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 running_instances = GetInstanceList([hv_name])
975 iname = instance.name
977 if iname not in running_instances:
978 logging.info("Instance %s not running, doing nothing", iname)
981 hyper = hypervisor.GetHypervisor(hv_name)
983 hyper.StopInstance(instance)
984 except errors.HypervisorError, err:
985 _Fail("Failed to stop instance %s: %s", iname, err)
987 # test every 10secs for 2min
991 if instance.name not in GetInstanceList([hv_name]):
995 # the shutdown did not succeed
996 logging.error("Shutdown of '%s' unsuccessful, using destroy", iname)
999 hyper.StopInstance(instance, force=True)
1000 except errors.HypervisorError, err:
1001 _Fail("Failed to force stop instance %s: %s", iname, err)
1004 if instance.name in GetInstanceList([hv_name]):
1005 _Fail("Could not shutdown instance %s even by destroy", iname)
1007 _RemoveBlockDevLinks(iname, instance.disks)
1010 def InstanceReboot(instance, reboot_type):
1011 """Reboot an instance.
1013 @type instance: L{objects.Instance}
1014 @param instance: the instance object to reboot
1015 @type reboot_type: str
1016 @param reboot_type: the type of reboot, one the following
1018 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1019 instance OS, do not recreate the VM
1020 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1021 restart the VM (at the hypervisor level)
1022 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1023 not accepted here, since that mode is handled differently, in
1024 cmdlib, and translates into full stop and start of the
1025 instance (instead of a call_instance_reboot RPC)
1029 running_instances = GetInstanceList([instance.hypervisor])
1031 if instance.name not in running_instances:
1032 _Fail("Cannot reboot instance %s that is not running", instance.name)
1034 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1035 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1037 hyper.RebootInstance(instance)
1038 except errors.HypervisorError, err:
1039 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1040 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1042 InstanceShutdown(instance)
1043 return StartInstance(instance)
1044 except errors.HypervisorError, err:
1045 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1047 _Fail("Invalid reboot_type received: %s", reboot_type)
1050 def MigrationInfo(instance):
1051 """Gather information about an instance to be migrated.
1053 @type instance: L{objects.Instance}
1054 @param instance: the instance definition
1057 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1059 info = hyper.MigrationInfo(instance)
1060 except errors.HypervisorError, err:
1061 _Fail("Failed to fetch migration information: %s", err, exc=True)
1065 def AcceptInstance(instance, info, target):
1066 """Prepare the node to accept an instance.
1068 @type instance: L{objects.Instance}
1069 @param instance: the instance definition
1070 @type info: string/data (opaque)
1071 @param info: migration information, from the source node
1072 @type target: string
1073 @param target: target host (usually ip), on this node
1076 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1078 hyper.AcceptInstance(instance, info, target)
1079 except errors.HypervisorError, err:
1080 _Fail("Failed to accept instance: %s", err, exc=True)
1083 def FinalizeMigration(instance, info, success):
1084 """Finalize any preparation to accept an instance.
1086 @type instance: L{objects.Instance}
1087 @param instance: the instance definition
1088 @type info: string/data (opaque)
1089 @param info: migration information, from the source node
1090 @type success: boolean
1091 @param success: whether the migration was a success or a failure
1094 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1096 hyper.FinalizeMigration(instance, info, success)
1097 except errors.HypervisorError, err:
1098 _Fail("Failed to finalize migration: %s", err, exc=True)
1101 def MigrateInstance(instance, target, live):
1102 """Migrates an instance to another node.
1104 @type instance: L{objects.Instance}
1105 @param instance: the instance definition
1106 @type target: string
1107 @param target: the target node name
1109 @param live: whether the migration should be done live or not (the
1110 interpretation of this parameter is left to the hypervisor)
1112 @return: a tuple of (success, msg) where:
1113 - succes is a boolean denoting the success/failure of the operation
1114 - msg is a string with details in case of failure
1117 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1120 hyper.MigrateInstance(instance.name, target, live)
1121 except errors.HypervisorError, err:
1122 _Fail("Failed to migrate instance: %s", err, exc=True)
1125 def BlockdevCreate(disk, size, owner, on_primary, info):
1126 """Creates a block device for an instance.
1128 @type disk: L{objects.Disk}
1129 @param disk: the object describing the disk we should create
1131 @param size: the size of the physical underlying device, in MiB
1133 @param owner: the name of the instance for which disk is created,
1134 used for device cache data
1135 @type on_primary: boolean
1136 @param on_primary: indicates if it is the primary node or not
1138 @param info: string that will be sent to the physical device
1139 creation, used for example to set (LVM) tags on LVs
1141 @return: the new unique_id of the device (this can sometime be
1142 computed only after creation), or None. On secondary nodes,
1143 it's not required to return anything.
1148 for child in disk.children:
1150 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1151 except errors.BlockDeviceError, err:
1152 _Fail("Can't assemble device %s: %s", child, err)
1153 if on_primary or disk.AssembleOnSecondary():
1154 # we need the children open in case the device itself has to
1158 except errors.BlockDeviceError, err:
1159 _Fail("Can't make child '%s' read-write: %s", child, err)
1163 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1164 except errors.BlockDeviceError, err:
1165 _Fail("Can't create block device: %s", err)
1167 if on_primary or disk.AssembleOnSecondary():
1170 except errors.BlockDeviceError, err:
1171 _Fail("Can't assemble device after creation, unusual event: %s", err)
1172 device.SetSyncSpeed(constants.SYNC_SPEED)
1173 if on_primary or disk.OpenOnSecondary():
1175 device.Open(force=True)
1176 except errors.BlockDeviceError, err:
1177 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1178 DevCacheManager.UpdateCache(device.dev_path, owner,
1179 on_primary, disk.iv_name)
1181 device.SetInfo(info)
1183 return device.unique_id
1186 def BlockdevRemove(disk):
1187 """Remove a block device.
1189 @note: This is intended to be called recursively.
1191 @type disk: L{objects.Disk}
1192 @param disk: the disk object we should remove
1194 @return: the success of the operation
1199 rdev = _RecursiveFindBD(disk)
1200 except errors.BlockDeviceError, err:
1201 # probably can't attach
1202 logging.info("Can't attach to device %s in remove", disk)
1204 if rdev is not None:
1205 r_path = rdev.dev_path
1208 except errors.BlockDeviceError, err:
1209 msgs.append(str(err))
1211 DevCacheManager.RemoveCache(r_path)
1214 for child in disk.children:
1216 BlockdevRemove(child)
1217 except RPCFail, err:
1218 msgs.append(str(err))
1221 _Fail("; ".join(msgs))
1224 def _RecursiveAssembleBD(disk, owner, as_primary):
1225 """Activate a block device for an instance.
1227 This is run on the primary and secondary nodes for an instance.
1229 @note: this function is called recursively.
1231 @type disk: L{objects.Disk}
1232 @param disk: the disk we try to assemble
1234 @param owner: the name of the instance which owns the disk
1235 @type as_primary: boolean
1236 @param as_primary: if we should make the block device
1239 @return: the assembled device or None (in case no device
1241 @raise errors.BlockDeviceError: in case there is an error
1242 during the activation of the children or the device
1248 mcn = disk.ChildrenNeeded()
1250 mcn = 0 # max number of Nones allowed
1252 mcn = len(disk.children) - mcn # max number of Nones
1253 for chld_disk in disk.children:
1255 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1256 except errors.BlockDeviceError, err:
1257 if children.count(None) >= mcn:
1260 logging.error("Error in child activation (but continuing): %s",
1262 children.append(cdev)
1264 if as_primary or disk.AssembleOnSecondary():
1265 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1266 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1268 if as_primary or disk.OpenOnSecondary():
1270 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1271 as_primary, disk.iv_name)
1278 def BlockdevAssemble(disk, owner, as_primary):
1279 """Activate a block device for an instance.
1281 This is a wrapper over _RecursiveAssembleBD.
1283 @rtype: str or boolean
1284 @return: a C{/dev/...} path for primary nodes, and
1285 C{True} for secondary nodes
1289 result = _RecursiveAssembleBD(disk, owner, as_primary)
1290 if isinstance(result, bdev.BlockDev):
1291 result = result.dev_path
1292 except errors.BlockDeviceError, err:
1293 _Fail("Error while assembling disk: %s", err, exc=True)
1298 def BlockdevShutdown(disk):
1299 """Shut down a block device.
1301 First, if the device is assembled (Attach() is successful), then
1302 the device is shutdown. Then the children of the device are
1305 This function is called recursively. Note that we don't cache the
1306 children or such, as oppossed to assemble, shutdown of different
1307 devices doesn't require that the upper device was active.
1309 @type disk: L{objects.Disk}
1310 @param disk: the description of the disk we should
1316 r_dev = _RecursiveFindBD(disk)
1317 if r_dev is not None:
1318 r_path = r_dev.dev_path
1321 DevCacheManager.RemoveCache(r_path)
1322 except errors.BlockDeviceError, err:
1323 msgs.append(str(err))
1326 for child in disk.children:
1328 BlockdevShutdown(child)
1329 except RPCFail, err:
1330 msgs.append(str(err))
1333 _Fail("; ".join(msgs))
1336 def BlockdevAddchildren(parent_cdev, new_cdevs):
1337 """Extend a mirrored block device.
1339 @type parent_cdev: L{objects.Disk}
1340 @param parent_cdev: the disk to which we should add children
1341 @type new_cdevs: list of L{objects.Disk}
1342 @param new_cdevs: the list of children which we should add
1346 parent_bdev = _RecursiveFindBD(parent_cdev)
1347 if parent_bdev is None:
1348 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1349 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1350 if new_bdevs.count(None) > 0:
1351 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1352 parent_bdev.AddChildren(new_bdevs)
1355 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1356 """Shrink a mirrored block device.
1358 @type parent_cdev: L{objects.Disk}
1359 @param parent_cdev: the disk from which we should remove children
1360 @type new_cdevs: list of L{objects.Disk}
1361 @param new_cdevs: the list of children which we should remove
1365 parent_bdev = _RecursiveFindBD(parent_cdev)
1366 if parent_bdev is None:
1367 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1369 for disk in new_cdevs:
1370 rpath = disk.StaticDevPath()
1372 bd = _RecursiveFindBD(disk)
1374 _Fail("Can't find device %s while removing children", disk)
1376 devs.append(bd.dev_path)
1379 parent_bdev.RemoveChildren(devs)
1382 def BlockdevGetmirrorstatus(disks):
1383 """Get the mirroring status of a list of devices.
1385 @type disks: list of L{objects.Disk}
1386 @param disks: the list of disks which we should query
1389 a list of (mirror_done, estimated_time) tuples, which
1390 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1391 @raise errors.BlockDeviceError: if any of the disks cannot be
1397 rbd = _RecursiveFindBD(dsk)
1399 _Fail("Can't find device %s", dsk)
1401 stats.append(rbd.CombinedSyncStatus())
1406 def _RecursiveFindBD(disk):
1407 """Check if a device is activated.
1409 If so, return information about the real device.
1411 @type disk: L{objects.Disk}
1412 @param disk: the disk object we need to find
1414 @return: None if the device can't be found,
1415 otherwise the device instance
1420 for chdisk in disk.children:
1421 children.append(_RecursiveFindBD(chdisk))
1423 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1426 def BlockdevFind(disk):
1427 """Check if a device is activated.
1429 If it is, return information about the real device.
1431 @type disk: L{objects.Disk}
1432 @param disk: the disk to find
1433 @rtype: None or objects.BlockDevStatus
1434 @return: None if the disk cannot be found, otherwise a the current
1439 rbd = _RecursiveFindBD(disk)
1440 except errors.BlockDeviceError, err:
1441 _Fail("Failed to find device: %s", err, exc=True)
1446 return rbd.GetSyncStatus()
1449 def BlockdevGetsize(disks):
1450 """Computes the size of the given disks.
1452 If a disk is not found, returns None instead.
1454 @type disks: list of L{objects.Disk}
1455 @param disks: the list of disk to compute the size for
1457 @return: list with elements None if the disk cannot be found,
1464 rbd = _RecursiveFindBD(cf)
1465 except errors.BlockDeviceError, err:
1471 result.append(rbd.GetActualSize())
1475 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1476 """Export a block device to a remote node.
1478 @type disk: L{objects.Disk}
1479 @param disk: the description of the disk to export
1480 @type dest_node: str
1481 @param dest_node: the destination node to export to
1482 @type dest_path: str
1483 @param dest_path: the destination path on the target node
1484 @type cluster_name: str
1485 @param cluster_name: the cluster name, needed for SSH hostalias
1489 real_disk = _RecursiveFindBD(disk)
1490 if real_disk is None:
1491 _Fail("Block device '%s' is not set up", disk)
1495 # the block size on the read dd is 1MiB to match our units
1496 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1497 "dd if=%s bs=1048576 count=%s",
1498 real_disk.dev_path, str(disk.size))
1500 # we set here a smaller block size as, due to ssh buffering, more
1501 # than 64-128k will mostly ignored; we use nocreat to fail if the
1502 # device is not already there or we pass a wrong path; we use
1503 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1504 # to not buffer too much memory; this means that at best, we flush
1505 # every 64k, which will not be very fast
1506 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1507 " oflag=dsync", dest_path)
1509 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1510 constants.GANETI_RUNAS,
1513 # all commands have been checked, so we're safe to combine them
1514 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1516 result = utils.RunCmd(["bash", "-c", command])
1519 _Fail("Disk copy command '%s' returned error: %s"
1520 " output: %s", command, result.fail_reason, result.output)
1523 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1524 """Write a file to the filesystem.
1526 This allows the master to overwrite(!) a file. It will only perform
1527 the operation if the file belongs to a list of configuration files.
1529 @type file_name: str
1530 @param file_name: the target file name
1532 @param data: the new contents of the file
1534 @param mode: the mode to give the file (can be None)
1536 @param uid: the owner of the file (can be -1 for default)
1538 @param gid: the group of the file (can be -1 for default)
1540 @param atime: the atime to set on the file (can be None)
1542 @param mtime: the mtime to set on the file (can be None)
1546 if not os.path.isabs(file_name):
1547 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1549 if file_name not in _ALLOWED_UPLOAD_FILES:
1550 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1553 raw_data = _Decompress(data)
1555 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1556 atime=atime, mtime=mtime)
1559 def WriteSsconfFiles(values):
1560 """Update all ssconf files.
1562 Wrapper around the SimpleStore.WriteFiles.
1565 ssconf.SimpleStore().WriteFiles(values)
1568 def _ErrnoOrStr(err):
1569 """Format an EnvironmentError exception.
1571 If the L{err} argument has an errno attribute, it will be looked up
1572 and converted into a textual C{E...} description. Otherwise the
1573 string representation of the error will be returned.
1575 @type err: L{EnvironmentError}
1576 @param err: the exception to format
1579 if hasattr(err, 'errno'):
1580 detail = errno.errorcode[err.errno]
1586 def _OSOndiskAPIVersion(name, os_dir):
1587 """Compute and return the API version of a given OS.
1589 This function will try to read the API version of the OS given by
1590 the 'name' parameter and residing in the 'os_dir' directory.
1593 @param name: the OS name we should look for
1595 @param os_dir: the directory inwhich we should look for the OS
1597 @return: tuple (status, data) with status denoting the validity and
1598 data holding either the vaid versions or an error message
1601 api_file = os.path.sep.join([os_dir, "ganeti_api_version"])
1604 st = os.stat(api_file)
1605 except EnvironmentError, err:
1606 return False, ("Required file 'ganeti_api_version' file not"
1607 " found under path %s: %s" % (os_dir, _ErrnoOrStr(err)))
1609 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1610 return False, ("File 'ganeti_api_version' file at %s is not"
1611 " a regular file" % os_dir)
1614 api_versions = utils.ReadFile(api_file).splitlines()
1615 except EnvironmentError, err:
1616 return False, ("Error while reading the API version file at %s: %s" %
1617 (api_file, _ErrnoOrStr(err)))
1620 api_versions = [int(version.strip()) for version in api_versions]
1621 except (TypeError, ValueError), err:
1622 return False, ("API version(s) can't be converted to integer: %s" %
1625 return True, api_versions
1628 def DiagnoseOS(top_dirs=None):
1629 """Compute the validity for all OSes.
1631 @type top_dirs: list
1632 @param top_dirs: the list of directories in which to
1633 search (if not given defaults to
1634 L{constants.OS_SEARCH_PATH})
1635 @rtype: list of L{objects.OS}
1636 @return: a list of tuples (name, path, status, diagnose)
1637 for all (potential) OSes under all search paths, where:
1638 - name is the (potential) OS name
1639 - path is the full path to the OS
1640 - status True/False is the validity of the OS
1641 - diagnose is the error message for an invalid OS, otherwise empty
1644 if top_dirs is None:
1645 top_dirs = constants.OS_SEARCH_PATH
1648 for dir_name in top_dirs:
1649 if os.path.isdir(dir_name):
1651 f_names = utils.ListVisibleFiles(dir_name)
1652 except EnvironmentError, err:
1653 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1655 for name in f_names:
1656 os_path = os.path.sep.join([dir_name, name])
1657 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1662 result.append((name, os_path, status, diagnose))
1667 def _TryOSFromDisk(name, base_dir=None):
1668 """Create an OS instance from disk.
1670 This function will return an OS instance if the given name is a
1673 @type base_dir: string
1674 @keyword base_dir: Base directory containing OS installations.
1675 Defaults to a search in all the OS_SEARCH_PATH dirs.
1677 @return: success and either the OS instance if we find a valid one,
1681 if base_dir is None:
1682 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1684 return False, "Directory for OS %s not found in search path" % name
1686 os_dir = os.path.sep.join([base_dir, name])
1688 status, api_versions = _OSOndiskAPIVersion(name, os_dir)
1691 return status, api_versions
1693 if not constants.OS_API_VERSIONS.intersection(api_versions):
1694 return False, ("API version mismatch for path '%s': found %s, want %s." %
1695 (os_dir, api_versions, constants.OS_API_VERSIONS))
1697 # OS Scripts dictionary, we will populate it with the actual script names
1698 os_scripts = dict.fromkeys(constants.OS_SCRIPTS)
1700 for script in os_scripts:
1701 os_scripts[script] = os.path.sep.join([os_dir, script])
1704 st = os.stat(os_scripts[script])
1705 except EnvironmentError, err:
1706 return False, ("Script '%s' under path '%s' is missing (%s)" %
1707 (script, os_dir, _ErrnoOrStr(err)))
1709 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1710 return False, ("Script '%s' under path '%s' is not executable" %
1713 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1714 return False, ("Script '%s' under path '%s' is not a regular file" %
1717 os_obj = objects.OS(name=name, path=os_dir,
1718 create_script=os_scripts[constants.OS_SCRIPT_CREATE],
1719 export_script=os_scripts[constants.OS_SCRIPT_EXPORT],
1720 import_script=os_scripts[constants.OS_SCRIPT_IMPORT],
1721 rename_script=os_scripts[constants.OS_SCRIPT_RENAME],
1722 api_versions=api_versions)
1726 def OSFromDisk(name, base_dir=None):
1727 """Create an OS instance from disk.
1729 This function will return an OS instance if the given name is a
1730 valid OS name. Otherwise, it will raise an appropriate
1731 L{RPCFail} exception, detailing why this is not a valid OS.
1733 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1734 an exception but returns true/false status data.
1736 @type base_dir: string
1737 @keyword base_dir: Base directory containing OS installations.
1738 Defaults to a search in all the OS_SEARCH_PATH dirs.
1739 @rtype: L{objects.OS}
1740 @return: the OS instance if we find a valid one
1741 @raise RPCFail: if we don't find a valid OS
1744 status, payload = _TryOSFromDisk(name, base_dir)
1752 def OSEnvironment(instance, os, debug=0):
1753 """Calculate the environment for an os script.
1755 @type instance: L{objects.Instance}
1756 @param instance: target instance for the os script run
1757 @type os: L{objects.OS}
1758 @param os: operating system for which the environment is being built
1759 @type debug: integer
1760 @param debug: debug level (0 or 1, for OS Api 10)
1762 @return: dict of environment variables
1763 @raise errors.BlockDeviceError: if the block device
1768 api_version = max(constants.OS_API_VERSIONS.intersection(os.api_versions))
1769 result['OS_API_VERSION'] = '%d' % api_version
1770 result['INSTANCE_NAME'] = instance.name
1771 result['INSTANCE_OS'] = instance.os
1772 result['HYPERVISOR'] = instance.hypervisor
1773 result['DISK_COUNT'] = '%d' % len(instance.disks)
1774 result['NIC_COUNT'] = '%d' % len(instance.nics)
1775 result['DEBUG_LEVEL'] = '%d' % debug
1776 for idx, disk in enumerate(instance.disks):
1777 real_disk = _RecursiveFindBD(disk)
1778 if real_disk is None:
1779 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1782 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1783 result['DISK_%d_ACCESS' % idx] = disk.mode
1784 if constants.HV_DISK_TYPE in instance.hvparams:
1785 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1786 instance.hvparams[constants.HV_DISK_TYPE]
1787 if disk.dev_type in constants.LDS_BLOCK:
1788 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1789 elif disk.dev_type == constants.LD_FILE:
1790 result['DISK_%d_BACKEND_TYPE' % idx] = \
1791 'file:%s' % disk.physical_id[0]
1792 for idx, nic in enumerate(instance.nics):
1793 result['NIC_%d_MAC' % idx] = nic.mac
1795 result['NIC_%d_IP' % idx] = nic.ip
1796 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1797 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1798 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1799 if nic.nicparams[constants.NIC_LINK]:
1800 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1801 if constants.HV_NIC_TYPE in instance.hvparams:
1802 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1803 instance.hvparams[constants.HV_NIC_TYPE]
1805 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1806 for key, value in source.items():
1807 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1811 def BlockdevGrow(disk, amount):
1812 """Grow a stack of block devices.
1814 This function is called recursively, with the childrens being the
1815 first ones to resize.
1817 @type disk: L{objects.Disk}
1818 @param disk: the disk to be grown
1819 @rtype: (status, result)
1820 @return: a tuple with the status of the operation
1821 (True/False), and the errors message if status
1825 r_dev = _RecursiveFindBD(disk)
1827 _Fail("Cannot find block device %s", disk)
1831 except errors.BlockDeviceError, err:
1832 _Fail("Failed to grow block device: %s", err, exc=True)
1835 def BlockdevSnapshot(disk):
1836 """Create a snapshot copy of a block device.
1838 This function is called recursively, and the snapshot is actually created
1839 just for the leaf lvm backend device.
1841 @type disk: L{objects.Disk}
1842 @param disk: the disk to be snapshotted
1844 @return: snapshot disk path
1848 if len(disk.children) == 1:
1849 # only one child, let's recurse on it
1850 return BlockdevSnapshot(disk.children[0])
1852 # more than one child, choose one that matches
1853 for child in disk.children:
1854 if child.size == disk.size:
1855 # return implies breaking the loop
1856 return BlockdevSnapshot(child)
1857 elif disk.dev_type == constants.LD_LV:
1858 r_dev = _RecursiveFindBD(disk)
1859 if r_dev is not None:
1860 # let's stay on the safe side and ask for the full size, for now
1861 return r_dev.Snapshot(disk.size)
1863 _Fail("Cannot find block device %s", disk)
1865 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1866 disk.unique_id, disk.dev_type)
1869 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1870 """Export a block device snapshot to a remote node.
1872 @type disk: L{objects.Disk}
1873 @param disk: the description of the disk to export
1874 @type dest_node: str
1875 @param dest_node: the destination node to export to
1876 @type instance: L{objects.Instance}
1877 @param instance: the instance object to whom the disk belongs
1878 @type cluster_name: str
1879 @param cluster_name: the cluster name, needed for SSH hostalias
1881 @param idx: the index of the disk in the instance's disk list,
1882 used to export to the OS scripts environment
1886 inst_os = OSFromDisk(instance.os)
1887 export_env = OSEnvironment(instance, inst_os)
1889 export_script = inst_os.export_script
1891 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1892 instance.name, int(time.time()))
1893 if not os.path.exists(constants.LOG_OS_DIR):
1894 os.mkdir(constants.LOG_OS_DIR, 0750)
1895 real_disk = _RecursiveFindBD(disk)
1896 if real_disk is None:
1897 _Fail("Block device '%s' is not set up", disk)
1901 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1902 export_env['EXPORT_INDEX'] = str(idx)
1904 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1905 destfile = disk.physical_id[1]
1907 # the target command is built out of three individual commands,
1908 # which are joined by pipes; we check each individual command for
1910 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
1911 inst_os.path, export_script, logfile)
1915 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1916 destdir, destdir, destfile)
1917 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1918 constants.GANETI_RUNAS,
1921 # all commands have been checked, so we're safe to combine them
1922 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1924 result = utils.RunCmd(["bash", "-c", command], env=export_env)
1927 _Fail("OS snapshot export command '%s' returned error: %s"
1928 " output: %s", command, result.fail_reason, result.output)
1931 def FinalizeExport(instance, snap_disks):
1932 """Write out the export configuration information.
1934 @type instance: L{objects.Instance}
1935 @param instance: the instance which we export, used for
1936 saving configuration
1937 @type snap_disks: list of L{objects.Disk}
1938 @param snap_disks: list of snapshot block devices, which
1939 will be used to get the actual name of the dump file
1944 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1945 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
1947 config = objects.SerializableConfigParser()
1949 config.add_section(constants.INISECT_EXP)
1950 config.set(constants.INISECT_EXP, 'version', '0')
1951 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
1952 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
1953 config.set(constants.INISECT_EXP, 'os', instance.os)
1954 config.set(constants.INISECT_EXP, 'compression', 'gzip')
1956 config.add_section(constants.INISECT_INS)
1957 config.set(constants.INISECT_INS, 'name', instance.name)
1958 config.set(constants.INISECT_INS, 'memory', '%d' %
1959 instance.beparams[constants.BE_MEMORY])
1960 config.set(constants.INISECT_INS, 'vcpus', '%d' %
1961 instance.beparams[constants.BE_VCPUS])
1962 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
1965 for nic_count, nic in enumerate(instance.nics):
1967 config.set(constants.INISECT_INS, 'nic%d_mac' %
1968 nic_count, '%s' % nic.mac)
1969 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
1970 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
1972 # TODO: redundant: on load can read nics until it doesn't exist
1973 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
1976 for disk_count, disk in enumerate(snap_disks):
1979 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
1980 ('%s' % disk.iv_name))
1981 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
1982 ('%s' % disk.physical_id[1]))
1983 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
1986 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
1988 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
1989 data=config.Dumps())
1990 shutil.rmtree(finaldestdir, True)
1991 shutil.move(destdir, finaldestdir)
1994 def ExportInfo(dest):
1995 """Get export configuration information.
1998 @param dest: directory containing the export
2000 @rtype: L{objects.SerializableConfigParser}
2001 @return: a serializable config file containing the
2005 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
2007 config = objects.SerializableConfigParser()
2010 if (not config.has_section(constants.INISECT_EXP) or
2011 not config.has_section(constants.INISECT_INS)):
2012 _Fail("Export info file doesn't have the required fields")
2014 return config.Dumps()
2017 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
2018 """Import an os image into an instance.
2020 @type instance: L{objects.Instance}
2021 @param instance: instance to import the disks into
2022 @type src_node: string
2023 @param src_node: source node for the disk images
2024 @type src_images: list of string
2025 @param src_images: absolute paths of the disk images
2026 @rtype: list of boolean
2027 @return: each boolean represent the success of importing the n-th disk
2030 inst_os = OSFromDisk(instance.os)
2031 import_env = OSEnvironment(instance, inst_os)
2032 import_script = inst_os.import_script
2034 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2035 instance.name, int(time.time()))
2036 if not os.path.exists(constants.LOG_OS_DIR):
2037 os.mkdir(constants.LOG_OS_DIR, 0750)
2040 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2041 import_script, logfile)
2044 for idx, image in enumerate(src_images):
2046 destcmd = utils.BuildShellCmd('cat %s', image)
2047 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2048 constants.GANETI_RUNAS,
2050 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2051 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2052 import_env['IMPORT_INDEX'] = str(idx)
2053 result = utils.RunCmd(command, env=import_env)
2055 logging.error("Disk import command '%s' returned error: %s"
2056 " output: %s", command, result.fail_reason,
2058 final_result.append("error importing disk %d: %s, %s" %
2059 (idx, result.fail_reason, result.output[-100]))
2062 _Fail("; ".join(final_result), log=False)
2066 """Return a list of exports currently available on this machine.
2069 @return: list of the exports
2072 if os.path.isdir(constants.EXPORT_DIR):
2073 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2075 _Fail("No exports directory")
2078 def RemoveExport(export):
2079 """Remove an existing export from the node.
2082 @param export: the name of the export to remove
2086 target = os.path.join(constants.EXPORT_DIR, export)
2089 shutil.rmtree(target)
2090 except EnvironmentError, err:
2091 _Fail("Error while removing the export: %s", err, exc=True)
2094 def BlockdevRename(devlist):
2095 """Rename a list of block devices.
2097 @type devlist: list of tuples
2098 @param devlist: list of tuples of the form (disk,
2099 new_logical_id, new_physical_id); disk is an
2100 L{objects.Disk} object describing the current disk,
2101 and new logical_id/physical_id is the name we
2104 @return: True if all renames succeeded, False otherwise
2109 for disk, unique_id in devlist:
2110 dev = _RecursiveFindBD(disk)
2112 msgs.append("Can't find device %s in rename" % str(disk))
2116 old_rpath = dev.dev_path
2117 dev.Rename(unique_id)
2118 new_rpath = dev.dev_path
2119 if old_rpath != new_rpath:
2120 DevCacheManager.RemoveCache(old_rpath)
2121 # FIXME: we should add the new cache information here, like:
2122 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2123 # but we don't have the owner here - maybe parse from existing
2124 # cache? for now, we only lose lvm data when we rename, which
2125 # is less critical than DRBD or MD
2126 except errors.BlockDeviceError, err:
2127 msgs.append("Can't rename device '%s' to '%s': %s" %
2128 (dev, unique_id, err))
2129 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2132 _Fail("; ".join(msgs))
2135 def _TransformFileStorageDir(file_storage_dir):
2136 """Checks whether given file_storage_dir is valid.
2138 Checks wheter the given file_storage_dir is within the cluster-wide
2139 default file_storage_dir stored in SimpleStore. Only paths under that
2140 directory are allowed.
2142 @type file_storage_dir: str
2143 @param file_storage_dir: the path to check
2145 @return: the normalized path if valid, None otherwise
2149 file_storage_dir = os.path.normpath(file_storage_dir)
2150 base_file_storage_dir = cfg.GetFileStorageDir()
2151 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2152 base_file_storage_dir):
2153 _Fail("File storage directory '%s' is not under base file"
2154 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2155 return file_storage_dir
2158 def CreateFileStorageDir(file_storage_dir):
2159 """Create file storage directory.
2161 @type file_storage_dir: str
2162 @param file_storage_dir: directory to create
2165 @return: tuple with first element a boolean indicating wheter dir
2166 creation was successful or not
2169 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2170 if os.path.exists(file_storage_dir):
2171 if not os.path.isdir(file_storage_dir):
2172 _Fail("Specified storage dir '%s' is not a directory",
2176 os.makedirs(file_storage_dir, 0750)
2177 except OSError, err:
2178 _Fail("Cannot create file storage directory '%s': %s",
2179 file_storage_dir, err, exc=True)
2182 def RemoveFileStorageDir(file_storage_dir):
2183 """Remove file storage directory.
2185 Remove it only if it's empty. If not log an error and return.
2187 @type file_storage_dir: str
2188 @param file_storage_dir: the directory we should cleanup
2189 @rtype: tuple (success,)
2190 @return: tuple of one element, C{success}, denoting
2191 whether the operation was successful
2194 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2195 if os.path.exists(file_storage_dir):
2196 if not os.path.isdir(file_storage_dir):
2197 _Fail("Specified Storage directory '%s' is not a directory",
2199 # deletes dir only if empty, otherwise we want to fail the rpc call
2201 os.rmdir(file_storage_dir)
2202 except OSError, err:
2203 _Fail("Cannot remove file storage directory '%s': %s",
2204 file_storage_dir, err)
2207 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2208 """Rename the file storage directory.
2210 @type old_file_storage_dir: str
2211 @param old_file_storage_dir: the current path
2212 @type new_file_storage_dir: str
2213 @param new_file_storage_dir: the name we should rename to
2214 @rtype: tuple (success,)
2215 @return: tuple of one element, C{success}, denoting
2216 whether the operation was successful
2219 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2220 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2221 if not os.path.exists(new_file_storage_dir):
2222 if os.path.isdir(old_file_storage_dir):
2224 os.rename(old_file_storage_dir, new_file_storage_dir)
2225 except OSError, err:
2226 _Fail("Cannot rename '%s' to '%s': %s",
2227 old_file_storage_dir, new_file_storage_dir, err)
2229 _Fail("Specified storage dir '%s' is not a directory",
2230 old_file_storage_dir)
2232 if os.path.exists(old_file_storage_dir):
2233 _Fail("Cannot rename '%s' to '%s': both locations exist",
2234 old_file_storage_dir, new_file_storage_dir)
2237 def _EnsureJobQueueFile(file_name):
2238 """Checks whether the given filename is in the queue directory.
2240 @type file_name: str
2241 @param file_name: the file name we should check
2243 @raises RPCFail: if the file is not valid
2246 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2247 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2250 _Fail("Passed job queue file '%s' does not belong to"
2251 " the queue directory '%s'", file_name, queue_dir)
2254 def JobQueueUpdate(file_name, content):
2255 """Updates a file in the queue directory.
2257 This is just a wrapper over L{utils.WriteFile}, with proper
2260 @type file_name: str
2261 @param file_name: the job file name
2263 @param content: the new job contents
2265 @return: the success of the operation
2268 _EnsureJobQueueFile(file_name)
2270 # Write and replace the file atomically
2271 utils.WriteFile(file_name, data=_Decompress(content))
2274 def JobQueueRename(old, new):
2275 """Renames a job queue file.
2277 This is just a wrapper over os.rename with proper checking.
2280 @param old: the old (actual) file name
2282 @param new: the desired file name
2284 @return: the success of the operation and payload
2287 _EnsureJobQueueFile(old)
2288 _EnsureJobQueueFile(new)
2290 utils.RenameFile(old, new, mkdir=True)
2293 def JobQueueSetDrainFlag(drain_flag):
2294 """Set the drain flag for the queue.
2296 This will set or unset the queue drain flag.
2298 @type drain_flag: boolean
2299 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2301 @return: always True, None
2302 @warning: the function always returns True
2306 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2308 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2311 def BlockdevClose(instance_name, disks):
2312 """Closes the given block devices.
2314 This means they will be switched to secondary mode (in case of
2317 @param instance_name: if the argument is not empty, the symlinks
2318 of this instance will be removed
2319 @type disks: list of L{objects.Disk}
2320 @param disks: the list of disks to be closed
2321 @rtype: tuple (success, message)
2322 @return: a tuple of success and message, where success
2323 indicates the succes of the operation, and message
2324 which will contain the error details in case we
2330 rd = _RecursiveFindBD(cf)
2332 _Fail("Can't find device %s", cf)
2339 except errors.BlockDeviceError, err:
2340 msg.append(str(err))
2342 _Fail("Can't make devices secondary: %s", ",".join(msg))
2345 _RemoveBlockDevLinks(instance_name, disks)
2348 def ValidateHVParams(hvname, hvparams):
2349 """Validates the given hypervisor parameters.
2351 @type hvname: string
2352 @param hvname: the hypervisor name
2353 @type hvparams: dict
2354 @param hvparams: the hypervisor parameters to be validated
2359 hv_type = hypervisor.GetHypervisor(hvname)
2360 hv_type.ValidateParameters(hvparams)
2361 except errors.HypervisorError, err:
2362 _Fail(str(err), log=False)
2366 """Demotes the current node from master candidate role.
2369 # try to ensure we're not the master by mistake
2370 master, myself = ssconf.GetMasterAndMyself()
2371 if master == myself:
2372 _Fail("ssconf status shows I'm the master node, will not demote")
2373 pid_file = utils.DaemonPidFileName(constants.MASTERD)
2374 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2375 _Fail("The master daemon is running, will not demote")
2377 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2378 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2379 except EnvironmentError, err:
2380 if err.errno != errno.ENOENT:
2381 _Fail("Error while backing up cluster file: %s", err, exc=True)
2382 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2385 def _FindDisks(nodes_ip, disks):
2386 """Sets the physical ID on disks and returns the block devices.
2389 # set the correct physical ID
2390 my_name = utils.HostInfo().name
2392 cf.SetPhysicalID(my_name, nodes_ip)
2397 rd = _RecursiveFindBD(cf)
2399 _Fail("Can't find device %s", cf)
2404 def DrbdDisconnectNet(nodes_ip, disks):
2405 """Disconnects the network on a list of drbd devices.
2408 bdevs = _FindDisks(nodes_ip, disks)
2414 except errors.BlockDeviceError, err:
2415 _Fail("Can't change network configuration to standalone mode: %s",
2419 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2420 """Attaches the network on a list of drbd devices.
2423 bdevs = _FindDisks(nodes_ip, disks)
2426 for idx, rd in enumerate(bdevs):
2428 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2429 except EnvironmentError, err:
2430 _Fail("Can't create symlink: %s", err)
2431 # reconnect disks, switch to new master configuration and if
2432 # needed primary mode
2435 rd.AttachNet(multimaster)
2436 except errors.BlockDeviceError, err:
2437 _Fail("Can't change network configuration: %s", err)
2438 # wait until the disks are connected; we need to retry the re-attach
2439 # if the device becomes standalone, as this might happen if the one
2440 # node disconnects and reconnects in a different mode before the
2441 # other node reconnects; in this case, one or both of the nodes will
2442 # decide it has wrong configuration and switch to standalone
2443 RECONNECT_TIMEOUT = 2 * 60
2444 sleep_time = 0.100 # start with 100 miliseconds
2445 timeout_limit = time.time() + RECONNECT_TIMEOUT
2446 while time.time() < timeout_limit:
2447 all_connected = True
2449 stats = rd.GetProcStatus()
2450 if not (stats.is_connected or stats.is_in_resync):
2451 all_connected = False
2452 if stats.is_standalone:
2453 # peer had different config info and this node became
2454 # standalone, even though this should not happen with the
2455 # new staged way of changing disk configs
2457 rd.AttachNet(multimaster)
2458 except errors.BlockDeviceError, err:
2459 _Fail("Can't change network configuration: %s", err)
2462 time.sleep(sleep_time)
2463 sleep_time = min(5, sleep_time * 1.5)
2464 if not all_connected:
2465 _Fail("Timeout in disk reconnecting")
2467 # change to primary mode
2471 except errors.BlockDeviceError, err:
2472 _Fail("Can't change to primary mode: %s", err)
2475 def DrbdWaitSync(nodes_ip, disks):
2476 """Wait until DRBDs have synchronized.
2479 bdevs = _FindDisks(nodes_ip, disks)
2484 stats = rd.GetProcStatus()
2485 if not (stats.is_connected or stats.is_in_resync):
2486 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2487 alldone = alldone and (not stats.is_in_resync)
2488 if stats.sync_percent is not None:
2489 min_resync = min(min_resync, stats.sync_percent)
2491 return (alldone, min_resync)
2494 def PowercycleNode(hypervisor_type):
2495 """Hard-powercycle the node.
2497 Because we need to return first, and schedule the powercycle in the
2498 background, we won't be able to report failures nicely.
2501 hyper = hypervisor.GetHypervisor(hypervisor_type)
2505 # if we can't fork, we'll pretend that we're in the child process
2508 return "Reboot scheduled in 5 seconds"
2510 hyper.PowercycleNode()
2513 class HooksRunner(object):
2516 This class is instantiated on the node side (ganeti-noded) and not
2520 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2522 def __init__(self, hooks_base_dir=None):
2523 """Constructor for hooks runner.
2525 @type hooks_base_dir: str or None
2526 @param hooks_base_dir: if not None, this overrides the
2527 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2530 if hooks_base_dir is None:
2531 hooks_base_dir = constants.HOOKS_BASE_DIR
2532 self._BASE_DIR = hooks_base_dir
2535 def ExecHook(script, env):
2536 """Exec one hook script.
2539 @param script: the full path to the script
2541 @param env: the environment with which to exec the script
2542 @rtype: tuple (success, message)
2543 @return: a tuple of success and message, where success
2544 indicates the succes of the operation, and message
2545 which will contain the error details in case we
2549 # exec the process using subprocess and log the output
2552 fdstdin = open("/dev/null", "r")
2553 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2554 stderr=subprocess.STDOUT, close_fds=True,
2555 shell=False, cwd="/", env=env)
2558 output = child.stdout.read(4096)
2559 child.stdout.close()
2560 except EnvironmentError, err:
2561 output += "Hook script error: %s" % str(err)
2565 result = child.wait()
2567 except EnvironmentError, err:
2568 if err.errno == errno.EINTR:
2572 # try not to leak fds
2573 for fd in (fdstdin, ):
2577 except EnvironmentError, err:
2578 # just log the error
2579 #logging.exception("Error while closing fd %s", fd)
2582 return result == 0, utils.SafeEncode(output.strip())
2584 def RunHooks(self, hpath, phase, env):
2585 """Run the scripts in the hooks directory.
2588 @param hpath: the path to the hooks directory which
2591 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2592 L{constants.HOOKS_PHASE_POST}
2594 @param env: dictionary with the environment for the hook
2596 @return: list of 3-element tuples:
2598 - script result, either L{constants.HKR_SUCCESS} or
2599 L{constants.HKR_FAIL}
2600 - output of the script
2602 @raise errors.ProgrammerError: for invalid input
2606 if phase == constants.HOOKS_PHASE_PRE:
2608 elif phase == constants.HOOKS_PHASE_POST:
2611 _Fail("Unknown hooks phase '%s'", phase)
2615 subdir = "%s-%s.d" % (hpath, suffix)
2616 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2618 dir_contents = utils.ListVisibleFiles(dir_name)
2620 # FIXME: must log output in case of failures
2623 # we use the standard python sort order,
2624 # so 00name is the recommended naming scheme
2626 for relname in dir_contents:
2627 fname = os.path.join(dir_name, relname)
2628 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2629 self.RE_MASK.match(relname) is not None):
2630 rrval = constants.HKR_SKIP
2633 result, output = self.ExecHook(fname, env)
2635 rrval = constants.HKR_FAIL
2637 rrval = constants.HKR_SUCCESS
2638 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2643 class IAllocatorRunner(object):
2644 """IAllocator runner.
2646 This class is instantiated on the node side (ganeti-noded) and not on
2650 def Run(self, name, idata):
2651 """Run an iallocator script.
2654 @param name: the iallocator script name
2656 @param idata: the allocator input data
2659 @return: two element tuple of:
2661 - either error message or stdout of allocator (for success)
2664 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2666 if alloc_script is None:
2667 _Fail("iallocator module '%s' not found in the search path", name)
2669 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2673 result = utils.RunCmd([alloc_script, fin_name])
2675 _Fail("iallocator module '%s' failed: %s, output '%s'",
2676 name, result.fail_reason, result.output)
2680 return result.stdout
2683 class DevCacheManager(object):
2684 """Simple class for managing a cache of block device information.
2687 _DEV_PREFIX = "/dev/"
2688 _ROOT_DIR = constants.BDEV_CACHE_DIR
2691 def _ConvertPath(cls, dev_path):
2692 """Converts a /dev/name path to the cache file name.
2694 This replaces slashes with underscores and strips the /dev
2695 prefix. It then returns the full path to the cache file.
2698 @param dev_path: the C{/dev/} path name
2700 @return: the converted path name
2703 if dev_path.startswith(cls._DEV_PREFIX):
2704 dev_path = dev_path[len(cls._DEV_PREFIX):]
2705 dev_path = dev_path.replace("/", "_")
2706 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2710 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2711 """Updates the cache information for a given device.
2714 @param dev_path: the pathname of the device
2716 @param owner: the owner (instance name) of the device
2717 @type on_primary: bool
2718 @param on_primary: whether this is the primary
2721 @param iv_name: the instance-visible name of the
2722 device, as in objects.Disk.iv_name
2727 if dev_path is None:
2728 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2730 fpath = cls._ConvertPath(dev_path)
2736 iv_name = "not_visible"
2737 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2739 utils.WriteFile(fpath, data=fdata)
2740 except EnvironmentError, err:
2741 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2744 def RemoveCache(cls, dev_path):
2745 """Remove data for a dev_path.
2747 This is just a wrapper over L{utils.RemoveFile} with a converted
2748 path name and logging.
2751 @param dev_path: the pathname of the device
2756 if dev_path is None:
2757 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2759 fpath = cls._ConvertPath(dev_path)
2761 utils.RemoveFile(fpath)
2762 except EnvironmentError, err:
2763 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)