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(["ip", "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(["ip", "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):
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.
942 # TODO: Use tempfile.mkstemp to create unique filename
943 base = ("%s-%s-%s-%s.log" %
944 (kind, os_name, instance, utils.TimestampForFilename()))
945 return utils.PathJoin(constants.LOG_OS_DIR, base)
948 def InstanceOsAdd(instance, reinstall, debug):
949 """Add an OS to an instance.
951 @type instance: L{objects.Instance}
952 @param instance: Instance whose OS is to be installed
953 @type reinstall: boolean
954 @param reinstall: whether this is an instance reinstall
956 @param debug: debug level, passed to the OS scripts
960 inst_os = OSFromDisk(instance.os)
962 create_env = OSEnvironment(instance, inst_os, debug)
964 create_env["INSTANCE_REINSTALL"] = "1"
966 logfile = _InstanceLogName("add", instance.os, instance.name)
968 result = utils.RunCmd([inst_os.create_script], env=create_env,
969 cwd=inst_os.path, output=logfile, reset_env=True)
971 logging.error("os create command '%s' returned error: %s, logfile: %s,"
972 " output: %s", result.cmd, result.fail_reason, logfile,
974 lines = [utils.SafeEncode(val)
975 for val in utils.TailFile(logfile, lines=20)]
976 _Fail("OS create script failed (%s), last lines in the"
977 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
980 def RunRenameInstance(instance, old_name, debug):
981 """Run the OS rename script for an instance.
983 @type instance: L{objects.Instance}
984 @param instance: Instance whose OS is to be installed
985 @type old_name: string
986 @param old_name: previous instance name
988 @param debug: debug level, passed to the OS scripts
990 @return: the success of the operation
993 inst_os = OSFromDisk(instance.os)
995 rename_env = OSEnvironment(instance, inst_os, debug)
996 rename_env["OLD_INSTANCE_NAME"] = old_name
998 logfile = _InstanceLogName("rename", instance.os,
999 "%s-%s" % (old_name, instance.name))
1001 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1002 cwd=inst_os.path, output=logfile, reset_env=True)
1005 logging.error("os create command '%s' returned error: %s output: %s",
1006 result.cmd, result.fail_reason, result.output)
1007 lines = [utils.SafeEncode(val)
1008 for val in utils.TailFile(logfile, lines=20)]
1009 _Fail("OS rename script failed (%s), last lines in the"
1010 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1013 def _GetBlockDevSymlinkPath(instance_name, idx):
1014 return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1015 (instance_name, constants.DISK_SEPARATOR, idx))
1018 def _SymlinkBlockDev(instance_name, device_path, idx):
1019 """Set up symlinks to a instance's block device.
1021 This is an auxiliary function run when an instance is start (on the primary
1022 node) or when an instance is migrated (on the target node).
1025 @param instance_name: the name of the target instance
1026 @param device_path: path of the physical block device, on the node
1027 @param idx: the disk index
1028 @return: absolute path to the disk's symlink
1031 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1033 os.symlink(device_path, link_name)
1034 except OSError, err:
1035 if err.errno == errno.EEXIST:
1036 if (not os.path.islink(link_name) or
1037 os.readlink(link_name) != device_path):
1038 os.remove(link_name)
1039 os.symlink(device_path, link_name)
1046 def _RemoveBlockDevLinks(instance_name, disks):
1047 """Remove the block device symlinks belonging to the given instance.
1050 for idx, _ in enumerate(disks):
1051 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1052 if os.path.islink(link_name):
1054 os.remove(link_name)
1056 logging.exception("Can't remove symlink '%s'", link_name)
1059 def _GatherAndLinkBlockDevs(instance):
1060 """Set up an instance's block device(s).
1062 This is run on the primary node at instance startup. The block
1063 devices must be already assembled.
1065 @type instance: L{objects.Instance}
1066 @param instance: the instance whose disks we shoul assemble
1068 @return: list of (disk_object, device_path)
1072 for idx, disk in enumerate(instance.disks):
1073 device = _RecursiveFindBD(disk)
1075 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1079 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1081 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1084 block_devices.append((disk, link_name))
1086 return block_devices
1089 def StartInstance(instance, startup_paused):
1090 """Start an instance.
1092 @type instance: L{objects.Instance}
1093 @param instance: the instance object
1094 @type startup_paused: bool
1095 @param instance: pause instance at startup?
1099 running_instances = GetInstanceList([instance.hypervisor])
1101 if instance.name in running_instances:
1102 logging.info("Instance %s already running, not starting", instance.name)
1106 block_devices = _GatherAndLinkBlockDevs(instance)
1107 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1108 hyper.StartInstance(instance, block_devices, startup_paused)
1109 except errors.BlockDeviceError, err:
1110 _Fail("Block device error: %s", err, exc=True)
1111 except errors.HypervisorError, err:
1112 _RemoveBlockDevLinks(instance.name, instance.disks)
1113 _Fail("Hypervisor error: %s", err, exc=True)
1116 def InstanceShutdown(instance, timeout):
1117 """Shut an instance down.
1119 @note: this functions uses polling with a hardcoded timeout.
1121 @type instance: L{objects.Instance}
1122 @param instance: the instance object
1123 @type timeout: integer
1124 @param timeout: maximum timeout for soft shutdown
1128 hv_name = instance.hypervisor
1129 hyper = hypervisor.GetHypervisor(hv_name)
1130 iname = instance.name
1132 if instance.name not in hyper.ListInstances():
1133 logging.info("Instance %s not running, doing nothing", iname)
1138 self.tried_once = False
1141 if iname not in hyper.ListInstances():
1145 hyper.StopInstance(instance, retry=self.tried_once)
1146 except errors.HypervisorError, err:
1147 if iname not in hyper.ListInstances():
1148 # if the instance is no longer existing, consider this a
1149 # success and go to cleanup
1152 _Fail("Failed to stop instance %s: %s", iname, err)
1154 self.tried_once = True
1156 raise utils.RetryAgain()
1159 utils.Retry(_TryShutdown(), 5, timeout)
1160 except utils.RetryTimeout:
1161 # the shutdown did not succeed
1162 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1165 hyper.StopInstance(instance, force=True)
1166 except errors.HypervisorError, err:
1167 if iname in hyper.ListInstances():
1168 # only raise an error if the instance still exists, otherwise
1169 # the error could simply be "instance ... unknown"!
1170 _Fail("Failed to force stop instance %s: %s", iname, err)
1174 if iname in hyper.ListInstances():
1175 _Fail("Could not shutdown instance %s even by destroy", iname)
1178 hyper.CleanupInstance(instance.name)
1179 except errors.HypervisorError, err:
1180 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1182 _RemoveBlockDevLinks(iname, instance.disks)
1185 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1186 """Reboot an instance.
1188 @type instance: L{objects.Instance}
1189 @param instance: the instance object to reboot
1190 @type reboot_type: str
1191 @param reboot_type: the type of reboot, one the following
1193 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1194 instance OS, do not recreate the VM
1195 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1196 restart the VM (at the hypervisor level)
1197 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1198 not accepted here, since that mode is handled differently, in
1199 cmdlib, and translates into full stop and start of the
1200 instance (instead of a call_instance_reboot RPC)
1201 @type shutdown_timeout: integer
1202 @param shutdown_timeout: maximum timeout for soft shutdown
1206 running_instances = GetInstanceList([instance.hypervisor])
1208 if instance.name not in running_instances:
1209 _Fail("Cannot reboot instance %s that is not running", instance.name)
1211 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1212 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1214 hyper.RebootInstance(instance)
1215 except errors.HypervisorError, err:
1216 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1217 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1219 InstanceShutdown(instance, shutdown_timeout)
1220 return StartInstance(instance, False)
1221 except errors.HypervisorError, err:
1222 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1224 _Fail("Invalid reboot_type received: %s", reboot_type)
1227 def MigrationInfo(instance):
1228 """Gather information about an instance to be migrated.
1230 @type instance: L{objects.Instance}
1231 @param instance: the instance definition
1234 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1236 info = hyper.MigrationInfo(instance)
1237 except errors.HypervisorError, err:
1238 _Fail("Failed to fetch migration information: %s", err, exc=True)
1242 def AcceptInstance(instance, info, target):
1243 """Prepare the node to accept an instance.
1245 @type instance: L{objects.Instance}
1246 @param instance: the instance definition
1247 @type info: string/data (opaque)
1248 @param info: migration information, from the source node
1249 @type target: string
1250 @param target: target host (usually ip), on this node
1253 # TODO: why is this required only for DTS_EXT_MIRROR?
1254 if instance.disk_template in constants.DTS_EXT_MIRROR:
1255 # Create the symlinks, as the disks are not active
1258 _GatherAndLinkBlockDevs(instance)
1259 except errors.BlockDeviceError, err:
1260 _Fail("Block device error: %s", err, exc=True)
1262 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1264 hyper.AcceptInstance(instance, info, target)
1265 except errors.HypervisorError, err:
1266 if instance.disk_template in constants.DTS_EXT_MIRROR:
1267 _RemoveBlockDevLinks(instance.name, instance.disks)
1268 _Fail("Failed to accept instance: %s", err, exc=True)
1271 def FinalizeMigration(instance, info, success):
1272 """Finalize any preparation to accept an instance.
1274 @type instance: L{objects.Instance}
1275 @param instance: the instance definition
1276 @type info: string/data (opaque)
1277 @param info: migration information, from the source node
1278 @type success: boolean
1279 @param success: whether the migration was a success or a failure
1282 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1284 hyper.FinalizeMigration(instance, info, success)
1285 except errors.HypervisorError, err:
1286 _Fail("Failed to finalize migration: %s", err, exc=True)
1289 def MigrateInstance(instance, target, live):
1290 """Migrates an instance to another node.
1292 @type instance: L{objects.Instance}
1293 @param instance: the instance definition
1294 @type target: string
1295 @param target: the target node name
1297 @param live: whether the migration should be done live or not (the
1298 interpretation of this parameter is left to the hypervisor)
1300 @return: a tuple of (success, msg) where:
1301 - succes is a boolean denoting the success/failure of the operation
1302 - msg is a string with details in case of failure
1305 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1308 hyper.MigrateInstance(instance, target, live)
1309 except errors.HypervisorError, err:
1310 _Fail("Failed to migrate instance: %s", err, exc=True)
1313 def BlockdevCreate(disk, size, owner, on_primary, info):
1314 """Creates a block device for an instance.
1316 @type disk: L{objects.Disk}
1317 @param disk: the object describing the disk we should create
1319 @param size: the size of the physical underlying device, in MiB
1321 @param owner: the name of the instance for which disk is created,
1322 used for device cache data
1323 @type on_primary: boolean
1324 @param on_primary: indicates if it is the primary node or not
1326 @param info: string that will be sent to the physical device
1327 creation, used for example to set (LVM) tags on LVs
1329 @return: the new unique_id of the device (this can sometime be
1330 computed only after creation), or None. On secondary nodes,
1331 it's not required to return anything.
1334 # TODO: remove the obsolete "size" argument
1335 # pylint: disable-msg=W0613
1338 for child in disk.children:
1340 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1341 except errors.BlockDeviceError, err:
1342 _Fail("Can't assemble device %s: %s", child, err)
1343 if on_primary or disk.AssembleOnSecondary():
1344 # we need the children open in case the device itself has to
1347 # pylint: disable-msg=E1103
1349 except errors.BlockDeviceError, err:
1350 _Fail("Can't make child '%s' read-write: %s", child, err)
1354 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1355 except errors.BlockDeviceError, err:
1356 _Fail("Can't create block device: %s", err)
1358 if on_primary or disk.AssembleOnSecondary():
1361 except errors.BlockDeviceError, err:
1362 _Fail("Can't assemble device after creation, unusual event: %s", err)
1363 device.SetSyncSpeed(constants.SYNC_SPEED)
1364 if on_primary or disk.OpenOnSecondary():
1366 device.Open(force=True)
1367 except errors.BlockDeviceError, err:
1368 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1369 DevCacheManager.UpdateCache(device.dev_path, owner,
1370 on_primary, disk.iv_name)
1372 device.SetInfo(info)
1374 return device.unique_id
1377 def _WipeDevice(path, offset, size):
1378 """This function actually wipes the device.
1380 @param path: The path to the device to wipe
1381 @param offset: The offset in MiB in the file
1382 @param size: The size in MiB to write
1385 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1386 "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1388 result = utils.RunCmd(cmd)
1391 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1392 result.fail_reason, result.output)
1395 def BlockdevWipe(disk, offset, size):
1396 """Wipes a block device.
1398 @type disk: L{objects.Disk}
1399 @param disk: the disk object we want to wipe
1401 @param offset: The offset in MiB in the file
1403 @param size: The size in MiB to write
1407 rdev = _RecursiveFindBD(disk)
1408 except errors.BlockDeviceError:
1412 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1414 # Do cross verify some of the parameters
1415 if offset > rdev.size:
1416 _Fail("Offset is bigger than device size")
1417 if (offset + size) > rdev.size:
1418 _Fail("The provided offset and size to wipe is bigger than device size")
1420 _WipeDevice(rdev.dev_path, offset, size)
1423 def BlockdevPauseResumeSync(disks, pause):
1424 """Pause or resume the sync of the block device.
1426 @type disks: list of L{objects.Disk}
1427 @param disks: the disks object we want to pause/resume
1429 @param pause: Wheater to pause or resume
1435 rdev = _RecursiveFindBD(disk)
1436 except errors.BlockDeviceError:
1440 success.append((False, ("Cannot change sync for device %s:"
1441 " device not found" % disk.iv_name)))
1444 result = rdev.PauseResumeSync(pause)
1447 success.append((result, None))
1453 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1458 def BlockdevRemove(disk):
1459 """Remove a block device.
1461 @note: This is intended to be called recursively.
1463 @type disk: L{objects.Disk}
1464 @param disk: the disk object we should remove
1466 @return: the success of the operation
1471 rdev = _RecursiveFindBD(disk)
1472 except errors.BlockDeviceError, err:
1473 # probably can't attach
1474 logging.info("Can't attach to device %s in remove", disk)
1476 if rdev is not None:
1477 r_path = rdev.dev_path
1480 except errors.BlockDeviceError, err:
1481 msgs.append(str(err))
1483 DevCacheManager.RemoveCache(r_path)
1486 for child in disk.children:
1488 BlockdevRemove(child)
1489 except RPCFail, err:
1490 msgs.append(str(err))
1493 _Fail("; ".join(msgs))
1496 def _RecursiveAssembleBD(disk, owner, as_primary):
1497 """Activate a block device for an instance.
1499 This is run on the primary and secondary nodes for an instance.
1501 @note: this function is called recursively.
1503 @type disk: L{objects.Disk}
1504 @param disk: the disk we try to assemble
1506 @param owner: the name of the instance which owns the disk
1507 @type as_primary: boolean
1508 @param as_primary: if we should make the block device
1511 @return: the assembled device or None (in case no device
1513 @raise errors.BlockDeviceError: in case there is an error
1514 during the activation of the children or the device
1520 mcn = disk.ChildrenNeeded()
1522 mcn = 0 # max number of Nones allowed
1524 mcn = len(disk.children) - mcn # max number of Nones
1525 for chld_disk in disk.children:
1527 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1528 except errors.BlockDeviceError, err:
1529 if children.count(None) >= mcn:
1532 logging.error("Error in child activation (but continuing): %s",
1534 children.append(cdev)
1536 if as_primary or disk.AssembleOnSecondary():
1537 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1538 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1540 if as_primary or disk.OpenOnSecondary():
1542 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1543 as_primary, disk.iv_name)
1550 def BlockdevAssemble(disk, owner, as_primary, idx):
1551 """Activate a block device for an instance.
1553 This is a wrapper over _RecursiveAssembleBD.
1555 @rtype: str or boolean
1556 @return: a C{/dev/...} path for primary nodes, and
1557 C{True} for secondary nodes
1561 result = _RecursiveAssembleBD(disk, owner, as_primary)
1562 if isinstance(result, bdev.BlockDev):
1563 # pylint: disable-msg=E1103
1564 result = result.dev_path
1566 _SymlinkBlockDev(owner, result, idx)
1567 except errors.BlockDeviceError, err:
1568 _Fail("Error while assembling disk: %s", err, exc=True)
1569 except OSError, err:
1570 _Fail("Error while symlinking disk: %s", err, exc=True)
1575 def BlockdevShutdown(disk):
1576 """Shut down a block device.
1578 First, if the device is assembled (Attach() is successful), then
1579 the device is shutdown. Then the children of the device are
1582 This function is called recursively. Note that we don't cache the
1583 children or such, as oppossed to assemble, shutdown of different
1584 devices doesn't require that the upper device was active.
1586 @type disk: L{objects.Disk}
1587 @param disk: the description of the disk we should
1593 r_dev = _RecursiveFindBD(disk)
1594 if r_dev is not None:
1595 r_path = r_dev.dev_path
1598 DevCacheManager.RemoveCache(r_path)
1599 except errors.BlockDeviceError, err:
1600 msgs.append(str(err))
1603 for child in disk.children:
1605 BlockdevShutdown(child)
1606 except RPCFail, err:
1607 msgs.append(str(err))
1610 _Fail("; ".join(msgs))
1613 def BlockdevAddchildren(parent_cdev, new_cdevs):
1614 """Extend a mirrored block device.
1616 @type parent_cdev: L{objects.Disk}
1617 @param parent_cdev: the disk to which we should add children
1618 @type new_cdevs: list of L{objects.Disk}
1619 @param new_cdevs: the list of children which we should add
1623 parent_bdev = _RecursiveFindBD(parent_cdev)
1624 if parent_bdev is None:
1625 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1626 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1627 if new_bdevs.count(None) > 0:
1628 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1629 parent_bdev.AddChildren(new_bdevs)
1632 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1633 """Shrink a mirrored block device.
1635 @type parent_cdev: L{objects.Disk}
1636 @param parent_cdev: the disk from which we should remove children
1637 @type new_cdevs: list of L{objects.Disk}
1638 @param new_cdevs: the list of children which we should remove
1642 parent_bdev = _RecursiveFindBD(parent_cdev)
1643 if parent_bdev is None:
1644 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1646 for disk in new_cdevs:
1647 rpath = disk.StaticDevPath()
1649 bd = _RecursiveFindBD(disk)
1651 _Fail("Can't find device %s while removing children", disk)
1653 devs.append(bd.dev_path)
1655 if not utils.IsNormAbsPath(rpath):
1656 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1658 parent_bdev.RemoveChildren(devs)
1661 def BlockdevGetmirrorstatus(disks):
1662 """Get the mirroring status of a list of devices.
1664 @type disks: list of L{objects.Disk}
1665 @param disks: the list of disks which we should query
1667 @return: List of L{objects.BlockDevStatus}, one for each disk
1668 @raise errors.BlockDeviceError: if any of the disks cannot be
1674 rbd = _RecursiveFindBD(dsk)
1676 _Fail("Can't find device %s", dsk)
1678 stats.append(rbd.CombinedSyncStatus())
1683 def BlockdevGetmirrorstatusMulti(disks):
1684 """Get the mirroring status of a list of devices.
1686 @type disks: list of L{objects.Disk}
1687 @param disks: the list of disks which we should query
1689 @return: List of tuples, (bool, status), one for each disk; bool denotes
1690 success/failure, status is L{objects.BlockDevStatus} on success, string
1697 rbd = _RecursiveFindBD(disk)
1699 result.append((False, "Can't find device %s" % disk))
1702 status = rbd.CombinedSyncStatus()
1703 except errors.BlockDeviceError, err:
1704 logging.exception("Error while getting disk status")
1705 result.append((False, str(err)))
1707 result.append((True, status))
1709 assert len(disks) == len(result)
1714 def _RecursiveFindBD(disk):
1715 """Check if a device is activated.
1717 If so, return information about the real device.
1719 @type disk: L{objects.Disk}
1720 @param disk: the disk object we need to find
1722 @return: None if the device can't be found,
1723 otherwise the device instance
1728 for chdisk in disk.children:
1729 children.append(_RecursiveFindBD(chdisk))
1731 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1734 def _OpenRealBD(disk):
1735 """Opens the underlying block device of a disk.
1737 @type disk: L{objects.Disk}
1738 @param disk: the disk object we want to open
1741 real_disk = _RecursiveFindBD(disk)
1742 if real_disk is None:
1743 _Fail("Block device '%s' is not set up", disk)
1750 def BlockdevFind(disk):
1751 """Check if a device is activated.
1753 If it is, return information about the real device.
1755 @type disk: L{objects.Disk}
1756 @param disk: the disk to find
1757 @rtype: None or objects.BlockDevStatus
1758 @return: None if the disk cannot be found, otherwise a the current
1763 rbd = _RecursiveFindBD(disk)
1764 except errors.BlockDeviceError, err:
1765 _Fail("Failed to find device: %s", err, exc=True)
1770 return rbd.GetSyncStatus()
1773 def BlockdevGetsize(disks):
1774 """Computes the size of the given disks.
1776 If a disk is not found, returns None instead.
1778 @type disks: list of L{objects.Disk}
1779 @param disks: the list of disk to compute the size for
1781 @return: list with elements None if the disk cannot be found,
1788 rbd = _RecursiveFindBD(cf)
1789 except errors.BlockDeviceError:
1795 result.append(rbd.GetActualSize())
1799 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1800 """Export a block device to a remote node.
1802 @type disk: L{objects.Disk}
1803 @param disk: the description of the disk to export
1804 @type dest_node: str
1805 @param dest_node: the destination node to export to
1806 @type dest_path: str
1807 @param dest_path: the destination path on the target node
1808 @type cluster_name: str
1809 @param cluster_name: the cluster name, needed for SSH hostalias
1813 real_disk = _OpenRealBD(disk)
1815 # the block size on the read dd is 1MiB to match our units
1816 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1817 "dd if=%s bs=1048576 count=%s",
1818 real_disk.dev_path, str(disk.size))
1820 # we set here a smaller block size as, due to ssh buffering, more
1821 # than 64-128k will mostly ignored; we use nocreat to fail if the
1822 # device is not already there or we pass a wrong path; we use
1823 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1824 # to not buffer too much memory; this means that at best, we flush
1825 # every 64k, which will not be very fast
1826 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1827 " oflag=dsync", dest_path)
1829 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1830 constants.GANETI_RUNAS,
1833 # all commands have been checked, so we're safe to combine them
1834 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1836 result = utils.RunCmd(["bash", "-c", command])
1839 _Fail("Disk copy command '%s' returned error: %s"
1840 " output: %s", command, result.fail_reason, result.output)
1843 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1844 """Write a file to the filesystem.
1846 This allows the master to overwrite(!) a file. It will only perform
1847 the operation if the file belongs to a list of configuration files.
1849 @type file_name: str
1850 @param file_name: the target file name
1852 @param data: the new contents of the file
1854 @param mode: the mode to give the file (can be None)
1856 @param uid: the owner of the file
1858 @param gid: the group of the file
1860 @param atime: the atime to set on the file (can be None)
1862 @param mtime: the mtime to set on the file (can be None)
1866 if not os.path.isabs(file_name):
1867 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1869 if file_name not in _ALLOWED_UPLOAD_FILES:
1870 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1873 raw_data = _Decompress(data)
1875 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
1876 _Fail("Invalid username/groupname type")
1878 getents = runtime.GetEnts()
1879 uid = getents.LookupUser(uid)
1880 gid = getents.LookupGroup(gid)
1882 utils.SafeWriteFile(file_name, None,
1883 data=raw_data, mode=mode, uid=uid, gid=gid,
1884 atime=atime, mtime=mtime)
1887 def RunOob(oob_program, command, node, timeout):
1888 """Executes oob_program with given command on given node.
1890 @param oob_program: The path to the executable oob_program
1891 @param command: The command to invoke on oob_program
1892 @param node: The node given as an argument to the program
1893 @param timeout: Timeout after which we kill the oob program
1896 @raise RPCFail: If execution fails for some reason
1899 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
1902 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
1903 result.fail_reason, result.output)
1905 return result.stdout
1908 def WriteSsconfFiles(values):
1909 """Update all ssconf files.
1911 Wrapper around the SimpleStore.WriteFiles.
1914 ssconf.SimpleStore().WriteFiles(values)
1917 def _ErrnoOrStr(err):
1918 """Format an EnvironmentError exception.
1920 If the L{err} argument has an errno attribute, it will be looked up
1921 and converted into a textual C{E...} description. Otherwise the
1922 string representation of the error will be returned.
1924 @type err: L{EnvironmentError}
1925 @param err: the exception to format
1928 if hasattr(err, "errno"):
1929 detail = errno.errorcode[err.errno]
1935 def _OSOndiskAPIVersion(os_dir):
1936 """Compute and return the API version of a given OS.
1938 This function will try to read the API version of the OS residing in
1939 the 'os_dir' directory.
1942 @param os_dir: the directory in which we should look for the OS
1944 @return: tuple (status, data) with status denoting the validity and
1945 data holding either the vaid versions or an error message
1948 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1951 st = os.stat(api_file)
1952 except EnvironmentError, err:
1953 return False, ("Required file '%s' not found under path %s: %s" %
1954 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1956 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1957 return False, ("File '%s' in %s is not a regular file" %
1958 (constants.OS_API_FILE, os_dir))
1961 api_versions = utils.ReadFile(api_file).splitlines()
1962 except EnvironmentError, err:
1963 return False, ("Error while reading the API version file at %s: %s" %
1964 (api_file, _ErrnoOrStr(err)))
1967 api_versions = [int(version.strip()) for version in api_versions]
1968 except (TypeError, ValueError), err:
1969 return False, ("API version(s) can't be converted to integer: %s" %
1972 return True, api_versions
1975 def DiagnoseOS(top_dirs=None):
1976 """Compute the validity for all OSes.
1978 @type top_dirs: list
1979 @param top_dirs: the list of directories in which to
1980 search (if not given defaults to
1981 L{constants.OS_SEARCH_PATH})
1982 @rtype: list of L{objects.OS}
1983 @return: a list of tuples (name, path, status, diagnose, variants,
1984 parameters, api_version) for all (potential) OSes under all
1985 search paths, where:
1986 - name is the (potential) OS name
1987 - path is the full path to the OS
1988 - status True/False is the validity of the OS
1989 - diagnose is the error message for an invalid OS, otherwise empty
1990 - variants is a list of supported OS variants, if any
1991 - parameters is a list of (name, help) parameters, if any
1992 - api_version is a list of support OS API versions
1995 if top_dirs is None:
1996 top_dirs = constants.OS_SEARCH_PATH
1999 for dir_name in top_dirs:
2000 if os.path.isdir(dir_name):
2002 f_names = utils.ListVisibleFiles(dir_name)
2003 except EnvironmentError, err:
2004 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2006 for name in f_names:
2007 os_path = utils.PathJoin(dir_name, name)
2008 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2011 variants = os_inst.supported_variants
2012 parameters = os_inst.supported_parameters
2013 api_versions = os_inst.api_versions
2016 variants = parameters = api_versions = []
2017 result.append((name, os_path, status, diagnose, variants,
2018 parameters, api_versions))
2023 def _TryOSFromDisk(name, base_dir=None):
2024 """Create an OS instance from disk.
2026 This function will return an OS instance if the given name is a
2029 @type base_dir: string
2030 @keyword base_dir: Base directory containing OS installations.
2031 Defaults to a search in all the OS_SEARCH_PATH dirs.
2033 @return: success and either the OS instance if we find a valid one,
2037 if base_dir is None:
2038 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2040 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2043 return False, "Directory for OS %s not found in search path" % name
2045 status, api_versions = _OSOndiskAPIVersion(os_dir)
2048 return status, api_versions
2050 if not constants.OS_API_VERSIONS.intersection(api_versions):
2051 return False, ("API version mismatch for path '%s': found %s, want %s." %
2052 (os_dir, api_versions, constants.OS_API_VERSIONS))
2054 # OS Files dictionary, we will populate it with the absolute path
2055 # names; if the value is True, then it is a required file, otherwise
2057 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2059 if max(api_versions) >= constants.OS_API_V15:
2060 os_files[constants.OS_VARIANTS_FILE] = False
2062 if max(api_versions) >= constants.OS_API_V20:
2063 os_files[constants.OS_PARAMETERS_FILE] = True
2065 del os_files[constants.OS_SCRIPT_VERIFY]
2067 for (filename, required) in os_files.items():
2068 os_files[filename] = utils.PathJoin(os_dir, filename)
2071 st = os.stat(os_files[filename])
2072 except EnvironmentError, err:
2073 if err.errno == errno.ENOENT and not required:
2074 del os_files[filename]
2076 return False, ("File '%s' under path '%s' is missing (%s)" %
2077 (filename, os_dir, _ErrnoOrStr(err)))
2079 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2080 return False, ("File '%s' under path '%s' is not a regular file" %
2083 if filename in constants.OS_SCRIPTS:
2084 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2085 return False, ("File '%s' under path '%s' is not executable" %
2089 if constants.OS_VARIANTS_FILE in os_files:
2090 variants_file = os_files[constants.OS_VARIANTS_FILE]
2092 variants = utils.ReadFile(variants_file).splitlines()
2093 except EnvironmentError, err:
2094 # we accept missing files, but not other errors
2095 if err.errno != errno.ENOENT:
2096 return False, ("Error while reading the OS variants file at %s: %s" %
2097 (variants_file, _ErrnoOrStr(err)))
2100 if constants.OS_PARAMETERS_FILE in os_files:
2101 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2103 parameters = utils.ReadFile(parameters_file).splitlines()
2104 except EnvironmentError, err:
2105 return False, ("Error while reading the OS parameters file at %s: %s" %
2106 (parameters_file, _ErrnoOrStr(err)))
2107 parameters = [v.split(None, 1) for v in parameters]
2109 os_obj = objects.OS(name=name, path=os_dir,
2110 create_script=os_files[constants.OS_SCRIPT_CREATE],
2111 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2112 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2113 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2114 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2116 supported_variants=variants,
2117 supported_parameters=parameters,
2118 api_versions=api_versions)
2122 def OSFromDisk(name, base_dir=None):
2123 """Create an OS instance from disk.
2125 This function will return an OS instance if the given name is a
2126 valid OS name. Otherwise, it will raise an appropriate
2127 L{RPCFail} exception, detailing why this is not a valid OS.
2129 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2130 an exception but returns true/false status data.
2132 @type base_dir: string
2133 @keyword base_dir: Base directory containing OS installations.
2134 Defaults to a search in all the OS_SEARCH_PATH dirs.
2135 @rtype: L{objects.OS}
2136 @return: the OS instance if we find a valid one
2137 @raise RPCFail: if we don't find a valid OS
2140 name_only = objects.OS.GetName(name)
2141 status, payload = _TryOSFromDisk(name_only, base_dir)
2149 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2150 """Calculate the basic environment for an os script.
2153 @param os_name: full operating system name (including variant)
2154 @type inst_os: L{objects.OS}
2155 @param inst_os: operating system for which the environment is being built
2156 @type os_params: dict
2157 @param os_params: the OS parameters
2158 @type debug: integer
2159 @param debug: debug level (0 or 1, for OS Api 10)
2161 @return: dict of environment variables
2162 @raise errors.BlockDeviceError: if the block device
2168 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2169 result["OS_API_VERSION"] = "%d" % api_version
2170 result["OS_NAME"] = inst_os.name
2171 result["DEBUG_LEVEL"] = "%d" % debug
2174 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2175 variant = objects.OS.GetVariant(os_name)
2177 variant = inst_os.supported_variants[0]
2180 result["OS_VARIANT"] = variant
2183 for pname, pvalue in os_params.items():
2184 result["OSP_%s" % pname.upper()] = pvalue
2189 def OSEnvironment(instance, inst_os, debug=0):
2190 """Calculate the environment for an os script.
2192 @type instance: L{objects.Instance}
2193 @param instance: target instance for the os script run
2194 @type inst_os: L{objects.OS}
2195 @param inst_os: operating system for which the environment is being built
2196 @type debug: integer
2197 @param debug: debug level (0 or 1, for OS Api 10)
2199 @return: dict of environment variables
2200 @raise errors.BlockDeviceError: if the block device
2204 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2206 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2207 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2209 result["HYPERVISOR"] = instance.hypervisor
2210 result["DISK_COUNT"] = "%d" % len(instance.disks)
2211 result["NIC_COUNT"] = "%d" % len(instance.nics)
2212 result["INSTANCE_SECONDARY_NODES"] = \
2213 ("%s" % " ".join(instance.secondary_nodes))
2216 for idx, disk in enumerate(instance.disks):
2217 real_disk = _OpenRealBD(disk)
2218 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2219 result["DISK_%d_ACCESS" % idx] = disk.mode
2220 if constants.HV_DISK_TYPE in instance.hvparams:
2221 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2222 instance.hvparams[constants.HV_DISK_TYPE]
2223 if disk.dev_type in constants.LDS_BLOCK:
2224 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2225 elif disk.dev_type == constants.LD_FILE:
2226 result["DISK_%d_BACKEND_TYPE" % idx] = \
2227 "file:%s" % disk.physical_id[0]
2230 for idx, nic in enumerate(instance.nics):
2231 result["NIC_%d_MAC" % idx] = nic.mac
2233 result["NIC_%d_IP" % idx] = nic.ip
2234 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2235 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2236 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2237 if nic.nicparams[constants.NIC_LINK]:
2238 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2239 if constants.HV_NIC_TYPE in instance.hvparams:
2240 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2241 instance.hvparams[constants.HV_NIC_TYPE]
2244 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2245 for key, value in source.items():
2246 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2251 def BlockdevGrow(disk, amount, dryrun):
2252 """Grow a stack of block devices.
2254 This function is called recursively, with the childrens being the
2255 first ones to resize.
2257 @type disk: L{objects.Disk}
2258 @param disk: the disk to be grown
2259 @type amount: integer
2260 @param amount: the amount (in mebibytes) to grow with
2261 @type dryrun: boolean
2262 @param dryrun: whether to execute the operation in simulation mode
2263 only, without actually increasing the size
2264 @rtype: (status, result)
2265 @return: a tuple with the status of the operation (True/False), and
2266 the errors message if status is False
2269 r_dev = _RecursiveFindBD(disk)
2271 _Fail("Cannot find block device %s", disk)
2274 r_dev.Grow(amount, dryrun)
2275 except errors.BlockDeviceError, err:
2276 _Fail("Failed to grow block device: %s", err, exc=True)
2279 def BlockdevSnapshot(disk):
2280 """Create a snapshot copy of a block device.
2282 This function is called recursively, and the snapshot is actually created
2283 just for the leaf lvm backend device.
2285 @type disk: L{objects.Disk}
2286 @param disk: the disk to be snapshotted
2288 @return: snapshot disk ID as (vg, lv)
2291 if disk.dev_type == constants.LD_DRBD8:
2292 if not disk.children:
2293 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2295 return BlockdevSnapshot(disk.children[0])
2296 elif disk.dev_type == constants.LD_LV:
2297 r_dev = _RecursiveFindBD(disk)
2298 if r_dev is not None:
2299 # FIXME: choose a saner value for the snapshot size
2300 # let's stay on the safe side and ask for the full size, for now
2301 return r_dev.Snapshot(disk.size)
2303 _Fail("Cannot find block device %s", disk)
2305 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2306 disk.unique_id, disk.dev_type)
2309 def FinalizeExport(instance, snap_disks):
2310 """Write out the export configuration information.
2312 @type instance: L{objects.Instance}
2313 @param instance: the instance which we export, used for
2314 saving configuration
2315 @type snap_disks: list of L{objects.Disk}
2316 @param snap_disks: list of snapshot block devices, which
2317 will be used to get the actual name of the dump file
2322 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2323 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2325 config = objects.SerializableConfigParser()
2327 config.add_section(constants.INISECT_EXP)
2328 config.set(constants.INISECT_EXP, "version", "0")
2329 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2330 config.set(constants.INISECT_EXP, "source", instance.primary_node)
2331 config.set(constants.INISECT_EXP, "os", instance.os)
2332 config.set(constants.INISECT_EXP, "compression", "none")
2334 config.add_section(constants.INISECT_INS)
2335 config.set(constants.INISECT_INS, "name", instance.name)
2336 config.set(constants.INISECT_INS, "memory", "%d" %
2337 instance.beparams[constants.BE_MEMORY])
2338 config.set(constants.INISECT_INS, "vcpus", "%d" %
2339 instance.beparams[constants.BE_VCPUS])
2340 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2341 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2342 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2345 for nic_count, nic in enumerate(instance.nics):
2347 config.set(constants.INISECT_INS, "nic%d_mac" %
2348 nic_count, "%s" % nic.mac)
2349 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2350 for param in constants.NICS_PARAMETER_TYPES:
2351 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2352 "%s" % nic.nicparams.get(param, None))
2353 # TODO: redundant: on load can read nics until it doesn't exist
2354 config.set(constants.INISECT_INS, "nic_count" , "%d" % nic_total)
2357 for disk_count, disk in enumerate(snap_disks):
2360 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2361 ("%s" % disk.iv_name))
2362 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2363 ("%s" % disk.physical_id[1]))
2364 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2367 config.set(constants.INISECT_INS, "disk_count" , "%d" % disk_total)
2369 # New-style hypervisor/backend parameters
2371 config.add_section(constants.INISECT_HYP)
2372 for name, value in instance.hvparams.items():
2373 if name not in constants.HVC_GLOBALS:
2374 config.set(constants.INISECT_HYP, name, str(value))
2376 config.add_section(constants.INISECT_BEP)
2377 for name, value in instance.beparams.items():
2378 config.set(constants.INISECT_BEP, name, str(value))
2380 config.add_section(constants.INISECT_OSP)
2381 for name, value in instance.osparams.items():
2382 config.set(constants.INISECT_OSP, name, str(value))
2384 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2385 data=config.Dumps())
2386 shutil.rmtree(finaldestdir, ignore_errors=True)
2387 shutil.move(destdir, finaldestdir)
2390 def ExportInfo(dest):
2391 """Get export configuration information.
2394 @param dest: directory containing the export
2396 @rtype: L{objects.SerializableConfigParser}
2397 @return: a serializable config file containing the
2401 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2403 config = objects.SerializableConfigParser()
2406 if (not config.has_section(constants.INISECT_EXP) or
2407 not config.has_section(constants.INISECT_INS)):
2408 _Fail("Export info file doesn't have the required fields")
2410 return config.Dumps()
2414 """Return a list of exports currently available on this machine.
2417 @return: list of the exports
2420 if os.path.isdir(constants.EXPORT_DIR):
2421 return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2423 _Fail("No exports directory")
2426 def RemoveExport(export):
2427 """Remove an existing export from the node.
2430 @param export: the name of the export to remove
2434 target = utils.PathJoin(constants.EXPORT_DIR, export)
2437 shutil.rmtree(target)
2438 except EnvironmentError, err:
2439 _Fail("Error while removing the export: %s", err, exc=True)
2442 def BlockdevRename(devlist):
2443 """Rename a list of block devices.
2445 @type devlist: list of tuples
2446 @param devlist: list of tuples of the form (disk,
2447 new_logical_id, new_physical_id); disk is an
2448 L{objects.Disk} object describing the current disk,
2449 and new logical_id/physical_id is the name we
2452 @return: True if all renames succeeded, False otherwise
2457 for disk, unique_id in devlist:
2458 dev = _RecursiveFindBD(disk)
2460 msgs.append("Can't find device %s in rename" % str(disk))
2464 old_rpath = dev.dev_path
2465 dev.Rename(unique_id)
2466 new_rpath = dev.dev_path
2467 if old_rpath != new_rpath:
2468 DevCacheManager.RemoveCache(old_rpath)
2469 # FIXME: we should add the new cache information here, like:
2470 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2471 # but we don't have the owner here - maybe parse from existing
2472 # cache? for now, we only lose lvm data when we rename, which
2473 # is less critical than DRBD or MD
2474 except errors.BlockDeviceError, err:
2475 msgs.append("Can't rename device '%s' to '%s': %s" %
2476 (dev, unique_id, err))
2477 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2480 _Fail("; ".join(msgs))
2483 def _TransformFileStorageDir(fs_dir):
2484 """Checks whether given file_storage_dir is valid.
2486 Checks wheter the given fs_dir is within the cluster-wide default
2487 file_storage_dir or the shared_file_storage_dir, which are stored in
2488 SimpleStore. Only paths under those directories are allowed.
2491 @param fs_dir: the path to check
2493 @return: the normalized path if valid, None otherwise
2496 if not constants.ENABLE_FILE_STORAGE:
2497 _Fail("File storage disabled at configure time")
2499 fs_dir = os.path.normpath(fs_dir)
2500 base_fstore = cfg.GetFileStorageDir()
2501 base_shared = cfg.GetSharedFileStorageDir()
2502 if ((os.path.commonprefix([fs_dir, base_fstore]) != base_fstore) and
2503 (os.path.commonprefix([fs_dir, base_shared]) != base_shared)):
2504 _Fail("File storage directory '%s' is not under base file"
2505 " storage directory '%s' or shared storage directory '%s'",
2506 fs_dir, base_fstore, base_shared)
2510 def CreateFileStorageDir(file_storage_dir):
2511 """Create file storage directory.
2513 @type file_storage_dir: str
2514 @param file_storage_dir: directory to create
2517 @return: tuple with first element a boolean indicating wheter dir
2518 creation was successful or not
2521 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2522 if os.path.exists(file_storage_dir):
2523 if not os.path.isdir(file_storage_dir):
2524 _Fail("Specified storage dir '%s' is not a directory",
2528 os.makedirs(file_storage_dir, 0750)
2529 except OSError, err:
2530 _Fail("Cannot create file storage directory '%s': %s",
2531 file_storage_dir, err, exc=True)
2534 def RemoveFileStorageDir(file_storage_dir):
2535 """Remove file storage directory.
2537 Remove it only if it's empty. If not log an error and return.
2539 @type file_storage_dir: str
2540 @param file_storage_dir: the directory we should cleanup
2541 @rtype: tuple (success,)
2542 @return: tuple of one element, C{success}, denoting
2543 whether the operation was successful
2546 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2547 if os.path.exists(file_storage_dir):
2548 if not os.path.isdir(file_storage_dir):
2549 _Fail("Specified Storage directory '%s' is not a directory",
2551 # deletes dir only if empty, otherwise we want to fail the rpc call
2553 os.rmdir(file_storage_dir)
2554 except OSError, err:
2555 _Fail("Cannot remove file storage directory '%s': %s",
2556 file_storage_dir, err)
2559 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2560 """Rename the file storage directory.
2562 @type old_file_storage_dir: str
2563 @param old_file_storage_dir: the current path
2564 @type new_file_storage_dir: str
2565 @param new_file_storage_dir: the name we should rename to
2566 @rtype: tuple (success,)
2567 @return: tuple of one element, C{success}, denoting
2568 whether the operation was successful
2571 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2572 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2573 if not os.path.exists(new_file_storage_dir):
2574 if os.path.isdir(old_file_storage_dir):
2576 os.rename(old_file_storage_dir, new_file_storage_dir)
2577 except OSError, err:
2578 _Fail("Cannot rename '%s' to '%s': %s",
2579 old_file_storage_dir, new_file_storage_dir, err)
2581 _Fail("Specified storage dir '%s' is not a directory",
2582 old_file_storage_dir)
2584 if os.path.exists(old_file_storage_dir):
2585 _Fail("Cannot rename '%s' to '%s': both locations exist",
2586 old_file_storage_dir, new_file_storage_dir)
2589 def _EnsureJobQueueFile(file_name):
2590 """Checks whether the given filename is in the queue directory.
2592 @type file_name: str
2593 @param file_name: the file name we should check
2595 @raises RPCFail: if the file is not valid
2598 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2599 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2602 _Fail("Passed job queue file '%s' does not belong to"
2603 " the queue directory '%s'", file_name, queue_dir)
2606 def JobQueueUpdate(file_name, content):
2607 """Updates a file in the queue directory.
2609 This is just a wrapper over L{utils.io.WriteFile}, with proper
2612 @type file_name: str
2613 @param file_name: the job file name
2615 @param content: the new job contents
2617 @return: the success of the operation
2620 _EnsureJobQueueFile(file_name)
2621 getents = runtime.GetEnts()
2623 # Write and replace the file atomically
2624 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2625 gid=getents.masterd_gid)
2628 def JobQueueRename(old, new):
2629 """Renames a job queue file.
2631 This is just a wrapper over os.rename with proper checking.
2634 @param old: the old (actual) file name
2636 @param new: the desired file name
2638 @return: the success of the operation and payload
2641 _EnsureJobQueueFile(old)
2642 _EnsureJobQueueFile(new)
2644 utils.RenameFile(old, new, mkdir=True)
2647 def BlockdevClose(instance_name, disks):
2648 """Closes the given block devices.
2650 This means they will be switched to secondary mode (in case of
2653 @param instance_name: if the argument is not empty, the symlinks
2654 of this instance will be removed
2655 @type disks: list of L{objects.Disk}
2656 @param disks: the list of disks to be closed
2657 @rtype: tuple (success, message)
2658 @return: a tuple of success and message, where success
2659 indicates the succes of the operation, and message
2660 which will contain the error details in case we
2666 rd = _RecursiveFindBD(cf)
2668 _Fail("Can't find device %s", cf)
2675 except errors.BlockDeviceError, err:
2676 msg.append(str(err))
2678 _Fail("Can't make devices secondary: %s", ",".join(msg))
2681 _RemoveBlockDevLinks(instance_name, disks)
2684 def ValidateHVParams(hvname, hvparams):
2685 """Validates the given hypervisor parameters.
2687 @type hvname: string
2688 @param hvname: the hypervisor name
2689 @type hvparams: dict
2690 @param hvparams: the hypervisor parameters to be validated
2695 hv_type = hypervisor.GetHypervisor(hvname)
2696 hv_type.ValidateParameters(hvparams)
2697 except errors.HypervisorError, err:
2698 _Fail(str(err), log=False)
2701 def _CheckOSPList(os_obj, parameters):
2702 """Check whether a list of parameters is supported by the OS.
2704 @type os_obj: L{objects.OS}
2705 @param os_obj: OS object to check
2706 @type parameters: list
2707 @param parameters: the list of parameters to check
2710 supported = [v[0] for v in os_obj.supported_parameters]
2711 delta = frozenset(parameters).difference(supported)
2713 _Fail("The following parameters are not supported"
2714 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2717 def ValidateOS(required, osname, checks, osparams):
2718 """Validate the given OS' parameters.
2720 @type required: boolean
2721 @param required: whether absence of the OS should translate into
2723 @type osname: string
2724 @param osname: the OS to be validated
2726 @param checks: list of the checks to run (currently only 'parameters')
2727 @type osparams: dict
2728 @param osparams: dictionary with OS parameters
2730 @return: True if the validation passed, or False if the OS was not
2731 found and L{required} was false
2734 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2735 _Fail("Unknown checks required for OS %s: %s", osname,
2736 set(checks).difference(constants.OS_VALIDATE_CALLS))
2738 name_only = objects.OS.GetName(osname)
2739 status, tbv = _TryOSFromDisk(name_only, None)
2747 if max(tbv.api_versions) < constants.OS_API_V20:
2750 if constants.OS_VALIDATE_PARAMETERS in checks:
2751 _CheckOSPList(tbv, osparams.keys())
2753 validate_env = OSCoreEnv(osname, tbv, osparams)
2754 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2755 cwd=tbv.path, reset_env=True)
2757 logging.error("os validate command '%s' returned error: %s output: %s",
2758 result.cmd, result.fail_reason, result.output)
2759 _Fail("OS validation script failed (%s), output: %s",
2760 result.fail_reason, result.output, log=False)
2766 """Demotes the current node from master candidate role.
2769 # try to ensure we're not the master by mistake
2770 master, myself = ssconf.GetMasterAndMyself()
2771 if master == myself:
2772 _Fail("ssconf status shows I'm the master node, will not demote")
2774 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2775 if not result.failed:
2776 _Fail("The master daemon is running, will not demote")
2779 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2780 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2781 except EnvironmentError, err:
2782 if err.errno != errno.ENOENT:
2783 _Fail("Error while backing up cluster file: %s", err, exc=True)
2785 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2788 def _GetX509Filenames(cryptodir, name):
2789 """Returns the full paths for the private key and certificate.
2792 return (utils.PathJoin(cryptodir, name),
2793 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2794 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2797 def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2798 """Creates a new X509 certificate for SSL/TLS.
2801 @param validity: Validity in seconds
2802 @rtype: tuple; (string, string)
2803 @return: Certificate name and public part
2806 (key_pem, cert_pem) = \
2807 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2808 min(validity, _MAX_SSL_CERT_VALIDITY))
2810 cert_dir = tempfile.mkdtemp(dir=cryptodir,
2811 prefix="x509-%s-" % utils.TimestampForFilename())
2813 name = os.path.basename(cert_dir)
2814 assert len(name) > 5
2816 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2818 utils.WriteFile(key_file, mode=0400, data=key_pem)
2819 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2821 # Never return private key as it shouldn't leave the node
2822 return (name, cert_pem)
2824 shutil.rmtree(cert_dir, ignore_errors=True)
2828 def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2829 """Removes a X509 certificate.
2832 @param name: Certificate name
2835 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2837 utils.RemoveFile(key_file)
2838 utils.RemoveFile(cert_file)
2842 except EnvironmentError, err:
2843 _Fail("Cannot remove certificate directory '%s': %s",
2847 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2848 """Returns the command for the requested input/output.
2850 @type instance: L{objects.Instance}
2851 @param instance: The instance object
2852 @param mode: Import/export mode
2853 @param ieio: Input/output type
2854 @param ieargs: Input/output arguments
2857 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2864 if ieio == constants.IEIO_FILE:
2865 (filename, ) = ieargs
2867 if not utils.IsNormAbsPath(filename):
2868 _Fail("Path '%s' is not normalized or absolute", filename)
2870 directory = os.path.normpath(os.path.dirname(filename))
2872 if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2873 constants.EXPORT_DIR):
2874 _Fail("File '%s' is not under exports directory '%s'",
2875 filename, constants.EXPORT_DIR)
2878 utils.Makedirs(directory, mode=0750)
2880 quoted_filename = utils.ShellQuote(filename)
2882 if mode == constants.IEM_IMPORT:
2883 suffix = "> %s" % quoted_filename
2884 elif mode == constants.IEM_EXPORT:
2885 suffix = "< %s" % quoted_filename
2887 # Retrieve file size
2889 st = os.stat(filename)
2890 except EnvironmentError, err:
2891 logging.error("Can't stat(2) %s: %s", filename, err)
2893 exp_size = utils.BytesToMebibyte(st.st_size)
2895 elif ieio == constants.IEIO_RAW_DISK:
2898 real_disk = _OpenRealBD(disk)
2900 if mode == constants.IEM_IMPORT:
2901 # we set here a smaller block size as, due to transport buffering, more
2902 # than 64-128k will mostly ignored; we use nocreat to fail if the device
2903 # is not already there or we pass a wrong path; we use notrunc to no
2904 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2905 # much memory; this means that at best, we flush every 64k, which will
2907 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2908 " bs=%s oflag=dsync"),
2912 elif mode == constants.IEM_EXPORT:
2913 # the block size on the read dd is 1MiB to match our units
2914 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2916 str(1024 * 1024), # 1 MB
2918 exp_size = disk.size
2920 elif ieio == constants.IEIO_SCRIPT:
2921 (disk, disk_index, ) = ieargs
2923 assert isinstance(disk_index, (int, long))
2925 real_disk = _OpenRealBD(disk)
2927 inst_os = OSFromDisk(instance.os)
2928 env = OSEnvironment(instance, inst_os)
2930 if mode == constants.IEM_IMPORT:
2931 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2932 env["IMPORT_INDEX"] = str(disk_index)
2933 script = inst_os.import_script
2935 elif mode == constants.IEM_EXPORT:
2936 env["EXPORT_DEVICE"] = real_disk.dev_path
2937 env["EXPORT_INDEX"] = str(disk_index)
2938 script = inst_os.export_script
2940 # TODO: Pass special environment only to script
2941 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2943 if mode == constants.IEM_IMPORT:
2944 suffix = "| %s" % script_cmd
2946 elif mode == constants.IEM_EXPORT:
2947 prefix = "%s |" % script_cmd
2949 # Let script predict size
2950 exp_size = constants.IE_CUSTOM_SIZE
2953 _Fail("Invalid %s I/O mode %r", mode, ieio)
2955 return (env, prefix, suffix, exp_size)
2958 def _CreateImportExportStatusDir(prefix):
2959 """Creates status directory for import/export.
2962 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2964 (prefix, utils.TimestampForFilename())))
2967 def StartImportExportDaemon(mode, opts, host, port, instance, component,
2969 """Starts an import or export daemon.
2971 @param mode: Import/output mode
2972 @type opts: L{objects.ImportExportOptions}
2973 @param opts: Daemon options
2975 @param host: Remote host for export (None for import)
2977 @param port: Remote port for export (None for import)
2978 @type instance: L{objects.Instance}
2979 @param instance: Instance object
2980 @type component: string
2981 @param component: which part of the instance is transferred now,
2983 @param ieio: Input/output type
2984 @param ieioargs: Input/output arguments
2987 if mode == constants.IEM_IMPORT:
2990 if not (host is None and port is None):
2991 _Fail("Can not specify host or port on import")
2993 elif mode == constants.IEM_EXPORT:
2996 if host is None or port is None:
2997 _Fail("Host and port must be specified for an export")
3000 _Fail("Invalid mode %r", mode)
3002 if (opts.key_name is None) ^ (opts.ca_pem is None):
3003 _Fail("Cluster certificate can only be used for both key and CA")
3005 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3006 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3008 if opts.key_name is None:
3010 key_path = constants.NODED_CERT_FILE
3011 cert_path = constants.NODED_CERT_FILE
3012 assert opts.ca_pem is None
3014 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
3016 assert opts.ca_pem is not None
3018 for i in [key_path, cert_path]:
3019 if not os.path.exists(i):
3020 _Fail("File '%s' does not exist" % i)
3022 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3024 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3025 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3026 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3028 if opts.ca_pem is None:
3030 ca = utils.ReadFile(constants.NODED_CERT_FILE)
3035 utils.WriteFile(ca_file, data=ca, mode=0400)
3038 constants.IMPORT_EXPORT_DAEMON,
3040 "--key=%s" % key_path,
3041 "--cert=%s" % cert_path,
3042 "--ca=%s" % ca_file,
3046 cmd.append("--host=%s" % host)
3049 cmd.append("--port=%s" % port)
3052 cmd.append("--ipv6")
3054 cmd.append("--ipv4")
3057 cmd.append("--compress=%s" % opts.compress)
3060 cmd.append("--magic=%s" % opts.magic)
3062 if exp_size is not None:
3063 cmd.append("--expected-size=%s" % exp_size)
3066 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3069 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3071 if mode == constants.IEM_EXPORT:
3072 # Retry connection a few times when connecting to remote peer
3073 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3074 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3075 elif opts.connect_timeout is not None:
3076 assert mode == constants.IEM_IMPORT
3077 # Overall timeout for establishing connection while listening
3078 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3080 logfile = _InstanceLogName(prefix, instance.os, instance.name)
3082 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3083 # support for receiving a file descriptor for output
3084 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3087 # The import/export name is simply the status directory name
3088 return os.path.basename(status_dir)
3091 shutil.rmtree(status_dir, ignore_errors=True)
3095 def GetImportExportStatus(names):
3096 """Returns import/export daemon status.
3098 @type names: sequence
3099 @param names: List of names
3100 @rtype: List of dicts
3101 @return: Returns a list of the state of each named import/export or None if a
3102 status couldn't be read
3108 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3112 data = utils.ReadFile(status_file)
3113 except EnvironmentError, err:
3114 if err.errno != errno.ENOENT:
3122 result.append(serializer.LoadJson(data))
3127 def AbortImportExport(name):
3128 """Sends SIGTERM to a running import/export daemon.
3131 logging.info("Abort import/export %s", name)
3133 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3134 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3137 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3139 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3142 def CleanupImportExport(name):
3143 """Cleanup after an import or export.
3145 If the import/export daemon is still running it's killed. Afterwards the
3146 whole status directory is removed.
3149 logging.info("Finalizing import/export %s", name)
3151 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3153 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3156 logging.info("Import/export %s is still running with PID %s",
3158 utils.KillProcess(pid, waitpid=False)
3160 shutil.rmtree(status_dir, ignore_errors=True)
3163 def _FindDisks(nodes_ip, disks):
3164 """Sets the physical ID on disks and returns the block devices.
3167 # set the correct physical ID
3168 my_name = netutils.Hostname.GetSysName()
3170 cf.SetPhysicalID(my_name, nodes_ip)
3175 rd = _RecursiveFindBD(cf)
3177 _Fail("Can't find device %s", cf)
3182 def DrbdDisconnectNet(nodes_ip, disks):
3183 """Disconnects the network on a list of drbd devices.
3186 bdevs = _FindDisks(nodes_ip, disks)
3192 except errors.BlockDeviceError, err:
3193 _Fail("Can't change network configuration to standalone mode: %s",
3197 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3198 """Attaches the network on a list of drbd devices.
3201 bdevs = _FindDisks(nodes_ip, disks)
3204 for idx, rd in enumerate(bdevs):
3206 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3207 except EnvironmentError, err:
3208 _Fail("Can't create symlink: %s", err)
3209 # reconnect disks, switch to new master configuration and if
3210 # needed primary mode
3213 rd.AttachNet(multimaster)
3214 except errors.BlockDeviceError, err:
3215 _Fail("Can't change network configuration: %s", err)
3217 # wait until the disks are connected; we need to retry the re-attach
3218 # if the device becomes standalone, as this might happen if the one
3219 # node disconnects and reconnects in a different mode before the
3220 # other node reconnects; in this case, one or both of the nodes will
3221 # decide it has wrong configuration and switch to standalone
3224 all_connected = True
3227 stats = rd.GetProcStatus()
3229 all_connected = (all_connected and
3230 (stats.is_connected or stats.is_in_resync))
3232 if stats.is_standalone:
3233 # peer had different config info and this node became
3234 # standalone, even though this should not happen with the
3235 # new staged way of changing disk configs
3237 rd.AttachNet(multimaster)
3238 except errors.BlockDeviceError, err:
3239 _Fail("Can't change network configuration: %s", err)
3241 if not all_connected:
3242 raise utils.RetryAgain()
3245 # Start with a delay of 100 miliseconds and go up to 5 seconds
3246 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3247 except utils.RetryTimeout:
3248 _Fail("Timeout in disk reconnecting")
3251 # change to primary mode
3255 except errors.BlockDeviceError, err:
3256 _Fail("Can't change to primary mode: %s", err)
3259 def DrbdWaitSync(nodes_ip, disks):
3260 """Wait until DRBDs have synchronized.
3264 stats = rd.GetProcStatus()
3265 if not (stats.is_connected or stats.is_in_resync):
3266 raise utils.RetryAgain()
3269 bdevs = _FindDisks(nodes_ip, disks)
3275 # poll each second for 15 seconds
3276 stats = utils.Retry(_helper, 1, 15, args=[rd])
3277 except utils.RetryTimeout:
3278 stats = rd.GetProcStatus()
3280 if not (stats.is_connected or stats.is_in_resync):
3281 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3282 alldone = alldone and (not stats.is_in_resync)
3283 if stats.sync_percent is not None:
3284 min_resync = min(min_resync, stats.sync_percent)
3286 return (alldone, min_resync)
3289 def GetDrbdUsermodeHelper():
3290 """Returns DRBD usermode helper currently configured.
3294 return bdev.BaseDRBD.GetUsermodeHelper()
3295 except errors.BlockDeviceError, err:
3299 def PowercycleNode(hypervisor_type):
3300 """Hard-powercycle the node.
3302 Because we need to return first, and schedule the powercycle in the
3303 background, we won't be able to report failures nicely.
3306 hyper = hypervisor.GetHypervisor(hypervisor_type)
3310 # if we can't fork, we'll pretend that we're in the child process
3313 return "Reboot scheduled in 5 seconds"
3314 # ensure the child is running on ram
3317 except Exception: # pylint: disable-msg=W0703
3320 hyper.PowercycleNode()
3323 class HooksRunner(object):
3326 This class is instantiated on the node side (ganeti-noded) and not
3330 def __init__(self, hooks_base_dir=None):
3331 """Constructor for hooks runner.
3333 @type hooks_base_dir: str or None
3334 @param hooks_base_dir: if not None, this overrides the
3335 L{constants.HOOKS_BASE_DIR} (useful for unittests)
3338 if hooks_base_dir is None:
3339 hooks_base_dir = constants.HOOKS_BASE_DIR
3340 # yeah, _BASE_DIR is not valid for attributes, we use it like a
3342 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3344 def RunHooks(self, hpath, phase, env):
3345 """Run the scripts in the hooks directory.
3348 @param hpath: the path to the hooks directory which
3351 @param phase: either L{constants.HOOKS_PHASE_PRE} or
3352 L{constants.HOOKS_PHASE_POST}
3354 @param env: dictionary with the environment for the hook
3356 @return: list of 3-element tuples:
3358 - script result, either L{constants.HKR_SUCCESS} or
3359 L{constants.HKR_FAIL}
3360 - output of the script
3362 @raise errors.ProgrammerError: for invalid input
3366 if phase == constants.HOOKS_PHASE_PRE:
3368 elif phase == constants.HOOKS_PHASE_POST:
3371 _Fail("Unknown hooks phase '%s'", phase)
3374 subdir = "%s-%s.d" % (hpath, suffix)
3375 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3379 if not os.path.isdir(dir_name):
3380 # for non-existing/non-dirs, we simply exit instead of logging a
3381 # warning at every operation
3384 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3386 for (relname, relstatus, runresult) in runparts_results:
3387 if relstatus == constants.RUNPARTS_SKIP:
3388 rrval = constants.HKR_SKIP
3390 elif relstatus == constants.RUNPARTS_ERR:
3391 rrval = constants.HKR_FAIL
3392 output = "Hook script execution error: %s" % runresult
3393 elif relstatus == constants.RUNPARTS_RUN:
3394 if runresult.failed:
3395 rrval = constants.HKR_FAIL
3397 rrval = constants.HKR_SUCCESS
3398 output = utils.SafeEncode(runresult.output.strip())
3399 results.append(("%s/%s" % (subdir, relname), rrval, output))
3404 class IAllocatorRunner(object):
3405 """IAllocator runner.
3407 This class is instantiated on the node side (ganeti-noded) and not on
3412 def Run(name, idata):
3413 """Run an iallocator script.
3416 @param name: the iallocator script name
3418 @param idata: the allocator input data
3421 @return: two element tuple of:
3423 - either error message or stdout of allocator (for success)
3426 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3428 if alloc_script is None:
3429 _Fail("iallocator module '%s' not found in the search path", name)
3431 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3435 result = utils.RunCmd([alloc_script, fin_name])
3437 _Fail("iallocator module '%s' failed: %s, output '%s'",
3438 name, result.fail_reason, result.output)
3442 return result.stdout
3445 class DevCacheManager(object):
3446 """Simple class for managing a cache of block device information.
3449 _DEV_PREFIX = "/dev/"
3450 _ROOT_DIR = constants.BDEV_CACHE_DIR
3453 def _ConvertPath(cls, dev_path):
3454 """Converts a /dev/name path to the cache file name.
3456 This replaces slashes with underscores and strips the /dev
3457 prefix. It then returns the full path to the cache file.
3460 @param dev_path: the C{/dev/} path name
3462 @return: the converted path name
3465 if dev_path.startswith(cls._DEV_PREFIX):
3466 dev_path = dev_path[len(cls._DEV_PREFIX):]
3467 dev_path = dev_path.replace("/", "_")
3468 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3472 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3473 """Updates the cache information for a given device.
3476 @param dev_path: the pathname of the device
3478 @param owner: the owner (instance name) of the device
3479 @type on_primary: bool
3480 @param on_primary: whether this is the primary
3483 @param iv_name: the instance-visible name of the
3484 device, as in objects.Disk.iv_name
3489 if dev_path is None:
3490 logging.error("DevCacheManager.UpdateCache got a None dev_path")
3492 fpath = cls._ConvertPath(dev_path)
3498 iv_name = "not_visible"
3499 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3501 utils.WriteFile(fpath, data=fdata)
3502 except EnvironmentError, err:
3503 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3506 def RemoveCache(cls, dev_path):
3507 """Remove data for a dev_path.
3509 This is just a wrapper over L{utils.io.RemoveFile} with a converted
3510 path name and logging.
3513 @param dev_path: the pathname of the device
3518 if dev_path is None:
3519 logging.error("DevCacheManager.RemoveCache got a None dev_path")
3521 fpath = cls._ConvertPath(dev_path)
3523 utils.RemoveFile(fpath)
3524 except EnvironmentError, err:
3525 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)