4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 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
26 @var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted
27 in the L{_CleanDirectory} function
31 # pylint: disable-msg=E1103
33 # E1103: %s %r has no %r member (but some types could not be
34 # inferred), because the _TryOSFromDisk returns either (True, os_obj)
35 # or (False, "string") which confuses pylint
52 from ganeti import errors
53 from ganeti import utils
54 from ganeti import ssh
55 from ganeti import hypervisor
56 from ganeti import constants
57 from ganeti import bdev
58 from ganeti import objects
59 from ganeti import ssconf
60 from ganeti import serializer
61 from ganeti import netutils
62 from ganeti import runtime
65 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
66 _ALLOWED_CLEAN_DIRS = frozenset([
68 constants.JOB_QUEUE_ARCHIVE_DIR,
70 constants.CRYPTO_KEYS_DIR,
72 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
73 _X509_KEY_FILE = "key"
74 _X509_CERT_FILE = "cert"
75 _IES_STATUS_FILE = "status"
79 #: Valid LVS output line regex
80 _LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
83 class RPCFail(Exception):
84 """Class denoting RPC failure.
86 Its argument is the error message.
91 def _Fail(msg, *args, **kwargs):
92 """Log an error and the raise an RPCFail exception.
94 This exception is then handled specially in the ganeti daemon and
95 turned into a 'failed' return type. As such, this function is a
96 useful shortcut for logging the error and returning it to the master
100 @param msg: the text of the exception
106 if "log" not in kwargs or kwargs["log"]: # if we should log this error
107 if "exc" in kwargs and kwargs["exc"]:
108 logging.exception(msg)
115 """Simple wrapper to return a SimpleStore.
117 @rtype: L{ssconf.SimpleStore}
118 @return: a SimpleStore instance
121 return ssconf.SimpleStore()
124 def _GetSshRunner(cluster_name):
125 """Simple wrapper to return an SshRunner.
127 @type cluster_name: str
128 @param cluster_name: the cluster name, which is needed
129 by the SshRunner constructor
130 @rtype: L{ssh.SshRunner}
131 @return: an SshRunner instance
134 return ssh.SshRunner(cluster_name)
137 def _Decompress(data):
138 """Unpacks data compressed by the RPC client.
140 @type data: list or tuple
141 @param data: Data sent by RPC client
143 @return: Decompressed data
146 assert isinstance(data, (list, tuple))
147 assert len(data) == 2
148 (encoding, content) = data
149 if encoding == constants.RPC_ENCODING_NONE:
151 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
152 return zlib.decompress(base64.b64decode(content))
154 raise AssertionError("Unknown data encoding")
157 def _CleanDirectory(path, exclude=None):
158 """Removes all regular files in a directory.
161 @param path: the directory to clean
163 @param exclude: list of files to be excluded, defaults
167 if path not in _ALLOWED_CLEAN_DIRS:
168 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
171 if not os.path.isdir(path):
176 # Normalize excluded paths
177 exclude = [os.path.normpath(i) for i in exclude]
179 for rel_name in utils.ListVisibleFiles(path):
180 full_name = utils.PathJoin(path, rel_name)
181 if full_name in exclude:
183 if os.path.isfile(full_name) and not os.path.islink(full_name):
184 utils.RemoveFile(full_name)
187 def _BuildUploadFileList():
188 """Build the list of allowed upload files.
190 This is abstracted so that it's built only once at module import time.
193 allowed_files = set([
194 constants.CLUSTER_CONF_FILE,
196 constants.SSH_KNOWN_HOSTS_FILE,
197 constants.VNC_PASSWORD_FILE,
198 constants.RAPI_CERT_FILE,
199 constants.RAPI_USERS_FILE,
200 constants.CONFD_HMAC_KEY,
201 constants.CLUSTER_DOMAIN_SECRET_FILE,
204 for hv_name in constants.HYPER_TYPES:
205 hv_class = hypervisor.GetHypervisorClass(hv_name)
206 allowed_files.update(hv_class.GetAncillaryFiles())
208 return frozenset(allowed_files)
211 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
215 """Removes job queue files and archived jobs.
221 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
222 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
226 """Returns master information.
228 This is an utility function to compute master information, either
229 for consumption here or from the node daemon.
232 @return: master_netdev, master_ip, master_name, primary_ip_family
233 @raise RPCFail: in case of errors
238 master_netdev = cfg.GetMasterNetdev()
239 master_ip = cfg.GetMasterIP()
240 master_node = cfg.GetMasterNode()
241 primary_ip_family = cfg.GetPrimaryIPFamily()
242 except errors.ConfigurationError, err:
243 _Fail("Cluster configuration incomplete: %s", err, exc=True)
244 return (master_netdev, master_ip, master_node, primary_ip_family)
247 def StartMaster(start_daemons, no_voting):
248 """Activate local node as master node.
250 The function will either try activate the IP address of the master
251 (unless someone else has it) or also start the master daemons, based
252 on the start_daemons parameter.
254 @type start_daemons: boolean
255 @param start_daemons: whether to start the master daemons
256 (ganeti-masterd and ganeti-rapi), or (if false) activate the
258 @type no_voting: boolean
259 @param no_voting: whether to start ganeti-masterd without a node vote
260 (if start_daemons is True), but still non-interactively
264 # GetMasterInfo will raise an exception if not able to return data
265 master_netdev, master_ip, _, family = GetMasterInfo()
268 # either start the master and rapi daemons
271 masterd_args = "--no-voting --yes-do-it"
276 "EXTRA_MASTERD_ARGS": masterd_args,
279 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
281 msg = "Can't start Ganeti master: %s" % result.output
286 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
287 if netutils.IPAddress.Own(master_ip):
288 # we already have the ip:
289 logging.debug("Master IP already configured, doing nothing")
291 msg = "Someone else has the master ip, not activating"
295 ipcls = netutils.IP4Address
296 if family == netutils.IP6Address.family:
297 ipcls = netutils.IP6Address
299 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
300 "%s/%d" % (master_ip, ipcls.iplen),
301 "dev", master_netdev, "label",
302 "%s:0" % master_netdev])
304 msg = "Can't activate master IP: %s" % result.output
308 # we ignore the exit code of the following cmds
309 if ipcls == netutils.IP4Address:
310 utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev, "-s",
311 master_ip, master_ip])
312 elif ipcls == netutils.IP6Address:
314 utils.RunCmd(["ndisc6", "-q", "-r 3", master_ip, master_netdev])
315 except errors.OpExecError:
316 # TODO: Better error reporting
317 logging.warning("Can't execute ndisc6, please install if missing")
320 _Fail("; ".join(err_msgs))
323 def StopMaster(stop_daemons):
324 """Deactivate this node as master.
326 The function will always try to deactivate the IP address of the
327 master. It will also stop the master daemons depending on the
328 stop_daemons parameter.
330 @type stop_daemons: boolean
331 @param stop_daemons: whether to also stop the master daemons
332 (ganeti-masterd and ganeti-rapi)
336 # TODO: log and report back to the caller the error failures; we
337 # need to decide in which case we fail the RPC for this
339 # GetMasterInfo will raise an exception if not able to return data
340 master_netdev, master_ip, _, family = GetMasterInfo()
342 ipcls = netutils.IP4Address
343 if family == netutils.IP6Address.family:
344 ipcls = netutils.IP6Address
346 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
347 "%s/%d" % (master_ip, ipcls.iplen),
348 "dev", master_netdev])
350 logging.error("Can't remove the master IP, error: %s", result.output)
351 # but otherwise ignore the failure
354 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
356 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
358 result.cmd, result.exit_code, result.output)
361 def EtcHostsModify(mode, host, ip):
362 """Modify a host entry in /etc/hosts.
364 @param mode: The mode to operate. Either add or remove entry
365 @param host: The host to operate on
366 @param ip: The ip associated with the entry
369 if mode == constants.ETC_HOSTS_ADD:
371 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
373 utils.AddHostToEtcHosts(host, ip)
374 elif mode == constants.ETC_HOSTS_REMOVE:
376 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
377 " parameter is present")
378 utils.RemoveHostFromEtcHosts(host)
380 RPCFail("Mode not supported")
383 def LeaveCluster(modify_ssh_setup):
384 """Cleans up and remove the current node.
386 This function cleans up and prepares the current node to be removed
389 If processing is successful, then it raises an
390 L{errors.QuitGanetiException} which is used as a special case to
391 shutdown the node daemon.
393 @param modify_ssh_setup: boolean
396 _CleanDirectory(constants.DATA_DIR)
397 _CleanDirectory(constants.CRYPTO_KEYS_DIR)
402 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
404 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
406 utils.RemoveFile(priv_key)
407 utils.RemoveFile(pub_key)
408 except errors.OpExecError:
409 logging.exception("Error while processing ssh files")
412 utils.RemoveFile(constants.CONFD_HMAC_KEY)
413 utils.RemoveFile(constants.RAPI_CERT_FILE)
414 utils.RemoveFile(constants.NODED_CERT_FILE)
415 except: # pylint: disable-msg=W0702
416 logging.exception("Error while removing cluster secrets")
418 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
420 logging.error("Command %s failed with exitcode %s and error %s",
421 result.cmd, result.exit_code, result.output)
423 # Raise a custom exception (handled in ganeti-noded)
424 raise errors.QuitGanetiException(True, "Shutdown scheduled")
427 def GetNodeInfo(vgname, hypervisor_type):
428 """Gives back a hash with different information about the node.
430 @type vgname: C{string}
431 @param vgname: the name of the volume group to ask for disk space information
432 @type hypervisor_type: C{str}
433 @param hypervisor_type: the name of the hypervisor to ask for
436 @return: dictionary with the following keys:
437 - vg_size is the size of the configured volume group in MiB
438 - vg_free is the free size of the volume group in MiB
439 - memory_dom0 is the memory allocated for domain0 in MiB
440 - memory_free is the currently available (free) ram in MiB
441 - memory_total is the total number of ram in MiB
446 if vgname is not None:
447 vginfo = bdev.LogicalVolume.GetVGInfo([vgname])
448 vg_free = vg_size = None
450 vg_free = int(round(vginfo[0][0], 0))
451 vg_size = int(round(vginfo[0][1], 0))
452 outputarray["vg_size"] = vg_size
453 outputarray["vg_free"] = vg_free
455 if hypervisor_type is not None:
456 hyper = hypervisor.GetHypervisor(hypervisor_type)
457 hyp_info = hyper.GetNodeInfo()
458 if hyp_info is not None:
459 outputarray.update(hyp_info)
461 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
466 def VerifyNode(what, cluster_name):
467 """Verify the status of the local node.
469 Based on the input L{what} parameter, various checks are done on the
472 If the I{filelist} key is present, this list of
473 files is checksummed and the file/checksum pairs are returned.
475 If the I{nodelist} key is present, we check that we have
476 connectivity via ssh with the target nodes (and check the hostname
479 If the I{node-net-test} key is present, we check that we have
480 connectivity to the given nodes via both primary IP and, if
481 applicable, secondary IPs.
484 @param what: a dictionary of things to check:
485 - filelist: list of files for which to compute checksums
486 - nodelist: list of nodes we should check ssh communication with
487 - node-net-test: list of nodes we should check node daemon port
489 - hypervisor: list with hypervisors to run the verify for
491 @return: a dictionary with the same keys as the input dict, and
492 values representing the result of the checks
496 my_name = netutils.Hostname.GetSysName()
497 port = netutils.GetDaemonPort(constants.NODED)
498 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
500 if constants.NV_HYPERVISOR in what and vm_capable:
501 result[constants.NV_HYPERVISOR] = tmp = {}
502 for hv_name in what[constants.NV_HYPERVISOR]:
504 val = hypervisor.GetHypervisor(hv_name).Verify()
505 except errors.HypervisorError, err:
506 val = "Error while checking hypervisor: %s" % str(err)
509 if constants.NV_HVPARAMS in what and vm_capable:
510 result[constants.NV_HVPARAMS] = tmp = []
511 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
513 logging.info("Validating hv %s, %s", hv_name, hvparms)
514 hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
515 except errors.HypervisorError, err:
516 tmp.append((source, hv_name, str(err)))
518 if constants.NV_FILELIST in what:
519 result[constants.NV_FILELIST] = utils.FingerprintFiles(
520 what[constants.NV_FILELIST])
522 if constants.NV_NODELIST in what:
523 result[constants.NV_NODELIST] = tmp = {}
524 random.shuffle(what[constants.NV_NODELIST])
525 for node in what[constants.NV_NODELIST]:
526 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
530 if constants.NV_NODENETTEST in what:
531 result[constants.NV_NODENETTEST] = tmp = {}
532 my_pip = my_sip = None
533 for name, pip, sip in what[constants.NV_NODENETTEST]:
539 tmp[my_name] = ("Can't find my own primary/secondary IP"
542 for name, pip, sip in what[constants.NV_NODENETTEST]:
544 if not netutils.TcpPing(pip, port, source=my_pip):
545 fail.append("primary")
547 if not netutils.TcpPing(sip, port, source=my_sip):
548 fail.append("secondary")
550 tmp[name] = ("failure using the %s interface(s)" %
553 if constants.NV_MASTERIP in what:
554 # FIXME: add checks on incoming data structures (here and in the
555 # rest of the function)
556 master_name, master_ip = what[constants.NV_MASTERIP]
557 if master_name == my_name:
558 source = constants.IP4_ADDRESS_LOCALHOST
561 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
564 if constants.NV_OOB_PATHS in what:
565 result[constants.NV_OOB_PATHS] = tmp = []
566 for path in what[constants.NV_OOB_PATHS]:
570 tmp.append("error stating out of band helper: %s" % err)
572 if stat.S_ISREG(st.st_mode):
573 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
576 tmp.append("out of band helper %s is not executable" % path)
578 tmp.append("out of band helper %s is not a file" % path)
580 if constants.NV_LVLIST in what and vm_capable:
582 val = GetVolumeList(utils.ListVolumeGroups().keys())
585 result[constants.NV_LVLIST] = val
587 if constants.NV_INSTANCELIST in what and vm_capable:
588 # GetInstanceList can fail
590 val = GetInstanceList(what[constants.NV_INSTANCELIST])
593 result[constants.NV_INSTANCELIST] = val
595 if constants.NV_VGLIST in what and vm_capable:
596 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
598 if constants.NV_PVLIST in what and vm_capable:
599 result[constants.NV_PVLIST] = \
600 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
601 filter_allocatable=False)
603 if constants.NV_VERSION in what:
604 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
605 constants.RELEASE_VERSION)
607 if constants.NV_HVINFO in what and vm_capable:
608 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
609 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
611 if constants.NV_DRBDLIST in what and vm_capable:
613 used_minors = bdev.DRBD8.GetUsedDevs().keys()
614 except errors.BlockDeviceError, err:
615 logging.warning("Can't get used minors list", exc_info=True)
616 used_minors = str(err)
617 result[constants.NV_DRBDLIST] = used_minors
619 if constants.NV_DRBDHELPER in what and vm_capable:
622 payload = bdev.BaseDRBD.GetUsermodeHelper()
623 except errors.BlockDeviceError, err:
624 logging.error("Can't get DRBD usermode helper: %s", str(err))
627 result[constants.NV_DRBDHELPER] = (status, payload)
629 if constants.NV_NODESETUP in what:
630 result[constants.NV_NODESETUP] = tmpr = []
631 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
632 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
633 " under /sys, missing required directories /sys/block"
634 " and /sys/class/net")
635 if (not os.path.isdir("/proc/sys") or
636 not os.path.isfile("/proc/sysrq-trigger")):
637 tmpr.append("The procfs filesystem doesn't seem to be mounted"
638 " under /proc, missing required directory /proc/sys and"
639 " the file /proc/sysrq-trigger")
641 if constants.NV_TIME in what:
642 result[constants.NV_TIME] = utils.SplitTime(time.time())
644 if constants.NV_OSLIST in what and vm_capable:
645 result[constants.NV_OSLIST] = DiagnoseOS()
647 if constants.NV_BRIDGES in what and vm_capable:
648 result[constants.NV_BRIDGES] = [bridge
649 for bridge in what[constants.NV_BRIDGES]
650 if not utils.BridgeExists(bridge)]
654 def GetBlockDevSizes(devices):
655 """Return the size of the given block devices
658 @param devices: list of block device nodes to query
661 dictionary of all block devices under /dev (key). The value is their
664 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
670 for devpath in devices:
671 if os.path.commonprefix([DEV_PREFIX, devpath]) != DEV_PREFIX:
675 st = os.stat(devpath)
676 except EnvironmentError, err:
677 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
680 if stat.S_ISBLK(st.st_mode):
681 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
683 # We don't want to fail, just do not list this device as available
684 logging.warning("Cannot get size for block device %s", devpath)
687 size = int(result.stdout) / (1024 * 1024)
688 blockdevs[devpath] = size
692 def GetVolumeList(vg_names):
693 """Compute list of logical volumes and their size.
696 @param vg_names: the volume groups whose LVs we should list, or
697 empty for all volume groups
700 dictionary of all partions (key) with value being a tuple of
701 their size (in MiB), inactive and online status::
703 {'xenvg/test1': ('20.06', True, True)}
705 in case of errors, a string is returned with the error
713 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
714 "--separator=%s" % sep,
715 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
717 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
719 for line in result.stdout.splitlines():
721 match = _LVSLINE_REGEX.match(line)
723 logging.error("Invalid line returned from lvs output: '%s'", line)
725 vg_name, name, size, attr = match.groups()
726 inactive = attr[4] == "-"
727 online = attr[5] == "o"
728 virtual = attr[0] == "v"
730 # we don't want to report such volumes as existing, since they
731 # don't really hold data
733 lvs[vg_name+"/"+name] = (size, inactive, online)
738 def ListVolumeGroups():
739 """List the volume groups and their size.
742 @return: dictionary with keys volume name and values the
746 return utils.ListVolumeGroups()
750 """List all volumes on this node.
754 A list of dictionaries, each having four keys:
755 - name: the logical volume name,
756 - size: the size of the logical volume
757 - dev: the physical device on which the LV lives
758 - vg: the volume group to which it belongs
760 In case of errors, we return an empty list and log the
763 Note that since a logical volume can live on multiple physical
764 volumes, the resulting list might include a logical volume
768 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
770 "--options=lv_name,lv_size,devices,vg_name"])
772 _Fail("Failed to list logical volumes, lvs output: %s",
776 return dev.split("(")[0]
779 return [parse_dev(x) for x in dev.split(",")]
782 line = [v.strip() for v in line]
783 return [{"name": line[0], "size": line[1],
784 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
787 for line in result.stdout.splitlines():
788 if line.count("|") >= 3:
789 all_devs.extend(map_line(line.split("|")))
791 logging.warning("Strange line in the output from lvs: '%s'", line)
795 def BridgesExist(bridges_list):
796 """Check if a list of bridges exist on the current node.
799 @return: C{True} if all of them exist, C{False} otherwise
803 for bridge in bridges_list:
804 if not utils.BridgeExists(bridge):
805 missing.append(bridge)
808 _Fail("Missing bridges %s", utils.CommaJoin(missing))
811 def GetInstanceList(hypervisor_list):
812 """Provides a list of instances.
814 @type hypervisor_list: list
815 @param hypervisor_list: the list of hypervisors to query information
818 @return: a list of all running instances on the current node
819 - instance1.example.com
820 - instance2.example.com
824 for hname in hypervisor_list:
826 names = hypervisor.GetHypervisor(hname).ListInstances()
827 results.extend(names)
828 except errors.HypervisorError, err:
829 _Fail("Error enumerating instances (hypervisor %s): %s",
830 hname, err, exc=True)
835 def GetInstanceInfo(instance, hname):
836 """Gives back the information about an instance as a dictionary.
838 @type instance: string
839 @param instance: the instance name
841 @param hname: the hypervisor type of the instance
844 @return: dictionary with the following keys:
845 - memory: memory size of instance (int)
846 - state: xen state of instance (string)
847 - time: cpu time of instance (float)
852 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
853 if iinfo is not None:
854 output["memory"] = iinfo[2]
855 output["state"] = iinfo[4]
856 output["time"] = iinfo[5]
861 def GetInstanceMigratable(instance):
862 """Gives whether an instance can be migrated.
864 @type instance: L{objects.Instance}
865 @param instance: object representing the instance to be checked.
868 @return: tuple of (result, description) where:
869 - result: whether the instance can be migrated or not
870 - description: a description of the issue, if relevant
873 hyper = hypervisor.GetHypervisor(instance.hypervisor)
874 iname = instance.name
875 if iname not in hyper.ListInstances():
876 _Fail("Instance %s is not running", iname)
878 for idx in range(len(instance.disks)):
879 link_name = _GetBlockDevSymlinkPath(iname, idx)
880 if not os.path.islink(link_name):
881 logging.warning("Instance %s is missing symlink %s for disk %d",
882 iname, link_name, idx)
885 def GetAllInstancesInfo(hypervisor_list):
886 """Gather data about all instances.
888 This is the equivalent of L{GetInstanceInfo}, except that it
889 computes data for all instances at once, thus being faster if one
890 needs data about more than one instance.
892 @type hypervisor_list: list
893 @param hypervisor_list: list of hypervisors to query for instance data
896 @return: dictionary of instance: data, with data having the following keys:
897 - memory: memory size of instance (int)
898 - state: xen state of instance (string)
899 - time: cpu time of instance (float)
900 - vcpus: the number of vcpus
905 for hname in hypervisor_list:
906 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
908 for name, _, memory, vcpus, state, times in iinfo:
916 # we only check static parameters, like memory and vcpus,
917 # and not state and time which can change between the
918 # invocations of the different hypervisors
919 for key in "memory", "vcpus":
920 if value[key] != output[name][key]:
921 _Fail("Instance %s is running twice"
922 " with different parameters", name)
928 def _InstanceLogName(kind, os_name, instance, component):
929 """Compute the OS log filename for a given instance and operation.
931 The instance name and os name are passed in as strings since not all
932 operations have these as part of an instance object.
935 @param kind: the operation type (e.g. add, import, etc.)
936 @type os_name: string
937 @param os_name: the os name
938 @type instance: string
939 @param instance: the name of the instance being imported/added/etc.
940 @type component: string or None
941 @param component: the name of the component of the instance being
945 # TODO: Use tempfile.mkstemp to create unique filename
947 assert "/" not in component
948 c_msg = "-%s" % component
951 base = ("%s-%s-%s%s-%s.log" %
952 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
953 return utils.PathJoin(constants.LOG_OS_DIR, base)
956 def InstanceOsAdd(instance, reinstall, debug):
957 """Add an OS to an instance.
959 @type instance: L{objects.Instance}
960 @param instance: Instance whose OS is to be installed
961 @type reinstall: boolean
962 @param reinstall: whether this is an instance reinstall
964 @param debug: debug level, passed to the OS scripts
968 inst_os = OSFromDisk(instance.os)
970 create_env = OSEnvironment(instance, inst_os, debug)
972 create_env["INSTANCE_REINSTALL"] = "1"
974 logfile = _InstanceLogName("add", instance.os, instance.name, None)
976 result = utils.RunCmd([inst_os.create_script], env=create_env,
977 cwd=inst_os.path, output=logfile, reset_env=True)
979 logging.error("os create command '%s' returned error: %s, logfile: %s,"
980 " output: %s", result.cmd, result.fail_reason, logfile,
982 lines = [utils.SafeEncode(val)
983 for val in utils.TailFile(logfile, lines=20)]
984 _Fail("OS create script failed (%s), last lines in the"
985 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
988 def RunRenameInstance(instance, old_name, debug):
989 """Run the OS rename script for an instance.
991 @type instance: L{objects.Instance}
992 @param instance: Instance whose OS is to be installed
993 @type old_name: string
994 @param old_name: previous instance name
996 @param debug: debug level, passed to the OS scripts
998 @return: the success of the operation
1001 inst_os = OSFromDisk(instance.os)
1003 rename_env = OSEnvironment(instance, inst_os, debug)
1004 rename_env["OLD_INSTANCE_NAME"] = old_name
1006 logfile = _InstanceLogName("rename", instance.os,
1007 "%s-%s" % (old_name, instance.name), None)
1009 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1010 cwd=inst_os.path, output=logfile, reset_env=True)
1013 logging.error("os create command '%s' returned error: %s output: %s",
1014 result.cmd, result.fail_reason, result.output)
1015 lines = [utils.SafeEncode(val)
1016 for val in utils.TailFile(logfile, lines=20)]
1017 _Fail("OS rename script failed (%s), last lines in the"
1018 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1021 def _GetBlockDevSymlinkPath(instance_name, idx):
1022 return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1023 (instance_name, constants.DISK_SEPARATOR, idx))
1026 def _SymlinkBlockDev(instance_name, device_path, idx):
1027 """Set up symlinks to a instance's block device.
1029 This is an auxiliary function run when an instance is start (on the primary
1030 node) or when an instance is migrated (on the target node).
1033 @param instance_name: the name of the target instance
1034 @param device_path: path of the physical block device, on the node
1035 @param idx: the disk index
1036 @return: absolute path to the disk's symlink
1039 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1041 os.symlink(device_path, link_name)
1042 except OSError, err:
1043 if err.errno == errno.EEXIST:
1044 if (not os.path.islink(link_name) or
1045 os.readlink(link_name) != device_path):
1046 os.remove(link_name)
1047 os.symlink(device_path, link_name)
1054 def _RemoveBlockDevLinks(instance_name, disks):
1055 """Remove the block device symlinks belonging to the given instance.
1058 for idx, _ in enumerate(disks):
1059 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1060 if os.path.islink(link_name):
1062 os.remove(link_name)
1064 logging.exception("Can't remove symlink '%s'", link_name)
1067 def _GatherAndLinkBlockDevs(instance):
1068 """Set up an instance's block device(s).
1070 This is run on the primary node at instance startup. The block
1071 devices must be already assembled.
1073 @type instance: L{objects.Instance}
1074 @param instance: the instance whose disks we shoul assemble
1076 @return: list of (disk_object, device_path)
1080 for idx, disk in enumerate(instance.disks):
1081 device = _RecursiveFindBD(disk)
1083 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1087 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1089 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1092 block_devices.append((disk, link_name))
1094 return block_devices
1097 def StartInstance(instance, startup_paused):
1098 """Start an instance.
1100 @type instance: L{objects.Instance}
1101 @param instance: the instance object
1102 @type startup_paused: bool
1103 @param instance: pause instance at startup?
1107 running_instances = GetInstanceList([instance.hypervisor])
1109 if instance.name in running_instances:
1110 logging.info("Instance %s already running, not starting", instance.name)
1114 block_devices = _GatherAndLinkBlockDevs(instance)
1115 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1116 hyper.StartInstance(instance, block_devices, startup_paused)
1117 except errors.BlockDeviceError, err:
1118 _Fail("Block device error: %s", err, exc=True)
1119 except errors.HypervisorError, err:
1120 _RemoveBlockDevLinks(instance.name, instance.disks)
1121 _Fail("Hypervisor error: %s", err, exc=True)
1124 def InstanceShutdown(instance, timeout):
1125 """Shut an instance down.
1127 @note: this functions uses polling with a hardcoded timeout.
1129 @type instance: L{objects.Instance}
1130 @param instance: the instance object
1131 @type timeout: integer
1132 @param timeout: maximum timeout for soft shutdown
1136 hv_name = instance.hypervisor
1137 hyper = hypervisor.GetHypervisor(hv_name)
1138 iname = instance.name
1140 if instance.name not in hyper.ListInstances():
1141 logging.info("Instance %s not running, doing nothing", iname)
1146 self.tried_once = False
1149 if iname not in hyper.ListInstances():
1153 hyper.StopInstance(instance, retry=self.tried_once)
1154 except errors.HypervisorError, err:
1155 if iname not in hyper.ListInstances():
1156 # if the instance is no longer existing, consider this a
1157 # success and go to cleanup
1160 _Fail("Failed to stop instance %s: %s", iname, err)
1162 self.tried_once = True
1164 raise utils.RetryAgain()
1167 utils.Retry(_TryShutdown(), 5, timeout)
1168 except utils.RetryTimeout:
1169 # the shutdown did not succeed
1170 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1173 hyper.StopInstance(instance, force=True)
1174 except errors.HypervisorError, err:
1175 if iname in hyper.ListInstances():
1176 # only raise an error if the instance still exists, otherwise
1177 # the error could simply be "instance ... unknown"!
1178 _Fail("Failed to force stop instance %s: %s", iname, err)
1182 if iname in hyper.ListInstances():
1183 _Fail("Could not shutdown instance %s even by destroy", iname)
1186 hyper.CleanupInstance(instance.name)
1187 except errors.HypervisorError, err:
1188 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1190 _RemoveBlockDevLinks(iname, instance.disks)
1193 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1194 """Reboot an instance.
1196 @type instance: L{objects.Instance}
1197 @param instance: the instance object to reboot
1198 @type reboot_type: str
1199 @param reboot_type: the type of reboot, one the following
1201 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1202 instance OS, do not recreate the VM
1203 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1204 restart the VM (at the hypervisor level)
1205 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1206 not accepted here, since that mode is handled differently, in
1207 cmdlib, and translates into full stop and start of the
1208 instance (instead of a call_instance_reboot RPC)
1209 @type shutdown_timeout: integer
1210 @param shutdown_timeout: maximum timeout for soft shutdown
1214 running_instances = GetInstanceList([instance.hypervisor])
1216 if instance.name not in running_instances:
1217 _Fail("Cannot reboot instance %s that is not running", instance.name)
1219 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1220 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1222 hyper.RebootInstance(instance)
1223 except errors.HypervisorError, err:
1224 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1225 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1227 InstanceShutdown(instance, shutdown_timeout)
1228 return StartInstance(instance, False)
1229 except errors.HypervisorError, err:
1230 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1232 _Fail("Invalid reboot_type received: %s", reboot_type)
1235 def MigrationInfo(instance):
1236 """Gather information about an instance to be migrated.
1238 @type instance: L{objects.Instance}
1239 @param instance: the instance definition
1242 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1244 info = hyper.MigrationInfo(instance)
1245 except errors.HypervisorError, err:
1246 _Fail("Failed to fetch migration information: %s", err, exc=True)
1250 def AcceptInstance(instance, info, target):
1251 """Prepare the node to accept an instance.
1253 @type instance: L{objects.Instance}
1254 @param instance: the instance definition
1255 @type info: string/data (opaque)
1256 @param info: migration information, from the source node
1257 @type target: string
1258 @param target: target host (usually ip), on this node
1261 # TODO: why is this required only for DTS_EXT_MIRROR?
1262 if instance.disk_template in constants.DTS_EXT_MIRROR:
1263 # Create the symlinks, as the disks are not active
1266 _GatherAndLinkBlockDevs(instance)
1267 except errors.BlockDeviceError, err:
1268 _Fail("Block device error: %s", err, exc=True)
1270 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1272 hyper.AcceptInstance(instance, info, target)
1273 except errors.HypervisorError, err:
1274 if instance.disk_template in constants.DTS_EXT_MIRROR:
1275 _RemoveBlockDevLinks(instance.name, instance.disks)
1276 _Fail("Failed to accept instance: %s", err, exc=True)
1279 def FinalizeMigration(instance, info, success):
1280 """Finalize any preparation to accept an instance.
1282 @type instance: L{objects.Instance}
1283 @param instance: the instance definition
1284 @type info: string/data (opaque)
1285 @param info: migration information, from the source node
1286 @type success: boolean
1287 @param success: whether the migration was a success or a failure
1290 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1292 hyper.FinalizeMigration(instance, info, success)
1293 except errors.HypervisorError, err:
1294 _Fail("Failed to finalize migration: %s", err, exc=True)
1297 def MigrateInstance(instance, target, live):
1298 """Migrates an instance to another node.
1300 @type instance: L{objects.Instance}
1301 @param instance: the instance definition
1302 @type target: string
1303 @param target: the target node name
1305 @param live: whether the migration should be done live or not (the
1306 interpretation of this parameter is left to the hypervisor)
1308 @return: a tuple of (success, msg) where:
1309 - succes is a boolean denoting the success/failure of the operation
1310 - msg is a string with details in case of failure
1313 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1316 hyper.MigrateInstance(instance, target, live)
1317 except errors.HypervisorError, err:
1318 _Fail("Failed to migrate instance: %s", err, exc=True)
1321 def BlockdevCreate(disk, size, owner, on_primary, info):
1322 """Creates a block device for an instance.
1324 @type disk: L{objects.Disk}
1325 @param disk: the object describing the disk we should create
1327 @param size: the size of the physical underlying device, in MiB
1329 @param owner: the name of the instance for which disk is created,
1330 used for device cache data
1331 @type on_primary: boolean
1332 @param on_primary: indicates if it is the primary node or not
1334 @param info: string that will be sent to the physical device
1335 creation, used for example to set (LVM) tags on LVs
1337 @return: the new unique_id of the device (this can sometime be
1338 computed only after creation), or None. On secondary nodes,
1339 it's not required to return anything.
1342 # TODO: remove the obsolete "size" argument
1343 # pylint: disable-msg=W0613
1346 for child in disk.children:
1348 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1349 except errors.BlockDeviceError, err:
1350 _Fail("Can't assemble device %s: %s", child, err)
1351 if on_primary or disk.AssembleOnSecondary():
1352 # we need the children open in case the device itself has to
1355 # pylint: disable-msg=E1103
1357 except errors.BlockDeviceError, err:
1358 _Fail("Can't make child '%s' read-write: %s", child, err)
1362 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1363 except errors.BlockDeviceError, err:
1364 _Fail("Can't create block device: %s", err)
1366 if on_primary or disk.AssembleOnSecondary():
1369 except errors.BlockDeviceError, err:
1370 _Fail("Can't assemble device after creation, unusual event: %s", err)
1371 device.SetSyncSpeed(constants.SYNC_SPEED)
1372 if on_primary or disk.OpenOnSecondary():
1374 device.Open(force=True)
1375 except errors.BlockDeviceError, err:
1376 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1377 DevCacheManager.UpdateCache(device.dev_path, owner,
1378 on_primary, disk.iv_name)
1380 device.SetInfo(info)
1382 return device.unique_id
1385 def _WipeDevice(path, offset, size):
1386 """This function actually wipes the device.
1388 @param path: The path to the device to wipe
1389 @param offset: The offset in MiB in the file
1390 @param size: The size in MiB to write
1393 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1394 "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1396 result = utils.RunCmd(cmd)
1399 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1400 result.fail_reason, result.output)
1403 def BlockdevWipe(disk, offset, size):
1404 """Wipes a block device.
1406 @type disk: L{objects.Disk}
1407 @param disk: the disk object we want to wipe
1409 @param offset: The offset in MiB in the file
1411 @param size: The size in MiB to write
1415 rdev = _RecursiveFindBD(disk)
1416 except errors.BlockDeviceError:
1420 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1422 # Do cross verify some of the parameters
1423 if offset > rdev.size:
1424 _Fail("Offset is bigger than device size")
1425 if (offset + size) > rdev.size:
1426 _Fail("The provided offset and size to wipe is bigger than device size")
1428 _WipeDevice(rdev.dev_path, offset, size)
1431 def BlockdevPauseResumeSync(disks, pause):
1432 """Pause or resume the sync of the block device.
1434 @type disks: list of L{objects.Disk}
1435 @param disks: the disks object we want to pause/resume
1437 @param pause: Wheater to pause or resume
1443 rdev = _RecursiveFindBD(disk)
1444 except errors.BlockDeviceError:
1448 success.append((False, ("Cannot change sync for device %s:"
1449 " device not found" % disk.iv_name)))
1452 result = rdev.PauseResumeSync(pause)
1455 success.append((result, None))
1461 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1466 def BlockdevRemove(disk):
1467 """Remove a block device.
1469 @note: This is intended to be called recursively.
1471 @type disk: L{objects.Disk}
1472 @param disk: the disk object we should remove
1474 @return: the success of the operation
1479 rdev = _RecursiveFindBD(disk)
1480 except errors.BlockDeviceError, err:
1481 # probably can't attach
1482 logging.info("Can't attach to device %s in remove", disk)
1484 if rdev is not None:
1485 r_path = rdev.dev_path
1488 except errors.BlockDeviceError, err:
1489 msgs.append(str(err))
1491 DevCacheManager.RemoveCache(r_path)
1494 for child in disk.children:
1496 BlockdevRemove(child)
1497 except RPCFail, err:
1498 msgs.append(str(err))
1501 _Fail("; ".join(msgs))
1504 def _RecursiveAssembleBD(disk, owner, as_primary):
1505 """Activate a block device for an instance.
1507 This is run on the primary and secondary nodes for an instance.
1509 @note: this function is called recursively.
1511 @type disk: L{objects.Disk}
1512 @param disk: the disk we try to assemble
1514 @param owner: the name of the instance which owns the disk
1515 @type as_primary: boolean
1516 @param as_primary: if we should make the block device
1519 @return: the assembled device or None (in case no device
1521 @raise errors.BlockDeviceError: in case there is an error
1522 during the activation of the children or the device
1528 mcn = disk.ChildrenNeeded()
1530 mcn = 0 # max number of Nones allowed
1532 mcn = len(disk.children) - mcn # max number of Nones
1533 for chld_disk in disk.children:
1535 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1536 except errors.BlockDeviceError, err:
1537 if children.count(None) >= mcn:
1540 logging.error("Error in child activation (but continuing): %s",
1542 children.append(cdev)
1544 if as_primary or disk.AssembleOnSecondary():
1545 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1546 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1548 if as_primary or disk.OpenOnSecondary():
1550 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1551 as_primary, disk.iv_name)
1558 def BlockdevAssemble(disk, owner, as_primary, idx):
1559 """Activate a block device for an instance.
1561 This is a wrapper over _RecursiveAssembleBD.
1563 @rtype: str or boolean
1564 @return: a C{/dev/...} path for primary nodes, and
1565 C{True} for secondary nodes
1569 result = _RecursiveAssembleBD(disk, owner, as_primary)
1570 if isinstance(result, bdev.BlockDev):
1571 # pylint: disable-msg=E1103
1572 result = result.dev_path
1574 _SymlinkBlockDev(owner, result, idx)
1575 except errors.BlockDeviceError, err:
1576 _Fail("Error while assembling disk: %s", err, exc=True)
1577 except OSError, err:
1578 _Fail("Error while symlinking disk: %s", err, exc=True)
1583 def BlockdevShutdown(disk):
1584 """Shut down a block device.
1586 First, if the device is assembled (Attach() is successful), then
1587 the device is shutdown. Then the children of the device are
1590 This function is called recursively. Note that we don't cache the
1591 children or such, as oppossed to assemble, shutdown of different
1592 devices doesn't require that the upper device was active.
1594 @type disk: L{objects.Disk}
1595 @param disk: the description of the disk we should
1601 r_dev = _RecursiveFindBD(disk)
1602 if r_dev is not None:
1603 r_path = r_dev.dev_path
1606 DevCacheManager.RemoveCache(r_path)
1607 except errors.BlockDeviceError, err:
1608 msgs.append(str(err))
1611 for child in disk.children:
1613 BlockdevShutdown(child)
1614 except RPCFail, err:
1615 msgs.append(str(err))
1618 _Fail("; ".join(msgs))
1621 def BlockdevAddchildren(parent_cdev, new_cdevs):
1622 """Extend a mirrored block device.
1624 @type parent_cdev: L{objects.Disk}
1625 @param parent_cdev: the disk to which we should add children
1626 @type new_cdevs: list of L{objects.Disk}
1627 @param new_cdevs: the list of children which we should add
1631 parent_bdev = _RecursiveFindBD(parent_cdev)
1632 if parent_bdev is None:
1633 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1634 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1635 if new_bdevs.count(None) > 0:
1636 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1637 parent_bdev.AddChildren(new_bdevs)
1640 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1641 """Shrink a mirrored block device.
1643 @type parent_cdev: L{objects.Disk}
1644 @param parent_cdev: the disk from which we should remove children
1645 @type new_cdevs: list of L{objects.Disk}
1646 @param new_cdevs: the list of children which we should remove
1650 parent_bdev = _RecursiveFindBD(parent_cdev)
1651 if parent_bdev is None:
1652 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1654 for disk in new_cdevs:
1655 rpath = disk.StaticDevPath()
1657 bd = _RecursiveFindBD(disk)
1659 _Fail("Can't find device %s while removing children", disk)
1661 devs.append(bd.dev_path)
1663 if not utils.IsNormAbsPath(rpath):
1664 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1666 parent_bdev.RemoveChildren(devs)
1669 def BlockdevGetmirrorstatus(disks):
1670 """Get the mirroring status of a list of devices.
1672 @type disks: list of L{objects.Disk}
1673 @param disks: the list of disks which we should query
1675 @return: List of L{objects.BlockDevStatus}, one for each disk
1676 @raise errors.BlockDeviceError: if any of the disks cannot be
1682 rbd = _RecursiveFindBD(dsk)
1684 _Fail("Can't find device %s", dsk)
1686 stats.append(rbd.CombinedSyncStatus())
1691 def BlockdevGetmirrorstatusMulti(disks):
1692 """Get the mirroring status of a list of devices.
1694 @type disks: list of L{objects.Disk}
1695 @param disks: the list of disks which we should query
1697 @return: List of tuples, (bool, status), one for each disk; bool denotes
1698 success/failure, status is L{objects.BlockDevStatus} on success, string
1705 rbd = _RecursiveFindBD(disk)
1707 result.append((False, "Can't find device %s" % disk))
1710 status = rbd.CombinedSyncStatus()
1711 except errors.BlockDeviceError, err:
1712 logging.exception("Error while getting disk status")
1713 result.append((False, str(err)))
1715 result.append((True, status))
1717 assert len(disks) == len(result)
1722 def _RecursiveFindBD(disk):
1723 """Check if a device is activated.
1725 If so, return information about the real device.
1727 @type disk: L{objects.Disk}
1728 @param disk: the disk object we need to find
1730 @return: None if the device can't be found,
1731 otherwise the device instance
1736 for chdisk in disk.children:
1737 children.append(_RecursiveFindBD(chdisk))
1739 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1742 def _OpenRealBD(disk):
1743 """Opens the underlying block device of a disk.
1745 @type disk: L{objects.Disk}
1746 @param disk: the disk object we want to open
1749 real_disk = _RecursiveFindBD(disk)
1750 if real_disk is None:
1751 _Fail("Block device '%s' is not set up", disk)
1758 def BlockdevFind(disk):
1759 """Check if a device is activated.
1761 If it is, return information about the real device.
1763 @type disk: L{objects.Disk}
1764 @param disk: the disk to find
1765 @rtype: None or objects.BlockDevStatus
1766 @return: None if the disk cannot be found, otherwise a the current
1771 rbd = _RecursiveFindBD(disk)
1772 except errors.BlockDeviceError, err:
1773 _Fail("Failed to find device: %s", err, exc=True)
1778 return rbd.GetSyncStatus()
1781 def BlockdevGetsize(disks):
1782 """Computes the size of the given disks.
1784 If a disk is not found, returns None instead.
1786 @type disks: list of L{objects.Disk}
1787 @param disks: the list of disk to compute the size for
1789 @return: list with elements None if the disk cannot be found,
1796 rbd = _RecursiveFindBD(cf)
1797 except errors.BlockDeviceError:
1803 result.append(rbd.GetActualSize())
1807 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1808 """Export a block device to a remote node.
1810 @type disk: L{objects.Disk}
1811 @param disk: the description of the disk to export
1812 @type dest_node: str
1813 @param dest_node: the destination node to export to
1814 @type dest_path: str
1815 @param dest_path: the destination path on the target node
1816 @type cluster_name: str
1817 @param cluster_name: the cluster name, needed for SSH hostalias
1821 real_disk = _OpenRealBD(disk)
1823 # the block size on the read dd is 1MiB to match our units
1824 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1825 "dd if=%s bs=1048576 count=%s",
1826 real_disk.dev_path, str(disk.size))
1828 # we set here a smaller block size as, due to ssh buffering, more
1829 # than 64-128k will mostly ignored; we use nocreat to fail if the
1830 # device is not already there or we pass a wrong path; we use
1831 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1832 # to not buffer too much memory; this means that at best, we flush
1833 # every 64k, which will not be very fast
1834 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1835 " oflag=dsync", dest_path)
1837 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1838 constants.GANETI_RUNAS,
1841 # all commands have been checked, so we're safe to combine them
1842 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1844 result = utils.RunCmd(["bash", "-c", command])
1847 _Fail("Disk copy command '%s' returned error: %s"
1848 " output: %s", command, result.fail_reason, result.output)
1851 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1852 """Write a file to the filesystem.
1854 This allows the master to overwrite(!) a file. It will only perform
1855 the operation if the file belongs to a list of configuration files.
1857 @type file_name: str
1858 @param file_name: the target file name
1860 @param data: the new contents of the file
1862 @param mode: the mode to give the file (can be None)
1864 @param uid: the owner of the file
1866 @param gid: the group of the file
1868 @param atime: the atime to set on the file (can be None)
1870 @param mtime: the mtime to set on the file (can be None)
1874 if not os.path.isabs(file_name):
1875 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1877 if file_name not in _ALLOWED_UPLOAD_FILES:
1878 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1881 raw_data = _Decompress(data)
1883 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
1884 _Fail("Invalid username/groupname type")
1886 getents = runtime.GetEnts()
1887 uid = getents.LookupUser(uid)
1888 gid = getents.LookupGroup(gid)
1890 utils.SafeWriteFile(file_name, None,
1891 data=raw_data, mode=mode, uid=uid, gid=gid,
1892 atime=atime, mtime=mtime)
1895 def RunOob(oob_program, command, node, timeout):
1896 """Executes oob_program with given command on given node.
1898 @param oob_program: The path to the executable oob_program
1899 @param command: The command to invoke on oob_program
1900 @param node: The node given as an argument to the program
1901 @param timeout: Timeout after which we kill the oob program
1904 @raise RPCFail: If execution fails for some reason
1907 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
1910 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
1911 result.fail_reason, result.output)
1913 return result.stdout
1916 def WriteSsconfFiles(values):
1917 """Update all ssconf files.
1919 Wrapper around the SimpleStore.WriteFiles.
1922 ssconf.SimpleStore().WriteFiles(values)
1925 def _ErrnoOrStr(err):
1926 """Format an EnvironmentError exception.
1928 If the L{err} argument has an errno attribute, it will be looked up
1929 and converted into a textual C{E...} description. Otherwise the
1930 string representation of the error will be returned.
1932 @type err: L{EnvironmentError}
1933 @param err: the exception to format
1936 if hasattr(err, "errno"):
1937 detail = errno.errorcode[err.errno]
1943 def _OSOndiskAPIVersion(os_dir):
1944 """Compute and return the API version of a given OS.
1946 This function will try to read the API version of the OS residing in
1947 the 'os_dir' directory.
1950 @param os_dir: the directory in which we should look for the OS
1952 @return: tuple (status, data) with status denoting the validity and
1953 data holding either the vaid versions or an error message
1956 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1959 st = os.stat(api_file)
1960 except EnvironmentError, err:
1961 return False, ("Required file '%s' not found under path %s: %s" %
1962 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1964 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1965 return False, ("File '%s' in %s is not a regular file" %
1966 (constants.OS_API_FILE, os_dir))
1969 api_versions = utils.ReadFile(api_file).splitlines()
1970 except EnvironmentError, err:
1971 return False, ("Error while reading the API version file at %s: %s" %
1972 (api_file, _ErrnoOrStr(err)))
1975 api_versions = [int(version.strip()) for version in api_versions]
1976 except (TypeError, ValueError), err:
1977 return False, ("API version(s) can't be converted to integer: %s" %
1980 return True, api_versions
1983 def DiagnoseOS(top_dirs=None):
1984 """Compute the validity for all OSes.
1986 @type top_dirs: list
1987 @param top_dirs: the list of directories in which to
1988 search (if not given defaults to
1989 L{constants.OS_SEARCH_PATH})
1990 @rtype: list of L{objects.OS}
1991 @return: a list of tuples (name, path, status, diagnose, variants,
1992 parameters, api_version) for all (potential) OSes under all
1993 search paths, where:
1994 - name is the (potential) OS name
1995 - path is the full path to the OS
1996 - status True/False is the validity of the OS
1997 - diagnose is the error message for an invalid OS, otherwise empty
1998 - variants is a list of supported OS variants, if any
1999 - parameters is a list of (name, help) parameters, if any
2000 - api_version is a list of support OS API versions
2003 if top_dirs is None:
2004 top_dirs = constants.OS_SEARCH_PATH
2007 for dir_name in top_dirs:
2008 if os.path.isdir(dir_name):
2010 f_names = utils.ListVisibleFiles(dir_name)
2011 except EnvironmentError, err:
2012 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2014 for name in f_names:
2015 os_path = utils.PathJoin(dir_name, name)
2016 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2019 variants = os_inst.supported_variants
2020 parameters = os_inst.supported_parameters
2021 api_versions = os_inst.api_versions
2024 variants = parameters = api_versions = []
2025 result.append((name, os_path, status, diagnose, variants,
2026 parameters, api_versions))
2031 def _TryOSFromDisk(name, base_dir=None):
2032 """Create an OS instance from disk.
2034 This function will return an OS instance if the given name is a
2037 @type base_dir: string
2038 @keyword base_dir: Base directory containing OS installations.
2039 Defaults to a search in all the OS_SEARCH_PATH dirs.
2041 @return: success and either the OS instance if we find a valid one,
2045 if base_dir is None:
2046 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2048 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2051 return False, "Directory for OS %s not found in search path" % name
2053 status, api_versions = _OSOndiskAPIVersion(os_dir)
2056 return status, api_versions
2058 if not constants.OS_API_VERSIONS.intersection(api_versions):
2059 return False, ("API version mismatch for path '%s': found %s, want %s." %
2060 (os_dir, api_versions, constants.OS_API_VERSIONS))
2062 # OS Files dictionary, we will populate it with the absolute path
2063 # names; if the value is True, then it is a required file, otherwise
2065 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2067 if max(api_versions) >= constants.OS_API_V15:
2068 os_files[constants.OS_VARIANTS_FILE] = False
2070 if max(api_versions) >= constants.OS_API_V20:
2071 os_files[constants.OS_PARAMETERS_FILE] = True
2073 del os_files[constants.OS_SCRIPT_VERIFY]
2075 for (filename, required) in os_files.items():
2076 os_files[filename] = utils.PathJoin(os_dir, filename)
2079 st = os.stat(os_files[filename])
2080 except EnvironmentError, err:
2081 if err.errno == errno.ENOENT and not required:
2082 del os_files[filename]
2084 return False, ("File '%s' under path '%s' is missing (%s)" %
2085 (filename, os_dir, _ErrnoOrStr(err)))
2087 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2088 return False, ("File '%s' under path '%s' is not a regular file" %
2091 if filename in constants.OS_SCRIPTS:
2092 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2093 return False, ("File '%s' under path '%s' is not executable" %
2097 if constants.OS_VARIANTS_FILE in os_files:
2098 variants_file = os_files[constants.OS_VARIANTS_FILE]
2100 variants = utils.ReadFile(variants_file).splitlines()
2101 except EnvironmentError, err:
2102 # we accept missing files, but not other errors
2103 if err.errno != errno.ENOENT:
2104 return False, ("Error while reading the OS variants file at %s: %s" %
2105 (variants_file, _ErrnoOrStr(err)))
2108 if constants.OS_PARAMETERS_FILE in os_files:
2109 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2111 parameters = utils.ReadFile(parameters_file).splitlines()
2112 except EnvironmentError, err:
2113 return False, ("Error while reading the OS parameters file at %s: %s" %
2114 (parameters_file, _ErrnoOrStr(err)))
2115 parameters = [v.split(None, 1) for v in parameters]
2117 os_obj = objects.OS(name=name, path=os_dir,
2118 create_script=os_files[constants.OS_SCRIPT_CREATE],
2119 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2120 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2121 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2122 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2124 supported_variants=variants,
2125 supported_parameters=parameters,
2126 api_versions=api_versions)
2130 def OSFromDisk(name, base_dir=None):
2131 """Create an OS instance from disk.
2133 This function will return an OS instance if the given name is a
2134 valid OS name. Otherwise, it will raise an appropriate
2135 L{RPCFail} exception, detailing why this is not a valid OS.
2137 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2138 an exception but returns true/false status data.
2140 @type base_dir: string
2141 @keyword base_dir: Base directory containing OS installations.
2142 Defaults to a search in all the OS_SEARCH_PATH dirs.
2143 @rtype: L{objects.OS}
2144 @return: the OS instance if we find a valid one
2145 @raise RPCFail: if we don't find a valid OS
2148 name_only = objects.OS.GetName(name)
2149 status, payload = _TryOSFromDisk(name_only, base_dir)
2157 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2158 """Calculate the basic environment for an os script.
2161 @param os_name: full operating system name (including variant)
2162 @type inst_os: L{objects.OS}
2163 @param inst_os: operating system for which the environment is being built
2164 @type os_params: dict
2165 @param os_params: the OS parameters
2166 @type debug: integer
2167 @param debug: debug level (0 or 1, for OS Api 10)
2169 @return: dict of environment variables
2170 @raise errors.BlockDeviceError: if the block device
2176 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2177 result["OS_API_VERSION"] = "%d" % api_version
2178 result["OS_NAME"] = inst_os.name
2179 result["DEBUG_LEVEL"] = "%d" % debug
2182 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2183 variant = objects.OS.GetVariant(os_name)
2185 variant = inst_os.supported_variants[0]
2188 result["OS_VARIANT"] = variant
2191 for pname, pvalue in os_params.items():
2192 result["OSP_%s" % pname.upper()] = pvalue
2197 def OSEnvironment(instance, inst_os, debug=0):
2198 """Calculate the environment for an os script.
2200 @type instance: L{objects.Instance}
2201 @param instance: target instance for the os script run
2202 @type inst_os: L{objects.OS}
2203 @param inst_os: operating system for which the environment is being built
2204 @type debug: integer
2205 @param debug: debug level (0 or 1, for OS Api 10)
2207 @return: dict of environment variables
2208 @raise errors.BlockDeviceError: if the block device
2212 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2214 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2215 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2217 result["HYPERVISOR"] = instance.hypervisor
2218 result["DISK_COUNT"] = "%d" % len(instance.disks)
2219 result["NIC_COUNT"] = "%d" % len(instance.nics)
2220 result["INSTANCE_SECONDARY_NODES"] = \
2221 ("%s" % " ".join(instance.secondary_nodes))
2224 for idx, disk in enumerate(instance.disks):
2225 real_disk = _OpenRealBD(disk)
2226 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2227 result["DISK_%d_ACCESS" % idx] = disk.mode
2228 if constants.HV_DISK_TYPE in instance.hvparams:
2229 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2230 instance.hvparams[constants.HV_DISK_TYPE]
2231 if disk.dev_type in constants.LDS_BLOCK:
2232 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2233 elif disk.dev_type == constants.LD_FILE:
2234 result["DISK_%d_BACKEND_TYPE" % idx] = \
2235 "file:%s" % disk.physical_id[0]
2238 for idx, nic in enumerate(instance.nics):
2239 result["NIC_%d_MAC" % idx] = nic.mac
2241 result["NIC_%d_IP" % idx] = nic.ip
2242 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2243 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2244 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2245 if nic.nicparams[constants.NIC_LINK]:
2246 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2247 if constants.HV_NIC_TYPE in instance.hvparams:
2248 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2249 instance.hvparams[constants.HV_NIC_TYPE]
2252 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2253 for key, value in source.items():
2254 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2259 def BlockdevGrow(disk, amount, dryrun):
2260 """Grow a stack of block devices.
2262 This function is called recursively, with the childrens being the
2263 first ones to resize.
2265 @type disk: L{objects.Disk}
2266 @param disk: the disk to be grown
2267 @type amount: integer
2268 @param amount: the amount (in mebibytes) to grow with
2269 @type dryrun: boolean
2270 @param dryrun: whether to execute the operation in simulation mode
2271 only, without actually increasing the size
2272 @rtype: (status, result)
2273 @return: a tuple with the status of the operation (True/False), and
2274 the errors message if status is False
2277 r_dev = _RecursiveFindBD(disk)
2279 _Fail("Cannot find block device %s", disk)
2282 r_dev.Grow(amount, dryrun)
2283 except errors.BlockDeviceError, err:
2284 _Fail("Failed to grow block device: %s", err, exc=True)
2287 def BlockdevSnapshot(disk):
2288 """Create a snapshot copy of a block device.
2290 This function is called recursively, and the snapshot is actually created
2291 just for the leaf lvm backend device.
2293 @type disk: L{objects.Disk}
2294 @param disk: the disk to be snapshotted
2296 @return: snapshot disk ID as (vg, lv)
2299 if disk.dev_type == constants.LD_DRBD8:
2300 if not disk.children:
2301 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2303 return BlockdevSnapshot(disk.children[0])
2304 elif disk.dev_type == constants.LD_LV:
2305 r_dev = _RecursiveFindBD(disk)
2306 if r_dev is not None:
2307 # FIXME: choose a saner value for the snapshot size
2308 # let's stay on the safe side and ask for the full size, for now
2309 return r_dev.Snapshot(disk.size)
2311 _Fail("Cannot find block device %s", disk)
2313 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2314 disk.unique_id, disk.dev_type)
2317 def FinalizeExport(instance, snap_disks):
2318 """Write out the export configuration information.
2320 @type instance: L{objects.Instance}
2321 @param instance: the instance which we export, used for
2322 saving configuration
2323 @type snap_disks: list of L{objects.Disk}
2324 @param snap_disks: list of snapshot block devices, which
2325 will be used to get the actual name of the dump file
2330 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2331 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2333 config = objects.SerializableConfigParser()
2335 config.add_section(constants.INISECT_EXP)
2336 config.set(constants.INISECT_EXP, "version", "0")
2337 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2338 config.set(constants.INISECT_EXP, "source", instance.primary_node)
2339 config.set(constants.INISECT_EXP, "os", instance.os)
2340 config.set(constants.INISECT_EXP, "compression", "none")
2342 config.add_section(constants.INISECT_INS)
2343 config.set(constants.INISECT_INS, "name", instance.name)
2344 config.set(constants.INISECT_INS, "memory", "%d" %
2345 instance.beparams[constants.BE_MEMORY])
2346 config.set(constants.INISECT_INS, "vcpus", "%d" %
2347 instance.beparams[constants.BE_VCPUS])
2348 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2349 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2350 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2353 for nic_count, nic in enumerate(instance.nics):
2355 config.set(constants.INISECT_INS, "nic%d_mac" %
2356 nic_count, "%s" % nic.mac)
2357 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2358 for param in constants.NICS_PARAMETER_TYPES:
2359 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2360 "%s" % nic.nicparams.get(param, None))
2361 # TODO: redundant: on load can read nics until it doesn't exist
2362 config.set(constants.INISECT_INS, "nic_count" , "%d" % nic_total)
2365 for disk_count, disk in enumerate(snap_disks):
2368 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2369 ("%s" % disk.iv_name))
2370 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2371 ("%s" % disk.physical_id[1]))
2372 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2375 config.set(constants.INISECT_INS, "disk_count" , "%d" % disk_total)
2377 # New-style hypervisor/backend parameters
2379 config.add_section(constants.INISECT_HYP)
2380 for name, value in instance.hvparams.items():
2381 if name not in constants.HVC_GLOBALS:
2382 config.set(constants.INISECT_HYP, name, str(value))
2384 config.add_section(constants.INISECT_BEP)
2385 for name, value in instance.beparams.items():
2386 config.set(constants.INISECT_BEP, name, str(value))
2388 config.add_section(constants.INISECT_OSP)
2389 for name, value in instance.osparams.items():
2390 config.set(constants.INISECT_OSP, name, str(value))
2392 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2393 data=config.Dumps())
2394 shutil.rmtree(finaldestdir, ignore_errors=True)
2395 shutil.move(destdir, finaldestdir)
2398 def ExportInfo(dest):
2399 """Get export configuration information.
2402 @param dest: directory containing the export
2404 @rtype: L{objects.SerializableConfigParser}
2405 @return: a serializable config file containing the
2409 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2411 config = objects.SerializableConfigParser()
2414 if (not config.has_section(constants.INISECT_EXP) or
2415 not config.has_section(constants.INISECT_INS)):
2416 _Fail("Export info file doesn't have the required fields")
2418 return config.Dumps()
2422 """Return a list of exports currently available on this machine.
2425 @return: list of the exports
2428 if os.path.isdir(constants.EXPORT_DIR):
2429 return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2431 _Fail("No exports directory")
2434 def RemoveExport(export):
2435 """Remove an existing export from the node.
2438 @param export: the name of the export to remove
2442 target = utils.PathJoin(constants.EXPORT_DIR, export)
2445 shutil.rmtree(target)
2446 except EnvironmentError, err:
2447 _Fail("Error while removing the export: %s", err, exc=True)
2450 def BlockdevRename(devlist):
2451 """Rename a list of block devices.
2453 @type devlist: list of tuples
2454 @param devlist: list of tuples of the form (disk,
2455 new_logical_id, new_physical_id); disk is an
2456 L{objects.Disk} object describing the current disk,
2457 and new logical_id/physical_id is the name we
2460 @return: True if all renames succeeded, False otherwise
2465 for disk, unique_id in devlist:
2466 dev = _RecursiveFindBD(disk)
2468 msgs.append("Can't find device %s in rename" % str(disk))
2472 old_rpath = dev.dev_path
2473 dev.Rename(unique_id)
2474 new_rpath = dev.dev_path
2475 if old_rpath != new_rpath:
2476 DevCacheManager.RemoveCache(old_rpath)
2477 # FIXME: we should add the new cache information here, like:
2478 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2479 # but we don't have the owner here - maybe parse from existing
2480 # cache? for now, we only lose lvm data when we rename, which
2481 # is less critical than DRBD or MD
2482 except errors.BlockDeviceError, err:
2483 msgs.append("Can't rename device '%s' to '%s': %s" %
2484 (dev, unique_id, err))
2485 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2488 _Fail("; ".join(msgs))
2491 def _TransformFileStorageDir(fs_dir):
2492 """Checks whether given file_storage_dir is valid.
2494 Checks wheter the given fs_dir is within the cluster-wide default
2495 file_storage_dir or the shared_file_storage_dir, which are stored in
2496 SimpleStore. Only paths under those directories are allowed.
2499 @param fs_dir: the path to check
2501 @return: the normalized path if valid, None otherwise
2504 if not constants.ENABLE_FILE_STORAGE:
2505 _Fail("File storage disabled at configure time")
2507 fs_dir = os.path.normpath(fs_dir)
2508 base_fstore = cfg.GetFileStorageDir()
2509 base_shared = cfg.GetSharedFileStorageDir()
2510 if ((os.path.commonprefix([fs_dir, base_fstore]) != base_fstore) and
2511 (os.path.commonprefix([fs_dir, base_shared]) != base_shared)):
2512 _Fail("File storage directory '%s' is not under base file"
2513 " storage directory '%s' or shared storage directory '%s'",
2514 fs_dir, base_fstore, base_shared)
2518 def CreateFileStorageDir(file_storage_dir):
2519 """Create file storage directory.
2521 @type file_storage_dir: str
2522 @param file_storage_dir: directory to create
2525 @return: tuple with first element a boolean indicating wheter dir
2526 creation was successful or not
2529 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2530 if os.path.exists(file_storage_dir):
2531 if not os.path.isdir(file_storage_dir):
2532 _Fail("Specified storage dir '%s' is not a directory",
2536 os.makedirs(file_storage_dir, 0750)
2537 except OSError, err:
2538 _Fail("Cannot create file storage directory '%s': %s",
2539 file_storage_dir, err, exc=True)
2542 def RemoveFileStorageDir(file_storage_dir):
2543 """Remove file storage directory.
2545 Remove it only if it's empty. If not log an error and return.
2547 @type file_storage_dir: str
2548 @param file_storage_dir: the directory we should cleanup
2549 @rtype: tuple (success,)
2550 @return: tuple of one element, C{success}, denoting
2551 whether the operation was successful
2554 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2555 if os.path.exists(file_storage_dir):
2556 if not os.path.isdir(file_storage_dir):
2557 _Fail("Specified Storage directory '%s' is not a directory",
2559 # deletes dir only if empty, otherwise we want to fail the rpc call
2561 os.rmdir(file_storage_dir)
2562 except OSError, err:
2563 _Fail("Cannot remove file storage directory '%s': %s",
2564 file_storage_dir, err)
2567 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2568 """Rename the file storage directory.
2570 @type old_file_storage_dir: str
2571 @param old_file_storage_dir: the current path
2572 @type new_file_storage_dir: str
2573 @param new_file_storage_dir: the name we should rename to
2574 @rtype: tuple (success,)
2575 @return: tuple of one element, C{success}, denoting
2576 whether the operation was successful
2579 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2580 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2581 if not os.path.exists(new_file_storage_dir):
2582 if os.path.isdir(old_file_storage_dir):
2584 os.rename(old_file_storage_dir, new_file_storage_dir)
2585 except OSError, err:
2586 _Fail("Cannot rename '%s' to '%s': %s",
2587 old_file_storage_dir, new_file_storage_dir, err)
2589 _Fail("Specified storage dir '%s' is not a directory",
2590 old_file_storage_dir)
2592 if os.path.exists(old_file_storage_dir):
2593 _Fail("Cannot rename '%s' to '%s': both locations exist",
2594 old_file_storage_dir, new_file_storage_dir)
2597 def _EnsureJobQueueFile(file_name):
2598 """Checks whether the given filename is in the queue directory.
2600 @type file_name: str
2601 @param file_name: the file name we should check
2603 @raises RPCFail: if the file is not valid
2606 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2607 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2610 _Fail("Passed job queue file '%s' does not belong to"
2611 " the queue directory '%s'", file_name, queue_dir)
2614 def JobQueueUpdate(file_name, content):
2615 """Updates a file in the queue directory.
2617 This is just a wrapper over L{utils.io.WriteFile}, with proper
2620 @type file_name: str
2621 @param file_name: the job file name
2623 @param content: the new job contents
2625 @return: the success of the operation
2628 _EnsureJobQueueFile(file_name)
2629 getents = runtime.GetEnts()
2631 # Write and replace the file atomically
2632 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2633 gid=getents.masterd_gid)
2636 def JobQueueRename(old, new):
2637 """Renames a job queue file.
2639 This is just a wrapper over os.rename with proper checking.
2642 @param old: the old (actual) file name
2644 @param new: the desired file name
2646 @return: the success of the operation and payload
2649 _EnsureJobQueueFile(old)
2650 _EnsureJobQueueFile(new)
2652 utils.RenameFile(old, new, mkdir=True)
2655 def BlockdevClose(instance_name, disks):
2656 """Closes the given block devices.
2658 This means they will be switched to secondary mode (in case of
2661 @param instance_name: if the argument is not empty, the symlinks
2662 of this instance will be removed
2663 @type disks: list of L{objects.Disk}
2664 @param disks: the list of disks to be closed
2665 @rtype: tuple (success, message)
2666 @return: a tuple of success and message, where success
2667 indicates the succes of the operation, and message
2668 which will contain the error details in case we
2674 rd = _RecursiveFindBD(cf)
2676 _Fail("Can't find device %s", cf)
2683 except errors.BlockDeviceError, err:
2684 msg.append(str(err))
2686 _Fail("Can't make devices secondary: %s", ",".join(msg))
2689 _RemoveBlockDevLinks(instance_name, disks)
2692 def ValidateHVParams(hvname, hvparams):
2693 """Validates the given hypervisor parameters.
2695 @type hvname: string
2696 @param hvname: the hypervisor name
2697 @type hvparams: dict
2698 @param hvparams: the hypervisor parameters to be validated
2703 hv_type = hypervisor.GetHypervisor(hvname)
2704 hv_type.ValidateParameters(hvparams)
2705 except errors.HypervisorError, err:
2706 _Fail(str(err), log=False)
2709 def _CheckOSPList(os_obj, parameters):
2710 """Check whether a list of parameters is supported by the OS.
2712 @type os_obj: L{objects.OS}
2713 @param os_obj: OS object to check
2714 @type parameters: list
2715 @param parameters: the list of parameters to check
2718 supported = [v[0] for v in os_obj.supported_parameters]
2719 delta = frozenset(parameters).difference(supported)
2721 _Fail("The following parameters are not supported"
2722 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2725 def ValidateOS(required, osname, checks, osparams):
2726 """Validate the given OS' parameters.
2728 @type required: boolean
2729 @param required: whether absence of the OS should translate into
2731 @type osname: string
2732 @param osname: the OS to be validated
2734 @param checks: list of the checks to run (currently only 'parameters')
2735 @type osparams: dict
2736 @param osparams: dictionary with OS parameters
2738 @return: True if the validation passed, or False if the OS was not
2739 found and L{required} was false
2742 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2743 _Fail("Unknown checks required for OS %s: %s", osname,
2744 set(checks).difference(constants.OS_VALIDATE_CALLS))
2746 name_only = objects.OS.GetName(osname)
2747 status, tbv = _TryOSFromDisk(name_only, None)
2755 if max(tbv.api_versions) < constants.OS_API_V20:
2758 if constants.OS_VALIDATE_PARAMETERS in checks:
2759 _CheckOSPList(tbv, osparams.keys())
2761 validate_env = OSCoreEnv(osname, tbv, osparams)
2762 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2763 cwd=tbv.path, reset_env=True)
2765 logging.error("os validate command '%s' returned error: %s output: %s",
2766 result.cmd, result.fail_reason, result.output)
2767 _Fail("OS validation script failed (%s), output: %s",
2768 result.fail_reason, result.output, log=False)
2774 """Demotes the current node from master candidate role.
2777 # try to ensure we're not the master by mistake
2778 master, myself = ssconf.GetMasterAndMyself()
2779 if master == myself:
2780 _Fail("ssconf status shows I'm the master node, will not demote")
2782 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2783 if not result.failed:
2784 _Fail("The master daemon is running, will not demote")
2787 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2788 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2789 except EnvironmentError, err:
2790 if err.errno != errno.ENOENT:
2791 _Fail("Error while backing up cluster file: %s", err, exc=True)
2793 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2796 def _GetX509Filenames(cryptodir, name):
2797 """Returns the full paths for the private key and certificate.
2800 return (utils.PathJoin(cryptodir, name),
2801 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2802 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2805 def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2806 """Creates a new X509 certificate for SSL/TLS.
2809 @param validity: Validity in seconds
2810 @rtype: tuple; (string, string)
2811 @return: Certificate name and public part
2814 (key_pem, cert_pem) = \
2815 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2816 min(validity, _MAX_SSL_CERT_VALIDITY))
2818 cert_dir = tempfile.mkdtemp(dir=cryptodir,
2819 prefix="x509-%s-" % utils.TimestampForFilename())
2821 name = os.path.basename(cert_dir)
2822 assert len(name) > 5
2824 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2826 utils.WriteFile(key_file, mode=0400, data=key_pem)
2827 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2829 # Never return private key as it shouldn't leave the node
2830 return (name, cert_pem)
2832 shutil.rmtree(cert_dir, ignore_errors=True)
2836 def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2837 """Removes a X509 certificate.
2840 @param name: Certificate name
2843 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2845 utils.RemoveFile(key_file)
2846 utils.RemoveFile(cert_file)
2850 except EnvironmentError, err:
2851 _Fail("Cannot remove certificate directory '%s': %s",
2855 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2856 """Returns the command for the requested input/output.
2858 @type instance: L{objects.Instance}
2859 @param instance: The instance object
2860 @param mode: Import/export mode
2861 @param ieio: Input/output type
2862 @param ieargs: Input/output arguments
2865 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2872 if ieio == constants.IEIO_FILE:
2873 (filename, ) = ieargs
2875 if not utils.IsNormAbsPath(filename):
2876 _Fail("Path '%s' is not normalized or absolute", filename)
2878 directory = os.path.normpath(os.path.dirname(filename))
2880 if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2881 constants.EXPORT_DIR):
2882 _Fail("File '%s' is not under exports directory '%s'",
2883 filename, constants.EXPORT_DIR)
2886 utils.Makedirs(directory, mode=0750)
2888 quoted_filename = utils.ShellQuote(filename)
2890 if mode == constants.IEM_IMPORT:
2891 suffix = "> %s" % quoted_filename
2892 elif mode == constants.IEM_EXPORT:
2893 suffix = "< %s" % quoted_filename
2895 # Retrieve file size
2897 st = os.stat(filename)
2898 except EnvironmentError, err:
2899 logging.error("Can't stat(2) %s: %s", filename, err)
2901 exp_size = utils.BytesToMebibyte(st.st_size)
2903 elif ieio == constants.IEIO_RAW_DISK:
2906 real_disk = _OpenRealBD(disk)
2908 if mode == constants.IEM_IMPORT:
2909 # we set here a smaller block size as, due to transport buffering, more
2910 # than 64-128k will mostly ignored; we use nocreat to fail if the device
2911 # is not already there or we pass a wrong path; we use notrunc to no
2912 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2913 # much memory; this means that at best, we flush every 64k, which will
2915 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2916 " bs=%s oflag=dsync"),
2920 elif mode == constants.IEM_EXPORT:
2921 # the block size on the read dd is 1MiB to match our units
2922 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2924 str(1024 * 1024), # 1 MB
2926 exp_size = disk.size
2928 elif ieio == constants.IEIO_SCRIPT:
2929 (disk, disk_index, ) = ieargs
2931 assert isinstance(disk_index, (int, long))
2933 real_disk = _OpenRealBD(disk)
2935 inst_os = OSFromDisk(instance.os)
2936 env = OSEnvironment(instance, inst_os)
2938 if mode == constants.IEM_IMPORT:
2939 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2940 env["IMPORT_INDEX"] = str(disk_index)
2941 script = inst_os.import_script
2943 elif mode == constants.IEM_EXPORT:
2944 env["EXPORT_DEVICE"] = real_disk.dev_path
2945 env["EXPORT_INDEX"] = str(disk_index)
2946 script = inst_os.export_script
2948 # TODO: Pass special environment only to script
2949 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2951 if mode == constants.IEM_IMPORT:
2952 suffix = "| %s" % script_cmd
2954 elif mode == constants.IEM_EXPORT:
2955 prefix = "%s |" % script_cmd
2957 # Let script predict size
2958 exp_size = constants.IE_CUSTOM_SIZE
2961 _Fail("Invalid %s I/O mode %r", mode, ieio)
2963 return (env, prefix, suffix, exp_size)
2966 def _CreateImportExportStatusDir(prefix):
2967 """Creates status directory for import/export.
2970 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2972 (prefix, utils.TimestampForFilename())))
2975 def StartImportExportDaemon(mode, opts, host, port, instance, component,
2977 """Starts an import or export daemon.
2979 @param mode: Import/output mode
2980 @type opts: L{objects.ImportExportOptions}
2981 @param opts: Daemon options
2983 @param host: Remote host for export (None for import)
2985 @param port: Remote port for export (None for import)
2986 @type instance: L{objects.Instance}
2987 @param instance: Instance object
2988 @type component: string
2989 @param component: which part of the instance is transferred now,
2991 @param ieio: Input/output type
2992 @param ieioargs: Input/output arguments
2995 if mode == constants.IEM_IMPORT:
2998 if not (host is None and port is None):
2999 _Fail("Can not specify host or port on import")
3001 elif mode == constants.IEM_EXPORT:
3004 if host is None or port is None:
3005 _Fail("Host and port must be specified for an export")
3008 _Fail("Invalid mode %r", mode)
3010 if (opts.key_name is None) ^ (opts.ca_pem is None):
3011 _Fail("Cluster certificate can only be used for both key and CA")
3013 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3014 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3016 if opts.key_name is None:
3018 key_path = constants.NODED_CERT_FILE
3019 cert_path = constants.NODED_CERT_FILE
3020 assert opts.ca_pem is None
3022 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
3024 assert opts.ca_pem is not None
3026 for i in [key_path, cert_path]:
3027 if not os.path.exists(i):
3028 _Fail("File '%s' does not exist" % i)
3030 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3032 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3033 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3034 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3036 if opts.ca_pem is None:
3038 ca = utils.ReadFile(constants.NODED_CERT_FILE)
3043 utils.WriteFile(ca_file, data=ca, mode=0400)
3046 constants.IMPORT_EXPORT_DAEMON,
3048 "--key=%s" % key_path,
3049 "--cert=%s" % cert_path,
3050 "--ca=%s" % ca_file,
3054 cmd.append("--host=%s" % host)
3057 cmd.append("--port=%s" % port)
3060 cmd.append("--ipv6")
3062 cmd.append("--ipv4")
3065 cmd.append("--compress=%s" % opts.compress)
3068 cmd.append("--magic=%s" % opts.magic)
3070 if exp_size is not None:
3071 cmd.append("--expected-size=%s" % exp_size)
3074 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3077 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3079 if mode == constants.IEM_EXPORT:
3080 # Retry connection a few times when connecting to remote peer
3081 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3082 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3083 elif opts.connect_timeout is not None:
3084 assert mode == constants.IEM_IMPORT
3085 # Overall timeout for establishing connection while listening
3086 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3088 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3090 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3091 # support for receiving a file descriptor for output
3092 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3095 # The import/export name is simply the status directory name
3096 return os.path.basename(status_dir)
3099 shutil.rmtree(status_dir, ignore_errors=True)
3103 def GetImportExportStatus(names):
3104 """Returns import/export daemon status.
3106 @type names: sequence
3107 @param names: List of names
3108 @rtype: List of dicts
3109 @return: Returns a list of the state of each named import/export or None if a
3110 status couldn't be read
3116 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3120 data = utils.ReadFile(status_file)
3121 except EnvironmentError, err:
3122 if err.errno != errno.ENOENT:
3130 result.append(serializer.LoadJson(data))
3135 def AbortImportExport(name):
3136 """Sends SIGTERM to a running import/export daemon.
3139 logging.info("Abort import/export %s", name)
3141 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3142 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3145 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3147 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3150 def CleanupImportExport(name):
3151 """Cleanup after an import or export.
3153 If the import/export daemon is still running it's killed. Afterwards the
3154 whole status directory is removed.
3157 logging.info("Finalizing import/export %s", name)
3159 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3161 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3164 logging.info("Import/export %s is still running with PID %s",
3166 utils.KillProcess(pid, waitpid=False)
3168 shutil.rmtree(status_dir, ignore_errors=True)
3171 def _FindDisks(nodes_ip, disks):
3172 """Sets the physical ID on disks and returns the block devices.
3175 # set the correct physical ID
3176 my_name = netutils.Hostname.GetSysName()
3178 cf.SetPhysicalID(my_name, nodes_ip)
3183 rd = _RecursiveFindBD(cf)
3185 _Fail("Can't find device %s", cf)
3190 def DrbdDisconnectNet(nodes_ip, disks):
3191 """Disconnects the network on a list of drbd devices.
3194 bdevs = _FindDisks(nodes_ip, disks)
3200 except errors.BlockDeviceError, err:
3201 _Fail("Can't change network configuration to standalone mode: %s",
3205 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3206 """Attaches the network on a list of drbd devices.
3209 bdevs = _FindDisks(nodes_ip, disks)
3212 for idx, rd in enumerate(bdevs):
3214 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3215 except EnvironmentError, err:
3216 _Fail("Can't create symlink: %s", err)
3217 # reconnect disks, switch to new master configuration and if
3218 # needed primary mode
3221 rd.AttachNet(multimaster)
3222 except errors.BlockDeviceError, err:
3223 _Fail("Can't change network configuration: %s", err)
3225 # wait until the disks are connected; we need to retry the re-attach
3226 # if the device becomes standalone, as this might happen if the one
3227 # node disconnects and reconnects in a different mode before the
3228 # other node reconnects; in this case, one or both of the nodes will
3229 # decide it has wrong configuration and switch to standalone
3232 all_connected = True
3235 stats = rd.GetProcStatus()
3237 all_connected = (all_connected and
3238 (stats.is_connected or stats.is_in_resync))
3240 if stats.is_standalone:
3241 # peer had different config info and this node became
3242 # standalone, even though this should not happen with the
3243 # new staged way of changing disk configs
3245 rd.AttachNet(multimaster)
3246 except errors.BlockDeviceError, err:
3247 _Fail("Can't change network configuration: %s", err)
3249 if not all_connected:
3250 raise utils.RetryAgain()
3253 # Start with a delay of 100 miliseconds and go up to 5 seconds
3254 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3255 except utils.RetryTimeout:
3256 _Fail("Timeout in disk reconnecting")
3259 # change to primary mode
3263 except errors.BlockDeviceError, err:
3264 _Fail("Can't change to primary mode: %s", err)
3267 def DrbdWaitSync(nodes_ip, disks):
3268 """Wait until DRBDs have synchronized.
3272 stats = rd.GetProcStatus()
3273 if not (stats.is_connected or stats.is_in_resync):
3274 raise utils.RetryAgain()
3277 bdevs = _FindDisks(nodes_ip, disks)
3283 # poll each second for 15 seconds
3284 stats = utils.Retry(_helper, 1, 15, args=[rd])
3285 except utils.RetryTimeout:
3286 stats = rd.GetProcStatus()
3288 if not (stats.is_connected or stats.is_in_resync):
3289 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3290 alldone = alldone and (not stats.is_in_resync)
3291 if stats.sync_percent is not None:
3292 min_resync = min(min_resync, stats.sync_percent)
3294 return (alldone, min_resync)
3297 def GetDrbdUsermodeHelper():
3298 """Returns DRBD usermode helper currently configured.
3302 return bdev.BaseDRBD.GetUsermodeHelper()
3303 except errors.BlockDeviceError, err:
3307 def PowercycleNode(hypervisor_type):
3308 """Hard-powercycle the node.
3310 Because we need to return first, and schedule the powercycle in the
3311 background, we won't be able to report failures nicely.
3314 hyper = hypervisor.GetHypervisor(hypervisor_type)
3318 # if we can't fork, we'll pretend that we're in the child process
3321 return "Reboot scheduled in 5 seconds"
3322 # ensure the child is running on ram
3325 except Exception: # pylint: disable-msg=W0703
3328 hyper.PowercycleNode()
3331 class HooksRunner(object):
3334 This class is instantiated on the node side (ganeti-noded) and not
3338 def __init__(self, hooks_base_dir=None):
3339 """Constructor for hooks runner.
3341 @type hooks_base_dir: str or None
3342 @param hooks_base_dir: if not None, this overrides the
3343 L{constants.HOOKS_BASE_DIR} (useful for unittests)
3346 if hooks_base_dir is None:
3347 hooks_base_dir = constants.HOOKS_BASE_DIR
3348 # yeah, _BASE_DIR is not valid for attributes, we use it like a
3350 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3352 def RunHooks(self, hpath, phase, env):
3353 """Run the scripts in the hooks directory.
3356 @param hpath: the path to the hooks directory which
3359 @param phase: either L{constants.HOOKS_PHASE_PRE} or
3360 L{constants.HOOKS_PHASE_POST}
3362 @param env: dictionary with the environment for the hook
3364 @return: list of 3-element tuples:
3366 - script result, either L{constants.HKR_SUCCESS} or
3367 L{constants.HKR_FAIL}
3368 - output of the script
3370 @raise errors.ProgrammerError: for invalid input
3374 if phase == constants.HOOKS_PHASE_PRE:
3376 elif phase == constants.HOOKS_PHASE_POST:
3379 _Fail("Unknown hooks phase '%s'", phase)
3382 subdir = "%s-%s.d" % (hpath, suffix)
3383 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3387 if not os.path.isdir(dir_name):
3388 # for non-existing/non-dirs, we simply exit instead of logging a
3389 # warning at every operation
3392 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3394 for (relname, relstatus, runresult) in runparts_results:
3395 if relstatus == constants.RUNPARTS_SKIP:
3396 rrval = constants.HKR_SKIP
3398 elif relstatus == constants.RUNPARTS_ERR:
3399 rrval = constants.HKR_FAIL
3400 output = "Hook script execution error: %s" % runresult
3401 elif relstatus == constants.RUNPARTS_RUN:
3402 if runresult.failed:
3403 rrval = constants.HKR_FAIL
3405 rrval = constants.HKR_SUCCESS
3406 output = utils.SafeEncode(runresult.output.strip())
3407 results.append(("%s/%s" % (subdir, relname), rrval, output))
3412 class IAllocatorRunner(object):
3413 """IAllocator runner.
3415 This class is instantiated on the node side (ganeti-noded) and not on
3420 def Run(name, idata):
3421 """Run an iallocator script.
3424 @param name: the iallocator script name
3426 @param idata: the allocator input data
3429 @return: two element tuple of:
3431 - either error message or stdout of allocator (for success)
3434 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3436 if alloc_script is None:
3437 _Fail("iallocator module '%s' not found in the search path", name)
3439 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3443 result = utils.RunCmd([alloc_script, fin_name])
3445 _Fail("iallocator module '%s' failed: %s, output '%s'",
3446 name, result.fail_reason, result.output)
3450 return result.stdout
3453 class DevCacheManager(object):
3454 """Simple class for managing a cache of block device information.
3457 _DEV_PREFIX = "/dev/"
3458 _ROOT_DIR = constants.BDEV_CACHE_DIR
3461 def _ConvertPath(cls, dev_path):
3462 """Converts a /dev/name path to the cache file name.
3464 This replaces slashes with underscores and strips the /dev
3465 prefix. It then returns the full path to the cache file.
3468 @param dev_path: the C{/dev/} path name
3470 @return: the converted path name
3473 if dev_path.startswith(cls._DEV_PREFIX):
3474 dev_path = dev_path[len(cls._DEV_PREFIX):]
3475 dev_path = dev_path.replace("/", "_")
3476 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3480 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3481 """Updates the cache information for a given device.
3484 @param dev_path: the pathname of the device
3486 @param owner: the owner (instance name) of the device
3487 @type on_primary: bool
3488 @param on_primary: whether this is the primary
3491 @param iv_name: the instance-visible name of the
3492 device, as in objects.Disk.iv_name
3497 if dev_path is None:
3498 logging.error("DevCacheManager.UpdateCache got a None dev_path")
3500 fpath = cls._ConvertPath(dev_path)
3506 iv_name = "not_visible"
3507 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3509 utils.WriteFile(fpath, data=fdata)
3510 except EnvironmentError, err:
3511 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3514 def RemoveCache(cls, dev_path):
3515 """Remove data for a dev_path.
3517 This is just a wrapper over L{utils.io.RemoveFile} with a converted
3518 path name and logging.
3521 @param dev_path: the pathname of the device
3526 if dev_path is None:
3527 logging.error("DevCacheManager.RemoveCache got a None dev_path")
3529 fpath = cls._ConvertPath(dev_path)
3531 utils.RemoveFile(fpath)
3532 except EnvironmentError, err:
3533 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)