4 # Copyright (C) 2006, 2007 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Functions used by the node daemon
24 @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
25 the L{UploadFile} function
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
64 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
65 _ALLOWED_CLEAN_DIRS = frozenset([
67 constants.JOB_QUEUE_ARCHIVE_DIR,
69 constants.CRYPTO_KEYS_DIR,
71 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
72 _X509_KEY_FILE = "key"
73 _X509_CERT_FILE = "cert"
74 _IES_STATUS_FILE = "status"
79 class RPCFail(Exception):
80 """Class denoting RPC failure.
82 Its argument is the error message.
87 def _Fail(msg, *args, **kwargs):
88 """Log an error and the raise an RPCFail exception.
90 This exception is then handled specially in the ganeti daemon and
91 turned into a 'failed' return type. As such, this function is a
92 useful shortcut for logging the error and returning it to the master
96 @param msg: the text of the exception
102 if "log" not in kwargs or kwargs["log"]: # if we should log this error
103 if "exc" in kwargs and kwargs["exc"]:
104 logging.exception(msg)
111 """Simple wrapper to return a SimpleStore.
113 @rtype: L{ssconf.SimpleStore}
114 @return: a SimpleStore instance
117 return ssconf.SimpleStore()
120 def _GetSshRunner(cluster_name):
121 """Simple wrapper to return an SshRunner.
123 @type cluster_name: str
124 @param cluster_name: the cluster name, which is needed
125 by the SshRunner constructor
126 @rtype: L{ssh.SshRunner}
127 @return: an SshRunner instance
130 return ssh.SshRunner(cluster_name)
133 def _Decompress(data):
134 """Unpacks data compressed by the RPC client.
136 @type data: list or tuple
137 @param data: Data sent by RPC client
139 @return: Decompressed data
142 assert isinstance(data, (list, tuple))
143 assert len(data) == 2
144 (encoding, content) = data
145 if encoding == constants.RPC_ENCODING_NONE:
147 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
148 return zlib.decompress(base64.b64decode(content))
150 raise AssertionError("Unknown data encoding")
153 def _CleanDirectory(path, exclude=None):
154 """Removes all regular files in a directory.
157 @param path: the directory to clean
159 @param exclude: list of files to be excluded, defaults
163 if path not in _ALLOWED_CLEAN_DIRS:
164 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
167 if not os.path.isdir(path):
172 # Normalize excluded paths
173 exclude = [os.path.normpath(i) for i in exclude]
175 for rel_name in utils.ListVisibleFiles(path):
176 full_name = utils.PathJoin(path, rel_name)
177 if full_name in exclude:
179 if os.path.isfile(full_name) and not os.path.islink(full_name):
180 utils.RemoveFile(full_name)
183 def _BuildUploadFileList():
184 """Build the list of allowed upload files.
186 This is abstracted so that it's built only once at module import time.
189 allowed_files = set([
190 constants.CLUSTER_CONF_FILE,
192 constants.SSH_KNOWN_HOSTS_FILE,
193 constants.VNC_PASSWORD_FILE,
194 constants.RAPI_CERT_FILE,
195 constants.RAPI_USERS_FILE,
196 constants.CONFD_HMAC_KEY,
197 constants.CLUSTER_DOMAIN_SECRET_FILE,
200 for hv_name in constants.HYPER_TYPES:
201 hv_class = hypervisor.GetHypervisorClass(hv_name)
202 allowed_files.update(hv_class.GetAncillaryFiles())
204 return frozenset(allowed_files)
207 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
211 """Removes job queue files and archived jobs.
217 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
218 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
222 """Returns master information.
224 This is an utility function to compute master information, either
225 for consumption here or from the node daemon.
228 @return: master_netdev, master_ip, master_name
229 @raise RPCFail: in case of errors
234 master_netdev = cfg.GetMasterNetdev()
235 master_ip = cfg.GetMasterIP()
236 master_node = cfg.GetMasterNode()
237 except errors.ConfigurationError, err:
238 _Fail("Cluster configuration incomplete: %s", err, exc=True)
239 return (master_netdev, master_ip, master_node)
242 def StartMaster(start_daemons, no_voting):
243 """Activate local node as master node.
245 The function will either try activate the IP address of the master
246 (unless someone else has it) or also start the master daemons, based
247 on the start_daemons parameter.
249 @type start_daemons: boolean
250 @param start_daemons: whether to start the master daemons
251 (ganeti-masterd and ganeti-rapi), or (if false) activate the
253 @type no_voting: boolean
254 @param no_voting: whether to start ganeti-masterd without a node vote
255 (if start_daemons is True), but still non-interactively
259 # GetMasterInfo will raise an exception if not able to return data
260 master_netdev, master_ip, _ = GetMasterInfo()
263 # either start the master and rapi daemons
266 masterd_args = "--no-voting --yes-do-it"
271 "EXTRA_MASTERD_ARGS": masterd_args,
274 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
276 msg = "Can't start Ganeti master: %s" % result.output
281 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
282 if netutils.OwnIpAddress(master_ip):
283 # we already have the ip:
284 logging.debug("Master IP already configured, doing nothing")
286 msg = "Someone else has the master ip, not activating"
290 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
291 "dev", master_netdev, "label",
292 "%s:0" % master_netdev])
294 msg = "Can't activate master IP: %s" % result.output
298 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
299 "-s", master_ip, master_ip])
300 # we'll ignore the exit code of arping
303 _Fail("; ".join(err_msgs))
306 def StopMaster(stop_daemons):
307 """Deactivate this node as master.
309 The function will always try to deactivate the IP address of the
310 master. It will also stop the master daemons depending on the
311 stop_daemons parameter.
313 @type stop_daemons: boolean
314 @param stop_daemons: whether to also stop the master daemons
315 (ganeti-masterd and ganeti-rapi)
319 # TODO: log and report back to the caller the error failures; we
320 # need to decide in which case we fail the RPC for this
322 # GetMasterInfo will raise an exception if not able to return data
323 master_netdev, master_ip, _ = GetMasterInfo()
325 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
326 "dev", master_netdev])
328 logging.error("Can't remove the master IP, error: %s", result.output)
329 # but otherwise ignore the failure
332 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
334 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
336 result.cmd, result.exit_code, result.output)
339 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
340 """Joins this node to the cluster.
342 This does the following:
343 - updates the hostkeys of the machine (rsa and dsa)
344 - adds the ssh private key to the user
345 - adds the ssh public key to the users' authorized_keys file
348 @param dsa: the DSA private key to write
350 @param dsapub: the DSA public key to write
352 @param rsa: the RSA private key to write
354 @param rsapub: the RSA public key to write
356 @param sshkey: the SSH private key to write
358 @param sshpub: the SSH public key to write
360 @return: the success of the operation
363 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
364 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
365 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
366 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
367 for name, content, mode in sshd_keys:
368 utils.WriteFile(name, data=content, mode=mode)
371 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
373 except errors.OpExecError, err:
374 _Fail("Error while processing user ssh files: %s", err, exc=True)
376 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
377 utils.WriteFile(name, data=content, mode=0600)
379 utils.AddAuthorizedKey(auth_keys, sshpub)
381 result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"])
383 _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)",
384 result.cmd, result.exit_code, result.output)
387 def LeaveCluster(modify_ssh_setup):
388 """Cleans up and remove the current node.
390 This function cleans up and prepares the current node to be removed
393 If processing is successful, then it raises an
394 L{errors.QuitGanetiException} which is used as a special case to
395 shutdown the node daemon.
397 @param modify_ssh_setup: boolean
400 _CleanDirectory(constants.DATA_DIR)
401 _CleanDirectory(constants.CRYPTO_KEYS_DIR)
406 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
408 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
410 utils.RemoveFile(priv_key)
411 utils.RemoveFile(pub_key)
412 except errors.OpExecError:
413 logging.exception("Error while processing ssh files")
416 utils.RemoveFile(constants.CONFD_HMAC_KEY)
417 utils.RemoveFile(constants.RAPI_CERT_FILE)
418 utils.RemoveFile(constants.NODED_CERT_FILE)
419 except: # pylint: disable-msg=W0702
420 logging.exception("Error while removing cluster secrets")
422 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
424 logging.error("Command %s failed with exitcode %s and error %s",
425 result.cmd, result.exit_code, result.output)
427 # Raise a custom exception (handled in ganeti-noded)
428 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
431 def GetNodeInfo(vgname, hypervisor_type):
432 """Gives back a hash with different information about the node.
434 @type vgname: C{string}
435 @param vgname: the name of the volume group to ask for disk space information
436 @type hypervisor_type: C{str}
437 @param hypervisor_type: the name of the hypervisor to ask for
440 @return: dictionary with the following keys:
441 - vg_size is the size of the configured volume group in MiB
442 - vg_free is the free size of the volume group in MiB
443 - memory_dom0 is the memory allocated for domain0 in MiB
444 - memory_free is the currently available (free) ram in MiB
445 - memory_total is the total number of ram in MiB
449 vginfo = _GetVGInfo(vgname)
450 outputarray['vg_size'] = vginfo['vg_size']
451 outputarray['vg_free'] = vginfo['vg_free']
453 hyper = hypervisor.GetHypervisor(hypervisor_type)
454 hyp_info = hyper.GetNodeInfo()
455 if hyp_info is not None:
456 outputarray.update(hyp_info)
458 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
463 def VerifyNode(what, cluster_name):
464 """Verify the status of the local node.
466 Based on the input L{what} parameter, various checks are done on the
469 If the I{filelist} key is present, this list of
470 files is checksummed and the file/checksum pairs are returned.
472 If the I{nodelist} key is present, we check that we have
473 connectivity via ssh with the target nodes (and check the hostname
476 If the I{node-net-test} key is present, we check that we have
477 connectivity to the given nodes via both primary IP and, if
478 applicable, secondary IPs.
481 @param what: a dictionary of things to check:
482 - filelist: list of files for which to compute checksums
483 - nodelist: list of nodes we should check ssh communication with
484 - node-net-test: list of nodes we should check node daemon port
486 - hypervisor: list with hypervisors to run the verify for
488 @return: a dictionary with the same keys as the input dict, and
489 values representing the result of the checks
493 my_name = netutils.HostInfo().name
494 port = netutils.GetDaemonPort(constants.NODED)
496 if constants.NV_HYPERVISOR in what:
497 result[constants.NV_HYPERVISOR] = tmp = {}
498 for hv_name in what[constants.NV_HYPERVISOR]:
500 val = hypervisor.GetHypervisor(hv_name).Verify()
501 except errors.HypervisorError, err:
502 val = "Error while checking hypervisor: %s" % str(err)
505 if constants.NV_FILELIST in what:
506 result[constants.NV_FILELIST] = utils.FingerprintFiles(
507 what[constants.NV_FILELIST])
509 if constants.NV_NODELIST in what:
510 result[constants.NV_NODELIST] = tmp = {}
511 random.shuffle(what[constants.NV_NODELIST])
512 for node in what[constants.NV_NODELIST]:
513 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
517 if constants.NV_NODENETTEST in what:
518 result[constants.NV_NODENETTEST] = tmp = {}
519 my_pip = my_sip = None
520 for name, pip, sip in what[constants.NV_NODENETTEST]:
526 tmp[my_name] = ("Can't find my own primary/secondary IP"
529 for name, pip, sip in what[constants.NV_NODENETTEST]:
531 if not netutils.TcpPing(pip, port, source=my_pip):
532 fail.append("primary")
534 if not netutils.TcpPing(sip, port, source=my_sip):
535 fail.append("secondary")
537 tmp[name] = ("failure using the %s interface(s)" %
540 if constants.NV_MASTERIP in what:
541 # FIXME: add checks on incoming data structures (here and in the
542 # rest of the function)
543 master_name, master_ip = what[constants.NV_MASTERIP]
544 if master_name == my_name:
545 source = constants.IP4_ADDRESS_LOCALHOST
548 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
551 if constants.NV_LVLIST in what:
553 val = GetVolumeList(what[constants.NV_LVLIST])
556 result[constants.NV_LVLIST] = val
558 if constants.NV_INSTANCELIST in what:
559 # GetInstanceList can fail
561 val = GetInstanceList(what[constants.NV_INSTANCELIST])
564 result[constants.NV_INSTANCELIST] = val
566 if constants.NV_VGLIST in what:
567 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
569 if constants.NV_PVLIST in what:
570 result[constants.NV_PVLIST] = \
571 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
572 filter_allocatable=False)
574 if constants.NV_VERSION in what:
575 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
576 constants.RELEASE_VERSION)
578 if constants.NV_HVINFO in what:
579 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
580 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
582 if constants.NV_DRBDLIST in what:
584 used_minors = bdev.DRBD8.GetUsedDevs().keys()
585 except errors.BlockDeviceError, err:
586 logging.warning("Can't get used minors list", exc_info=True)
587 used_minors = str(err)
588 result[constants.NV_DRBDLIST] = used_minors
590 if constants.NV_DRBDHELPER in what:
593 payload = bdev.BaseDRBD.GetUsermodeHelper()
594 except errors.BlockDeviceError, err:
595 logging.error("Can't get DRBD usermode helper: %s", str(err))
598 result[constants.NV_DRBDHELPER] = (status, payload)
600 if constants.NV_NODESETUP in what:
601 result[constants.NV_NODESETUP] = tmpr = []
602 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
603 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
604 " under /sys, missing required directories /sys/block"
605 " and /sys/class/net")
606 if (not os.path.isdir("/proc/sys") or
607 not os.path.isfile("/proc/sysrq-trigger")):
608 tmpr.append("The procfs filesystem doesn't seem to be mounted"
609 " under /proc, missing required directory /proc/sys and"
610 " the file /proc/sysrq-trigger")
612 if constants.NV_TIME in what:
613 result[constants.NV_TIME] = utils.SplitTime(time.time())
615 if constants.NV_OSLIST in what:
616 result[constants.NV_OSLIST] = DiagnoseOS()
621 def GetVolumeList(vg_name):
622 """Compute list of logical volumes and their size.
625 @param vg_name: the volume group whose LVs we should list
628 dictionary of all partions (key) with value being a tuple of
629 their size (in MiB), inactive and online status::
631 {'test1': ('20.06', True, True)}
633 in case of errors, a string is returned with the error
639 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
640 "--separator=%s" % sep,
641 "-olv_name,lv_size,lv_attr", vg_name])
643 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
645 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
646 for line in result.stdout.splitlines():
648 match = valid_line_re.match(line)
650 logging.error("Invalid line returned from lvs output: '%s'", line)
652 name, size, attr = match.groups()
653 inactive = attr[4] == '-'
654 online = attr[5] == 'o'
655 virtual = attr[0] == 'v'
657 # we don't want to report such volumes as existing, since they
658 # don't really hold data
660 lvs[name] = (size, inactive, online)
665 def ListVolumeGroups():
666 """List the volume groups and their size.
669 @return: dictionary with keys volume name and values the
673 return utils.ListVolumeGroups()
677 """List all volumes on this node.
681 A list of dictionaries, each having four keys:
682 - name: the logical volume name,
683 - size: the size of the logical volume
684 - dev: the physical device on which the LV lives
685 - vg: the volume group to which it belongs
687 In case of errors, we return an empty list and log the
690 Note that since a logical volume can live on multiple physical
691 volumes, the resulting list might include a logical volume
695 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
697 "--options=lv_name,lv_size,devices,vg_name"])
699 _Fail("Failed to list logical volumes, lvs output: %s",
703 return dev.split('(')[0]
706 return [parse_dev(x) for x in dev.split(",")]
709 line = [v.strip() for v in line]
710 return [{'name': line[0], 'size': line[1],
711 'dev': dev, 'vg': line[3]} for dev in handle_dev(line[2])]
714 for line in result.stdout.splitlines():
715 if line.count('|') >= 3:
716 all_devs.extend(map_line(line.split('|')))
718 logging.warning("Strange line in the output from lvs: '%s'", line)
722 def BridgesExist(bridges_list):
723 """Check if a list of bridges exist on the current node.
726 @return: C{True} if all of them exist, C{False} otherwise
730 for bridge in bridges_list:
731 if not utils.BridgeExists(bridge):
732 missing.append(bridge)
735 _Fail("Missing bridges %s", utils.CommaJoin(missing))
738 def GetInstanceList(hypervisor_list):
739 """Provides a list of instances.
741 @type hypervisor_list: list
742 @param hypervisor_list: the list of hypervisors to query information
745 @return: a list of all running instances on the current node
746 - instance1.example.com
747 - instance2.example.com
751 for hname in hypervisor_list:
753 names = hypervisor.GetHypervisor(hname).ListInstances()
754 results.extend(names)
755 except errors.HypervisorError, err:
756 _Fail("Error enumerating instances (hypervisor %s): %s",
757 hname, err, exc=True)
762 def GetInstanceInfo(instance, hname):
763 """Gives back the information about an instance as a dictionary.
765 @type instance: string
766 @param instance: the instance name
768 @param hname: the hypervisor type of the instance
771 @return: dictionary with the following keys:
772 - memory: memory size of instance (int)
773 - state: xen state of instance (string)
774 - time: cpu time of instance (float)
779 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
780 if iinfo is not None:
781 output['memory'] = iinfo[2]
782 output['state'] = iinfo[4]
783 output['time'] = iinfo[5]
788 def GetInstanceMigratable(instance):
789 """Gives whether an instance can be migrated.
791 @type instance: L{objects.Instance}
792 @param instance: object representing the instance to be checked.
795 @return: tuple of (result, description) where:
796 - result: whether the instance can be migrated or not
797 - description: a description of the issue, if relevant
800 hyper = hypervisor.GetHypervisor(instance.hypervisor)
801 iname = instance.name
802 if iname not in hyper.ListInstances():
803 _Fail("Instance %s is not running", iname)
805 for idx in range(len(instance.disks)):
806 link_name = _GetBlockDevSymlinkPath(iname, idx)
807 if not os.path.islink(link_name):
808 logging.warning("Instance %s is missing symlink %s for disk %d",
809 iname, link_name, idx)
812 def GetAllInstancesInfo(hypervisor_list):
813 """Gather data about all instances.
815 This is the equivalent of L{GetInstanceInfo}, except that it
816 computes data for all instances at once, thus being faster if one
817 needs data about more than one instance.
819 @type hypervisor_list: list
820 @param hypervisor_list: list of hypervisors to query for instance data
823 @return: dictionary of instance: data, with data having the following keys:
824 - memory: memory size of instance (int)
825 - state: xen state of instance (string)
826 - time: cpu time of instance (float)
827 - vcpus: the number of vcpus
832 for hname in hypervisor_list:
833 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
835 for name, _, memory, vcpus, state, times in iinfo:
843 # we only check static parameters, like memory and vcpus,
844 # and not state and time which can change between the
845 # invocations of the different hypervisors
846 for key in 'memory', 'vcpus':
847 if value[key] != output[name][key]:
848 _Fail("Instance %s is running twice"
849 " with different parameters", name)
855 def _InstanceLogName(kind, os_name, instance):
856 """Compute the OS log filename for a given instance and operation.
858 The instance name and os name are passed in as strings since not all
859 operations have these as part of an instance object.
862 @param kind: the operation type (e.g. add, import, etc.)
863 @type os_name: string
864 @param os_name: the os name
865 @type instance: string
866 @param instance: the name of the instance being imported/added/etc.
869 # TODO: Use tempfile.mkstemp to create unique filename
870 base = ("%s-%s-%s-%s.log" %
871 (kind, os_name, instance, utils.TimestampForFilename()))
872 return utils.PathJoin(constants.LOG_OS_DIR, base)
875 def InstanceOsAdd(instance, reinstall, debug):
876 """Add an OS to an instance.
878 @type instance: L{objects.Instance}
879 @param instance: Instance whose OS is to be installed
880 @type reinstall: boolean
881 @param reinstall: whether this is an instance reinstall
883 @param debug: debug level, passed to the OS scripts
887 inst_os = OSFromDisk(instance.os)
889 create_env = OSEnvironment(instance, inst_os, debug)
891 create_env['INSTANCE_REINSTALL'] = "1"
893 logfile = _InstanceLogName("add", instance.os, instance.name)
895 result = utils.RunCmd([inst_os.create_script], env=create_env,
896 cwd=inst_os.path, output=logfile,)
898 logging.error("os create command '%s' returned error: %s, logfile: %s,"
899 " output: %s", result.cmd, result.fail_reason, logfile,
901 lines = [utils.SafeEncode(val)
902 for val in utils.TailFile(logfile, lines=20)]
903 _Fail("OS create script failed (%s), last lines in the"
904 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
907 def RunRenameInstance(instance, old_name, debug):
908 """Run the OS rename script for an instance.
910 @type instance: L{objects.Instance}
911 @param instance: Instance whose OS is to be installed
912 @type old_name: string
913 @param old_name: previous instance name
915 @param debug: debug level, passed to the OS scripts
917 @return: the success of the operation
920 inst_os = OSFromDisk(instance.os)
922 rename_env = OSEnvironment(instance, inst_os, debug)
923 rename_env['OLD_INSTANCE_NAME'] = old_name
925 logfile = _InstanceLogName("rename", instance.os,
926 "%s-%s" % (old_name, instance.name))
928 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
929 cwd=inst_os.path, output=logfile)
932 logging.error("os create command '%s' returned error: %s output: %s",
933 result.cmd, result.fail_reason, result.output)
934 lines = [utils.SafeEncode(val)
935 for val in utils.TailFile(logfile, lines=20)]
936 _Fail("OS rename script failed (%s), last lines in the"
937 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
940 def _GetVGInfo(vg_name):
941 """Get information about the volume group.
944 @param vg_name: the volume group which we query
947 A dictionary with the following keys:
948 - C{vg_size} is the total size of the volume group in MiB
949 - C{vg_free} is the free size of the volume group in MiB
950 - C{pv_count} are the number of physical disks in that VG
952 If an error occurs during gathering of data, we return the same dict
953 with keys all set to None.
956 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
958 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
959 "--nosuffix", "--units=m", "--separator=:", vg_name])
962 logging.error("volume group %s not present", vg_name)
964 valarr = retval.stdout.strip().rstrip(':').split(':')
968 "vg_size": int(round(float(valarr[0]), 0)),
969 "vg_free": int(round(float(valarr[1]), 0)),
970 "pv_count": int(valarr[2]),
972 except (TypeError, ValueError), err:
973 logging.exception("Fail to parse vgs output: %s", err)
975 logging.error("vgs output has the wrong number of fields (expected"
976 " three): %s", str(valarr))
980 def _GetBlockDevSymlinkPath(instance_name, idx):
981 return utils.PathJoin(constants.DISK_LINKS_DIR,
982 "%s:%d" % (instance_name, idx))
985 def _SymlinkBlockDev(instance_name, device_path, idx):
986 """Set up symlinks to a instance's block device.
988 This is an auxiliary function run when an instance is start (on the primary
989 node) or when an instance is migrated (on the target node).
992 @param instance_name: the name of the target instance
993 @param device_path: path of the physical block device, on the node
994 @param idx: the disk index
995 @return: absolute path to the disk's symlink
998 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1000 os.symlink(device_path, link_name)
1001 except OSError, err:
1002 if err.errno == errno.EEXIST:
1003 if (not os.path.islink(link_name) or
1004 os.readlink(link_name) != device_path):
1005 os.remove(link_name)
1006 os.symlink(device_path, link_name)
1013 def _RemoveBlockDevLinks(instance_name, disks):
1014 """Remove the block device symlinks belonging to the given instance.
1017 for idx, _ in enumerate(disks):
1018 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1019 if os.path.islink(link_name):
1021 os.remove(link_name)
1023 logging.exception("Can't remove symlink '%s'", link_name)
1026 def _GatherAndLinkBlockDevs(instance):
1027 """Set up an instance's block device(s).
1029 This is run on the primary node at instance startup. The block
1030 devices must be already assembled.
1032 @type instance: L{objects.Instance}
1033 @param instance: the instance whose disks we shoul assemble
1035 @return: list of (disk_object, device_path)
1039 for idx, disk in enumerate(instance.disks):
1040 device = _RecursiveFindBD(disk)
1042 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1046 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1048 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1051 block_devices.append((disk, link_name))
1053 return block_devices
1056 def StartInstance(instance):
1057 """Start an instance.
1059 @type instance: L{objects.Instance}
1060 @param instance: the instance object
1064 running_instances = GetInstanceList([instance.hypervisor])
1066 if instance.name in running_instances:
1067 logging.info("Instance %s already running, not starting", instance.name)
1071 block_devices = _GatherAndLinkBlockDevs(instance)
1072 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1073 hyper.StartInstance(instance, block_devices)
1074 except errors.BlockDeviceError, err:
1075 _Fail("Block device error: %s", err, exc=True)
1076 except errors.HypervisorError, err:
1077 _RemoveBlockDevLinks(instance.name, instance.disks)
1078 _Fail("Hypervisor error: %s", err, exc=True)
1081 def InstanceShutdown(instance, timeout):
1082 """Shut an instance down.
1084 @note: this functions uses polling with a hardcoded timeout.
1086 @type instance: L{objects.Instance}
1087 @param instance: the instance object
1088 @type timeout: integer
1089 @param timeout: maximum timeout for soft shutdown
1093 hv_name = instance.hypervisor
1094 hyper = hypervisor.GetHypervisor(hv_name)
1095 iname = instance.name
1097 if instance.name not in hyper.ListInstances():
1098 logging.info("Instance %s not running, doing nothing", iname)
1103 self.tried_once = False
1106 if iname not in hyper.ListInstances():
1110 hyper.StopInstance(instance, retry=self.tried_once)
1111 except errors.HypervisorError, err:
1112 if iname not in hyper.ListInstances():
1113 # if the instance is no longer existing, consider this a
1114 # success and go to cleanup
1117 _Fail("Failed to stop instance %s: %s", iname, err)
1119 self.tried_once = True
1121 raise utils.RetryAgain()
1124 utils.Retry(_TryShutdown(), 5, timeout)
1125 except utils.RetryTimeout:
1126 # the shutdown did not succeed
1127 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1130 hyper.StopInstance(instance, force=True)
1131 except errors.HypervisorError, err:
1132 if iname in hyper.ListInstances():
1133 # only raise an error if the instance still exists, otherwise
1134 # the error could simply be "instance ... unknown"!
1135 _Fail("Failed to force stop instance %s: %s", iname, err)
1139 if iname in hyper.ListInstances():
1140 _Fail("Could not shutdown instance %s even by destroy", iname)
1143 hyper.CleanupInstance(instance.name)
1144 except errors.HypervisorError, err:
1145 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1147 _RemoveBlockDevLinks(iname, instance.disks)
1150 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1151 """Reboot an instance.
1153 @type instance: L{objects.Instance}
1154 @param instance: the instance object to reboot
1155 @type reboot_type: str
1156 @param reboot_type: the type of reboot, one the following
1158 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1159 instance OS, do not recreate the VM
1160 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1161 restart the VM (at the hypervisor level)
1162 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1163 not accepted here, since that mode is handled differently, in
1164 cmdlib, and translates into full stop and start of the
1165 instance (instead of a call_instance_reboot RPC)
1166 @type shutdown_timeout: integer
1167 @param shutdown_timeout: maximum timeout for soft shutdown
1171 running_instances = GetInstanceList([instance.hypervisor])
1173 if instance.name not in running_instances:
1174 _Fail("Cannot reboot instance %s that is not running", instance.name)
1176 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1177 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1179 hyper.RebootInstance(instance)
1180 except errors.HypervisorError, err:
1181 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1182 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1184 InstanceShutdown(instance, shutdown_timeout)
1185 return StartInstance(instance)
1186 except errors.HypervisorError, err:
1187 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1189 _Fail("Invalid reboot_type received: %s", reboot_type)
1192 def MigrationInfo(instance):
1193 """Gather information about an instance to be migrated.
1195 @type instance: L{objects.Instance}
1196 @param instance: the instance definition
1199 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1201 info = hyper.MigrationInfo(instance)
1202 except errors.HypervisorError, err:
1203 _Fail("Failed to fetch migration information: %s", err, exc=True)
1207 def AcceptInstance(instance, info, target):
1208 """Prepare the node to accept an instance.
1210 @type instance: L{objects.Instance}
1211 @param instance: the instance definition
1212 @type info: string/data (opaque)
1213 @param info: migration information, from the source node
1214 @type target: string
1215 @param target: target host (usually ip), on this node
1218 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1220 hyper.AcceptInstance(instance, info, target)
1221 except errors.HypervisorError, err:
1222 _Fail("Failed to accept instance: %s", err, exc=True)
1225 def FinalizeMigration(instance, info, success):
1226 """Finalize any preparation to accept an instance.
1228 @type instance: L{objects.Instance}
1229 @param instance: the instance definition
1230 @type info: string/data (opaque)
1231 @param info: migration information, from the source node
1232 @type success: boolean
1233 @param success: whether the migration was a success or a failure
1236 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1238 hyper.FinalizeMigration(instance, info, success)
1239 except errors.HypervisorError, err:
1240 _Fail("Failed to finalize migration: %s", err, exc=True)
1243 def MigrateInstance(instance, target, live):
1244 """Migrates an instance to another node.
1246 @type instance: L{objects.Instance}
1247 @param instance: the instance definition
1248 @type target: string
1249 @param target: the target node name
1251 @param live: whether the migration should be done live or not (the
1252 interpretation of this parameter is left to the hypervisor)
1254 @return: a tuple of (success, msg) where:
1255 - succes is a boolean denoting the success/failure of the operation
1256 - msg is a string with details in case of failure
1259 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1262 hyper.MigrateInstance(instance, target, live)
1263 except errors.HypervisorError, err:
1264 _Fail("Failed to migrate instance: %s", err, exc=True)
1267 def BlockdevCreate(disk, size, owner, on_primary, info):
1268 """Creates a block device for an instance.
1270 @type disk: L{objects.Disk}
1271 @param disk: the object describing the disk we should create
1273 @param size: the size of the physical underlying device, in MiB
1275 @param owner: the name of the instance for which disk is created,
1276 used for device cache data
1277 @type on_primary: boolean
1278 @param on_primary: indicates if it is the primary node or not
1280 @param info: string that will be sent to the physical device
1281 creation, used for example to set (LVM) tags on LVs
1283 @return: the new unique_id of the device (this can sometime be
1284 computed only after creation), or None. On secondary nodes,
1285 it's not required to return anything.
1288 # TODO: remove the obsolete 'size' argument
1289 # pylint: disable-msg=W0613
1292 for child in disk.children:
1294 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1295 except errors.BlockDeviceError, err:
1296 _Fail("Can't assemble device %s: %s", child, err)
1297 if on_primary or disk.AssembleOnSecondary():
1298 # we need the children open in case the device itself has to
1301 # pylint: disable-msg=E1103
1303 except errors.BlockDeviceError, err:
1304 _Fail("Can't make child '%s' read-write: %s", child, err)
1308 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1309 except errors.BlockDeviceError, err:
1310 _Fail("Can't create block device: %s", err)
1312 if on_primary or disk.AssembleOnSecondary():
1315 except errors.BlockDeviceError, err:
1316 _Fail("Can't assemble device after creation, unusual event: %s", err)
1317 device.SetSyncSpeed(constants.SYNC_SPEED)
1318 if on_primary or disk.OpenOnSecondary():
1320 device.Open(force=True)
1321 except errors.BlockDeviceError, err:
1322 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1323 DevCacheManager.UpdateCache(device.dev_path, owner,
1324 on_primary, disk.iv_name)
1326 device.SetInfo(info)
1328 return device.unique_id
1331 def BlockdevRemove(disk):
1332 """Remove a block device.
1334 @note: This is intended to be called recursively.
1336 @type disk: L{objects.Disk}
1337 @param disk: the disk object we should remove
1339 @return: the success of the operation
1344 rdev = _RecursiveFindBD(disk)
1345 except errors.BlockDeviceError, err:
1346 # probably can't attach
1347 logging.info("Can't attach to device %s in remove", disk)
1349 if rdev is not None:
1350 r_path = rdev.dev_path
1353 except errors.BlockDeviceError, err:
1354 msgs.append(str(err))
1356 DevCacheManager.RemoveCache(r_path)
1359 for child in disk.children:
1361 BlockdevRemove(child)
1362 except RPCFail, err:
1363 msgs.append(str(err))
1366 _Fail("; ".join(msgs))
1369 def _RecursiveAssembleBD(disk, owner, as_primary):
1370 """Activate a block device for an instance.
1372 This is run on the primary and secondary nodes for an instance.
1374 @note: this function is called recursively.
1376 @type disk: L{objects.Disk}
1377 @param disk: the disk we try to assemble
1379 @param owner: the name of the instance which owns the disk
1380 @type as_primary: boolean
1381 @param as_primary: if we should make the block device
1384 @return: the assembled device or None (in case no device
1386 @raise errors.BlockDeviceError: in case there is an error
1387 during the activation of the children or the device
1393 mcn = disk.ChildrenNeeded()
1395 mcn = 0 # max number of Nones allowed
1397 mcn = len(disk.children) - mcn # max number of Nones
1398 for chld_disk in disk.children:
1400 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1401 except errors.BlockDeviceError, err:
1402 if children.count(None) >= mcn:
1405 logging.error("Error in child activation (but continuing): %s",
1407 children.append(cdev)
1409 if as_primary or disk.AssembleOnSecondary():
1410 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1411 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1413 if as_primary or disk.OpenOnSecondary():
1415 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1416 as_primary, disk.iv_name)
1423 def BlockdevAssemble(disk, owner, as_primary):
1424 """Activate a block device for an instance.
1426 This is a wrapper over _RecursiveAssembleBD.
1428 @rtype: str or boolean
1429 @return: a C{/dev/...} path for primary nodes, and
1430 C{True} for secondary nodes
1434 result = _RecursiveAssembleBD(disk, owner, as_primary)
1435 if isinstance(result, bdev.BlockDev):
1436 # pylint: disable-msg=E1103
1437 result = result.dev_path
1438 except errors.BlockDeviceError, err:
1439 _Fail("Error while assembling disk: %s", err, exc=True)
1444 def BlockdevShutdown(disk):
1445 """Shut down a block device.
1447 First, if the device is assembled (Attach() is successful), then
1448 the device is shutdown. Then the children of the device are
1451 This function is called recursively. Note that we don't cache the
1452 children or such, as oppossed to assemble, shutdown of different
1453 devices doesn't require that the upper device was active.
1455 @type disk: L{objects.Disk}
1456 @param disk: the description of the disk we should
1462 r_dev = _RecursiveFindBD(disk)
1463 if r_dev is not None:
1464 r_path = r_dev.dev_path
1467 DevCacheManager.RemoveCache(r_path)
1468 except errors.BlockDeviceError, err:
1469 msgs.append(str(err))
1472 for child in disk.children:
1474 BlockdevShutdown(child)
1475 except RPCFail, err:
1476 msgs.append(str(err))
1479 _Fail("; ".join(msgs))
1482 def BlockdevAddchildren(parent_cdev, new_cdevs):
1483 """Extend a mirrored block device.
1485 @type parent_cdev: L{objects.Disk}
1486 @param parent_cdev: the disk to which we should add children
1487 @type new_cdevs: list of L{objects.Disk}
1488 @param new_cdevs: the list of children which we should add
1492 parent_bdev = _RecursiveFindBD(parent_cdev)
1493 if parent_bdev is None:
1494 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1495 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1496 if new_bdevs.count(None) > 0:
1497 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1498 parent_bdev.AddChildren(new_bdevs)
1501 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1502 """Shrink a mirrored block device.
1504 @type parent_cdev: L{objects.Disk}
1505 @param parent_cdev: the disk from which we should remove children
1506 @type new_cdevs: list of L{objects.Disk}
1507 @param new_cdevs: the list of children which we should remove
1511 parent_bdev = _RecursiveFindBD(parent_cdev)
1512 if parent_bdev is None:
1513 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1515 for disk in new_cdevs:
1516 rpath = disk.StaticDevPath()
1518 bd = _RecursiveFindBD(disk)
1520 _Fail("Can't find device %s while removing children", disk)
1522 devs.append(bd.dev_path)
1524 if not utils.IsNormAbsPath(rpath):
1525 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1527 parent_bdev.RemoveChildren(devs)
1530 def BlockdevGetmirrorstatus(disks):
1531 """Get the mirroring status of a list of devices.
1533 @type disks: list of L{objects.Disk}
1534 @param disks: the list of disks which we should query
1537 a list of (mirror_done, estimated_time) tuples, which
1538 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1539 @raise errors.BlockDeviceError: if any of the disks cannot be
1545 rbd = _RecursiveFindBD(dsk)
1547 _Fail("Can't find device %s", dsk)
1549 stats.append(rbd.CombinedSyncStatus())
1554 def _RecursiveFindBD(disk):
1555 """Check if a device is activated.
1557 If so, return information about the real device.
1559 @type disk: L{objects.Disk}
1560 @param disk: the disk object we need to find
1562 @return: None if the device can't be found,
1563 otherwise the device instance
1568 for chdisk in disk.children:
1569 children.append(_RecursiveFindBD(chdisk))
1571 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1574 def _OpenRealBD(disk):
1575 """Opens the underlying block device of a disk.
1577 @type disk: L{objects.Disk}
1578 @param disk: the disk object we want to open
1581 real_disk = _RecursiveFindBD(disk)
1582 if real_disk is None:
1583 _Fail("Block device '%s' is not set up", disk)
1590 def BlockdevFind(disk):
1591 """Check if a device is activated.
1593 If it is, return information about the real device.
1595 @type disk: L{objects.Disk}
1596 @param disk: the disk to find
1597 @rtype: None or objects.BlockDevStatus
1598 @return: None if the disk cannot be found, otherwise a the current
1603 rbd = _RecursiveFindBD(disk)
1604 except errors.BlockDeviceError, err:
1605 _Fail("Failed to find device: %s", err, exc=True)
1610 return rbd.GetSyncStatus()
1613 def BlockdevGetsize(disks):
1614 """Computes the size of the given disks.
1616 If a disk is not found, returns None instead.
1618 @type disks: list of L{objects.Disk}
1619 @param disks: the list of disk to compute the size for
1621 @return: list with elements None if the disk cannot be found,
1628 rbd = _RecursiveFindBD(cf)
1629 except errors.BlockDeviceError:
1635 result.append(rbd.GetActualSize())
1639 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1640 """Export a block device to a remote node.
1642 @type disk: L{objects.Disk}
1643 @param disk: the description of the disk to export
1644 @type dest_node: str
1645 @param dest_node: the destination node to export to
1646 @type dest_path: str
1647 @param dest_path: the destination path on the target node
1648 @type cluster_name: str
1649 @param cluster_name: the cluster name, needed for SSH hostalias
1653 real_disk = _OpenRealBD(disk)
1655 # the block size on the read dd is 1MiB to match our units
1656 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1657 "dd if=%s bs=1048576 count=%s",
1658 real_disk.dev_path, str(disk.size))
1660 # we set here a smaller block size as, due to ssh buffering, more
1661 # than 64-128k will mostly ignored; we use nocreat to fail if the
1662 # device is not already there or we pass a wrong path; we use
1663 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1664 # to not buffer too much memory; this means that at best, we flush
1665 # every 64k, which will not be very fast
1666 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1667 " oflag=dsync", dest_path)
1669 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1670 constants.GANETI_RUNAS,
1673 # all commands have been checked, so we're safe to combine them
1674 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1676 result = utils.RunCmd(["bash", "-c", command])
1679 _Fail("Disk copy command '%s' returned error: %s"
1680 " output: %s", command, result.fail_reason, result.output)
1683 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1684 """Write a file to the filesystem.
1686 This allows the master to overwrite(!) a file. It will only perform
1687 the operation if the file belongs to a list of configuration files.
1689 @type file_name: str
1690 @param file_name: the target file name
1692 @param data: the new contents of the file
1694 @param mode: the mode to give the file (can be None)
1696 @param uid: the owner of the file (can be -1 for default)
1698 @param gid: the group of the file (can be -1 for default)
1700 @param atime: the atime to set on the file (can be None)
1702 @param mtime: the mtime to set on the file (can be None)
1706 if not os.path.isabs(file_name):
1707 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1709 if file_name not in _ALLOWED_UPLOAD_FILES:
1710 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1713 raw_data = _Decompress(data)
1715 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1716 atime=atime, mtime=mtime)
1719 def WriteSsconfFiles(values):
1720 """Update all ssconf files.
1722 Wrapper around the SimpleStore.WriteFiles.
1725 ssconf.SimpleStore().WriteFiles(values)
1728 def _ErrnoOrStr(err):
1729 """Format an EnvironmentError exception.
1731 If the L{err} argument has an errno attribute, it will be looked up
1732 and converted into a textual C{E...} description. Otherwise the
1733 string representation of the error will be returned.
1735 @type err: L{EnvironmentError}
1736 @param err: the exception to format
1739 if hasattr(err, 'errno'):
1740 detail = errno.errorcode[err.errno]
1746 def _OSOndiskAPIVersion(os_dir):
1747 """Compute and return the API version of a given OS.
1749 This function will try to read the API version of the OS residing in
1750 the 'os_dir' directory.
1753 @param os_dir: the directory in which we should look for the OS
1755 @return: tuple (status, data) with status denoting the validity and
1756 data holding either the vaid versions or an error message
1759 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1762 st = os.stat(api_file)
1763 except EnvironmentError, err:
1764 return False, ("Required file '%s' not found under path %s: %s" %
1765 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1767 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1768 return False, ("File '%s' in %s is not a regular file" %
1769 (constants.OS_API_FILE, os_dir))
1772 api_versions = utils.ReadFile(api_file).splitlines()
1773 except EnvironmentError, err:
1774 return False, ("Error while reading the API version file at %s: %s" %
1775 (api_file, _ErrnoOrStr(err)))
1778 api_versions = [int(version.strip()) for version in api_versions]
1779 except (TypeError, ValueError), err:
1780 return False, ("API version(s) can't be converted to integer: %s" %
1783 return True, api_versions
1786 def DiagnoseOS(top_dirs=None):
1787 """Compute the validity for all OSes.
1789 @type top_dirs: list
1790 @param top_dirs: the list of directories in which to
1791 search (if not given defaults to
1792 L{constants.OS_SEARCH_PATH})
1793 @rtype: list of L{objects.OS}
1794 @return: a list of tuples (name, path, status, diagnose, variants,
1795 parameters, api_version) for all (potential) OSes under all
1796 search paths, where:
1797 - name is the (potential) OS name
1798 - path is the full path to the OS
1799 - status True/False is the validity of the OS
1800 - diagnose is the error message for an invalid OS, otherwise empty
1801 - variants is a list of supported OS variants, if any
1802 - parameters is a list of (name, help) parameters, if any
1803 - api_version is a list of support OS API versions
1806 if top_dirs is None:
1807 top_dirs = constants.OS_SEARCH_PATH
1810 for dir_name in top_dirs:
1811 if os.path.isdir(dir_name):
1813 f_names = utils.ListVisibleFiles(dir_name)
1814 except EnvironmentError, err:
1815 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1817 for name in f_names:
1818 os_path = utils.PathJoin(dir_name, name)
1819 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1822 variants = os_inst.supported_variants
1823 parameters = os_inst.supported_parameters
1824 api_versions = os_inst.api_versions
1827 variants = parameters = api_versions = []
1828 result.append((name, os_path, status, diagnose, variants,
1829 parameters, api_versions))
1834 def _TryOSFromDisk(name, base_dir=None):
1835 """Create an OS instance from disk.
1837 This function will return an OS instance if the given name is a
1840 @type base_dir: string
1841 @keyword base_dir: Base directory containing OS installations.
1842 Defaults to a search in all the OS_SEARCH_PATH dirs.
1844 @return: success and either the OS instance if we find a valid one,
1848 if base_dir is None:
1849 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1851 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1854 return False, "Directory for OS %s not found in search path" % name
1856 status, api_versions = _OSOndiskAPIVersion(os_dir)
1859 return status, api_versions
1861 if not constants.OS_API_VERSIONS.intersection(api_versions):
1862 return False, ("API version mismatch for path '%s': found %s, want %s." %
1863 (os_dir, api_versions, constants.OS_API_VERSIONS))
1865 # OS Files dictionary, we will populate it with the absolute path names
1866 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1868 if max(api_versions) >= constants.OS_API_V15:
1869 os_files[constants.OS_VARIANTS_FILE] = ''
1871 if max(api_versions) >= constants.OS_API_V20:
1872 os_files[constants.OS_PARAMETERS_FILE] = ''
1874 del os_files[constants.OS_SCRIPT_VERIFY]
1876 for filename in os_files:
1877 os_files[filename] = utils.PathJoin(os_dir, filename)
1880 st = os.stat(os_files[filename])
1881 except EnvironmentError, err:
1882 return False, ("File '%s' under path '%s' is missing (%s)" %
1883 (filename, os_dir, _ErrnoOrStr(err)))
1885 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1886 return False, ("File '%s' under path '%s' is not a regular file" %
1889 if filename in constants.OS_SCRIPTS:
1890 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1891 return False, ("File '%s' under path '%s' is not executable" %
1895 if constants.OS_VARIANTS_FILE in os_files:
1896 variants_file = os_files[constants.OS_VARIANTS_FILE]
1898 variants = utils.ReadFile(variants_file).splitlines()
1899 except EnvironmentError, err:
1900 return False, ("Error while reading the OS variants file at %s: %s" %
1901 (variants_file, _ErrnoOrStr(err)))
1903 return False, ("No supported os variant found")
1906 if constants.OS_PARAMETERS_FILE in os_files:
1907 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
1909 parameters = utils.ReadFile(parameters_file).splitlines()
1910 except EnvironmentError, err:
1911 return False, ("Error while reading the OS parameters file at %s: %s" %
1912 (parameters_file, _ErrnoOrStr(err)))
1913 parameters = [v.split(None, 1) for v in parameters]
1915 os_obj = objects.OS(name=name, path=os_dir,
1916 create_script=os_files[constants.OS_SCRIPT_CREATE],
1917 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1918 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1919 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1920 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
1922 supported_variants=variants,
1923 supported_parameters=parameters,
1924 api_versions=api_versions)
1928 def OSFromDisk(name, base_dir=None):
1929 """Create an OS instance from disk.
1931 This function will return an OS instance if the given name is a
1932 valid OS name. Otherwise, it will raise an appropriate
1933 L{RPCFail} exception, detailing why this is not a valid OS.
1935 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1936 an exception but returns true/false status data.
1938 @type base_dir: string
1939 @keyword base_dir: Base directory containing OS installations.
1940 Defaults to a search in all the OS_SEARCH_PATH dirs.
1941 @rtype: L{objects.OS}
1942 @return: the OS instance if we find a valid one
1943 @raise RPCFail: if we don't find a valid OS
1946 name_only = name.split("+", 1)[0]
1947 status, payload = _TryOSFromDisk(name_only, base_dir)
1955 def OSCoreEnv(inst_os, os_params, debug=0):
1956 """Calculate the basic environment for an os script.
1958 @type inst_os: L{objects.OS}
1959 @param inst_os: operating system for which the environment is being built
1960 @type os_params: dict
1961 @param os_params: the OS parameters
1962 @type debug: integer
1963 @param debug: debug level (0 or 1, for OS Api 10)
1965 @return: dict of environment variables
1966 @raise errors.BlockDeviceError: if the block device
1972 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1973 result['OS_API_VERSION'] = '%d' % api_version
1974 result['OS_NAME'] = inst_os.name
1975 result['DEBUG_LEVEL'] = '%d' % debug
1978 if api_version >= constants.OS_API_V15:
1980 variant = inst_os.name.split('+', 1)[1]
1982 variant = inst_os.supported_variants[0]
1983 result['OS_VARIANT'] = variant
1986 for pname, pvalue in os_params.items():
1987 result['OSP_%s' % pname.upper()] = pvalue
1992 def OSEnvironment(instance, inst_os, debug=0):
1993 """Calculate the environment for an os script.
1995 @type instance: L{objects.Instance}
1996 @param instance: target instance for the os script run
1997 @type inst_os: L{objects.OS}
1998 @param inst_os: operating system for which the environment is being built
1999 @type debug: integer
2000 @param debug: debug level (0 or 1, for OS Api 10)
2002 @return: dict of environment variables
2003 @raise errors.BlockDeviceError: if the block device
2007 result = OSCoreEnv(inst_os, instance.osparams, debug=debug)
2009 result['INSTANCE_NAME'] = instance.name
2010 result['INSTANCE_OS'] = instance.os
2011 result['HYPERVISOR'] = instance.hypervisor
2012 result['DISK_COUNT'] = '%d' % len(instance.disks)
2013 result['NIC_COUNT'] = '%d' % len(instance.nics)
2016 for idx, disk in enumerate(instance.disks):
2017 real_disk = _OpenRealBD(disk)
2018 result['DISK_%d_PATH' % idx] = real_disk.dev_path
2019 result['DISK_%d_ACCESS' % idx] = disk.mode
2020 if constants.HV_DISK_TYPE in instance.hvparams:
2021 result['DISK_%d_FRONTEND_TYPE' % idx] = \
2022 instance.hvparams[constants.HV_DISK_TYPE]
2023 if disk.dev_type in constants.LDS_BLOCK:
2024 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
2025 elif disk.dev_type == constants.LD_FILE:
2026 result['DISK_%d_BACKEND_TYPE' % idx] = \
2027 'file:%s' % disk.physical_id[0]
2030 for idx, nic in enumerate(instance.nics):
2031 result['NIC_%d_MAC' % idx] = nic.mac
2033 result['NIC_%d_IP' % idx] = nic.ip
2034 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
2035 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2036 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
2037 if nic.nicparams[constants.NIC_LINK]:
2038 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
2039 if constants.HV_NIC_TYPE in instance.hvparams:
2040 result['NIC_%d_FRONTEND_TYPE' % idx] = \
2041 instance.hvparams[constants.HV_NIC_TYPE]
2044 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2045 for key, value in source.items():
2046 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2051 def BlockdevGrow(disk, amount):
2052 """Grow a stack of block devices.
2054 This function is called recursively, with the childrens being the
2055 first ones to resize.
2057 @type disk: L{objects.Disk}
2058 @param disk: the disk to be grown
2059 @rtype: (status, result)
2060 @return: a tuple with the status of the operation
2061 (True/False), and the errors message if status
2065 r_dev = _RecursiveFindBD(disk)
2067 _Fail("Cannot find block device %s", disk)
2071 except errors.BlockDeviceError, err:
2072 _Fail("Failed to grow block device: %s", err, exc=True)
2075 def BlockdevSnapshot(disk):
2076 """Create a snapshot copy of a block device.
2078 This function is called recursively, and the snapshot is actually created
2079 just for the leaf lvm backend device.
2081 @type disk: L{objects.Disk}
2082 @param disk: the disk to be snapshotted
2084 @return: snapshot disk path
2087 if disk.dev_type == constants.LD_DRBD8:
2088 if not disk.children:
2089 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2091 return BlockdevSnapshot(disk.children[0])
2092 elif disk.dev_type == constants.LD_LV:
2093 r_dev = _RecursiveFindBD(disk)
2094 if r_dev is not None:
2095 # FIXME: choose a saner value for the snapshot size
2096 # let's stay on the safe side and ask for the full size, for now
2097 return r_dev.Snapshot(disk.size)
2099 _Fail("Cannot find block device %s", disk)
2101 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2102 disk.unique_id, disk.dev_type)
2105 def FinalizeExport(instance, snap_disks):
2106 """Write out the export configuration information.
2108 @type instance: L{objects.Instance}
2109 @param instance: the instance which we export, used for
2110 saving configuration
2111 @type snap_disks: list of L{objects.Disk}
2112 @param snap_disks: list of snapshot block devices, which
2113 will be used to get the actual name of the dump file
2118 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2119 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2121 config = objects.SerializableConfigParser()
2123 config.add_section(constants.INISECT_EXP)
2124 config.set(constants.INISECT_EXP, 'version', '0')
2125 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2126 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2127 config.set(constants.INISECT_EXP, 'os', instance.os)
2128 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2130 config.add_section(constants.INISECT_INS)
2131 config.set(constants.INISECT_INS, 'name', instance.name)
2132 config.set(constants.INISECT_INS, 'memory', '%d' %
2133 instance.beparams[constants.BE_MEMORY])
2134 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2135 instance.beparams[constants.BE_VCPUS])
2136 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2137 config.set(constants.INISECT_INS, 'hypervisor', instance.hypervisor)
2140 for nic_count, nic in enumerate(instance.nics):
2142 config.set(constants.INISECT_INS, 'nic%d_mac' %
2143 nic_count, '%s' % nic.mac)
2144 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2145 for param in constants.NICS_PARAMETER_TYPES:
2146 config.set(constants.INISECT_INS, 'nic%d_%s' % (nic_count, param),
2147 '%s' % nic.nicparams.get(param, None))
2148 # TODO: redundant: on load can read nics until it doesn't exist
2149 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2152 for disk_count, disk in enumerate(snap_disks):
2155 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2156 ('%s' % disk.iv_name))
2157 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2158 ('%s' % disk.physical_id[1]))
2159 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2162 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2164 # New-style hypervisor/backend parameters
2166 config.add_section(constants.INISECT_HYP)
2167 for name, value in instance.hvparams.items():
2168 if name not in constants.HVC_GLOBALS:
2169 config.set(constants.INISECT_HYP, name, str(value))
2171 config.add_section(constants.INISECT_BEP)
2172 for name, value in instance.beparams.items():
2173 config.set(constants.INISECT_BEP, name, str(value))
2175 config.add_section(constants.INISECT_OSP)
2176 for name, value in instance.osparams.items():
2177 config.set(constants.INISECT_OSP, name, str(value))
2179 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2180 data=config.Dumps())
2181 shutil.rmtree(finaldestdir, ignore_errors=True)
2182 shutil.move(destdir, finaldestdir)
2185 def ExportInfo(dest):
2186 """Get export configuration information.
2189 @param dest: directory containing the export
2191 @rtype: L{objects.SerializableConfigParser}
2192 @return: a serializable config file containing the
2196 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2198 config = objects.SerializableConfigParser()
2201 if (not config.has_section(constants.INISECT_EXP) or
2202 not config.has_section(constants.INISECT_INS)):
2203 _Fail("Export info file doesn't have the required fields")
2205 return config.Dumps()
2209 """Return a list of exports currently available on this machine.
2212 @return: list of the exports
2215 if os.path.isdir(constants.EXPORT_DIR):
2216 return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2218 _Fail("No exports directory")
2221 def RemoveExport(export):
2222 """Remove an existing export from the node.
2225 @param export: the name of the export to remove
2229 target = utils.PathJoin(constants.EXPORT_DIR, export)
2232 shutil.rmtree(target)
2233 except EnvironmentError, err:
2234 _Fail("Error while removing the export: %s", err, exc=True)
2237 def BlockdevRename(devlist):
2238 """Rename a list of block devices.
2240 @type devlist: list of tuples
2241 @param devlist: list of tuples of the form (disk,
2242 new_logical_id, new_physical_id); disk is an
2243 L{objects.Disk} object describing the current disk,
2244 and new logical_id/physical_id is the name we
2247 @return: True if all renames succeeded, False otherwise
2252 for disk, unique_id in devlist:
2253 dev = _RecursiveFindBD(disk)
2255 msgs.append("Can't find device %s in rename" % str(disk))
2259 old_rpath = dev.dev_path
2260 dev.Rename(unique_id)
2261 new_rpath = dev.dev_path
2262 if old_rpath != new_rpath:
2263 DevCacheManager.RemoveCache(old_rpath)
2264 # FIXME: we should add the new cache information here, like:
2265 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2266 # but we don't have the owner here - maybe parse from existing
2267 # cache? for now, we only lose lvm data when we rename, which
2268 # is less critical than DRBD or MD
2269 except errors.BlockDeviceError, err:
2270 msgs.append("Can't rename device '%s' to '%s': %s" %
2271 (dev, unique_id, err))
2272 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2275 _Fail("; ".join(msgs))
2278 def _TransformFileStorageDir(file_storage_dir):
2279 """Checks whether given file_storage_dir is valid.
2281 Checks wheter the given file_storage_dir is within the cluster-wide
2282 default file_storage_dir stored in SimpleStore. Only paths under that
2283 directory are allowed.
2285 @type file_storage_dir: str
2286 @param file_storage_dir: the path to check
2288 @return: the normalized path if valid, None otherwise
2291 if not constants.ENABLE_FILE_STORAGE:
2292 _Fail("File storage disabled at configure time")
2294 file_storage_dir = os.path.normpath(file_storage_dir)
2295 base_file_storage_dir = cfg.GetFileStorageDir()
2296 if (os.path.commonprefix([file_storage_dir, base_file_storage_dir]) !=
2297 base_file_storage_dir):
2298 _Fail("File storage directory '%s' is not under base file"
2299 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2300 return file_storage_dir
2303 def CreateFileStorageDir(file_storage_dir):
2304 """Create file storage directory.
2306 @type file_storage_dir: str
2307 @param file_storage_dir: directory to create
2310 @return: tuple with first element a boolean indicating wheter dir
2311 creation was successful or not
2314 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2315 if os.path.exists(file_storage_dir):
2316 if not os.path.isdir(file_storage_dir):
2317 _Fail("Specified storage dir '%s' is not a directory",
2321 os.makedirs(file_storage_dir, 0750)
2322 except OSError, err:
2323 _Fail("Cannot create file storage directory '%s': %s",
2324 file_storage_dir, err, exc=True)
2327 def RemoveFileStorageDir(file_storage_dir):
2328 """Remove file storage directory.
2330 Remove it only if it's empty. If not log an error and return.
2332 @type file_storage_dir: str
2333 @param file_storage_dir: the directory we should cleanup
2334 @rtype: tuple (success,)
2335 @return: tuple of one element, C{success}, denoting
2336 whether the operation was successful
2339 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2340 if os.path.exists(file_storage_dir):
2341 if not os.path.isdir(file_storage_dir):
2342 _Fail("Specified Storage directory '%s' is not a directory",
2344 # deletes dir only if empty, otherwise we want to fail the rpc call
2346 os.rmdir(file_storage_dir)
2347 except OSError, err:
2348 _Fail("Cannot remove file storage directory '%s': %s",
2349 file_storage_dir, err)
2352 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2353 """Rename the file storage directory.
2355 @type old_file_storage_dir: str
2356 @param old_file_storage_dir: the current path
2357 @type new_file_storage_dir: str
2358 @param new_file_storage_dir: the name we should rename to
2359 @rtype: tuple (success,)
2360 @return: tuple of one element, C{success}, denoting
2361 whether the operation was successful
2364 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2365 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2366 if not os.path.exists(new_file_storage_dir):
2367 if os.path.isdir(old_file_storage_dir):
2369 os.rename(old_file_storage_dir, new_file_storage_dir)
2370 except OSError, err:
2371 _Fail("Cannot rename '%s' to '%s': %s",
2372 old_file_storage_dir, new_file_storage_dir, err)
2374 _Fail("Specified storage dir '%s' is not a directory",
2375 old_file_storage_dir)
2377 if os.path.exists(old_file_storage_dir):
2378 _Fail("Cannot rename '%s' to '%s': both locations exist",
2379 old_file_storage_dir, new_file_storage_dir)
2382 def _EnsureJobQueueFile(file_name):
2383 """Checks whether the given filename is in the queue directory.
2385 @type file_name: str
2386 @param file_name: the file name we should check
2388 @raises RPCFail: if the file is not valid
2391 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2392 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2395 _Fail("Passed job queue file '%s' does not belong to"
2396 " the queue directory '%s'", file_name, queue_dir)
2399 def JobQueueUpdate(file_name, content):
2400 """Updates a file in the queue directory.
2402 This is just a wrapper over L{utils.WriteFile}, with proper
2405 @type file_name: str
2406 @param file_name: the job file name
2408 @param content: the new job contents
2410 @return: the success of the operation
2413 _EnsureJobQueueFile(file_name)
2415 # Write and replace the file atomically
2416 utils.WriteFile(file_name, data=_Decompress(content))
2419 def JobQueueRename(old, new):
2420 """Renames a job queue file.
2422 This is just a wrapper over os.rename with proper checking.
2425 @param old: the old (actual) file name
2427 @param new: the desired file name
2429 @return: the success of the operation and payload
2432 _EnsureJobQueueFile(old)
2433 _EnsureJobQueueFile(new)
2435 utils.RenameFile(old, new, mkdir=True)
2438 def BlockdevClose(instance_name, disks):
2439 """Closes the given block devices.
2441 This means they will be switched to secondary mode (in case of
2444 @param instance_name: if the argument is not empty, the symlinks
2445 of this instance will be removed
2446 @type disks: list of L{objects.Disk}
2447 @param disks: the list of disks to be closed
2448 @rtype: tuple (success, message)
2449 @return: a tuple of success and message, where success
2450 indicates the succes of the operation, and message
2451 which will contain the error details in case we
2457 rd = _RecursiveFindBD(cf)
2459 _Fail("Can't find device %s", cf)
2466 except errors.BlockDeviceError, err:
2467 msg.append(str(err))
2469 _Fail("Can't make devices secondary: %s", ",".join(msg))
2472 _RemoveBlockDevLinks(instance_name, disks)
2475 def ValidateHVParams(hvname, hvparams):
2476 """Validates the given hypervisor parameters.
2478 @type hvname: string
2479 @param hvname: the hypervisor name
2480 @type hvparams: dict
2481 @param hvparams: the hypervisor parameters to be validated
2486 hv_type = hypervisor.GetHypervisor(hvname)
2487 hv_type.ValidateParameters(hvparams)
2488 except errors.HypervisorError, err:
2489 _Fail(str(err), log=False)
2492 def _CheckOSPList(os_obj, parameters):
2493 """Check whether a list of parameters is supported by the OS.
2495 @type os_obj: L{objects.OS}
2496 @param os_obj: OS object to check
2497 @type parameters: list
2498 @param parameters: the list of parameters to check
2501 supported = [v[0] for v in os_obj.supported_parameters]
2502 delta = frozenset(parameters).difference(supported)
2504 _Fail("The following parameters are not supported"
2505 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2508 def ValidateOS(required, osname, checks, osparams):
2509 """Validate the given OS' parameters.
2511 @type required: boolean
2512 @param required: whether absence of the OS should translate into
2514 @type osname: string
2515 @param osname: the OS to be validated
2517 @param checks: list of the checks to run (currently only 'parameters')
2518 @type osparams: dict
2519 @param osparams: dictionary with OS parameters
2521 @return: True if the validation passed, or False if the OS was not
2522 found and L{required} was false
2525 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2526 _Fail("Unknown checks required for OS %s: %s", osname,
2527 set(checks).difference(constants.OS_VALIDATE_CALLS))
2529 name_only = osname.split("+", 1)[0]
2530 status, tbv = _TryOSFromDisk(name_only, None)
2538 if max(tbv.api_versions) < constants.OS_API_V20:
2541 if constants.OS_VALIDATE_PARAMETERS in checks:
2542 _CheckOSPList(tbv, osparams.keys())
2544 validate_env = OSCoreEnv(tbv, osparams)
2545 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2548 logging.error("os validate command '%s' returned error: %s output: %s",
2549 result.cmd, result.fail_reason, result.output)
2550 _Fail("OS validation script failed (%s), output: %s",
2551 result.fail_reason, result.output, log=False)
2557 """Demotes the current node from master candidate role.
2560 # try to ensure we're not the master by mistake
2561 master, myself = ssconf.GetMasterAndMyself()
2562 if master == myself:
2563 _Fail("ssconf status shows I'm the master node, will not demote")
2565 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2566 if not result.failed:
2567 _Fail("The master daemon is running, will not demote")
2570 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2571 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2572 except EnvironmentError, err:
2573 if err.errno != errno.ENOENT:
2574 _Fail("Error while backing up cluster file: %s", err, exc=True)
2576 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2579 def _GetX509Filenames(cryptodir, name):
2580 """Returns the full paths for the private key and certificate.
2583 return (utils.PathJoin(cryptodir, name),
2584 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2585 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2588 def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2589 """Creates a new X509 certificate for SSL/TLS.
2592 @param validity: Validity in seconds
2593 @rtype: tuple; (string, string)
2594 @return: Certificate name and public part
2597 (key_pem, cert_pem) = \
2598 utils.GenerateSelfSignedX509Cert(netutils.HostInfo.SysName(),
2599 min(validity, _MAX_SSL_CERT_VALIDITY))
2601 cert_dir = tempfile.mkdtemp(dir=cryptodir,
2602 prefix="x509-%s-" % utils.TimestampForFilename())
2604 name = os.path.basename(cert_dir)
2605 assert len(name) > 5
2607 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2609 utils.WriteFile(key_file, mode=0400, data=key_pem)
2610 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2612 # Never return private key as it shouldn't leave the node
2613 return (name, cert_pem)
2615 shutil.rmtree(cert_dir, ignore_errors=True)
2619 def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2620 """Removes a X509 certificate.
2623 @param name: Certificate name
2626 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2628 utils.RemoveFile(key_file)
2629 utils.RemoveFile(cert_file)
2633 except EnvironmentError, err:
2634 _Fail("Cannot remove certificate directory '%s': %s",
2638 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2639 """Returns the command for the requested input/output.
2641 @type instance: L{objects.Instance}
2642 @param instance: The instance object
2643 @param mode: Import/export mode
2644 @param ieio: Input/output type
2645 @param ieargs: Input/output arguments
2648 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2655 if ieio == constants.IEIO_FILE:
2656 (filename, ) = ieargs
2658 if not utils.IsNormAbsPath(filename):
2659 _Fail("Path '%s' is not normalized or absolute", filename)
2661 directory = os.path.normpath(os.path.dirname(filename))
2663 if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2664 constants.EXPORT_DIR):
2665 _Fail("File '%s' is not under exports directory '%s'",
2666 filename, constants.EXPORT_DIR)
2669 utils.Makedirs(directory, mode=0750)
2671 quoted_filename = utils.ShellQuote(filename)
2673 if mode == constants.IEM_IMPORT:
2674 suffix = "> %s" % quoted_filename
2675 elif mode == constants.IEM_EXPORT:
2676 suffix = "< %s" % quoted_filename
2678 # Retrieve file size
2680 st = os.stat(filename)
2681 except EnvironmentError, err:
2682 logging.error("Can't stat(2) %s: %s", filename, err)
2684 exp_size = utils.BytesToMebibyte(st.st_size)
2686 elif ieio == constants.IEIO_RAW_DISK:
2689 real_disk = _OpenRealBD(disk)
2691 if mode == constants.IEM_IMPORT:
2692 # we set here a smaller block size as, due to transport buffering, more
2693 # than 64-128k will mostly ignored; we use nocreat to fail if the device
2694 # is not already there or we pass a wrong path; we use notrunc to no
2695 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2696 # much memory; this means that at best, we flush every 64k, which will
2698 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2699 " bs=%s oflag=dsync"),
2703 elif mode == constants.IEM_EXPORT:
2704 # the block size on the read dd is 1MiB to match our units
2705 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2707 str(1024 * 1024), # 1 MB
2709 exp_size = disk.size
2711 elif ieio == constants.IEIO_SCRIPT:
2712 (disk, disk_index, ) = ieargs
2714 assert isinstance(disk_index, (int, long))
2716 real_disk = _OpenRealBD(disk)
2718 inst_os = OSFromDisk(instance.os)
2719 env = OSEnvironment(instance, inst_os)
2721 if mode == constants.IEM_IMPORT:
2722 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2723 env["IMPORT_INDEX"] = str(disk_index)
2724 script = inst_os.import_script
2726 elif mode == constants.IEM_EXPORT:
2727 env["EXPORT_DEVICE"] = real_disk.dev_path
2728 env["EXPORT_INDEX"] = str(disk_index)
2729 script = inst_os.export_script
2731 # TODO: Pass special environment only to script
2732 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2734 if mode == constants.IEM_IMPORT:
2735 suffix = "| %s" % script_cmd
2737 elif mode == constants.IEM_EXPORT:
2738 prefix = "%s |" % script_cmd
2740 # Let script predict size
2741 exp_size = constants.IE_CUSTOM_SIZE
2744 _Fail("Invalid %s I/O mode %r", mode, ieio)
2746 return (env, prefix, suffix, exp_size)
2749 def _CreateImportExportStatusDir(prefix):
2750 """Creates status directory for import/export.
2753 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2755 (prefix, utils.TimestampForFilename())))
2758 def StartImportExportDaemon(mode, opts, host, port, instance, ieio, ieioargs):
2759 """Starts an import or export daemon.
2761 @param mode: Import/output mode
2762 @type opts: L{objects.ImportExportOptions}
2763 @param opts: Daemon options
2765 @param host: Remote host for export (None for import)
2767 @param port: Remote port for export (None for import)
2768 @type instance: L{objects.Instance}
2769 @param instance: Instance object
2770 @param ieio: Input/output type
2771 @param ieioargs: Input/output arguments
2774 if mode == constants.IEM_IMPORT:
2777 if not (host is None and port is None):
2778 _Fail("Can not specify host or port on import")
2780 elif mode == constants.IEM_EXPORT:
2783 if host is None or port is None:
2784 _Fail("Host and port must be specified for an export")
2787 _Fail("Invalid mode %r", mode)
2789 if (opts.key_name is None) ^ (opts.ca_pem is None):
2790 _Fail("Cluster certificate can only be used for both key and CA")
2792 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
2793 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
2795 if opts.key_name is None:
2797 key_path = constants.NODED_CERT_FILE
2798 cert_path = constants.NODED_CERT_FILE
2799 assert opts.ca_pem is None
2801 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
2803 assert opts.ca_pem is not None
2805 for i in [key_path, cert_path]:
2806 if not os.path.exists(i):
2807 _Fail("File '%s' does not exist" % i)
2809 status_dir = _CreateImportExportStatusDir(prefix)
2811 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
2812 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
2813 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
2815 if opts.ca_pem is None:
2817 ca = utils.ReadFile(constants.NODED_CERT_FILE)
2822 utils.WriteFile(ca_file, data=ca, mode=0400)
2825 constants.IMPORT_EXPORT_DAEMON,
2827 "--key=%s" % key_path,
2828 "--cert=%s" % cert_path,
2829 "--ca=%s" % ca_file,
2833 cmd.append("--host=%s" % host)
2836 cmd.append("--port=%s" % port)
2839 cmd.append("--compress=%s" % opts.compress)
2842 cmd.append("--magic=%s" % opts.magic)
2844 if exp_size is not None:
2845 cmd.append("--expected-size=%s" % exp_size)
2848 cmd.append("--cmd-prefix=%s" % cmd_prefix)
2851 cmd.append("--cmd-suffix=%s" % cmd_suffix)
2853 logfile = _InstanceLogName(prefix, instance.os, instance.name)
2855 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
2856 # support for receiving a file descriptor for output
2857 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
2860 # The import/export name is simply the status directory name
2861 return os.path.basename(status_dir)
2864 shutil.rmtree(status_dir, ignore_errors=True)
2868 def GetImportExportStatus(names):
2869 """Returns import/export daemon status.
2871 @type names: sequence
2872 @param names: List of names
2873 @rtype: List of dicts
2874 @return: Returns a list of the state of each named import/export or None if a
2875 status couldn't be read
2881 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
2885 data = utils.ReadFile(status_file)
2886 except EnvironmentError, err:
2887 if err.errno != errno.ENOENT:
2895 result.append(serializer.LoadJson(data))
2900 def AbortImportExport(name):
2901 """Sends SIGTERM to a running import/export daemon.
2904 logging.info("Abort import/export %s", name)
2906 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
2907 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
2910 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
2912 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
2915 def CleanupImportExport(name):
2916 """Cleanup after an import or export.
2918 If the import/export daemon is still running it's killed. Afterwards the
2919 whole status directory is removed.
2922 logging.info("Finalizing import/export %s", name)
2924 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
2926 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
2929 logging.info("Import/export %s is still running with PID %s",
2931 utils.KillProcess(pid, waitpid=False)
2933 shutil.rmtree(status_dir, ignore_errors=True)
2936 def _FindDisks(nodes_ip, disks):
2937 """Sets the physical ID on disks and returns the block devices.
2940 # set the correct physical ID
2941 my_name = netutils.HostInfo().name
2943 cf.SetPhysicalID(my_name, nodes_ip)
2948 rd = _RecursiveFindBD(cf)
2950 _Fail("Can't find device %s", cf)
2955 def DrbdDisconnectNet(nodes_ip, disks):
2956 """Disconnects the network on a list of drbd devices.
2959 bdevs = _FindDisks(nodes_ip, disks)
2965 except errors.BlockDeviceError, err:
2966 _Fail("Can't change network configuration to standalone mode: %s",
2970 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2971 """Attaches the network on a list of drbd devices.
2974 bdevs = _FindDisks(nodes_ip, disks)
2977 for idx, rd in enumerate(bdevs):
2979 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2980 except EnvironmentError, err:
2981 _Fail("Can't create symlink: %s", err)
2982 # reconnect disks, switch to new master configuration and if
2983 # needed primary mode
2986 rd.AttachNet(multimaster)
2987 except errors.BlockDeviceError, err:
2988 _Fail("Can't change network configuration: %s", err)
2990 # wait until the disks are connected; we need to retry the re-attach
2991 # if the device becomes standalone, as this might happen if the one
2992 # node disconnects and reconnects in a different mode before the
2993 # other node reconnects; in this case, one or both of the nodes will
2994 # decide it has wrong configuration and switch to standalone
2997 all_connected = True
3000 stats = rd.GetProcStatus()
3002 all_connected = (all_connected and
3003 (stats.is_connected or stats.is_in_resync))
3005 if stats.is_standalone:
3006 # peer had different config info and this node became
3007 # standalone, even though this should not happen with the
3008 # new staged way of changing disk configs
3010 rd.AttachNet(multimaster)
3011 except errors.BlockDeviceError, err:
3012 _Fail("Can't change network configuration: %s", err)
3014 if not all_connected:
3015 raise utils.RetryAgain()
3018 # Start with a delay of 100 miliseconds and go up to 5 seconds
3019 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3020 except utils.RetryTimeout:
3021 _Fail("Timeout in disk reconnecting")
3024 # change to primary mode
3028 except errors.BlockDeviceError, err:
3029 _Fail("Can't change to primary mode: %s", err)
3032 def DrbdWaitSync(nodes_ip, disks):
3033 """Wait until DRBDs have synchronized.
3037 stats = rd.GetProcStatus()
3038 if not (stats.is_connected or stats.is_in_resync):
3039 raise utils.RetryAgain()
3042 bdevs = _FindDisks(nodes_ip, disks)
3048 # poll each second for 15 seconds
3049 stats = utils.Retry(_helper, 1, 15, args=[rd])
3050 except utils.RetryTimeout:
3051 stats = rd.GetProcStatus()
3053 if not (stats.is_connected or stats.is_in_resync):
3054 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3055 alldone = alldone and (not stats.is_in_resync)
3056 if stats.sync_percent is not None:
3057 min_resync = min(min_resync, stats.sync_percent)
3059 return (alldone, min_resync)
3062 def GetDrbdUsermodeHelper():
3063 """Returns DRBD usermode helper currently configured.
3067 return bdev.BaseDRBD.GetUsermodeHelper()
3068 except errors.BlockDeviceError, err:
3072 def PowercycleNode(hypervisor_type):
3073 """Hard-powercycle the node.
3075 Because we need to return first, and schedule the powercycle in the
3076 background, we won't be able to report failures nicely.
3079 hyper = hypervisor.GetHypervisor(hypervisor_type)
3083 # if we can't fork, we'll pretend that we're in the child process
3086 return "Reboot scheduled in 5 seconds"
3087 # ensure the child is running on ram
3090 except Exception: # pylint: disable-msg=W0703
3093 hyper.PowercycleNode()
3096 class HooksRunner(object):
3099 This class is instantiated on the node side (ganeti-noded) and not
3103 def __init__(self, hooks_base_dir=None):
3104 """Constructor for hooks runner.
3106 @type hooks_base_dir: str or None
3107 @param hooks_base_dir: if not None, this overrides the
3108 L{constants.HOOKS_BASE_DIR} (useful for unittests)
3111 if hooks_base_dir is None:
3112 hooks_base_dir = constants.HOOKS_BASE_DIR
3113 # yeah, _BASE_DIR is not valid for attributes, we use it like a
3115 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3117 def RunHooks(self, hpath, phase, env):
3118 """Run the scripts in the hooks directory.
3121 @param hpath: the path to the hooks directory which
3124 @param phase: either L{constants.HOOKS_PHASE_PRE} or
3125 L{constants.HOOKS_PHASE_POST}
3127 @param env: dictionary with the environment for the hook
3129 @return: list of 3-element tuples:
3131 - script result, either L{constants.HKR_SUCCESS} or
3132 L{constants.HKR_FAIL}
3133 - output of the script
3135 @raise errors.ProgrammerError: for invalid input
3139 if phase == constants.HOOKS_PHASE_PRE:
3141 elif phase == constants.HOOKS_PHASE_POST:
3144 _Fail("Unknown hooks phase '%s'", phase)
3147 subdir = "%s-%s.d" % (hpath, suffix)
3148 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3152 if not os.path.isdir(dir_name):
3153 # for non-existing/non-dirs, we simply exit instead of logging a
3154 # warning at every operation
3157 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3159 for (relname, relstatus, runresult) in runparts_results:
3160 if relstatus == constants.RUNPARTS_SKIP:
3161 rrval = constants.HKR_SKIP
3163 elif relstatus == constants.RUNPARTS_ERR:
3164 rrval = constants.HKR_FAIL
3165 output = "Hook script execution error: %s" % runresult
3166 elif relstatus == constants.RUNPARTS_RUN:
3167 if runresult.failed:
3168 rrval = constants.HKR_FAIL
3170 rrval = constants.HKR_SUCCESS
3171 output = utils.SafeEncode(runresult.output.strip())
3172 results.append(("%s/%s" % (subdir, relname), rrval, output))
3177 class IAllocatorRunner(object):
3178 """IAllocator runner.
3180 This class is instantiated on the node side (ganeti-noded) and not on
3185 def Run(name, idata):
3186 """Run an iallocator script.
3189 @param name: the iallocator script name
3191 @param idata: the allocator input data
3194 @return: two element tuple of:
3196 - either error message or stdout of allocator (for success)
3199 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3201 if alloc_script is None:
3202 _Fail("iallocator module '%s' not found in the search path", name)
3204 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3208 result = utils.RunCmd([alloc_script, fin_name])
3210 _Fail("iallocator module '%s' failed: %s, output '%s'",
3211 name, result.fail_reason, result.output)
3215 return result.stdout
3218 class DevCacheManager(object):
3219 """Simple class for managing a cache of block device information.
3222 _DEV_PREFIX = "/dev/"
3223 _ROOT_DIR = constants.BDEV_CACHE_DIR
3226 def _ConvertPath(cls, dev_path):
3227 """Converts a /dev/name path to the cache file name.
3229 This replaces slashes with underscores and strips the /dev
3230 prefix. It then returns the full path to the cache file.
3233 @param dev_path: the C{/dev/} path name
3235 @return: the converted path name
3238 if dev_path.startswith(cls._DEV_PREFIX):
3239 dev_path = dev_path[len(cls._DEV_PREFIX):]
3240 dev_path = dev_path.replace("/", "_")
3241 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3245 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3246 """Updates the cache information for a given device.
3249 @param dev_path: the pathname of the device
3251 @param owner: the owner (instance name) of the device
3252 @type on_primary: bool
3253 @param on_primary: whether this is the primary
3256 @param iv_name: the instance-visible name of the
3257 device, as in objects.Disk.iv_name
3262 if dev_path is None:
3263 logging.error("DevCacheManager.UpdateCache got a None dev_path")
3265 fpath = cls._ConvertPath(dev_path)
3271 iv_name = "not_visible"
3272 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3274 utils.WriteFile(fpath, data=fdata)
3275 except EnvironmentError, err:
3276 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3279 def RemoveCache(cls, dev_path):
3280 """Remove data for a dev_path.
3282 This is just a wrapper over L{utils.RemoveFile} with a converted
3283 path name and logging.
3286 @param dev_path: the pathname of the device
3291 if dev_path is None:
3292 logging.error("DevCacheManager.RemoveCache got a None dev_path")
3294 fpath = cls._ConvertPath(dev_path)
3296 utils.RemoveFile(fpath)
3297 except EnvironmentError, err:
3298 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)