parse_dev(line[2].strip()),
620 'vg': line[3].strip(),
623 return [map_line(line.split('|')) for line in result.stdout.splitlines()
624 if line.count('|') >= 3]
627 def BridgesExist(bridges_list):
628 """Check if a list of bridges exist on the current node.
631 @return: C{True} if all of them exist, C{False} otherwise
635 for bridge in bridges_list:
636 if not utils.BridgeExists(bridge):
637 missing.append(bridge)
640 _Fail("Missing bridges %s", ", ".join(missing))
643 def GetInstanceList(hypervisor_list):
644 """Provides a list of instances.
646 @type hypervisor_list: list
647 @param hypervisor_list: the list of hypervisors to query information
650 @return: a list of all running instances on the current node
651 - instance1.example.com
652 - instance2.example.com
656 for hname in hypervisor_list:
658 names = hypervisor.GetHypervisor(hname).ListInstances()
659 results.extend(names)
660 except errors.HypervisorError, err:
661 _Fail("Error enumerating instances (hypervisor %s): %s",
662 hname, err, exc=True)
667 def GetInstanceInfo(instance, hname):
668 """Gives back the information about an instance as a dictionary.
670 @type instance: string
671 @param instance: the instance name
673 @param hname: the hypervisor type of the instance
676 @return: dictionary with the following keys:
677 - memory: memory size of instance (int)
678 - state: xen state of instance (string)
679 - time: cpu time of instance (float)
684 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
685 if iinfo is not None:
686 output['memory'] = iinfo[2]
687 output['state'] = iinfo[4]
688 output['time'] = iinfo[5]
693 def GetInstanceMigratable(instance):
694 """Gives whether an instance can be migrated.
696 @type instance: L{objects.Instance}
697 @param instance: object representing the instance to be checked.
700 @return: tuple of (result, description) where:
701 - result: whether the instance can be migrated or not
702 - description: a description of the issue, if relevant
705 hyper = hypervisor.GetHypervisor(instance.hypervisor)
706 iname = instance.name
707 if iname not in hyper.ListInstances():
708 _Fail("Instance %s is not running", iname)
710 for idx in range(len(instance.disks)):
711 link_name = _GetBlockDevSymlinkPath(iname, idx)
712 if not os.path.islink(link_name):
713 _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
716 def GetAllInstancesInfo(hypervisor_list):
717 """Gather data about all instances.
719 This is the equivalent of L{GetInstanceInfo}, except that it
720 computes data for all instances at once, thus being faster if one
721 needs data about more than one instance.
723 @type hypervisor_list: list
724 @param hypervisor_list: list of hypervisors to query for instance data
727 @return: dictionary of instance: data, with data having the following keys:
728 - memory: memory size of instance (int)
729 - state: xen state of instance (string)
730 - time: cpu time of instance (float)
731 - vcpus: the number of vcpus
736 for hname in hypervisor_list:
737 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
739 for name, _, memory, vcpus, state, times in iinfo:
747 # we only check static parameters, like memory and vcpus,
748 # and not state and time which can change between the
749 # invocations of the different hypervisors
750 for key in 'memory', 'vcpus':
751 if value[key] != output[name][key]:
752 _Fail("Instance %s is running twice"
753 " with different parameters", name)
759 def InstanceOsAdd(instance, reinstall):
760 """Add an OS to an instance.
762 @type instance: L{objects.Instance}
763 @param instance: Instance whose OS is to be installed
764 @type reinstall: boolean
765 @param reinstall: whether this is an instance reinstall
769 inst_os = OSFromDisk(instance.os)
771 create_env = OSEnvironment(instance, inst_os)
773 create_env['INSTANCE_REINSTALL'] = "1"
775 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
776 instance.name, int(time.time()))
778 result = utils.RunCmd([inst_os.create_script], env=create_env,
779 cwd=inst_os.path, output=logfile,)
781 logging.error("os create command '%s' returned error: %s, logfile: %s,"
782 " output: %s", result.cmd, result.fail_reason, logfile,
784 lines = [utils.SafeEncode(val)
785 for val in utils.TailFile(logfile, lines=20)]
786 _Fail("OS create script failed (%s), last lines in the"
787 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
790 def RunRenameInstance(instance, old_name):
791 """Run the OS rename script for an instance.
793 @type instance: L{objects.Instance}
794 @param instance: Instance whose OS is to be installed
795 @type old_name: string
796 @param old_name: previous instance name
798 @return: the success of the operation
801 inst_os = OSFromDisk(instance.os)
803 rename_env = OSEnvironment(instance, inst_os)
804 rename_env['OLD_INSTANCE_NAME'] = old_name
806 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
808 instance.name, int(time.time()))
810 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
811 cwd=inst_os.path, output=logfile)
814 logging.error("os create command '%s' returned error: %s output: %s",
815 result.cmd, result.fail_reason, result.output)
816 lines = [utils.SafeEncode(val)
817 for val in utils.TailFile(logfile, lines=20)]
818 _Fail("OS rename script failed (%s), last lines in the"
819 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
822 def _GetVGInfo(vg_name):
823 """Get information about the volume group.
826 @param vg_name: the volume group which we query
829 A dictionary with the following keys:
830 - C{vg_size} is the total size of the volume group in MiB
831 - C{vg_free} is the free size of the volume group in MiB
832 - C{pv_count} are the number of physical disks in that VG
834 If an error occurs during gathering of data, we return the same dict
835 with keys all set to None.
838 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
840 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
841 "--nosuffix", "--units=m", "--separator=:", vg_name])
844 logging.error("volume group %s not present", vg_name)
846 valarr = retval.stdout.strip().rstrip(':').split(':')
850 "vg_size": int(round(float(valarr[0]), 0)),
851 "vg_free": int(round(float(valarr[1]), 0)),
852 "pv_count": int(valarr[2]),
854 except ValueError, err:
855 logging.exception("Fail to parse vgs output: %s", err)
857 logging.error("vgs output has the wrong number of fields (expected"
858 " three): %s", str(valarr))
862 def _GetBlockDevSymlinkPath(instance_name, idx):
863 return os.path.join(constants.DISK_LINKS_DIR,
864 "%s:%d" % (instance_name, idx))
867 def _SymlinkBlockDev(instance_name, device_path, idx):
868 """Set up symlinks to a instance's block device.
870 This is an auxiliary function run when an instance is start (on the primary
871 node) or when an instance is migrated (on the target node).
874 @param instance_name: the name of the target instance
875 @param device_path: path of the physical block device, on the node
876 @param idx: the disk index
877 @return: absolute path to the disk's symlink
880 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
882 os.symlink(device_path, link_name)
884 if err.errno == errno.EEXIST:
885 if (not os.path.islink(link_name) or
886 os.readlink(link_name) != device_path):
888 os.symlink(device_path, link_name)
895 def _RemoveBlockDevLinks(instance_name, disks):
896 """Remove the block device symlinks belonging to the given instance.
899 for idx, _ in enumerate(disks):
900 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
901 if os.path.islink(link_name):
905 logging.exception("Can't remove symlink '%s'", link_name)
908 def _GatherAndLinkBlockDevs(instance):
909 """Set up an instance's block device(s).
911 This is run on the primary node at instance startup. The block
912 devices must be already assembled.
914 @type instance: L{objects.Instance}
915 @param instance: the instance whose disks we shoul assemble
917 @return: list of (disk_object, device_path)
921 for idx, disk in enumerate(instance.disks):
922 device = _RecursiveFindBD(disk)
924 raise errors.BlockDeviceError("Block device '%s' is not set up." %
928 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
930 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
933 block_devices.append((disk, link_name))
938 def StartInstance(instance):
939 """Start an instance.
941 @type instance: L{objects.Instance}
942 @param instance: the instance object
946 running_instances = GetInstanceList([instance.hypervisor])
948 if instance.name in running_instances:
949 logging.info("Instance %s already running, not starting", instance.name)
953 block_devices = _GatherAndLinkBlockDevs(instance)
954 hyper = hypervisor.GetHypervisor(instance.hypervisor)
955 hyper.StartInstance(instance, block_devices)
956 except errors.BlockDeviceError, err:
957 _Fail("Block device error: %s", err, exc=True)
958 except errors.HypervisorError, err:
959 _RemoveBlockDevLinks(instance.name, instance.disks)
960 _Fail("Hypervisor error: %s", err, exc=True)
963 def InstanceShutdown(instance):
964 """Shut an instance down.
966 @note: this functions uses polling with a hardcoded timeout.
968 @type instance: L{objects.Instance}
969 @param instance: the instance object
973 hv_name = instance.hypervisor
974 hyper = hypervisor.GetHypervisor(hv_name)
975 running_instances = hyper.ListInstances()
976 iname = instance.name
977 timeout = constants.DEFAULT_SHUTDOWN_TIMEOUT
979 if iname not in running_instances:
980 logging.info("Instance %s not running, doing nothing", iname)
984 end = start + timeout
988 while not tried_once and time.time() < end:
990 hyper.StopInstance(instance, retry=tried_once)
991 except errors.HypervisorError, err:
992 _Fail("Failed to stop instance %s: %s", iname, err)
994 time.sleep(sleep_time)
995 if instance.name not in hyper.ListInstances():
998 # 1.2 behaves particularly good for our case:
999 # it gives us 10 increasing steps and caps just slightly above 5 seconds
1002 # the shutdown did not succeed
1003 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1006 hyper.StopInstance(instance, force=True)
1007 except errors.HypervisorError, err:
1008 _Fail("Failed to force stop instance %s: %s", iname, err)
1011 if instance.name in GetInstanceList([hv_name]):
1012 _Fail("Could not shutdown instance %s even by destroy", iname)
1014 _RemoveBlockDevLinks(iname, instance.disks)
1017 def InstanceReboot(instance, reboot_type):
1018 """Reboot an instance.
1020 @type instance: L{objects.Instance}
1021 @param instance: the instance object to reboot
1022 @type reboot_type: str
1023 @param reboot_type: the type of reboot, one the following
1025 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1026 instance OS, do not recreate the VM
1027 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1028 restart the VM (at the hypervisor level)
1029 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1030 not accepted here, since that mode is handled differently, in
1031 cmdlib, and translates into full stop and start of the
1032 instance (instead of a call_instance_reboot RPC)
1036 running_instances = GetInstanceList([instance.hypervisor])
1038 if instance.name not in running_instances:
1039 _Fail("Cannot reboot instance %s that is not running", instance.name)
1041 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1042 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1044 hyper.RebootInstance(instance)
1045 except errors.HypervisorError, err:
1046 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1047 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1049 InstanceShutdown(instance)
1050 return StartInstance(instance)
1051 except errors.HypervisorError, err:
1052 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1054 _Fail("Invalid reboot_type received: %s", reboot_type)
1057 def MigrationInfo(instance):
1058 """Gather information about an instance to be migrated.
1060 @type instance: L{objects.Instance}
1061 @param instance: the instance definition
1064 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1066 info = hyper.MigrationInfo(instance)
1067 except errors.HypervisorError, err:
1068 _Fail("Failed to fetch migration information: %s", err, exc=True)
1072 def AcceptInstance(instance, info, target):
1073 """Prepare the node to accept an instance.
1075 @type instance: L{objects.Instance}
1076 @param instance: the instance definition
1077 @type info: string/data (opaque)
1078 @param info: migration information, from the source node
1079 @type target: string
1080 @param target: target host (usually ip), on this node
1083 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1085 hyper.AcceptInstance(instance, info, target)
1086 except errors.HypervisorError, err:
1087 _Fail("Failed to accept instance: %s", err, exc=True)
1090 def FinalizeMigration(instance, info, success):
1091 """Finalize any preparation to accept an instance.
1093 @type instance: L{objects.Instance}
1094 @param instance: the instance definition
1095 @type info: string/data (opaque)
1096 @param info: migration information, from the source node
1097 @type success: boolean
1098 @param success: whether the migration was a success or a failure
1101 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1103 hyper.FinalizeMigration(instance, info, success)
1104 except errors.HypervisorError, err:
1105 _Fail("Failed to finalize migration: %s", err, exc=True)
1108 def MigrateInstance(instance, target, live):
1109 """Migrates an instance to another node.
1111 @type instance: L{objects.Instance}
1112 @param instance: the instance definition
1113 @type target: string
1114 @param target: the target node name
1116 @param live: whether the migration should be done live or not (the
1117 interpretation of this parameter is left to the hypervisor)
1119 @return: a tuple of (success, msg) where:
1120 - succes is a boolean denoting the success/failure of the operation
1121 - msg is a string with details in case of failure
1124 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1127 hyper.MigrateInstance(instance.name, target, live)
1128 except errors.HypervisorError, err:
1129 _Fail("Failed to migrate instance: %s", err, exc=True)
1132 def BlockdevCreate(disk, size, owner, on_primary, info):
1133 """Creates a block device for an instance.
1135 @type disk: L{objects.Disk}
1136 @param disk: the object describing the disk we should create
1138 @param size: the size of the physical underlying device, in MiB
1140 @param owner: the name of the instance for which disk is created,
1141 used for device cache data
1142 @type on_primary: boolean
1143 @param on_primary: indicates if it is the primary node or not
1145 @param info: string that will be sent to the physical device
1146 creation, used for example to set (LVM) tags on LVs
1148 @return: the new unique_id of the device (this can sometime be
1149 computed only after creation), or None. On secondary nodes,
1150 it's not required to return anything.
1155 for child in disk.children:
1157 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1158 except errors.BlockDeviceError, err:
1159 _Fail("Can't assemble device %s: %s", child, err)
1160 if on_primary or disk.AssembleOnSecondary():
1161 # we need the children open in case the device itself has to
1165 except errors.BlockDeviceError, err:
1166 _Fail("Can't make child '%s' read-write: %s", child, err)
1170 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1171 except errors.BlockDeviceError, err:
1172 _Fail("Can't create block device: %s", err)
1174 if on_primary or disk.AssembleOnSecondary():
1177 except errors.BlockDeviceError, err:
1178 _Fail("Can't assemble device after creation, unusual event: %s", err)
1179 device.SetSyncSpeed(constants.SYNC_SPEED)
1180 if on_primary or disk.OpenOnSecondary():
1182 device.Open(force=True)
1183 except errors.BlockDeviceError, err:
1184 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1185 DevCacheManager.UpdateCache(device.dev_path, owner,
1186 on_primary, disk.iv_name)
1188 device.SetInfo(info)
1190 return device.unique_id
1193 def BlockdevRemove(disk):
1194 """Remove a block device.
1196 @note: This is intended to be called recursively.
1198 @type disk: L{objects.Disk}
1199 @param disk: the disk object we should remove
1201 @return: the success of the operation
1206 rdev = _RecursiveFindBD(disk)
1207 except errors.BlockDeviceError, err:
1208 # probably can't attach
1209 logging.info("Can't attach to device %s in remove", disk)
1211 if rdev is not None:
1212 r_path = rdev.dev_path
1215 except errors.BlockDeviceError, err:
1216 msgs.append(str(err))
1218 DevCacheManager.RemoveCache(r_path)
1221 for child in disk.children:
1223 BlockdevRemove(child)
1224 except RPCFail, err:
1225 msgs.append(str(err))
1228 _Fail("; ".join(msgs))
1231 def _RecursiveAssembleBD(disk, owner, as_primary):
1232 """Activate a block device for an instance.
1234 This is run on the primary and secondary nodes for an instance.
1236 @note: this function is called recursively.
1238 @type disk: L{objects.Disk}
1239 @param disk: the disk we try to assemble
1241 @param owner: the name of the instance which owns the disk
1242 @type as_primary: boolean
1243 @param as_primary: if we should make the block device
1246 @return: the assembled device or None (in case no device
1248 @raise errors.BlockDeviceError: in case there is an error
1249 during the activation of the children or the device
1255 mcn = disk.ChildrenNeeded()
1257 mcn = 0 # max number of Nones allowed
1259 mcn = len(disk.children) - mcn # max number of Nones
1260 for chld_disk in disk.children:
1262 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1263 except errors.BlockDeviceError, err:
1264 if children.count(None) >= mcn:
1267 logging.error("Error in child activation (but continuing): %s",
1269 children.append(cdev)
1271 if as_primary or disk.AssembleOnSecondary():
1272 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1273 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1275 if as_primary or disk.OpenOnSecondary():
1277 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1278 as_primary, disk.iv_name)
1285 def BlockdevAssemble(disk, owner, as_primary):
1286 """Activate a block device for an instance.
1288 This is a wrapper over _RecursiveAssembleBD.
1290 @rtype: str or boolean
1291 @return: a C{/dev/...} path for primary nodes, and
1292 C{True} for secondary nodes
1296 result = _RecursiveAssembleBD(disk, owner, as_primary)
1297 if isinstance(result, bdev.BlockDev):
1298 result = result.dev_path
1299 except errors.BlockDeviceError, err:
1300 _Fail("Error while assembling disk: %s", err, exc=True)
1305 def BlockdevShutdown(disk):
1306 """Shut down a block device.
1308 First, if the device is assembled (Attach() is successful), then
1309 the device is shutdown. Then the children of the device are
1312 This function is called recursively. Note that we don't cache the
1313 children or such, as oppossed to assemble, shutdown of different
1314 devices doesn't require that the upper device was active.
1316 @type disk: L{objects.Disk}
1317 @param disk: the description of the disk we should
1323 r_dev = _RecursiveFindBD(disk)
1324 if r_dev is not None:
1325 r_path = r_dev.dev_path
1328 DevCacheManager.RemoveCache(r_path)
1329 except errors.BlockDeviceError, err:
1330 msgs.append(str(err))
1333 for child in disk.children:
1335 BlockdevShutdown(child)
1336 except RPCFail, err:
1337 msgs.append(str(err))
1340 _Fail("; ".join(msgs))
1343 def BlockdevAddchildren(parent_cdev, new_cdevs):
1344 """Extend a mirrored block device.
1346 @type parent_cdev: L{objects.Disk}
1347 @param parent_cdev: the disk to which we should add children
1348 @type new_cdevs: list of L{objects.Disk}
1349 @param new_cdevs: the list of children which we should add
1353 parent_bdev = _RecursiveFindBD(parent_cdev)
1354 if parent_bdev is None:
1355 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1356 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1357 if new_bdevs.count(None) > 0:
1358 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1359 parent_bdev.AddChildren(new_bdevs)
1362 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1363 """Shrink a mirrored block device.
1365 @type parent_cdev: L{objects.Disk}
1366 @param parent_cdev: the disk from which we should remove children
1367 @type new_cdevs: list of L{objects.Disk}
1368 @param new_cdevs: the list of children which we should remove
1372 parent_bdev = _RecursiveFindBD(parent_cdev)
1373 if parent_bdev is None:
1374 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1376 for disk in new_cdevs:
1377 rpath = disk.StaticDevPath()
1379 bd = _RecursiveFindBD(disk)
1381 _Fail("Can't find device %s while removing children", disk)
1383 devs.append(bd.dev_path)
1386 parent_bdev.RemoveChildren(devs)
1389 def BlockdevGetmirrorstatus(disks):
1390 """Get the mirroring status of a list of devices.
1392 @type disks: list of L{objects.Disk}
1393 @param disks: the list of disks which we should query
1396 a list of (mirror_done, estimated_time) tuples, which
1397 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1398 @raise errors.BlockDeviceError: if any of the disks cannot be
1404 rbd = _RecursiveFindBD(dsk)
1406 _Fail("Can't find device %s", dsk)
1408 stats.append(rbd.CombinedSyncStatus())
1413 def _RecursiveFindBD(disk):
1414 """Check if a device is activated.
1416 If so, return information about the real device.
1418 @type disk: L{objects.Disk}
1419 @param disk: the disk object we need to find
1421 @return: None if the device can't be found,
1422 otherwise the device instance
1427 for chdisk in disk.children:
1428 children.append(_RecursiveFindBD(chdisk))
1430 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1433 def BlockdevFind(disk):
1434 """Check if a device is activated.
1436 If it is, return information about the real device.
1438 @type disk: L{objects.Disk}
1439 @param disk: the disk to find
1440 @rtype: None or objects.BlockDevStatus
1441 @return: None if the disk cannot be found, otherwise a the current
1446 rbd = _RecursiveFindBD(disk)
1447 except errors.BlockDeviceError, err:
1448 _Fail("Failed to find device: %s", err, exc=True)
1453 return rbd.GetSyncStatus()
1456 def BlockdevGetsize(disks):
1457 """Computes the size of the given disks.
1459 If a disk is not found, returns None instead.
1461 @type disks: list of L{objects.Disk}
1462 @param disks: the list of disk to compute the size for
1464 @return: list with elements None if the disk cannot be found,
1471 rbd = _RecursiveFindBD(cf)
1472 except errors.BlockDeviceError, err:
1478 result.append(rbd.GetActualSize())
1482 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1483 """Export a block device to a remote node.
1485 @type disk: L{objects.Disk}
1486 @param disk: the description of the disk to export
1487 @type dest_node: str
1488 @param dest_node: the destination node to export to
1489 @type dest_path: str
1490 @param dest_path: the destination path on the target node
1491 @type cluster_name: str
1492 @param cluster_name: the cluster name, needed for SSH hostalias
1496 real_disk = _RecursiveFindBD(disk)
1497 if real_disk is None:
1498 _Fail("Block device '%s' is not set up", disk)
1502 # the block size on the read dd is 1MiB to match our units
1503 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1504 "dd if=%s bs=1048576 count=%s",
1505 real_disk.dev_path, str(disk.size))
1507 # we set here a smaller block size as, due to ssh buffering, more
1508 # than 64-128k will mostly ignored; we use nocreat to fail if the
1509 # device is not already there or we pass a wrong path; we use
1510 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1511 # to not buffer too much memory; this means that at best, we flush
1512 # every 64k, which will not be very fast
1513 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1514 " oflag=dsync", dest_path)
1516 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1517 constants.GANETI_RUNAS,
1520 # all commands have been checked, so we're safe to combine them
1521 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1523 result = utils.RunCmd(["bash", "-c", command])
1526 _Fail("Disk copy command '%s' returned error: %s"
1527 " output: %s", command, result.fail_reason, result.output)
1530 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1531 """Write a file to the filesystem.
1533 This allows the master to overwrite(!) a file. It will only perform
1534 the operation if the file belongs to a list of configuration files.
1536 @type file_name: str
1537 @param file_name: the target file name
1539 @param data: the new contents of the file
1541 @param mode: the mode to give the file (can be None)
1543 @param uid: the owner of the file (can be -1 for default)
1545 @param gid: the group of the file (can be -1 for default)
1547 @param atime: the atime to set on the file (can be None)
1549 @param mtime: the mtime to set on the file (can be None)
1553 if not os.path.isabs(file_name):
1554 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1556 if file_name not in _ALLOWED_UPLOAD_FILES:
1557 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1560 raw_data = _Decompress(data)
1562 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1563 atime=atime, mtime=mtime)
1566 def WriteSsconfFiles(values):
1567 """Update all ssconf files.
1569 Wrapper around the SimpleStore.WriteFiles.
1572 ssconf.SimpleStore().WriteFiles(values)
1575 def _ErrnoOrStr(err):
1576 """Format an EnvironmentError exception.
1578 If the L{err} argument has an errno attribute, it will be looked up
1579 and converted into a textual C{E...} description. Otherwise the
1580 string representation of the error will be returned.
1582 @type err: L{EnvironmentError}
1583 @param err: the exception to format
1586 if hasattr(err, 'errno'):
1587 detail = errno.errorcode[err.errno]
1593 def _OSOndiskAPIVersion(name, os_dir):
1594 """Compute and return the API version of a given OS.
1596 This function will try to read the API version of the OS given by
1597 the 'name' parameter and residing in the 'os_dir' directory.
1600 @param name: the OS name we should look for
1602 @param os_dir: the directory inwhich we should look for the OS
1604 @return: tuple (status, data) with status denoting the validity and
1605 data holding either the vaid versions or an error message
1608 api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
1611 st = os.stat(api_file)
1612 except EnvironmentError, err:
1613 return False, ("Required file '%s' not found under path %s: %s" %
1614 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1616 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1617 return False, ("File '%s' in %s is not a regular file" %
1618 (constants.OS_API_FILE, os_dir))
1621 api_versions = utils.ReadFile(api_file).splitlines()
1622 except EnvironmentError, err:
1623 return False, ("Error while reading the API version file at %s: %s" %
1624 (api_file, _ErrnoOrStr(err)))
1627 api_versions = [int(version.strip()) for version in api_versions]
1628 except (TypeError, ValueError), err:
1629 return False, ("API version(s) can't be converted to integer: %s" %
1632 return True, api_versions
1635 def DiagnoseOS(top_dirs=None):
1636 """Compute the validity for all OSes.
1638 @type top_dirs: list
1639 @param top_dirs: the list of directories in which to
1640 search (if not given defaults to
1641 L{constants.OS_SEARCH_PATH})
1642 @rtype: list of L{objects.OS}
1643 @return: a list of tuples (name, path, status, diagnose, variants)
1644 for all (potential) OSes under all search paths, where:
1645 - name is the (potential) OS name
1646 - path is the full path to the OS
1647 - status True/False is the validity of the OS
1648 - diagnose is the error message for an invalid OS, otherwise empty
1649 - variants is a list of supported OS variants, if any
1652 if top_dirs is None:
1653 top_dirs = constants.OS_SEARCH_PATH
1656 for dir_name in top_dirs:
1657 if os.path.isdir(dir_name):
1659 f_names = utils.ListVisibleFiles(dir_name)
1660 except EnvironmentError, err:
1661 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1663 for name in f_names:
1664 os_path = os.path.sep.join([dir_name, name])
1665 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1668 variants = os_inst.supported_variants
1672 result.append((name, os_path, status, diagnose, variants))
1677 def _TryOSFromDisk(name, base_dir=None):
1678 """Create an OS instance from disk.
1680 This function will return an OS instance if the given name is a
1683 @type base_dir: string
1684 @keyword base_dir: Base directory containing OS installations.
1685 Defaults to a search in all the OS_SEARCH_PATH dirs.
1687 @return: success and either the OS instance if we find a valid one,
1691 if base_dir is None:
1692 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1694 return False, "Directory for OS %s not found in search path" % name
1696 os_dir = os.path.sep.join([base_dir, name])
1698 status, api_versions = _OSOndiskAPIVersion(name, os_dir)
1701 return status, api_versions
1703 if not constants.OS_API_VERSIONS.intersection(api_versions):
1704 return False, ("API version mismatch for path '%s': found %s, want %s." %
1705 (os_dir, api_versions, constants.OS_API_VERSIONS))
1707 # OS Files dictionary, we will populate it with the absolute path names
1708 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1710 if max(api_versions) >= constants.OS_API_V15:
1711 os_files[constants.OS_VARIANTS_FILE] = ''
1713 for name in os_files:
1714 os_files[name] = os.path.sep.join([os_dir, name])
1717 st = os.stat(os_files[name])
1718 except EnvironmentError, err:
1719 return False, ("File '%s' under path '%s' is missing (%s)" %
1720 (name, os_dir, _ErrnoOrStr(err)))
1722 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1723 return False, ("File '%s' under path '%s' is not a regular file" %
1726 if name in constants.OS_SCRIPTS:
1727 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1728 return False, ("File '%s' under path '%s' is not executable" %
1732 if constants.OS_VARIANTS_FILE in os_files:
1733 variants_file = os_files[constants.OS_VARIANTS_FILE]
1735 variants = utils.ReadFile(variants_file).splitlines()
1736 except EnvironmentError, err:
1737 return False, ("Error while reading the OS variants file at %s: %s" %
1738 (variants_file, _ErrnoOrStr(err)))
1740 return False, ("No supported os variant found")
1742 os_obj = objects.OS(name=name, path=os_dir,
1743 create_script=os_files[constants.OS_SCRIPT_CREATE],
1744 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1745 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1746 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1747 supported_variants=variants,
1748 api_versions=api_versions)
1752 def OSFromDisk(name, base_dir=None):
1753 """Create an OS instance from disk.
1755 This function will return an OS instance if the given name is a
1756 valid OS name. Otherwise, it will raise an appropriate
1757 L{RPCFail} exception, detailing why this is not a valid OS.
1759 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1760 an exception but returns true/false status data.
1762 @type base_dir: string
1763 @keyword base_dir: Base directory containing OS installations.
1764 Defaults to a search in all the OS_SEARCH_PATH dirs.
1765 @rtype: L{objects.OS}
1766 @return: the OS instance if we find a valid one
1767 @raise RPCFail: if we don't find a valid OS
1770 name_only = name.split('+',1)[0]
1771 status, payload = _TryOSFromDisk(name_only, base_dir)
1779 def OSEnvironment(instance, os, debug=0):
1780 """Calculate the environment for an os script.
1782 @type instance: L{objects.Instance}
1783 @param instance: target instance for the os script run
1784 @type os: L{objects.OS}
1785 @param os: operating system for which the environment is being built
1786 @type debug: integer
1787 @param debug: debug level (0 or 1, for OS Api 10)
1789 @return: dict of environment variables
1790 @raise errors.BlockDeviceError: if the block device
1795 api_version = max(constants.OS_API_VERSIONS.intersection(os.api_versions))
1796 result['OS_API_VERSION'] = '%d' % api_version
1797 result['INSTANCE_NAME'] = instance.name
1798 result['INSTANCE_OS'] = instance.os
1799 result['HYPERVISOR'] = instance.hypervisor
1800 result['DISK_COUNT'] = '%d' % len(instance.disks)
1801 result['NIC_COUNT'] = '%d' % len(instance.nics)
1802 result['DEBUG_LEVEL'] = '%d' % debug
1803 if api_version >= constants.OS_API_V15:
1805 variant = instance.os.split('+', 1)[1]
1807 variant = os.supported_variants[0]
1808 result['OS_VARIANT'] = variant
1809 for idx, disk in enumerate(instance.disks):
1810 real_disk = _RecursiveFindBD(disk)
1811 if real_disk is None:
1812 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1815 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1816 result['DISK_%d_ACCESS' % idx] = disk.mode
1817 if constants.HV_DISK_TYPE in instance.hvparams:
1818 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1819 instance.hvparams[constants.HV_DISK_TYPE]
1820 if disk.dev_type in constants.LDS_BLOCK:
1821 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1822 elif disk.dev_type == constants.LD_FILE:
1823 result['DISK_%d_BACKEND_TYPE' % idx] = \
1824 'file:%s' % disk.physical_id[0]
1825 for idx, nic in enumerate(instance.nics):
1826 result['NIC_%d_MAC' % idx] = nic.mac
1828 result['NIC_%d_IP' % idx] = nic.ip
1829 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1830 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1831 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1832 if nic.nicparams[constants.NIC_LINK]:
1833 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1834 if constants.HV_NIC_TYPE in instance.hvparams:
1835 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1836 instance.hvparams[constants.HV_NIC_TYPE]
1838 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1839 for key, value in source.items():
1840 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1844 def BlockdevGrow(disk, amount):
1845 """Grow a stack of block devices.
1847 This function is called recursively, with the childrens being the
1848 first ones to resize.
1850 @type disk: L{objects.Disk}
1851 @param disk: the disk to be grown
1852 @rtype: (status, result)
1853 @return: a tuple with the status of the operation
1854 (True/False), and the errors message if status
1858 r_dev = _RecursiveFindBD(disk)
1860 _Fail("Cannot find block device %s", disk)
1864 except errors.BlockDeviceError, err:
1865 _Fail("Failed to grow block device: %s", err, exc=True)
1868 def BlockdevSnapshot(disk):
1869 """Create a snapshot copy of a block device.
1871 This function is called recursively, and the snapshot is actually created
1872 just for the leaf lvm backend device.
1874 @type disk: L{objects.Disk}
1875 @param disk: the disk to be snapshotted
1877 @return: snapshot disk path
1881 if len(disk.children) == 1:
1882 # only one child, let's recurse on it
1883 return BlockdevSnapshot(disk.children[0])
1885 # more than one child, choose one that matches
1886 for child in disk.children:
1887 if child.size == disk.size:
1888 # return implies breaking the loop
1889 return BlockdevSnapshot(child)
1890 elif disk.dev_type == constants.LD_LV:
1891 r_dev = _RecursiveFindBD(disk)
1892 if r_dev is not None:
1893 # let's stay on the safe side and ask for the full size, for now
1894 return r_dev.Snapshot(disk.size)
1896 _Fail("Cannot find block device %s", disk)
1898 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1899 disk.unique_id, disk.dev_type)
1902 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1903 """Export a block device snapshot to a remote node.
1905 @type disk: L{objects.Disk}
1906 @param disk: the description of the disk to export
1907 @type dest_node: str
1908 @param dest_node: the destination node to export to
1909 @type instance: L{objects.Instance}
1910 @param instance: the instance object to whom the disk belongs
1911 @type cluster_name: str
1912 @param cluster_name: the cluster name, needed for SSH hostalias
1914 @param idx: the index of the disk in the instance's disk list,
1915 used to export to the OS scripts environment
1919 inst_os = OSFromDisk(instance.os)
1920 export_env = OSEnvironment(instance, inst_os)
1922 export_script = inst_os.export_script
1924 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1925 instance.name, int(time.time()))
1926 if not os.path.exists(constants.LOG_OS_DIR):
1927 os.mkdir(constants.LOG_OS_DIR, 0750)
1928 real_disk = _RecursiveFindBD(disk)
1929 if real_disk is None:
1930 _Fail("Block device '%s' is not set up", disk)
1934 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1935 export_env['EXPORT_INDEX'] = str(idx)
1937 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1938 destfile = disk.physical_id[1]
1940 # the target command is built out of three individual commands,
1941 # which are joined by pipes; we check each individual command for
1943 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
1944 inst_os.path, export_script, logfile)
1948 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1949 destdir, destdir, destfile)
1950 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1951 constants.GANETI_RUNAS,
1954 # all commands have been checked, so we're safe to combine them
1955 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1957 result = utils.RunCmd(["bash", "-c", command], env=export_env)
1960 _Fail("OS snapshot export command '%s' returned error: %s"
1961 " output: %s", command, result.fail_reason, result.output)
1964 def FinalizeExport(instance, snap_disks):
1965 """Write out the export configuration information.
1967 @type instance: L{objects.Instance}
1968 @param instance: the instance which we export, used for
1969 saving configuration
1970 @type snap_disks: list of L{objects.Disk}
1971 @param snap_disks: list of snapshot block devices, which
1972 will be used to get the actual name of the dump file
1977 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1978 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1980 config = objects.SerializableConfigParser()
1982 config.add_section(constants.INISECT_EXP)
1983 config.set(constants.INISECT_EXP, 'version', '0')
1984 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1985 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1986 config.set(constants.INISECT_EXP, 'os', instance.os)
1987 config.set(constants.INISECT_EXP, 'compression', 'gzip')
1989 config.add_section(constants.INISECT_INS)
1990 config.set(constants.INISECT_INS, 'name', instance.name)
1991 config.set(constants.INISECT_INS, 'memory', '%d' %
1992 instance.beparams[constants.BE_MEMORY])
1993 config.set(constants.INISECT_INS, 'vcpus', '%d' %
1994 instance.beparams[constants.BE_VCPUS])
1995 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1998 for nic_count, nic in enumerate(instance.nics):
2000 config.set(constants.INISECT_INS, 'nic%d_mac' %
2001 nic_count, '%s' % nic.mac)
2002 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2003 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
2005 # TODO: redundant: on load can read nics until it doesn't exist
2006 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2009 for disk_count, disk in enumerate(snap_disks):
2012 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2013 ('%s' % disk.iv_name))
2014 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2015 ('%s' % disk.physical_id[1]))
2016 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2019 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2021 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
2022 data=config.Dumps())
2023 shutil.rmtree(finaldestdir, True)
2024 shutil.move(destdir, finaldestdir)
2027 def ExportInfo(dest):
2028 """Get export configuration information.
2031 @param dest: directory containing the export
2033 @rtype: L{objects.SerializableConfigParser}
2034 @return: a serializable config file containing the
2038 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
2040 config = objects.SerializableConfigParser()
2043 if (not config.has_section(constants.INISECT_EXP) or
2044 not config.has_section(constants.INISECT_INS)):
2045 _Fail("Export info file doesn't have the required fields")
2047 return config.Dumps()
2050 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
2051 """Import an os image into an instance.
2053 @type instance: L{objects.Instance}
2054 @param instance: instance to import the disks into
2055 @type src_node: string
2056 @param src_node: source node for the disk images
2057 @type src_images: list of string
2058 @param src_images: absolute paths of the disk images
2059 @rtype: list of boolean
2060 @return: each boolean represent the success of importing the n-th disk
2063 inst_os = OSFromDisk(instance.os)
2064 import_env = OSEnvironment(instance, inst_os)
2065 import_script = inst_os.import_script
2067 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2068 instance.name, int(time.time()))
2069 if not os.path.exists(constants.LOG_OS_DIR):
2070 os.mkdir(constants.LOG_OS_DIR, 0750)
2073 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2074 import_script, logfile)
2077 for idx, image in enumerate(src_images):
2079 destcmd = utils.BuildShellCmd('cat %s', image)
2080 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2081 constants.GANETI_RUNAS,
2083 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2084 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2085 import_env['IMPORT_INDEX'] = str(idx)
2086 result = utils.RunCmd(command, env=import_env)
2088 logging.error("Disk import command '%s' returned error: %s"
2089 " output: %s", command, result.fail_reason,
2091 final_result.append("error importing disk %d: %s, %s" %
2092 (idx, result.fail_reason, result.output[-100]))
2095 _Fail("; ".join(final_result), log=False)
2099 """Return a list of exports currently available on this machine.
2102 @return: list of the exports
2105 if os.path.isdir(constants.EXPORT_DIR):
2106 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2108 _Fail("No exports directory")
2111 def RemoveExport(export):
2112 """Remove an existing export from the node.
2115 @param export: the name of the export to remove
2119 target = os.path.join(constants.EXPORT_DIR, export)
2122 shutil.rmtree(target)
2123 except EnvironmentError, err:
2124 _Fail("Error while removing the export: %s", err, exc=True)
2127 def BlockdevRename(devlist):
2128 """Rename a list of block devices.
2130 @type devlist: list of tuples
2131 @param devlist: list of tuples of the form (disk,
2132 new_logical_id, new_physical_id); disk is an
2133 L{objects.Disk} object describing the current disk,
2134 and new logical_id/physical_id is the name we
2137 @return: True if all renames succeeded, False otherwise
2142 for disk, unique_id in devlist:
2143 dev = _RecursiveFindBD(disk)
2145 msgs.append("Can't find device %s in rename" % str(disk))
2149 old_rpath = dev.dev_path
2150 dev.Rename(unique_id)
2151 new_rpath = dev.dev_path
2152 if old_rpath != new_rpath:
2153 DevCacheManager.RemoveCache(old_rpath)
2154 # FIXME: we should add the new cache information here, like:
2155 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2156 # but we don't have the owner here - maybe parse from existing
2157 # cache? for now, we only lose lvm data when we rename, which
2158 # is less critical than DRBD or MD
2159 except errors.BlockDeviceError, err:
2160 msgs.append("Can't rename device '%s' to '%s': %s" %
2161 (dev, unique_id, err))
2162 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2165 _Fail("; ".join(msgs))
2168 def _TransformFileStorageDir(file_storage_dir):
2169 """Checks whether given file_storage_dir is valid.
2171 Checks wheter the given file_storage_dir is within the cluster-wide
2172 default file_storage_dir stored in SimpleStore. Only paths under that
2173 directory are allowed.
2175 @type file_storage_dir: str
2176 @param file_storage_dir: the path to check
2178 @return: the normalized path if valid, None otherwise
2182 file_storage_dir = os.path.normpath(file_storage_dir)
2183 base_file_storage_dir = cfg.GetFileStorageDir()
2184 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2185 base_file_storage_dir):
2186 _Fail("File storage directory '%s' is not under base file"
2187 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2188 return file_storage_dir
2191 def CreateFileStorageDir(file_storage_dir):
2192 """Create file storage directory.
2194 @type file_storage_dir: str
2195 @param file_storage_dir: directory to create
2198 @return: tuple with first element a boolean indicating wheter dir
2199 creation was successful or not
2202 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2203 if os.path.exists(file_storage_dir):
2204 if not os.path.isdir(file_storage_dir):
2205 _Fail("Specified storage dir '%s' is not a directory",
2209 os.makedirs(file_storage_dir, 0750)
2210 except OSError, err:
2211 _Fail("Cannot create file storage directory '%s': %s",
2212 file_storage_dir, err, exc=True)
2215 def RemoveFileStorageDir(file_storage_dir):
2216 """Remove file storage directory.
2218 Remove it only if it's empty. If not log an error and return.
2220 @type file_storage_dir: str
2221 @param file_storage_dir: the directory we should cleanup
2222 @rtype: tuple (success,)
2223 @return: tuple of one element, C{success}, denoting
2224 whether the operation was successful
2227 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2228 if os.path.exists(file_storage_dir):
2229 if not os.path.isdir(file_storage_dir):
2230 _Fail("Specified Storage directory '%s' is not a directory",
2232 # deletes dir only if empty, otherwise we want to fail the rpc call
2234 os.rmdir(file_storage_dir)
2235 except OSError, err:
2236 _Fail("Cannot remove file storage directory '%s': %s",
2237 file_storage_dir, err)
2240 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2241 """Rename the file storage directory.
2243 @type old_file_storage_dir: str
2244 @param old_file_storage_dir: the current path
2245 @type new_file_storage_dir: str
2246 @param new_file_storage_dir: the name we should rename to
2247 @rtype: tuple (success,)
2248 @return: tuple of one element, C{success}, denoting
2249 whether the operation was successful
2252 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2253 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2254 if not os.path.exists(new_file_storage_dir):
2255 if os.path.isdir(old_file_storage_dir):
2257 os.rename(old_file_storage_dir, new_file_storage_dir)
2258 except OSError, err:
2259 _Fail("Cannot rename '%s' to '%s': %s",
2260 old_file_storage_dir, new_file_storage_dir, err)
2262 _Fail("Specified storage dir '%s' is not a directory",
2263 old_file_storage_dir)
2265 if os.path.exists(old_file_storage_dir):
2266 _Fail("Cannot rename '%s' to '%s': both locations exist",
2267 old_file_storage_dir, new_file_storage_dir)
2270 def _EnsureJobQueueFile(file_name):
2271 """Checks whether the given filename is in the queue directory.
2273 @type file_name: str
2274 @param file_name: the file name we should check
2276 @raises RPCFail: if the file is not valid
2279 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2280 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2283 _Fail("Passed job queue file '%s' does not belong to"
2284 " the queue directory '%s'", file_name, queue_dir)
2287 def JobQueueUpdate(file_name, content):
2288 """Updates a file in the queue directory.
2290 This is just a wrapper over L{utils.WriteFile}, with proper
2293 @type file_name: str
2294 @param file_name: the job file name
2296 @param content: the new job contents
2298 @return: the success of the operation
2301 _EnsureJobQueueFile(file_name)
2303 # Write and replace the file atomically
2304 utils.WriteFile(file_name, data=_Decompress(content))
2307 def JobQueueRename(old, new):
2308 """Renames a job queue file.
2310 This is just a wrapper over os.rename with proper checking.
2313 @param old: the old (actual) file name
2315 @param new: the desired file name
2317 @return: the success of the operation and payload
2320 _EnsureJobQueueFile(old)
2321 _EnsureJobQueueFile(new)
2323 utils.RenameFile(old, new, mkdir=True)
2326 def JobQueueSetDrainFlag(drain_flag):
2327 """Set the drain flag for the queue.
2329 This will set or unset the queue drain flag.
2331 @type drain_flag: boolean
2332 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2334 @return: always True, None
2335 @warning: the function always returns True
2339 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2341 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2344 def BlockdevClose(instance_name, disks):
2345 """Closes the given block devices.
2347 This means they will be switched to secondary mode (in case of
2350 @param instance_name: if the argument is not empty, the symlinks
2351 of this instance will be removed
2352 @type disks: list of L{objects.Disk}
2353 @param disks: the list of disks to be closed
2354 @rtype: tuple (success, message)
2355 @return: a tuple of success and message, where success
2356 indicates the succes of the operation, and message
2357 which will contain the error details in case we
2363 rd = _RecursiveFindBD(cf)
2365 _Fail("Can't find device %s", cf)
2372 except errors.BlockDeviceError, err:
2373 msg.append(str(err))
2375 _Fail("Can't make devices secondary: %s", ",".join(msg))
2378 _RemoveBlockDevLinks(instance_name, disks)
2381 def ValidateHVParams(hvname, hvparams):
2382 """Validates the given hypervisor parameters.
2384 @type hvname: string
2385 @param hvname: the hypervisor name
2386 @type hvparams: dict
2387 @param hvparams: the hypervisor parameters to be validated
2392 hv_type = hypervisor.GetHypervisor(hvname)
2393 hv_type.ValidateParameters(hvparams)
2394 except errors.HypervisorError, err:
2395 _Fail(str(err), log=False)
2399 """Demotes the current node from master candidate role.
2402 # try to ensure we're not the master by mistake
2403 master, myself = ssconf.GetMasterAndMyself()
2404 if master == myself:
2405 _Fail("ssconf status shows I'm the master node, will not demote")
2406 pid_file = utils.DaemonPidFileName(constants.MASTERD)
2407 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2408 _Fail("The master daemon is running, will not demote")
2410 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2411 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2412 except EnvironmentError, err:
2413 if err.errno != errno.ENOENT:
2414 _Fail("Error while backing up cluster file: %s", err, exc=True)
2415 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2418 def _FindDisks(nodes_ip, disks):
2419 """Sets the physical ID on disks and returns the block devices.
2422 # set the correct physical ID
2423 my_name = utils.HostInfo().name
2425 cf.SetPhysicalID(my_name, nodes_ip)
2430 rd = _RecursiveFindBD(cf)
2432 _Fail("Can't find device %s", cf)
2437 def DrbdDisconnectNet(nodes_ip, disks):
2438 """Disconnects the network on a list of drbd devices.
2441 bdevs = _FindDisks(nodes_ip, disks)
2447 except errors.BlockDeviceError, err:
2448 _Fail("Can't change network configuration to standalone mode: %s",
2452 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2453 """Attaches the network on a list of drbd devices.
2456 bdevs = _FindDisks(nodes_ip, disks)
2459 for idx, rd in enumerate(bdevs):
2461 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2462 except EnvironmentError, err:
2463 _Fail("Can't create symlink: %s", err)
2464 # reconnect disks, switch to new master configuration and if
2465 # needed primary mode
2468 rd.AttachNet(multimaster)
2469 except errors.BlockDeviceError, err:
2470 _Fail("Can't change network configuration: %s", err)
2471 # wait until the disks are connected; we need to retry the re-attach
2472 # if the device becomes standalone, as this might happen if the one
2473 # node disconnects and reconnects in a different mode before the
2474 # other node reconnects; in this case, one or both of the nodes will
2475 # decide it has wrong configuration and switch to standalone
2476 RECONNECT_TIMEOUT = 2 * 60
2477 sleep_time = 0.100 # start with 100 miliseconds
2478 timeout_limit = time.time() + RECONNECT_TIMEOUT
2479 while time.time() < timeout_limit:
2480 all_connected = True
2482 stats = rd.GetProcStatus()
2483 if not (stats.is_connected or stats.is_in_resync):
2484 all_connected = False
2485 if stats.is_standalone:
2486 # peer had different config info and this node became
2487 # standalone, even though this should not happen with the
2488 # new staged way of changing disk configs
2490 rd.AttachNet(multimaster)
2491 except errors.BlockDeviceError, err:
2492 _Fail("Can't change network configuration: %s", err)
2495 time.sleep(sleep_time)
2496 sleep_time = min(5, sleep_time * 1.5)
2497 if not all_connected:
2498 _Fail("Timeout in disk reconnecting")
2500 # change to primary mode
2504 except errors.BlockDeviceError, err:
2505 _Fail("Can't change to primary mode: %s", err)
2508 def DrbdWaitSync(nodes_ip, disks):
2509 """Wait until DRBDs have synchronized.
2512 bdevs = _FindDisks(nodes_ip, disks)
2517 stats = rd.GetProcStatus()
2518 if not (stats.is_connected or stats.is_in_resync):
2519 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2520 alldone = alldone and (not stats.is_in_resync)
2521 if stats.sync_percent is not None:
2522 min_resync = min(min_resync, stats.sync_percent)
2524 return (alldone, min_resync)
2527 def PowercycleNode(hypervisor_type):
2528 """Hard-powercycle the node.
2530 Because we need to return first, and schedule the powercycle in the
2531 background, we won't be able to report failures nicely.
2534 hyper = hypervisor.GetHypervisor(hypervisor_type)
2538 # if we can't fork, we'll pretend that we're in the child process
2541 return "Reboot scheduled in 5 seconds"
2543 hyper.PowercycleNode()
2546 class HooksRunner(object):
2549 This class is instantiated on the node side (ganeti-noded) and not
2553 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2555 def __init__(self, hooks_base_dir=None):
2556 """Constructor for hooks runner.
2558 @type hooks_base_dir: str or None
2559 @param hooks_base_dir: if not None, this overrides the
2560 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2563 if hooks_base_dir is None:
2564 hooks_base_dir = constants.HOOKS_BASE_DIR
2565 self._BASE_DIR = hooks_base_dir
2568 def ExecHook(script, env):
2569 """Exec one hook script.
2572 @param script: the full path to the script
2574 @param env: the environment with which to exec the script
2575 @rtype: tuple (success, message)
2576 @return: a tuple of success and message, where success
2577 indicates the succes of the operation, and message
2578 which will contain the error details in case we
2582 # exec the process using subprocess and log the output
2585 fdstdin = open("/dev/null", "r")
2586 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2587 stderr=subprocess.STDOUT, close_fds=True,
2588 shell=False, cwd="/", env=env)
2591 output = child.stdout.read(4096)
2592 child.stdout.close()
2593 except EnvironmentError, err:
2594 output += "Hook script error: %s" % str(err)
2598 result = child.wait()
2600 except EnvironmentError, err:
2601 if err.errno == errno.EINTR:
2605 # try not to leak fds
2606 for fd in (fdstdin, ):
2610 except EnvironmentError, err:
2611 # just log the error
2612 #logging.exception("Error while closing fd %s", fd)
2615 return result == 0, utils.SafeEncode(output.strip())
2617 def RunHooks(self, hpath, phase, env):
2618 """Run the scripts in the hooks directory.
2621 @param hpath: the path to the hooks directory which
2624 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2625 L{constants.HOOKS_PHASE_POST}
2627 @param env: dictionary with the environment for the hook
2629 @return: list of 3-element tuples:
2631 - script result, either L{constants.HKR_SUCCESS} or
2632 L{constants.HKR_FAIL}
2633 - output of the script
2635 @raise errors.ProgrammerError: for invalid input
2639 if phase == constants.HOOKS_PHASE_PRE:
2641 elif phase == constants.HOOKS_PHASE_POST:
2644 _Fail("Unknown hooks phase '%s'", phase)
2648 subdir = "%s-%s.d" % (hpath, suffix)
2649 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2651 dir_contents = utils.ListVisibleFiles(dir_name)
2653 # FIXME: must log output in case of failures
2656 # we use the standard python sort order,
2657 # so 00name is the recommended naming scheme
2659 for relname in dir_contents:
2660 fname = os.path.join(dir_name, relname)
2661 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2662 self.RE_MASK.match(relname) is not None):
2663 rrval = constants.HKR_SKIP
2666 result, output = self.ExecHook(fname, env)
2668 rrval = constants.HKR_FAIL
2670 rrval = constants.HKR_SUCCESS
2671 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2676 class IAllocatorRunner(object):
2677 """IAllocator runner.
2679 This class is instantiated on the node side (ganeti-noded) and not on
2683 def Run(self, name, idata):
2684 """Run an iallocator script.
2687 @param name: the iallocator script name
2689 @param idata: the allocator input data
2692 @return: two element tuple of:
2694 - either error message or stdout of allocator (for success)
2697 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2699 if alloc_script is None:
2700 _Fail("iallocator module '%s' not found in the search path", name)
2702 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2706 result = utils.RunCmd([alloc_script, fin_name])
2708 _Fail("iallocator module '%s' failed: %s, output '%s'",
2709 name, result.fail_reason, result.output)
2713 return result.stdout
2716 class DevCacheManager(object):
2717 """Simple class for managing a cache of block device information.
2720 _DEV_PREFIX = "/dev/"
2721 _ROOT_DIR = constants.BDEV_CACHE_DIR
2724 def _ConvertPath(cls, dev_path):
2725 """Converts a /dev/name path to the cache file name.
2727 This replaces slashes with underscores and strips the /dev
2728 prefix. It then returns the full path to the cache file.
2731 @param dev_path: the C{/dev/} path name
2733 @return: the converted path name
2736 if dev_path.startswith(cls._DEV_PREFIX):
2737 dev_path = dev_path[len(cls._DEV_PREFIX):]
2738 dev_path = dev_path.replace("/", "_")
2739 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2743 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2744 """Updates the cache information for a given device.
2747 @param dev_path: the pathname of the device
2749 @param owner: the owner (instance name) of the device
2750 @type on_primary: bool
2751 @param on_primary: whether this is the primary
2754 @param iv_name: the instance-visible name of the
2755 device, as in objects.Disk.iv_name
2760 if dev_path is None:
2761 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2763 fpath = cls._ConvertPath(dev_path)
2769 iv_name = "not_visible"
2770 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2772 utils.WriteFile(fpath, data=fdata)
2773 except EnvironmentError, err:
2774 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2777 def RemoveCache(cls, dev_path):
2778 """Remove data for a dev_path.
2780 This is just a wrapper over L{utils.RemoveFile} with a converted
2781 path name and logging.
2784 @param dev_path: the pathname of the device
2789 if dev_path is None:
2790 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2792 fpath = cls._ConvertPath(dev_path)
2794 utils.RemoveFile(fpath)
2795 except EnvironmentError, err:
2796 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)