4 # Copyright (C) 2006, 2007, 2008, 2009, 2010 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"
80 class RPCFail(Exception):
81 """Class denoting RPC failure.
83 Its argument is the error message.
88 def _Fail(msg, *args, **kwargs):
89 """Log an error and the raise an RPCFail exception.
91 This exception is then handled specially in the ganeti daemon and
92 turned into a 'failed' return type. As such, this function is a
93 useful shortcut for logging the error and returning it to the master
97 @param msg: the text of the exception
103 if "log" not in kwargs or kwargs["log"]: # if we should log this error
104 if "exc" in kwargs and kwargs["exc"]:
105 logging.exception(msg)
112 """Simple wrapper to return a SimpleStore.
114 @rtype: L{ssconf.SimpleStore}
115 @return: a SimpleStore instance
118 return ssconf.SimpleStore()
121 def _GetSshRunner(cluster_name):
122 """Simple wrapper to return an SshRunner.
124 @type cluster_name: str
125 @param cluster_name: the cluster name, which is needed
126 by the SshRunner constructor
127 @rtype: L{ssh.SshRunner}
128 @return: an SshRunner instance
131 return ssh.SshRunner(cluster_name)
134 def _Decompress(data):
135 """Unpacks data compressed by the RPC client.
137 @type data: list or tuple
138 @param data: Data sent by RPC client
140 @return: Decompressed data
143 assert isinstance(data, (list, tuple))
144 assert len(data) == 2
145 (encoding, content) = data
146 if encoding == constants.RPC_ENCODING_NONE:
148 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
149 return zlib.decompress(base64.b64decode(content))
151 raise AssertionError("Unknown data encoding")
154 def _CleanDirectory(path, exclude=None):
155 """Removes all regular files in a directory.
158 @param path: the directory to clean
160 @param exclude: list of files to be excluded, defaults
164 if path not in _ALLOWED_CLEAN_DIRS:
165 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
168 if not os.path.isdir(path):
173 # Normalize excluded paths
174 exclude = [os.path.normpath(i) for i in exclude]
176 for rel_name in utils.ListVisibleFiles(path):
177 full_name = utils.PathJoin(path, rel_name)
178 if full_name in exclude:
180 if os.path.isfile(full_name) and not os.path.islink(full_name):
181 utils.RemoveFile(full_name)
184 def _BuildUploadFileList():
185 """Build the list of allowed upload files.
187 This is abstracted so that it's built only once at module import time.
190 allowed_files = set([
191 constants.CLUSTER_CONF_FILE,
193 constants.SSH_KNOWN_HOSTS_FILE,
194 constants.VNC_PASSWORD_FILE,
195 constants.RAPI_CERT_FILE,
196 constants.RAPI_USERS_FILE,
197 constants.CONFD_HMAC_KEY,
198 constants.CLUSTER_DOMAIN_SECRET_FILE,
201 for hv_name in constants.HYPER_TYPES:
202 hv_class = hypervisor.GetHypervisorClass(hv_name)
203 allowed_files.update(hv_class.GetAncillaryFiles())
205 return frozenset(allowed_files)
208 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
212 """Removes job queue files and archived jobs.
218 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
219 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
223 """Returns master information.
225 This is an utility function to compute master information, either
226 for consumption here or from the node daemon.
229 @return: master_netdev, master_ip, master_name, primary_ip_family
230 @raise RPCFail: in case of errors
235 master_netdev = cfg.GetMasterNetdev()
236 master_ip = cfg.GetMasterIP()
237 master_node = cfg.GetMasterNode()
238 primary_ip_family = cfg.GetPrimaryIPFamily()
239 except errors.ConfigurationError, err:
240 _Fail("Cluster configuration incomplete: %s", err, exc=True)
241 return (master_netdev, master_ip, master_node, primary_ip_family)
244 def StartMaster(start_daemons, no_voting):
245 """Activate local node as master node.
247 The function will either try activate the IP address of the master
248 (unless someone else has it) or also start the master daemons, based
249 on the start_daemons parameter.
251 @type start_daemons: boolean
252 @param start_daemons: whether to start the master daemons
253 (ganeti-masterd and ganeti-rapi), or (if false) activate the
255 @type no_voting: boolean
256 @param no_voting: whether to start ganeti-masterd without a node vote
257 (if start_daemons is True), but still non-interactively
261 # GetMasterInfo will raise an exception if not able to return data
262 master_netdev, master_ip, _, family = GetMasterInfo()
265 # either start the master and rapi daemons
268 masterd_args = "--no-voting --yes-do-it"
273 "EXTRA_MASTERD_ARGS": masterd_args,
276 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
278 msg = "Can't start Ganeti master: %s" % result.output
283 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
284 if netutils.IPAddress.Own(master_ip):
285 # we already have the ip:
286 logging.debug("Master IP already configured, doing nothing")
288 msg = "Someone else has the master ip, not activating"
292 ipcls = netutils.IP4Address
293 if family == netutils.IP6Address.family:
294 ipcls = netutils.IP6Address
296 result = utils.RunCmd(["ip", "address", "add",
297 "%s/%d" % (master_ip, ipcls.iplen),
298 "dev", master_netdev, "label",
299 "%s:0" % master_netdev])
301 msg = "Can't activate master IP: %s" % result.output
305 # we ignore the exit code of the following cmds
306 if ipcls == netutils.IP4Address:
307 utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev, "-s",
308 master_ip, master_ip])
309 elif ipcls == netutils.IP6Address:
311 utils.RunCmd(["ndisc6", "-q", "-r 3", master_ip, master_netdev])
312 except errors.OpExecError:
313 # TODO: Better error reporting
314 logging.warning("Can't execute ndisc6, please install if missing")
317 _Fail("; ".join(err_msgs))
320 def StopMaster(stop_daemons):
321 """Deactivate this node as master.
323 The function will always try to deactivate the IP address of the
324 master. It will also stop the master daemons depending on the
325 stop_daemons parameter.
327 @type stop_daemons: boolean
328 @param stop_daemons: whether to also stop the master daemons
329 (ganeti-masterd and ganeti-rapi)
333 # TODO: log and report back to the caller the error failures; we
334 # need to decide in which case we fail the RPC for this
336 # GetMasterInfo will raise an exception if not able to return data
337 master_netdev, master_ip, _, family = GetMasterInfo()
339 ipcls = netutils.IP4Address
340 if family == netutils.IP6Address.family:
341 ipcls = netutils.IP6Address
343 result = utils.RunCmd(["ip", "address", "del",
344 "%s/%d" % (master_ip, ipcls.iplen),
345 "dev", master_netdev])
347 logging.error("Can't remove the master IP, error: %s", result.output)
348 # but otherwise ignore the failure
351 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
353 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
355 result.cmd, result.exit_code, result.output)
358 def EtcHostsModify(mode, host, ip):
359 """Modify a host entry in /etc/hosts.
361 @param mode: The mode to operate. Either add or remove entry
362 @param host: The host to operate on
363 @param ip: The ip associated with the entry
366 if mode == constants.ETC_HOSTS_ADD:
368 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
370 utils.AddHostToEtcHosts(host, ip)
371 elif mode == constants.ETC_HOSTS_REMOVE:
373 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
374 " parameter is present")
375 utils.RemoveHostFromEtcHosts(host)
377 RPCFail("Mode not supported")
380 def LeaveCluster(modify_ssh_setup):
381 """Cleans up and remove the current node.
383 This function cleans up and prepares the current node to be removed
386 If processing is successful, then it raises an
387 L{errors.QuitGanetiException} which is used as a special case to
388 shutdown the node daemon.
390 @param modify_ssh_setup: boolean
393 _CleanDirectory(constants.DATA_DIR)
394 _CleanDirectory(constants.CRYPTO_KEYS_DIR)
399 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
401 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
403 utils.RemoveFile(priv_key)
404 utils.RemoveFile(pub_key)
405 except errors.OpExecError:
406 logging.exception("Error while processing ssh files")
409 utils.RemoveFile(constants.CONFD_HMAC_KEY)
410 utils.RemoveFile(constants.RAPI_CERT_FILE)
411 utils.RemoveFile(constants.NODED_CERT_FILE)
412 except: # pylint: disable-msg=W0702
413 logging.exception("Error while removing cluster secrets")
415 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
417 logging.error("Command %s failed with exitcode %s and error %s",
418 result.cmd, result.exit_code, result.output)
420 # Raise a custom exception (handled in ganeti-noded)
421 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
424 def GetNodeInfo(vgname, hypervisor_type):
425 """Gives back a hash with different information about the node.
427 @type vgname: C{string}
428 @param vgname: the name of the volume group to ask for disk space information
429 @type hypervisor_type: C{str}
430 @param hypervisor_type: the name of the hypervisor to ask for
433 @return: dictionary with the following keys:
434 - vg_size is the size of the configured volume group in MiB
435 - vg_free is the free size of the volume group in MiB
436 - memory_dom0 is the memory allocated for domain0 in MiB
437 - memory_free is the currently available (free) ram in MiB
438 - memory_total is the total number of ram in MiB
443 vginfo = bdev.LogicalVolume.GetVGInfo([vgname])
444 vg_free = vg_size = None
446 vg_free = int(round(vginfo[0][0], 0))
447 vg_size = int(round(vginfo[0][1], 0))
449 outputarray['vg_size'] = vg_size
450 outputarray['vg_free'] = vg_free
452 hyper = hypervisor.GetHypervisor(hypervisor_type)
453 hyp_info = hyper.GetNodeInfo()
454 if hyp_info is not None:
455 outputarray.update(hyp_info)
457 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
462 def VerifyNode(what, cluster_name):
463 """Verify the status of the local node.
465 Based on the input L{what} parameter, various checks are done on the
468 If the I{filelist} key is present, this list of
469 files is checksummed and the file/checksum pairs are returned.
471 If the I{nodelist} key is present, we check that we have
472 connectivity via ssh with the target nodes (and check the hostname
475 If the I{node-net-test} key is present, we check that we have
476 connectivity to the given nodes via both primary IP and, if
477 applicable, secondary IPs.
480 @param what: a dictionary of things to check:
481 - filelist: list of files for which to compute checksums
482 - nodelist: list of nodes we should check ssh communication with
483 - node-net-test: list of nodes we should check node daemon port
485 - hypervisor: list with hypervisors to run the verify for
487 @return: a dictionary with the same keys as the input dict, and
488 values representing the result of the checks
492 my_name = netutils.Hostname.GetSysName()
493 port = netutils.GetDaemonPort(constants.NODED)
494 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
496 if constants.NV_HYPERVISOR in what and vm_capable:
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 and vm_capable:
553 val = GetVolumeList(what[constants.NV_LVLIST])
556 result[constants.NV_LVLIST] = val
558 if constants.NV_INSTANCELIST in what and vm_capable:
559 # GetInstanceList can fail
561 val = GetInstanceList(what[constants.NV_INSTANCELIST])
564 result[constants.NV_INSTANCELIST] = val
566 if constants.NV_VGLIST in what and vm_capable:
567 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
569 if constants.NV_PVLIST in what and vm_capable:
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 and vm_capable:
579 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
580 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
582 if constants.NV_DRBDLIST in what and vm_capable:
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 and vm_capable:
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 and vm_capable:
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 _GetBlockDevSymlinkPath(instance_name, idx):
941 return utils.PathJoin(constants.DISK_LINKS_DIR,
942 "%s:%d" % (instance_name, idx))
945 def _SymlinkBlockDev(instance_name, device_path, idx):
946 """Set up symlinks to a instance's block device.
948 This is an auxiliary function run when an instance is start (on the primary
949 node) or when an instance is migrated (on the target node).
952 @param instance_name: the name of the target instance
953 @param device_path: path of the physical block device, on the node
954 @param idx: the disk index
955 @return: absolute path to the disk's symlink
958 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
960 os.symlink(device_path, link_name)
962 if err.errno == errno.EEXIST:
963 if (not os.path.islink(link_name) or
964 os.readlink(link_name) != device_path):
966 os.symlink(device_path, link_name)
973 def _RemoveBlockDevLinks(instance_name, disks):
974 """Remove the block device symlinks belonging to the given instance.
977 for idx, _ in enumerate(disks):
978 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
979 if os.path.islink(link_name):
983 logging.exception("Can't remove symlink '%s'", link_name)
986 def _GatherAndLinkBlockDevs(instance):
987 """Set up an instance's block device(s).
989 This is run on the primary node at instance startup. The block
990 devices must be already assembled.
992 @type instance: L{objects.Instance}
993 @param instance: the instance whose disks we shoul assemble
995 @return: list of (disk_object, device_path)
999 for idx, disk in enumerate(instance.disks):
1000 device = _RecursiveFindBD(disk)
1002 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1006 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1008 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1011 block_devices.append((disk, link_name))
1013 return block_devices
1016 def StartInstance(instance):
1017 """Start an instance.
1019 @type instance: L{objects.Instance}
1020 @param instance: the instance object
1024 running_instances = GetInstanceList([instance.hypervisor])
1026 if instance.name in running_instances:
1027 logging.info("Instance %s already running, not starting", instance.name)
1031 block_devices = _GatherAndLinkBlockDevs(instance)
1032 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1033 hyper.StartInstance(instance, block_devices)
1034 except errors.BlockDeviceError, err:
1035 _Fail("Block device error: %s", err, exc=True)
1036 except errors.HypervisorError, err:
1037 _RemoveBlockDevLinks(instance.name, instance.disks)
1038 _Fail("Hypervisor error: %s", err, exc=True)
1041 def InstanceShutdown(instance, timeout):
1042 """Shut an instance down.
1044 @note: this functions uses polling with a hardcoded timeout.
1046 @type instance: L{objects.Instance}
1047 @param instance: the instance object
1048 @type timeout: integer
1049 @param timeout: maximum timeout for soft shutdown
1053 hv_name = instance.hypervisor
1054 hyper = hypervisor.GetHypervisor(hv_name)
1055 iname = instance.name
1057 if instance.name not in hyper.ListInstances():
1058 logging.info("Instance %s not running, doing nothing", iname)
1063 self.tried_once = False
1066 if iname not in hyper.ListInstances():
1070 hyper.StopInstance(instance, retry=self.tried_once)
1071 except errors.HypervisorError, err:
1072 if iname not in hyper.ListInstances():
1073 # if the instance is no longer existing, consider this a
1074 # success and go to cleanup
1077 _Fail("Failed to stop instance %s: %s", iname, err)
1079 self.tried_once = True
1081 raise utils.RetryAgain()
1084 utils.Retry(_TryShutdown(), 5, timeout)
1085 except utils.RetryTimeout:
1086 # the shutdown did not succeed
1087 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1090 hyper.StopInstance(instance, force=True)
1091 except errors.HypervisorError, err:
1092 if iname in hyper.ListInstances():
1093 # only raise an error if the instance still exists, otherwise
1094 # the error could simply be "instance ... unknown"!
1095 _Fail("Failed to force stop instance %s: %s", iname, err)
1099 if iname in hyper.ListInstances():
1100 _Fail("Could not shutdown instance %s even by destroy", iname)
1103 hyper.CleanupInstance(instance.name)
1104 except errors.HypervisorError, err:
1105 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1107 _RemoveBlockDevLinks(iname, instance.disks)
1110 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1111 """Reboot an instance.
1113 @type instance: L{objects.Instance}
1114 @param instance: the instance object to reboot
1115 @type reboot_type: str
1116 @param reboot_type: the type of reboot, one the following
1118 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1119 instance OS, do not recreate the VM
1120 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1121 restart the VM (at the hypervisor level)
1122 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1123 not accepted here, since that mode is handled differently, in
1124 cmdlib, and translates into full stop and start of the
1125 instance (instead of a call_instance_reboot RPC)
1126 @type shutdown_timeout: integer
1127 @param shutdown_timeout: maximum timeout for soft shutdown
1131 running_instances = GetInstanceList([instance.hypervisor])
1133 if instance.name not in running_instances:
1134 _Fail("Cannot reboot instance %s that is not running", instance.name)
1136 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1137 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1139 hyper.RebootInstance(instance)
1140 except errors.HypervisorError, err:
1141 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1142 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1144 InstanceShutdown(instance, shutdown_timeout)
1145 return StartInstance(instance)
1146 except errors.HypervisorError, err:
1147 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1149 _Fail("Invalid reboot_type received: %s", reboot_type)
1152 def MigrationInfo(instance):
1153 """Gather information about an instance to be migrated.
1155 @type instance: L{objects.Instance}
1156 @param instance: the instance definition
1159 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1161 info = hyper.MigrationInfo(instance)
1162 except errors.HypervisorError, err:
1163 _Fail("Failed to fetch migration information: %s", err, exc=True)
1167 def AcceptInstance(instance, info, target):
1168 """Prepare the node to accept an instance.
1170 @type instance: L{objects.Instance}
1171 @param instance: the instance definition
1172 @type info: string/data (opaque)
1173 @param info: migration information, from the source node
1174 @type target: string
1175 @param target: target host (usually ip), on this node
1178 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1180 hyper.AcceptInstance(instance, info, target)
1181 except errors.HypervisorError, err:
1182 _Fail("Failed to accept instance: %s", err, exc=True)
1185 def FinalizeMigration(instance, info, success):
1186 """Finalize any preparation to accept an instance.
1188 @type instance: L{objects.Instance}
1189 @param instance: the instance definition
1190 @type info: string/data (opaque)
1191 @param info: migration information, from the source node
1192 @type success: boolean
1193 @param success: whether the migration was a success or a failure
1196 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1198 hyper.FinalizeMigration(instance, info, success)
1199 except errors.HypervisorError, err:
1200 _Fail("Failed to finalize migration: %s", err, exc=True)
1203 def MigrateInstance(instance, target, live):
1204 """Migrates an instance to another node.
1206 @type instance: L{objects.Instance}
1207 @param instance: the instance definition
1208 @type target: string
1209 @param target: the target node name
1211 @param live: whether the migration should be done live or not (the
1212 interpretation of this parameter is left to the hypervisor)
1214 @return: a tuple of (success, msg) where:
1215 - succes is a boolean denoting the success/failure of the operation
1216 - msg is a string with details in case of failure
1219 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1222 hyper.MigrateInstance(instance, target, live)
1223 except errors.HypervisorError, err:
1224 _Fail("Failed to migrate instance: %s", err, exc=True)
1227 def BlockdevCreate(disk, size, owner, on_primary, info):
1228 """Creates a block device for an instance.
1230 @type disk: L{objects.Disk}
1231 @param disk: the object describing the disk we should create
1233 @param size: the size of the physical underlying device, in MiB
1235 @param owner: the name of the instance for which disk is created,
1236 used for device cache data
1237 @type on_primary: boolean
1238 @param on_primary: indicates if it is the primary node or not
1240 @param info: string that will be sent to the physical device
1241 creation, used for example to set (LVM) tags on LVs
1243 @return: the new unique_id of the device (this can sometime be
1244 computed only after creation), or None. On secondary nodes,
1245 it's not required to return anything.
1248 # TODO: remove the obsolete 'size' argument
1249 # pylint: disable-msg=W0613
1252 for child in disk.children:
1254 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1255 except errors.BlockDeviceError, err:
1256 _Fail("Can't assemble device %s: %s", child, err)
1257 if on_primary or disk.AssembleOnSecondary():
1258 # we need the children open in case the device itself has to
1261 # pylint: disable-msg=E1103
1263 except errors.BlockDeviceError, err:
1264 _Fail("Can't make child '%s' read-write: %s", child, err)
1268 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1269 except errors.BlockDeviceError, err:
1270 _Fail("Can't create block device: %s", err)
1272 if on_primary or disk.AssembleOnSecondary():
1275 except errors.BlockDeviceError, err:
1276 _Fail("Can't assemble device after creation, unusual event: %s", err)
1277 device.SetSyncSpeed(constants.SYNC_SPEED)
1278 if on_primary or disk.OpenOnSecondary():
1280 device.Open(force=True)
1281 except errors.BlockDeviceError, err:
1282 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1283 DevCacheManager.UpdateCache(device.dev_path, owner,
1284 on_primary, disk.iv_name)
1286 device.SetInfo(info)
1288 return device.unique_id
1291 def _WipeDevice(path, offset, size):
1292 """This function actually wipes the device.
1294 @param path: The path to the device to wipe
1295 @param offset: The offset in MiB in the file
1296 @param size: The size in MiB to write
1299 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1300 "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1302 result = utils.RunCmd(cmd)
1305 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1306 result.fail_reason, result.output)
1309 def BlockdevWipe(disk, offset, size):
1310 """Wipes a block device.
1312 @type disk: L{objects.Disk}
1313 @param disk: the disk object we want to wipe
1315 @param offset: The offset in MiB in the file
1317 @param size: The size in MiB to write
1321 rdev = _RecursiveFindBD(disk)
1322 except errors.BlockDeviceError:
1326 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1328 # Do cross verify some of the parameters
1329 if offset > rdev.size:
1330 _Fail("Offset is bigger than device size")
1331 if (offset + size) > rdev.size:
1332 _Fail("The provided offset and size to wipe is bigger than device size")
1334 _WipeDevice(rdev.dev_path, offset, size)
1337 def BlockdevRemove(disk):
1338 """Remove a block device.
1340 @note: This is intended to be called recursively.
1342 @type disk: L{objects.Disk}
1343 @param disk: the disk object we should remove
1345 @return: the success of the operation
1350 rdev = _RecursiveFindBD(disk)
1351 except errors.BlockDeviceError, err:
1352 # probably can't attach
1353 logging.info("Can't attach to device %s in remove", disk)
1355 if rdev is not None:
1356 r_path = rdev.dev_path
1359 except errors.BlockDeviceError, err:
1360 msgs.append(str(err))
1362 DevCacheManager.RemoveCache(r_path)
1365 for child in disk.children:
1367 BlockdevRemove(child)
1368 except RPCFail, err:
1369 msgs.append(str(err))
1372 _Fail("; ".join(msgs))
1375 def _RecursiveAssembleBD(disk, owner, as_primary):
1376 """Activate a block device for an instance.
1378 This is run on the primary and secondary nodes for an instance.
1380 @note: this function is called recursively.
1382 @type disk: L{objects.Disk}
1383 @param disk: the disk we try to assemble
1385 @param owner: the name of the instance which owns the disk
1386 @type as_primary: boolean
1387 @param as_primary: if we should make the block device
1390 @return: the assembled device or None (in case no device
1392 @raise errors.BlockDeviceError: in case there is an error
1393 during the activation of the children or the device
1399 mcn = disk.ChildrenNeeded()
1401 mcn = 0 # max number of Nones allowed
1403 mcn = len(disk.children) - mcn # max number of Nones
1404 for chld_disk in disk.children:
1406 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1407 except errors.BlockDeviceError, err:
1408 if children.count(None) >= mcn:
1411 logging.error("Error in child activation (but continuing): %s",
1413 children.append(cdev)
1415 if as_primary or disk.AssembleOnSecondary():
1416 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1417 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1419 if as_primary or disk.OpenOnSecondary():
1421 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1422 as_primary, disk.iv_name)
1429 def BlockdevAssemble(disk, owner, as_primary):
1430 """Activate a block device for an instance.
1432 This is a wrapper over _RecursiveAssembleBD.
1434 @rtype: str or boolean
1435 @return: a C{/dev/...} path for primary nodes, and
1436 C{True} for secondary nodes
1440 result = _RecursiveAssembleBD(disk, owner, as_primary)
1441 if isinstance(result, bdev.BlockDev):
1442 # pylint: disable-msg=E1103
1443 result = result.dev_path
1444 except errors.BlockDeviceError, err:
1445 _Fail("Error while assembling disk: %s", err, exc=True)
1450 def BlockdevShutdown(disk):
1451 """Shut down a block device.
1453 First, if the device is assembled (Attach() is successful), then
1454 the device is shutdown. Then the children of the device are
1457 This function is called recursively. Note that we don't cache the
1458 children or such, as oppossed to assemble, shutdown of different
1459 devices doesn't require that the upper device was active.
1461 @type disk: L{objects.Disk}
1462 @param disk: the description of the disk we should
1468 r_dev = _RecursiveFindBD(disk)
1469 if r_dev is not None:
1470 r_path = r_dev.dev_path
1473 DevCacheManager.RemoveCache(r_path)
1474 except errors.BlockDeviceError, err:
1475 msgs.append(str(err))
1478 for child in disk.children:
1480 BlockdevShutdown(child)
1481 except RPCFail, err:
1482 msgs.append(str(err))
1485 _Fail("; ".join(msgs))
1488 def BlockdevAddchildren(parent_cdev, new_cdevs):
1489 """Extend a mirrored block device.
1491 @type parent_cdev: L{objects.Disk}
1492 @param parent_cdev: the disk to which we should add children
1493 @type new_cdevs: list of L{objects.Disk}
1494 @param new_cdevs: the list of children which we should add
1498 parent_bdev = _RecursiveFindBD(parent_cdev)
1499 if parent_bdev is None:
1500 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1501 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1502 if new_bdevs.count(None) > 0:
1503 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1504 parent_bdev.AddChildren(new_bdevs)
1507 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1508 """Shrink a mirrored block device.
1510 @type parent_cdev: L{objects.Disk}
1511 @param parent_cdev: the disk from which we should remove children
1512 @type new_cdevs: list of L{objects.Disk}
1513 @param new_cdevs: the list of children which we should remove
1517 parent_bdev = _RecursiveFindBD(parent_cdev)
1518 if parent_bdev is None:
1519 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1521 for disk in new_cdevs:
1522 rpath = disk.StaticDevPath()
1524 bd = _RecursiveFindBD(disk)
1526 _Fail("Can't find device %s while removing children", disk)
1528 devs.append(bd.dev_path)
1530 if not utils.IsNormAbsPath(rpath):
1531 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1533 parent_bdev.RemoveChildren(devs)
1536 def BlockdevGetmirrorstatus(disks):
1537 """Get the mirroring status of a list of devices.
1539 @type disks: list of L{objects.Disk}
1540 @param disks: the list of disks which we should query
1542 @return: List of L{objects.BlockDevStatus}, one for each disk
1543 @raise errors.BlockDeviceError: if any of the disks cannot be
1549 rbd = _RecursiveFindBD(dsk)
1551 _Fail("Can't find device %s", dsk)
1553 stats.append(rbd.CombinedSyncStatus())
1558 def BlockdevGetmirrorstatusMulti(disks):
1559 """Get the mirroring status of a list of devices.
1561 @type disks: list of L{objects.Disk}
1562 @param disks: the list of disks which we should query
1564 @return: List of tuples, (bool, status), one for each disk; bool denotes
1565 success/failure, status is L{objects.BlockDevStatus} on success, string
1572 rbd = _RecursiveFindBD(disk)
1574 result.append((False, "Can't find device %s" % disk))
1577 status = rbd.CombinedSyncStatus()
1578 except errors.BlockDeviceError, err:
1579 logging.exception("Error while getting disk status")
1580 result.append((False, str(err)))
1582 result.append((True, status))
1584 assert len(disks) == len(result)
1589 def _RecursiveFindBD(disk):
1590 """Check if a device is activated.
1592 If so, return information about the real device.
1594 @type disk: L{objects.Disk}
1595 @param disk: the disk object we need to find
1597 @return: None if the device can't be found,
1598 otherwise the device instance
1603 for chdisk in disk.children:
1604 children.append(_RecursiveFindBD(chdisk))
1606 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1609 def _OpenRealBD(disk):
1610 """Opens the underlying block device of a disk.
1612 @type disk: L{objects.Disk}
1613 @param disk: the disk object we want to open
1616 real_disk = _RecursiveFindBD(disk)
1617 if real_disk is None:
1618 _Fail("Block device '%s' is not set up", disk)
1625 def BlockdevFind(disk):
1626 """Check if a device is activated.
1628 If it is, return information about the real device.
1630 @type disk: L{objects.Disk}
1631 @param disk: the disk to find
1632 @rtype: None or objects.BlockDevStatus
1633 @return: None if the disk cannot be found, otherwise a the current
1638 rbd = _RecursiveFindBD(disk)
1639 except errors.BlockDeviceError, err:
1640 _Fail("Failed to find device: %s", err, exc=True)
1645 return rbd.GetSyncStatus()
1648 def BlockdevGetsize(disks):
1649 """Computes the size of the given disks.
1651 If a disk is not found, returns None instead.
1653 @type disks: list of L{objects.Disk}
1654 @param disks: the list of disk to compute the size for
1656 @return: list with elements None if the disk cannot be found,
1663 rbd = _RecursiveFindBD(cf)
1664 except errors.BlockDeviceError:
1670 result.append(rbd.GetActualSize())
1674 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1675 """Export a block device to a remote node.
1677 @type disk: L{objects.Disk}
1678 @param disk: the description of the disk to export
1679 @type dest_node: str
1680 @param dest_node: the destination node to export to
1681 @type dest_path: str
1682 @param dest_path: the destination path on the target node
1683 @type cluster_name: str
1684 @param cluster_name: the cluster name, needed for SSH hostalias
1688 real_disk = _OpenRealBD(disk)
1690 # the block size on the read dd is 1MiB to match our units
1691 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1692 "dd if=%s bs=1048576 count=%s",
1693 real_disk.dev_path, str(disk.size))
1695 # we set here a smaller block size as, due to ssh buffering, more
1696 # than 64-128k will mostly ignored; we use nocreat to fail if the
1697 # device is not already there or we pass a wrong path; we use
1698 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1699 # to not buffer too much memory; this means that at best, we flush
1700 # every 64k, which will not be very fast
1701 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1702 " oflag=dsync", dest_path)
1704 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1705 constants.GANETI_RUNAS,
1708 # all commands have been checked, so we're safe to combine them
1709 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1711 result = utils.RunCmd(["bash", "-c", command])
1714 _Fail("Disk copy command '%s' returned error: %s"
1715 " output: %s", command, result.fail_reason, result.output)
1718 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1719 """Write a file to the filesystem.
1721 This allows the master to overwrite(!) a file. It will only perform
1722 the operation if the file belongs to a list of configuration files.
1724 @type file_name: str
1725 @param file_name: the target file name
1727 @param data: the new contents of the file
1729 @param mode: the mode to give the file (can be None)
1731 @param uid: the owner of the file (can be -1 for default)
1733 @param gid: the group of the file (can be -1 for default)
1735 @param atime: the atime to set on the file (can be None)
1737 @param mtime: the mtime to set on the file (can be None)
1741 if not os.path.isabs(file_name):
1742 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1744 if file_name not in _ALLOWED_UPLOAD_FILES:
1745 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1748 raw_data = _Decompress(data)
1750 utils.SafeWriteFile(file_name, None,
1751 data=raw_data, mode=mode, uid=uid, gid=gid,
1752 atime=atime, mtime=mtime)
1755 def WriteSsconfFiles(values):
1756 """Update all ssconf files.
1758 Wrapper around the SimpleStore.WriteFiles.
1761 ssconf.SimpleStore().WriteFiles(values)
1764 def _ErrnoOrStr(err):
1765 """Format an EnvironmentError exception.
1767 If the L{err} argument has an errno attribute, it will be looked up
1768 and converted into a textual C{E...} description. Otherwise the
1769 string representation of the error will be returned.
1771 @type err: L{EnvironmentError}
1772 @param err: the exception to format
1775 if hasattr(err, 'errno'):
1776 detail = errno.errorcode[err.errno]
1782 def _OSOndiskAPIVersion(os_dir):
1783 """Compute and return the API version of a given OS.
1785 This function will try to read the API version of the OS residing in
1786 the 'os_dir' directory.
1789 @param os_dir: the directory in which we should look for the OS
1791 @return: tuple (status, data) with status denoting the validity and
1792 data holding either the vaid versions or an error message
1795 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1798 st = os.stat(api_file)
1799 except EnvironmentError, err:
1800 return False, ("Required file '%s' not found under path %s: %s" %
1801 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1803 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1804 return False, ("File '%s' in %s is not a regular file" %
1805 (constants.OS_API_FILE, os_dir))
1808 api_versions = utils.ReadFile(api_file).splitlines()
1809 except EnvironmentError, err:
1810 return False, ("Error while reading the API version file at %s: %s" %
1811 (api_file, _ErrnoOrStr(err)))
1814 api_versions = [int(version.strip()) for version in api_versions]
1815 except (TypeError, ValueError), err:
1816 return False, ("API version(s) can't be converted to integer: %s" %
1819 return True, api_versions
1822 def DiagnoseOS(top_dirs=None):
1823 """Compute the validity for all OSes.
1825 @type top_dirs: list
1826 @param top_dirs: the list of directories in which to
1827 search (if not given defaults to
1828 L{constants.OS_SEARCH_PATH})
1829 @rtype: list of L{objects.OS}
1830 @return: a list of tuples (name, path, status, diagnose, variants,
1831 parameters, api_version) for all (potential) OSes under all
1832 search paths, where:
1833 - name is the (potential) OS name
1834 - path is the full path to the OS
1835 - status True/False is the validity of the OS
1836 - diagnose is the error message for an invalid OS, otherwise empty
1837 - variants is a list of supported OS variants, if any
1838 - parameters is a list of (name, help) parameters, if any
1839 - api_version is a list of support OS API versions
1842 if top_dirs is None:
1843 top_dirs = constants.OS_SEARCH_PATH
1846 for dir_name in top_dirs:
1847 if os.path.isdir(dir_name):
1849 f_names = utils.ListVisibleFiles(dir_name)
1850 except EnvironmentError, err:
1851 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1853 for name in f_names:
1854 os_path = utils.PathJoin(dir_name, name)
1855 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1858 variants = os_inst.supported_variants
1859 parameters = os_inst.supported_parameters
1860 api_versions = os_inst.api_versions
1863 variants = parameters = api_versions = []
1864 result.append((name, os_path, status, diagnose, variants,
1865 parameters, api_versions))
1870 def _TryOSFromDisk(name, base_dir=None):
1871 """Create an OS instance from disk.
1873 This function will return an OS instance if the given name is a
1876 @type base_dir: string
1877 @keyword base_dir: Base directory containing OS installations.
1878 Defaults to a search in all the OS_SEARCH_PATH dirs.
1880 @return: success and either the OS instance if we find a valid one,
1884 if base_dir is None:
1885 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1887 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1890 return False, "Directory for OS %s not found in search path" % name
1892 status, api_versions = _OSOndiskAPIVersion(os_dir)
1895 return status, api_versions
1897 if not constants.OS_API_VERSIONS.intersection(api_versions):
1898 return False, ("API version mismatch for path '%s': found %s, want %s." %
1899 (os_dir, api_versions, constants.OS_API_VERSIONS))
1901 # OS Files dictionary, we will populate it with the absolute path names
1902 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1904 if max(api_versions) >= constants.OS_API_V15:
1905 os_files[constants.OS_VARIANTS_FILE] = ''
1907 if max(api_versions) >= constants.OS_API_V20:
1908 os_files[constants.OS_PARAMETERS_FILE] = ''
1910 del os_files[constants.OS_SCRIPT_VERIFY]
1912 for filename in os_files:
1913 os_files[filename] = utils.PathJoin(os_dir, filename)
1916 st = os.stat(os_files[filename])
1917 except EnvironmentError, err:
1918 return False, ("File '%s' under path '%s' is missing (%s)" %
1919 (filename, os_dir, _ErrnoOrStr(err)))
1921 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1922 return False, ("File '%s' under path '%s' is not a regular file" %
1925 if filename in constants.OS_SCRIPTS:
1926 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1927 return False, ("File '%s' under path '%s' is not executable" %
1931 if constants.OS_VARIANTS_FILE in os_files:
1932 variants_file = os_files[constants.OS_VARIANTS_FILE]
1934 variants = utils.ReadFile(variants_file).splitlines()
1935 except EnvironmentError, err:
1936 return False, ("Error while reading the OS variants file at %s: %s" %
1937 (variants_file, _ErrnoOrStr(err)))
1939 return False, ("No supported os variant found")
1942 if constants.OS_PARAMETERS_FILE in os_files:
1943 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
1945 parameters = utils.ReadFile(parameters_file).splitlines()
1946 except EnvironmentError, err:
1947 return False, ("Error while reading the OS parameters file at %s: %s" %
1948 (parameters_file, _ErrnoOrStr(err)))
1949 parameters = [v.split(None, 1) for v in parameters]
1951 os_obj = objects.OS(name=name, path=os_dir,
1952 create_script=os_files[constants.OS_SCRIPT_CREATE],
1953 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1954 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1955 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1956 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
1958 supported_variants=variants,
1959 supported_parameters=parameters,
1960 api_versions=api_versions)
1964 def OSFromDisk(name, base_dir=None):
1965 """Create an OS instance from disk.
1967 This function will return an OS instance if the given name is a
1968 valid OS name. Otherwise, it will raise an appropriate
1969 L{RPCFail} exception, detailing why this is not a valid OS.
1971 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1972 an exception but returns true/false status data.
1974 @type base_dir: string
1975 @keyword base_dir: Base directory containing OS installations.
1976 Defaults to a search in all the OS_SEARCH_PATH dirs.
1977 @rtype: L{objects.OS}
1978 @return: the OS instance if we find a valid one
1979 @raise RPCFail: if we don't find a valid OS
1982 name_only = objects.OS.GetName(name)
1983 status, payload = _TryOSFromDisk(name_only, base_dir)
1991 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
1992 """Calculate the basic environment for an os script.
1995 @param os_name: full operating system name (including variant)
1996 @type inst_os: L{objects.OS}
1997 @param inst_os: operating system for which the environment is being built
1998 @type os_params: dict
1999 @param os_params: the OS parameters
2000 @type debug: integer
2001 @param debug: debug level (0 or 1, for OS Api 10)
2003 @return: dict of environment variables
2004 @raise errors.BlockDeviceError: if the block device
2010 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2011 result['OS_API_VERSION'] = '%d' % api_version
2012 result['OS_NAME'] = inst_os.name
2013 result['DEBUG_LEVEL'] = '%d' % debug
2016 if api_version >= constants.OS_API_V15:
2017 variant = objects.OS.GetVariant(os_name)
2019 variant = inst_os.supported_variants[0]
2020 result['OS_VARIANT'] = variant
2023 for pname, pvalue in os_params.items():
2024 result['OSP_%s' % pname.upper()] = pvalue
2029 def OSEnvironment(instance, inst_os, debug=0):
2030 """Calculate the environment for an os script.
2032 @type instance: L{objects.Instance}
2033 @param instance: target instance for the os script run
2034 @type inst_os: L{objects.OS}
2035 @param inst_os: operating system for which the environment is being built
2036 @type debug: integer
2037 @param debug: debug level (0 or 1, for OS Api 10)
2039 @return: dict of environment variables
2040 @raise errors.BlockDeviceError: if the block device
2044 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2046 for attr in ["name", "os", "uuid", "ctime", "mtime"]:
2047 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2049 result['HYPERVISOR'] = instance.hypervisor
2050 result['DISK_COUNT'] = '%d' % len(instance.disks)
2051 result['NIC_COUNT'] = '%d' % len(instance.nics)
2054 for idx, disk in enumerate(instance.disks):
2055 real_disk = _OpenRealBD(disk)
2056 result['DISK_%d_PATH' % idx] = real_disk.dev_path
2057 result['DISK_%d_ACCESS' % idx] = disk.mode
2058 if constants.HV_DISK_TYPE in instance.hvparams:
2059 result['DISK_%d_FRONTEND_TYPE' % idx] = \
2060 instance.hvparams[constants.HV_DISK_TYPE]
2061 if disk.dev_type in constants.LDS_BLOCK:
2062 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
2063 elif disk.dev_type == constants.LD_FILE:
2064 result['DISK_%d_BACKEND_TYPE' % idx] = \
2065 'file:%s' % disk.physical_id[0]
2068 for idx, nic in enumerate(instance.nics):
2069 result['NIC_%d_MAC' % idx] = nic.mac
2071 result['NIC_%d_IP' % idx] = nic.ip
2072 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
2073 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2074 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
2075 if nic.nicparams[constants.NIC_LINK]:
2076 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
2077 if constants.HV_NIC_TYPE in instance.hvparams:
2078 result['NIC_%d_FRONTEND_TYPE' % idx] = \
2079 instance.hvparams[constants.HV_NIC_TYPE]
2082 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2083 for key, value in source.items():
2084 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2089 def BlockdevGrow(disk, amount):
2090 """Grow a stack of block devices.
2092 This function is called recursively, with the childrens being the
2093 first ones to resize.
2095 @type disk: L{objects.Disk}
2096 @param disk: the disk to be grown
2097 @rtype: (status, result)
2098 @return: a tuple with the status of the operation
2099 (True/False), and the errors message if status
2103 r_dev = _RecursiveFindBD(disk)
2105 _Fail("Cannot find block device %s", disk)
2109 except errors.BlockDeviceError, err:
2110 _Fail("Failed to grow block device: %s", err, exc=True)
2113 def BlockdevSnapshot(disk):
2114 """Create a snapshot copy of a block device.
2116 This function is called recursively, and the snapshot is actually created
2117 just for the leaf lvm backend device.
2119 @type disk: L{objects.Disk}
2120 @param disk: the disk to be snapshotted
2122 @return: snapshot disk path
2125 if disk.dev_type == constants.LD_DRBD8:
2126 if not disk.children:
2127 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2129 return BlockdevSnapshot(disk.children[0])
2130 elif disk.dev_type == constants.LD_LV:
2131 r_dev = _RecursiveFindBD(disk)
2132 if r_dev is not None:
2133 # FIXME: choose a saner value for the snapshot size
2134 # let's stay on the safe side and ask for the full size, for now
2135 return r_dev.Snapshot(disk.size)
2137 _Fail("Cannot find block device %s", disk)
2139 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2140 disk.unique_id, disk.dev_type)
2143 def FinalizeExport(instance, snap_disks):
2144 """Write out the export configuration information.
2146 @type instance: L{objects.Instance}
2147 @param instance: the instance which we export, used for
2148 saving configuration
2149 @type snap_disks: list of L{objects.Disk}
2150 @param snap_disks: list of snapshot block devices, which
2151 will be used to get the actual name of the dump file
2156 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2157 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2159 config = objects.SerializableConfigParser()
2161 config.add_section(constants.INISECT_EXP)
2162 config.set(constants.INISECT_EXP, 'version', '0')
2163 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2164 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2165 config.set(constants.INISECT_EXP, 'os', instance.os)
2166 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2168 config.add_section(constants.INISECT_INS)
2169 config.set(constants.INISECT_INS, 'name', instance.name)
2170 config.set(constants.INISECT_INS, 'memory', '%d' %
2171 instance.beparams[constants.BE_MEMORY])
2172 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2173 instance.beparams[constants.BE_VCPUS])
2174 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2175 config.set(constants.INISECT_INS, 'hypervisor', instance.hypervisor)
2178 for nic_count, nic in enumerate(instance.nics):
2180 config.set(constants.INISECT_INS, 'nic%d_mac' %
2181 nic_count, '%s' % nic.mac)
2182 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2183 for param in constants.NICS_PARAMETER_TYPES:
2184 config.set(constants.INISECT_INS, 'nic%d_%s' % (nic_count, param),
2185 '%s' % nic.nicparams.get(param, None))
2186 # TODO: redundant: on load can read nics until it doesn't exist
2187 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2190 for disk_count, disk in enumerate(snap_disks):
2193 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2194 ('%s' % disk.iv_name))
2195 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2196 ('%s' % disk.physical_id[1]))
2197 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2200 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2202 # New-style hypervisor/backend parameters
2204 config.add_section(constants.INISECT_HYP)
2205 for name, value in instance.hvparams.items():
2206 if name not in constants.HVC_GLOBALS:
2207 config.set(constants.INISECT_HYP, name, str(value))
2209 config.add_section(constants.INISECT_BEP)
2210 for name, value in instance.beparams.items():
2211 config.set(constants.INISECT_BEP, name, str(value))
2213 config.add_section(constants.INISECT_OSP)
2214 for name, value in instance.osparams.items():
2215 config.set(constants.INISECT_OSP, name, str(value))
2217 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2218 data=config.Dumps())
2219 shutil.rmtree(finaldestdir, ignore_errors=True)
2220 shutil.move(destdir, finaldestdir)
2223 def ExportInfo(dest):
2224 """Get export configuration information.
2227 @param dest: directory containing the export
2229 @rtype: L{objects.SerializableConfigParser}
2230 @return: a serializable config file containing the
2234 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2236 config = objects.SerializableConfigParser()
2239 if (not config.has_section(constants.INISECT_EXP) or
2240 not config.has_section(constants.INISECT_INS)):
2241 _Fail("Export info file doesn't have the required fields")
2243 return config.Dumps()
2247 """Return a list of exports currently available on this machine.
2250 @return: list of the exports
2253 if os.path.isdir(constants.EXPORT_DIR):
2254 return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2256 _Fail("No exports directory")
2259 def RemoveExport(export):
2260 """Remove an existing export from the node.
2263 @param export: the name of the export to remove
2267 target = utils.PathJoin(constants.EXPORT_DIR, export)
2270 shutil.rmtree(target)
2271 except EnvironmentError, err:
2272 _Fail("Error while removing the export: %s", err, exc=True)
2275 def BlockdevRename(devlist):
2276 """Rename a list of block devices.
2278 @type devlist: list of tuples
2279 @param devlist: list of tuples of the form (disk,
2280 new_logical_id, new_physical_id); disk is an
2281 L{objects.Disk} object describing the current disk,
2282 and new logical_id/physical_id is the name we
2285 @return: True if all renames succeeded, False otherwise
2290 for disk, unique_id in devlist:
2291 dev = _RecursiveFindBD(disk)
2293 msgs.append("Can't find device %s in rename" % str(disk))
2297 old_rpath = dev.dev_path
2298 dev.Rename(unique_id)
2299 new_rpath = dev.dev_path
2300 if old_rpath != new_rpath:
2301 DevCacheManager.RemoveCache(old_rpath)
2302 # FIXME: we should add the new cache information here, like:
2303 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2304 # but we don't have the owner here - maybe parse from existing
2305 # cache? for now, we only lose lvm data when we rename, which
2306 # is less critical than DRBD or MD
2307 except errors.BlockDeviceError, err:
2308 msgs.append("Can't rename device '%s' to '%s': %s" %
2309 (dev, unique_id, err))
2310 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2313 _Fail("; ".join(msgs))
2316 def _TransformFileStorageDir(file_storage_dir):
2317 """Checks whether given file_storage_dir is valid.
2319 Checks wheter the given file_storage_dir is within the cluster-wide
2320 default file_storage_dir stored in SimpleStore. Only paths under that
2321 directory are allowed.
2323 @type file_storage_dir: str
2324 @param file_storage_dir: the path to check
2326 @return: the normalized path if valid, None otherwise
2329 if not constants.ENABLE_FILE_STORAGE:
2330 _Fail("File storage disabled at configure time")
2332 file_storage_dir = os.path.normpath(file_storage_dir)
2333 base_file_storage_dir = cfg.GetFileStorageDir()
2334 if (os.path.commonprefix([file_storage_dir, base_file_storage_dir]) !=
2335 base_file_storage_dir):
2336 _Fail("File storage directory '%s' is not under base file"
2337 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2338 return file_storage_dir
2341 def CreateFileStorageDir(file_storage_dir):
2342 """Create file storage directory.
2344 @type file_storage_dir: str
2345 @param file_storage_dir: directory to create
2348 @return: tuple with first element a boolean indicating wheter dir
2349 creation was successful or not
2352 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2353 if os.path.exists(file_storage_dir):
2354 if not os.path.isdir(file_storage_dir):
2355 _Fail("Specified storage dir '%s' is not a directory",
2359 os.makedirs(file_storage_dir, 0750)
2360 except OSError, err:
2361 _Fail("Cannot create file storage directory '%s': %s",
2362 file_storage_dir, err, exc=True)
2365 def RemoveFileStorageDir(file_storage_dir):
2366 """Remove file storage directory.
2368 Remove it only if it's empty. If not log an error and return.
2370 @type file_storage_dir: str
2371 @param file_storage_dir: the directory we should cleanup
2372 @rtype: tuple (success,)
2373 @return: tuple of one element, C{success}, denoting
2374 whether the operation was successful
2377 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2378 if os.path.exists(file_storage_dir):
2379 if not os.path.isdir(file_storage_dir):
2380 _Fail("Specified Storage directory '%s' is not a directory",
2382 # deletes dir only if empty, otherwise we want to fail the rpc call
2384 os.rmdir(file_storage_dir)
2385 except OSError, err:
2386 _Fail("Cannot remove file storage directory '%s': %s",
2387 file_storage_dir, err)
2390 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2391 """Rename the file storage directory.
2393 @type old_file_storage_dir: str
2394 @param old_file_storage_dir: the current path
2395 @type new_file_storage_dir: str
2396 @param new_file_storage_dir: the name we should rename to
2397 @rtype: tuple (success,)
2398 @return: tuple of one element, C{success}, denoting
2399 whether the operation was successful
2402 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2403 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2404 if not os.path.exists(new_file_storage_dir):
2405 if os.path.isdir(old_file_storage_dir):
2407 os.rename(old_file_storage_dir, new_file_storage_dir)
2408 except OSError, err:
2409 _Fail("Cannot rename '%s' to '%s': %s",
2410 old_file_storage_dir, new_file_storage_dir, err)
2412 _Fail("Specified storage dir '%s' is not a directory",
2413 old_file_storage_dir)
2415 if os.path.exists(old_file_storage_dir):
2416 _Fail("Cannot rename '%s' to '%s': both locations exist",
2417 old_file_storage_dir, new_file_storage_dir)
2420 def _EnsureJobQueueFile(file_name):
2421 """Checks whether the given filename is in the queue directory.
2423 @type file_name: str
2424 @param file_name: the file name we should check
2426 @raises RPCFail: if the file is not valid
2429 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2430 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2433 _Fail("Passed job queue file '%s' does not belong to"
2434 " the queue directory '%s'", file_name, queue_dir)
2437 def JobQueueUpdate(file_name, content):
2438 """Updates a file in the queue directory.
2440 This is just a wrapper over L{utils.WriteFile}, with proper
2443 @type file_name: str
2444 @param file_name: the job file name
2446 @param content: the new job contents
2448 @return: the success of the operation
2451 _EnsureJobQueueFile(file_name)
2452 getents = runtime.GetEnts()
2454 # Write and replace the file atomically
2455 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2456 gid=getents.masterd_gid)
2459 def JobQueueRename(old, new):
2460 """Renames a job queue file.
2462 This is just a wrapper over os.rename with proper checking.
2465 @param old: the old (actual) file name
2467 @param new: the desired file name
2469 @return: the success of the operation and payload
2472 _EnsureJobQueueFile(old)
2473 _EnsureJobQueueFile(new)
2475 utils.RenameFile(old, new, mkdir=True)
2478 def BlockdevClose(instance_name, disks):
2479 """Closes the given block devices.
2481 This means they will be switched to secondary mode (in case of
2484 @param instance_name: if the argument is not empty, the symlinks
2485 of this instance will be removed
2486 @type disks: list of L{objects.Disk}
2487 @param disks: the list of disks to be closed
2488 @rtype: tuple (success, message)
2489 @return: a tuple of success and message, where success
2490 indicates the succes of the operation, and message
2491 which will contain the error details in case we
2497 rd = _RecursiveFindBD(cf)
2499 _Fail("Can't find device %s", cf)
2506 except errors.BlockDeviceError, err:
2507 msg.append(str(err))
2509 _Fail("Can't make devices secondary: %s", ",".join(msg))
2512 _RemoveBlockDevLinks(instance_name, disks)
2515 def ValidateHVParams(hvname, hvparams):
2516 """Validates the given hypervisor parameters.
2518 @type hvname: string
2519 @param hvname: the hypervisor name
2520 @type hvparams: dict
2521 @param hvparams: the hypervisor parameters to be validated
2526 hv_type = hypervisor.GetHypervisor(hvname)
2527 hv_type.ValidateParameters(hvparams)
2528 except errors.HypervisorError, err:
2529 _Fail(str(err), log=False)
2532 def _CheckOSPList(os_obj, parameters):
2533 """Check whether a list of parameters is supported by the OS.
2535 @type os_obj: L{objects.OS}
2536 @param os_obj: OS object to check
2537 @type parameters: list
2538 @param parameters: the list of parameters to check
2541 supported = [v[0] for v in os_obj.supported_parameters]
2542 delta = frozenset(parameters).difference(supported)
2544 _Fail("The following parameters are not supported"
2545 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2548 def ValidateOS(required, osname, checks, osparams):
2549 """Validate the given OS' parameters.
2551 @type required: boolean
2552 @param required: whether absence of the OS should translate into
2554 @type osname: string
2555 @param osname: the OS to be validated
2557 @param checks: list of the checks to run (currently only 'parameters')
2558 @type osparams: dict
2559 @param osparams: dictionary with OS parameters
2561 @return: True if the validation passed, or False if the OS was not
2562 found and L{required} was false
2565 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2566 _Fail("Unknown checks required for OS %s: %s", osname,
2567 set(checks).difference(constants.OS_VALIDATE_CALLS))
2569 name_only = objects.OS.GetName(osname)
2570 status, tbv = _TryOSFromDisk(name_only, None)
2578 if max(tbv.api_versions) < constants.OS_API_V20:
2581 if constants.OS_VALIDATE_PARAMETERS in checks:
2582 _CheckOSPList(tbv, osparams.keys())
2584 validate_env = OSCoreEnv(osname, tbv, osparams)
2585 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2588 logging.error("os validate command '%s' returned error: %s output: %s",
2589 result.cmd, result.fail_reason, result.output)
2590 _Fail("OS validation script failed (%s), output: %s",
2591 result.fail_reason, result.output, log=False)
2597 """Demotes the current node from master candidate role.
2600 # try to ensure we're not the master by mistake
2601 master, myself = ssconf.GetMasterAndMyself()
2602 if master == myself:
2603 _Fail("ssconf status shows I'm the master node, will not demote")
2605 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2606 if not result.failed:
2607 _Fail("The master daemon is running, will not demote")
2610 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2611 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2612 except EnvironmentError, err:
2613 if err.errno != errno.ENOENT:
2614 _Fail("Error while backing up cluster file: %s", err, exc=True)
2616 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2619 def _GetX509Filenames(cryptodir, name):
2620 """Returns the full paths for the private key and certificate.
2623 return (utils.PathJoin(cryptodir, name),
2624 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2625 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2628 def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2629 """Creates a new X509 certificate for SSL/TLS.
2632 @param validity: Validity in seconds
2633 @rtype: tuple; (string, string)
2634 @return: Certificate name and public part
2637 (key_pem, cert_pem) = \
2638 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
2639 min(validity, _MAX_SSL_CERT_VALIDITY))
2641 cert_dir = tempfile.mkdtemp(dir=cryptodir,
2642 prefix="x509-%s-" % utils.TimestampForFilename())
2644 name = os.path.basename(cert_dir)
2645 assert len(name) > 5
2647 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2649 utils.WriteFile(key_file, mode=0400, data=key_pem)
2650 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2652 # Never return private key as it shouldn't leave the node
2653 return (name, cert_pem)
2655 shutil.rmtree(cert_dir, ignore_errors=True)
2659 def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2660 """Removes a X509 certificate.
2663 @param name: Certificate name
2666 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2668 utils.RemoveFile(key_file)
2669 utils.RemoveFile(cert_file)
2673 except EnvironmentError, err:
2674 _Fail("Cannot remove certificate directory '%s': %s",
2678 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2679 """Returns the command for the requested input/output.
2681 @type instance: L{objects.Instance}
2682 @param instance: The instance object
2683 @param mode: Import/export mode
2684 @param ieio: Input/output type
2685 @param ieargs: Input/output arguments
2688 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2695 if ieio == constants.IEIO_FILE:
2696 (filename, ) = ieargs
2698 if not utils.IsNormAbsPath(filename):
2699 _Fail("Path '%s' is not normalized or absolute", filename)
2701 directory = os.path.normpath(os.path.dirname(filename))
2703 if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2704 constants.EXPORT_DIR):
2705 _Fail("File '%s' is not under exports directory '%s'",
2706 filename, constants.EXPORT_DIR)
2709 utils.Makedirs(directory, mode=0750)
2711 quoted_filename = utils.ShellQuote(filename)
2713 if mode == constants.IEM_IMPORT:
2714 suffix = "> %s" % quoted_filename
2715 elif mode == constants.IEM_EXPORT:
2716 suffix = "< %s" % quoted_filename
2718 # Retrieve file size
2720 st = os.stat(filename)
2721 except EnvironmentError, err:
2722 logging.error("Can't stat(2) %s: %s", filename, err)
2724 exp_size = utils.BytesToMebibyte(st.st_size)
2726 elif ieio == constants.IEIO_RAW_DISK:
2729 real_disk = _OpenRealBD(disk)
2731 if mode == constants.IEM_IMPORT:
2732 # we set here a smaller block size as, due to transport buffering, more
2733 # than 64-128k will mostly ignored; we use nocreat to fail if the device
2734 # is not already there or we pass a wrong path; we use notrunc to no
2735 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2736 # much memory; this means that at best, we flush every 64k, which will
2738 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2739 " bs=%s oflag=dsync"),
2743 elif mode == constants.IEM_EXPORT:
2744 # the block size on the read dd is 1MiB to match our units
2745 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2747 str(1024 * 1024), # 1 MB
2749 exp_size = disk.size
2751 elif ieio == constants.IEIO_SCRIPT:
2752 (disk, disk_index, ) = ieargs
2754 assert isinstance(disk_index, (int, long))
2756 real_disk = _OpenRealBD(disk)
2758 inst_os = OSFromDisk(instance.os)
2759 env = OSEnvironment(instance, inst_os)
2761 if mode == constants.IEM_IMPORT:
2762 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2763 env["IMPORT_INDEX"] = str(disk_index)
2764 script = inst_os.import_script
2766 elif mode == constants.IEM_EXPORT:
2767 env["EXPORT_DEVICE"] = real_disk.dev_path
2768 env["EXPORT_INDEX"] = str(disk_index)
2769 script = inst_os.export_script
2771 # TODO: Pass special environment only to script
2772 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2774 if mode == constants.IEM_IMPORT:
2775 suffix = "| %s" % script_cmd
2777 elif mode == constants.IEM_EXPORT:
2778 prefix = "%s |" % script_cmd
2780 # Let script predict size
2781 exp_size = constants.IE_CUSTOM_SIZE
2784 _Fail("Invalid %s I/O mode %r", mode, ieio)
2786 return (env, prefix, suffix, exp_size)
2789 def _CreateImportExportStatusDir(prefix):
2790 """Creates status directory for import/export.
2793 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2795 (prefix, utils.TimestampForFilename())))
2798 def StartImportExportDaemon(mode, opts, host, port, instance, ieio, ieioargs):
2799 """Starts an import or export daemon.
2801 @param mode: Import/output mode
2802 @type opts: L{objects.ImportExportOptions}
2803 @param opts: Daemon options
2805 @param host: Remote host for export (None for import)
2807 @param port: Remote port for export (None for import)
2808 @type instance: L{objects.Instance}
2809 @param instance: Instance object
2810 @param ieio: Input/output type
2811 @param ieioargs: Input/output arguments
2814 if mode == constants.IEM_IMPORT:
2817 if not (host is None and port is None):
2818 _Fail("Can not specify host or port on import")
2820 elif mode == constants.IEM_EXPORT:
2823 if host is None or port is None:
2824 _Fail("Host and port must be specified for an export")
2827 _Fail("Invalid mode %r", mode)
2829 if (opts.key_name is None) ^ (opts.ca_pem is None):
2830 _Fail("Cluster certificate can only be used for both key and CA")
2832 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
2833 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
2835 if opts.key_name is None:
2837 key_path = constants.NODED_CERT_FILE
2838 cert_path = constants.NODED_CERT_FILE
2839 assert opts.ca_pem is None
2841 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
2843 assert opts.ca_pem is not None
2845 for i in [key_path, cert_path]:
2846 if not os.path.exists(i):
2847 _Fail("File '%s' does not exist" % i)
2849 status_dir = _CreateImportExportStatusDir(prefix)
2851 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
2852 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
2853 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
2855 if opts.ca_pem is None:
2857 ca = utils.ReadFile(constants.NODED_CERT_FILE)
2862 utils.WriteFile(ca_file, data=ca, mode=0400)
2865 constants.IMPORT_EXPORT_DAEMON,
2867 "--key=%s" % key_path,
2868 "--cert=%s" % cert_path,
2869 "--ca=%s" % ca_file,
2873 cmd.append("--host=%s" % host)
2876 cmd.append("--port=%s" % port)
2879 cmd.append("--compress=%s" % opts.compress)
2882 cmd.append("--magic=%s" % opts.magic)
2884 if exp_size is not None:
2885 cmd.append("--expected-size=%s" % exp_size)
2888 cmd.append("--cmd-prefix=%s" % cmd_prefix)
2891 cmd.append("--cmd-suffix=%s" % cmd_suffix)
2893 logfile = _InstanceLogName(prefix, instance.os, instance.name)
2895 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
2896 # support for receiving a file descriptor for output
2897 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
2900 # The import/export name is simply the status directory name
2901 return os.path.basename(status_dir)
2904 shutil.rmtree(status_dir, ignore_errors=True)
2908 def GetImportExportStatus(names):
2909 """Returns import/export daemon status.
2911 @type names: sequence
2912 @param names: List of names
2913 @rtype: List of dicts
2914 @return: Returns a list of the state of each named import/export or None if a
2915 status couldn't be read
2921 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
2925 data = utils.ReadFile(status_file)
2926 except EnvironmentError, err:
2927 if err.errno != errno.ENOENT:
2935 result.append(serializer.LoadJson(data))
2940 def AbortImportExport(name):
2941 """Sends SIGTERM to a running import/export daemon.
2944 logging.info("Abort import/export %s", name)
2946 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
2947 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
2950 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
2952 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
2955 def CleanupImportExport(name):
2956 """Cleanup after an import or export.
2958 If the import/export daemon is still running it's killed. Afterwards the
2959 whole status directory is removed.
2962 logging.info("Finalizing import/export %s", name)
2964 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
2966 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
2969 logging.info("Import/export %s is still running with PID %s",
2971 utils.KillProcess(pid, waitpid=False)
2973 shutil.rmtree(status_dir, ignore_errors=True)
2976 def _FindDisks(nodes_ip, disks):
2977 """Sets the physical ID on disks and returns the block devices.
2980 # set the correct physical ID
2981 my_name = netutils.Hostname.GetSysName()
2983 cf.SetPhysicalID(my_name, nodes_ip)
2988 rd = _RecursiveFindBD(cf)
2990 _Fail("Can't find device %s", cf)
2995 def DrbdDisconnectNet(nodes_ip, disks):
2996 """Disconnects the network on a list of drbd devices.
2999 bdevs = _FindDisks(nodes_ip, disks)
3005 except errors.BlockDeviceError, err:
3006 _Fail("Can't change network configuration to standalone mode: %s",
3010 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3011 """Attaches the network on a list of drbd devices.
3014 bdevs = _FindDisks(nodes_ip, disks)
3017 for idx, rd in enumerate(bdevs):
3019 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3020 except EnvironmentError, err:
3021 _Fail("Can't create symlink: %s", err)
3022 # reconnect disks, switch to new master configuration and if
3023 # needed primary mode
3026 rd.AttachNet(multimaster)
3027 except errors.BlockDeviceError, err:
3028 _Fail("Can't change network configuration: %s", err)
3030 # wait until the disks are connected; we need to retry the re-attach
3031 # if the device becomes standalone, as this might happen if the one
3032 # node disconnects and reconnects in a different mode before the
3033 # other node reconnects; in this case, one or both of the nodes will
3034 # decide it has wrong configuration and switch to standalone
3037 all_connected = True
3040 stats = rd.GetProcStatus()
3042 all_connected = (all_connected and
3043 (stats.is_connected or stats.is_in_resync))
3045 if stats.is_standalone:
3046 # peer had different config info and this node became
3047 # standalone, even though this should not happen with the
3048 # new staged way of changing disk configs
3050 rd.AttachNet(multimaster)
3051 except errors.BlockDeviceError, err:
3052 _Fail("Can't change network configuration: %s", err)
3054 if not all_connected:
3055 raise utils.RetryAgain()
3058 # Start with a delay of 100 miliseconds and go up to 5 seconds
3059 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3060 except utils.RetryTimeout:
3061 _Fail("Timeout in disk reconnecting")
3064 # change to primary mode
3068 except errors.BlockDeviceError, err:
3069 _Fail("Can't change to primary mode: %s", err)
3072 def DrbdWaitSync(nodes_ip, disks):
3073 """Wait until DRBDs have synchronized.
3077 stats = rd.GetProcStatus()
3078 if not (stats.is_connected or stats.is_in_resync):
3079 raise utils.RetryAgain()
3082 bdevs = _FindDisks(nodes_ip, disks)
3088 # poll each second for 15 seconds
3089 stats = utils.Retry(_helper, 1, 15, args=[rd])
3090 except utils.RetryTimeout:
3091 stats = rd.GetProcStatus()
3093 if not (stats.is_connected or stats.is_in_resync):
3094 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3095 alldone = alldone and (not stats.is_in_resync)
3096 if stats.sync_percent is not None:
3097 min_resync = min(min_resync, stats.sync_percent)
3099 return (alldone, min_resync)
3102 def GetDrbdUsermodeHelper():
3103 """Returns DRBD usermode helper currently configured.
3107 return bdev.BaseDRBD.GetUsermodeHelper()
3108 except errors.BlockDeviceError, err:
3112 def PowercycleNode(hypervisor_type):
3113 """Hard-powercycle the node.
3115 Because we need to return first, and schedule the powercycle in the
3116 background, we won't be able to report failures nicely.
3119 hyper = hypervisor.GetHypervisor(hypervisor_type)
3123 # if we can't fork, we'll pretend that we're in the child process
3126 return "Reboot scheduled in 5 seconds"
3127 # ensure the child is running on ram
3130 except Exception: # pylint: disable-msg=W0703
3133 hyper.PowercycleNode()
3136 class HooksRunner(object):
3139 This class is instantiated on the node side (ganeti-noded) and not
3143 def __init__(self, hooks_base_dir=None):
3144 """Constructor for hooks runner.
3146 @type hooks_base_dir: str or None
3147 @param hooks_base_dir: if not None, this overrides the
3148 L{constants.HOOKS_BASE_DIR} (useful for unittests)
3151 if hooks_base_dir is None:
3152 hooks_base_dir = constants.HOOKS_BASE_DIR
3153 # yeah, _BASE_DIR is not valid for attributes, we use it like a
3155 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
3157 def RunHooks(self, hpath, phase, env):
3158 """Run the scripts in the hooks directory.
3161 @param hpath: the path to the hooks directory which
3164 @param phase: either L{constants.HOOKS_PHASE_PRE} or
3165 L{constants.HOOKS_PHASE_POST}
3167 @param env: dictionary with the environment for the hook
3169 @return: list of 3-element tuples:
3171 - script result, either L{constants.HKR_SUCCESS} or
3172 L{constants.HKR_FAIL}
3173 - output of the script
3175 @raise errors.ProgrammerError: for invalid input
3179 if phase == constants.HOOKS_PHASE_PRE:
3181 elif phase == constants.HOOKS_PHASE_POST:
3184 _Fail("Unknown hooks phase '%s'", phase)
3187 subdir = "%s-%s.d" % (hpath, suffix)
3188 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3192 if not os.path.isdir(dir_name):
3193 # for non-existing/non-dirs, we simply exit instead of logging a
3194 # warning at every operation
3197 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3199 for (relname, relstatus, runresult) in runparts_results:
3200 if relstatus == constants.RUNPARTS_SKIP:
3201 rrval = constants.HKR_SKIP
3203 elif relstatus == constants.RUNPARTS_ERR:
3204 rrval = constants.HKR_FAIL
3205 output = "Hook script execution error: %s" % runresult
3206 elif relstatus == constants.RUNPARTS_RUN:
3207 if runresult.failed:
3208 rrval = constants.HKR_FAIL
3210 rrval = constants.HKR_SUCCESS
3211 output = utils.SafeEncode(runresult.output.strip())
3212 results.append(("%s/%s" % (subdir, relname), rrval, output))
3217 class IAllocatorRunner(object):
3218 """IAllocator runner.
3220 This class is instantiated on the node side (ganeti-noded) and not on
3225 def Run(name, idata):
3226 """Run an iallocator script.
3229 @param name: the iallocator script name
3231 @param idata: the allocator input data
3234 @return: two element tuple of:
3236 - either error message or stdout of allocator (for success)
3239 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3241 if alloc_script is None:
3242 _Fail("iallocator module '%s' not found in the search path", name)
3244 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3248 result = utils.RunCmd([alloc_script, fin_name])
3250 _Fail("iallocator module '%s' failed: %s, output '%s'",
3251 name, result.fail_reason, result.output)
3255 return result.stdout
3258 class DevCacheManager(object):
3259 """Simple class for managing a cache of block device information.
3262 _DEV_PREFIX = "/dev/"
3263 _ROOT_DIR = constants.BDEV_CACHE_DIR
3266 def _ConvertPath(cls, dev_path):
3267 """Converts a /dev/name path to the cache file name.
3269 This replaces slashes with underscores and strips the /dev
3270 prefix. It then returns the full path to the cache file.
3273 @param dev_path: the C{/dev/} path name
3275 @return: the converted path name
3278 if dev_path.startswith(cls._DEV_PREFIX):
3279 dev_path = dev_path[len(cls._DEV_PREFIX):]
3280 dev_path = dev_path.replace("/", "_")
3281 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3285 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3286 """Updates the cache information for a given device.
3289 @param dev_path: the pathname of the device
3291 @param owner: the owner (instance name) of the device
3292 @type on_primary: bool
3293 @param on_primary: whether this is the primary
3296 @param iv_name: the instance-visible name of the
3297 device, as in objects.Disk.iv_name
3302 if dev_path is None:
3303 logging.error("DevCacheManager.UpdateCache got a None dev_path")
3305 fpath = cls._ConvertPath(dev_path)
3311 iv_name = "not_visible"
3312 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3314 utils.WriteFile(fpath, data=fdata)
3315 except EnvironmentError, err:
3316 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3319 def RemoveCache(cls, dev_path):
3320 """Remove data for a dev_path.
3322 This is just a wrapper over L{utils.RemoveFile} with a converted
3323 path name and logging.
3326 @param dev_path: the pathname of the device
3331 if dev_path is None:
3332 logging.error("DevCacheManager.RemoveCache got a None dev_path")
3334 fpath = cls._ConvertPath(dev_path)
3336 utils.RemoveFile(fpath)
3337 except EnvironmentError, err:
3338 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)