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
63 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
64 _ALLOWED_CLEAN_DIRS = frozenset([
66 constants.JOB_QUEUE_ARCHIVE_DIR,
68 constants.CRYPTO_KEYS_DIR,
70 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
71 _X509_KEY_FILE = "key"
72 _X509_CERT_FILE = "cert"
73 _IES_STATUS_FILE = "status"
78 class RPCFail(Exception):
79 """Class denoting RPC failure.
81 Its argument is the error message.
86 def _Fail(msg, *args, **kwargs):
87 """Log an error and the raise an RPCFail exception.
89 This exception is then handled specially in the ganeti daemon and
90 turned into a 'failed' return type. As such, this function is a
91 useful shortcut for logging the error and returning it to the master
95 @param msg: the text of the exception
101 if "log" not in kwargs or kwargs["log"]: # if we should log this error
102 if "exc" in kwargs and kwargs["exc"]:
103 logging.exception(msg)
110 """Simple wrapper to return a SimpleStore.
112 @rtype: L{ssconf.SimpleStore}
113 @return: a SimpleStore instance
116 return ssconf.SimpleStore()
119 def _GetSshRunner(cluster_name):
120 """Simple wrapper to return an SshRunner.
122 @type cluster_name: str
123 @param cluster_name: the cluster name, which is needed
124 by the SshRunner constructor
125 @rtype: L{ssh.SshRunner}
126 @return: an SshRunner instance
129 return ssh.SshRunner(cluster_name)
132 def _Decompress(data):
133 """Unpacks data compressed by the RPC client.
135 @type data: list or tuple
136 @param data: Data sent by RPC client
138 @return: Decompressed data
141 assert isinstance(data, (list, tuple))
142 assert len(data) == 2
143 (encoding, content) = data
144 if encoding == constants.RPC_ENCODING_NONE:
146 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
147 return zlib.decompress(base64.b64decode(content))
149 raise AssertionError("Unknown data encoding")
152 def _CleanDirectory(path, exclude=None):
153 """Removes all regular files in a directory.
156 @param path: the directory to clean
158 @param exclude: list of files to be excluded, defaults
162 if path not in _ALLOWED_CLEAN_DIRS:
163 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
166 if not os.path.isdir(path):
171 # Normalize excluded paths
172 exclude = [os.path.normpath(i) for i in exclude]
174 for rel_name in utils.ListVisibleFiles(path):
175 full_name = utils.PathJoin(path, rel_name)
176 if full_name in exclude:
178 if os.path.isfile(full_name) and not os.path.islink(full_name):
179 utils.RemoveFile(full_name)
182 def _BuildUploadFileList():
183 """Build the list of allowed upload files.
185 This is abstracted so that it's built only once at module import time.
188 allowed_files = set([
189 constants.CLUSTER_CONF_FILE,
191 constants.SSH_KNOWN_HOSTS_FILE,
192 constants.VNC_PASSWORD_FILE,
193 constants.RAPI_CERT_FILE,
194 constants.RAPI_USERS_FILE,
195 constants.CONFD_HMAC_KEY,
196 constants.CLUSTER_DOMAIN_SECRET_FILE,
199 for hv_name in constants.HYPER_TYPES:
200 hv_class = hypervisor.GetHypervisorClass(hv_name)
201 allowed_files.update(hv_class.GetAncillaryFiles())
203 return frozenset(allowed_files)
206 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
210 """Removes job queue files and archived jobs.
216 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
217 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
221 """Returns master information.
223 This is an utility function to compute master information, either
224 for consumption here or from the node daemon.
227 @return: master_netdev, master_ip, master_name
228 @raise RPCFail: in case of errors
233 master_netdev = cfg.GetMasterNetdev()
234 master_ip = cfg.GetMasterIP()
235 master_node = cfg.GetMasterNode()
236 except errors.ConfigurationError, err:
237 _Fail("Cluster configuration incomplete: %s", err, exc=True)
238 return (master_netdev, master_ip, master_node)
241 def StartMaster(start_daemons, no_voting):
242 """Activate local node as master node.
244 The function will always try activate the IP address of the master
245 (unless someone else has it). It will also start the master daemons,
246 based on the start_daemons parameter.
248 @type start_daemons: boolean
249 @param start_daemons: whether to also start the master
250 daemons (ganeti-masterd and ganeti-rapi)
251 @type no_voting: boolean
252 @param no_voting: whether to start ganeti-masterd without a node vote
253 (if start_daemons is True), but still non-interactively
257 # GetMasterInfo will raise an exception if not able to return data
258 master_netdev, master_ip, _ = GetMasterInfo()
261 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
262 if utils.OwnIpAddress(master_ip):
263 # we already have the ip:
264 logging.debug("Master IP already configured, doing nothing")
266 msg = "Someone else has the master ip, not activating"
270 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
271 "dev", master_netdev, "label",
272 "%s:0" % master_netdev])
274 msg = "Can't activate master IP: %s" % result.output
278 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
279 "-s", master_ip, master_ip])
280 # we'll ignore the exit code of arping
282 # and now start the master and rapi daemons
285 masterd_args = "--no-voting --yes-do-it"
290 "EXTRA_MASTERD_ARGS": masterd_args,
293 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
295 msg = "Can't start Ganeti master: %s" % result.output
300 _Fail("; ".join(err_msgs))
303 def StopMaster(stop_daemons):
304 """Deactivate this node as master.
306 The function will always try to deactivate the IP address of the
307 master. It will also stop the master daemons depending on the
308 stop_daemons parameter.
310 @type stop_daemons: boolean
311 @param stop_daemons: whether to also stop the master daemons
312 (ganeti-masterd and ganeti-rapi)
316 # TODO: log and report back to the caller the error failures; we
317 # need to decide in which case we fail the RPC for this
319 # GetMasterInfo will raise an exception if not able to return data
320 master_netdev, master_ip, _ = GetMasterInfo()
322 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
323 "dev", master_netdev])
325 logging.error("Can't remove the master IP, error: %s", result.output)
326 # but otherwise ignore the failure
329 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
331 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
333 result.cmd, result.exit_code, result.output)
336 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
337 """Joins this node to the cluster.
339 This does the following:
340 - updates the hostkeys of the machine (rsa and dsa)
341 - adds the ssh private key to the user
342 - adds the ssh public key to the users' authorized_keys file
345 @param dsa: the DSA private key to write
347 @param dsapub: the DSA public key to write
349 @param rsa: the RSA private key to write
351 @param rsapub: the RSA public key to write
353 @param sshkey: the SSH private key to write
355 @param sshpub: the SSH public key to write
357 @return: the success of the operation
360 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
361 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
362 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
363 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
364 for name, content, mode in sshd_keys:
365 utils.WriteFile(name, data=content, mode=mode)
368 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
370 except errors.OpExecError, err:
371 _Fail("Error while processing user ssh files: %s", err, exc=True)
373 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
374 utils.WriteFile(name, data=content, mode=0600)
376 utils.AddAuthorizedKey(auth_keys, sshpub)
378 result = utils.RunCmd([constants.DAEMON_UTIL, "reload-ssh-keys"])
380 _Fail("Unable to reload SSH keys (command %r, exit code %s, output %r)",
381 result.cmd, result.exit_code, result.output)
384 def LeaveCluster(modify_ssh_setup):
385 """Cleans up and remove the current node.
387 This function cleans up and prepares the current node to be removed
390 If processing is successful, then it raises an
391 L{errors.QuitGanetiException} which is used as a special case to
392 shutdown the node daemon.
394 @param modify_ssh_setup: boolean
397 _CleanDirectory(constants.DATA_DIR)
398 _CleanDirectory(constants.CRYPTO_KEYS_DIR)
403 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
405 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
407 utils.RemoveFile(priv_key)
408 utils.RemoveFile(pub_key)
409 except errors.OpExecError:
410 logging.exception("Error while processing ssh files")
413 utils.RemoveFile(constants.CONFD_HMAC_KEY)
414 utils.RemoveFile(constants.RAPI_CERT_FILE)
415 utils.RemoveFile(constants.NODED_CERT_FILE)
416 except: # pylint: disable-msg=W0702
417 logging.exception("Error while removing cluster secrets")
419 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
421 logging.error("Command %s failed with exitcode %s and error %s",
422 result.cmd, result.exit_code, result.output)
424 # Raise a custom exception (handled in ganeti-noded)
425 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
428 def GetNodeInfo(vgname, hypervisor_type):
429 """Gives back a hash with different information about the node.
431 @type vgname: C{string}
432 @param vgname: the name of the volume group to ask for disk space information
433 @type hypervisor_type: C{str}
434 @param hypervisor_type: the name of the hypervisor to ask for
437 @return: dictionary with the following keys:
438 - vg_size is the size of the configured volume group in MiB
439 - vg_free is the free size of the volume group in MiB
440 - memory_dom0 is the memory allocated for domain0 in MiB
441 - memory_free is the currently available (free) ram in MiB
442 - memory_total is the total number of ram in MiB
446 vginfo = _GetVGInfo(vgname)
447 outputarray['vg_size'] = vginfo['vg_size']
448 outputarray['vg_free'] = vginfo['vg_free']
450 hyper = hypervisor.GetHypervisor(hypervisor_type)
451 hyp_info = hyper.GetNodeInfo()
452 if hyp_info is not None:
453 outputarray.update(hyp_info)
455 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
460 def VerifyNode(what, cluster_name):
461 """Verify the status of the local node.
463 Based on the input L{what} parameter, various checks are done on the
466 If the I{filelist} key is present, this list of
467 files is checksummed and the file/checksum pairs are returned.
469 If the I{nodelist} key is present, we check that we have
470 connectivity via ssh with the target nodes (and check the hostname
473 If the I{node-net-test} key is present, we check that we have
474 connectivity to the given nodes via both primary IP and, if
475 applicable, secondary IPs.
478 @param what: a dictionary of things to check:
479 - filelist: list of files for which to compute checksums
480 - nodelist: list of nodes we should check ssh communication with
481 - node-net-test: list of nodes we should check node daemon port
483 - hypervisor: list with hypervisors to run the verify for
485 @return: a dictionary with the same keys as the input dict, and
486 values representing the result of the checks
490 my_name = utils.HostInfo().name
491 port = utils.GetDaemonPort(constants.NODED)
493 if constants.NV_HYPERVISOR in what:
494 result[constants.NV_HYPERVISOR] = tmp = {}
495 for hv_name in what[constants.NV_HYPERVISOR]:
497 val = hypervisor.GetHypervisor(hv_name).Verify()
498 except errors.HypervisorError, err:
499 val = "Error while checking hypervisor: %s" % str(err)
502 if constants.NV_FILELIST in what:
503 result[constants.NV_FILELIST] = utils.FingerprintFiles(
504 what[constants.NV_FILELIST])
506 if constants.NV_NODELIST in what:
507 result[constants.NV_NODELIST] = tmp = {}
508 random.shuffle(what[constants.NV_NODELIST])
509 for node in what[constants.NV_NODELIST]:
510 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
514 if constants.NV_NODENETTEST in what:
515 result[constants.NV_NODENETTEST] = tmp = {}
516 my_pip = my_sip = None
517 for name, pip, sip in what[constants.NV_NODENETTEST]:
523 tmp[my_name] = ("Can't find my own primary/secondary IP"
526 for name, pip, sip in what[constants.NV_NODENETTEST]:
528 if not utils.TcpPing(pip, port, source=my_pip):
529 fail.append("primary")
531 if not utils.TcpPing(sip, port, source=my_sip):
532 fail.append("secondary")
534 tmp[name] = ("failure using the %s interface(s)" %
537 if constants.NV_MASTERIP in what:
538 # FIXME: add checks on incoming data structures (here and in the
539 # rest of the function)
540 master_name, master_ip = what[constants.NV_MASTERIP]
541 if master_name == my_name:
542 source = constants.LOCALHOST_IP_ADDRESS
545 result[constants.NV_MASTERIP] = utils.TcpPing(master_ip, port,
548 if constants.NV_LVLIST in what:
550 val = GetVolumeList(what[constants.NV_LVLIST])
553 result[constants.NV_LVLIST] = val
555 if constants.NV_INSTANCELIST in what:
556 # GetInstanceList can fail
558 val = GetInstanceList(what[constants.NV_INSTANCELIST])
561 result[constants.NV_INSTANCELIST] = val
563 if constants.NV_VGLIST in what:
564 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
566 if constants.NV_PVLIST in what:
567 result[constants.NV_PVLIST] = \
568 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
569 filter_allocatable=False)
571 if constants.NV_VERSION in what:
572 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
573 constants.RELEASE_VERSION)
575 if constants.NV_HVINFO in what:
576 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
577 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
579 if constants.NV_DRBDLIST in what:
581 used_minors = bdev.DRBD8.GetUsedDevs().keys()
582 except errors.BlockDeviceError, err:
583 logging.warning("Can't get used minors list", exc_info=True)
584 used_minors = str(err)
585 result[constants.NV_DRBDLIST] = used_minors
587 if constants.NV_NODESETUP in what:
588 result[constants.NV_NODESETUP] = tmpr = []
589 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
590 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
591 " under /sys, missing required directories /sys/block"
592 " and /sys/class/net")
593 if (not os.path.isdir("/proc/sys") or
594 not os.path.isfile("/proc/sysrq-trigger")):
595 tmpr.append("The procfs filesystem doesn't seem to be mounted"
596 " under /proc, missing required directory /proc/sys and"
597 " the file /proc/sysrq-trigger")
599 if constants.NV_TIME in what:
600 result[constants.NV_TIME] = utils.SplitTime(time.time())
605 def GetVolumeList(vg_name):
606 """Compute list of logical volumes and their size.
609 @param vg_name: the volume group whose LVs we should list
612 dictionary of all partions (key) with value being a tuple of
613 their size (in MiB), inactive and online status::
615 {'test1': ('20.06', True, True)}
617 in case of errors, a string is returned with the error
623 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
624 "--separator=%s" % sep,
625 "-olv_name,lv_size,lv_attr", vg_name])
627 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
629 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
630 for line in result.stdout.splitlines():
632 match = valid_line_re.match(line)
634 logging.error("Invalid line returned from lvs output: '%s'", line)
636 name, size, attr = match.groups()
637 inactive = attr[4] == '-'
638 online = attr[5] == 'o'
639 virtual = attr[0] == 'v'
641 # we don't want to report such volumes as existing, since they
642 # don't really hold data
644 lvs[name] = (size, inactive, online)
649 def ListVolumeGroups():
650 """List the volume groups and their size.
653 @return: dictionary with keys volume name and values the
657 return utils.ListVolumeGroups()
661 """List all volumes on this node.
665 A list of dictionaries, each having four keys:
666 - name: the logical volume name,
667 - size: the size of the logical volume
668 - dev: the physical device on which the LV lives
669 - vg: the volume group to which it belongs
671 In case of errors, we return an empty list and log the
674 Note that since a logical volume can live on multiple physical
675 volumes, the resulting list might include a logical volume
679 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
681 "--options=lv_name,lv_size,devices,vg_name"])
683 _Fail("Failed to list logical volumes, lvs output: %s",
687 return dev.split('(')[0]
690 return [parse_dev(x) for x in dev.split(",")]
693 line = [v.strip() for v in line]
694 return [{'name': line[0], 'size': line[1],
695 'dev': dev, 'vg': line[3]} for dev in handle_dev(line[2])]
698 for line in result.stdout.splitlines():
699 if line.count('|') >= 3:
700 all_devs.extend(map_line(line.split('|')))
702 logging.warning("Strange line in the output from lvs: '%s'", line)
706 def BridgesExist(bridges_list):
707 """Check if a list of bridges exist on the current node.
710 @return: C{True} if all of them exist, C{False} otherwise
714 for bridge in bridges_list:
715 if not utils.BridgeExists(bridge):
716 missing.append(bridge)
719 _Fail("Missing bridges %s", utils.CommaJoin(missing))
722 def GetInstanceList(hypervisor_list):
723 """Provides a list of instances.
725 @type hypervisor_list: list
726 @param hypervisor_list: the list of hypervisors to query information
729 @return: a list of all running instances on the current node
730 - instance1.example.com
731 - instance2.example.com
735 for hname in hypervisor_list:
737 names = hypervisor.GetHypervisor(hname).ListInstances()
738 results.extend(names)
739 except errors.HypervisorError, err:
740 _Fail("Error enumerating instances (hypervisor %s): %s",
741 hname, err, exc=True)
746 def GetInstanceInfo(instance, hname):
747 """Gives back the information about an instance as a dictionary.
749 @type instance: string
750 @param instance: the instance name
752 @param hname: the hypervisor type of the instance
755 @return: dictionary with the following keys:
756 - memory: memory size of instance (int)
757 - state: xen state of instance (string)
758 - time: cpu time of instance (float)
763 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
764 if iinfo is not None:
765 output['memory'] = iinfo[2]
766 output['state'] = iinfo[4]
767 output['time'] = iinfo[5]
772 def GetInstanceMigratable(instance):
773 """Gives whether an instance can be migrated.
775 @type instance: L{objects.Instance}
776 @param instance: object representing the instance to be checked.
779 @return: tuple of (result, description) where:
780 - result: whether the instance can be migrated or not
781 - description: a description of the issue, if relevant
784 hyper = hypervisor.GetHypervisor(instance.hypervisor)
785 iname = instance.name
786 if iname not in hyper.ListInstances():
787 _Fail("Instance %s is not running", iname)
789 for idx in range(len(instance.disks)):
790 link_name = _GetBlockDevSymlinkPath(iname, idx)
791 if not os.path.islink(link_name):
792 _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
795 def GetAllInstancesInfo(hypervisor_list):
796 """Gather data about all instances.
798 This is the equivalent of L{GetInstanceInfo}, except that it
799 computes data for all instances at once, thus being faster if one
800 needs data about more than one instance.
802 @type hypervisor_list: list
803 @param hypervisor_list: list of hypervisors to query for instance data
806 @return: dictionary of instance: data, with data having the following keys:
807 - memory: memory size of instance (int)
808 - state: xen state of instance (string)
809 - time: cpu time of instance (float)
810 - vcpus: the number of vcpus
815 for hname in hypervisor_list:
816 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
818 for name, _, memory, vcpus, state, times in iinfo:
826 # we only check static parameters, like memory and vcpus,
827 # and not state and time which can change between the
828 # invocations of the different hypervisors
829 for key in 'memory', 'vcpus':
830 if value[key] != output[name][key]:
831 _Fail("Instance %s is running twice"
832 " with different parameters", name)
838 def _InstanceLogName(kind, os_name, instance):
839 """Compute the OS log filename for a given instance and operation.
841 The instance name and os name are passed in as strings since not all
842 operations have these as part of an instance object.
845 @param kind: the operation type (e.g. add, import, etc.)
846 @type os_name: string
847 @param os_name: the os name
848 @type instance: string
849 @param instance: the name of the instance being imported/added/etc.
852 # TODO: Use tempfile.mkstemp to create unique filename
853 base = ("%s-%s-%s-%s.log" %
854 (kind, os_name, instance, utils.TimestampForFilename()))
855 return utils.PathJoin(constants.LOG_OS_DIR, base)
858 def InstanceOsAdd(instance, reinstall, debug):
859 """Add an OS to an instance.
861 @type instance: L{objects.Instance}
862 @param instance: Instance whose OS is to be installed
863 @type reinstall: boolean
864 @param reinstall: whether this is an instance reinstall
866 @param debug: debug level, passed to the OS scripts
870 inst_os = OSFromDisk(instance.os)
872 create_env = OSEnvironment(instance, inst_os, debug)
874 create_env['INSTANCE_REINSTALL'] = "1"
876 logfile = _InstanceLogName("add", instance.os, instance.name)
878 result = utils.RunCmd([inst_os.create_script], env=create_env,
879 cwd=inst_os.path, output=logfile,)
881 logging.error("os create command '%s' returned error: %s, logfile: %s,"
882 " output: %s", result.cmd, result.fail_reason, logfile,
884 lines = [utils.SafeEncode(val)
885 for val in utils.TailFile(logfile, lines=20)]
886 _Fail("OS create script failed (%s), last lines in the"
887 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
890 def RunRenameInstance(instance, old_name, debug):
891 """Run the OS rename script for an instance.
893 @type instance: L{objects.Instance}
894 @param instance: Instance whose OS is to be installed
895 @type old_name: string
896 @param old_name: previous instance name
898 @param debug: debug level, passed to the OS scripts
900 @return: the success of the operation
903 inst_os = OSFromDisk(instance.os)
905 rename_env = OSEnvironment(instance, inst_os, debug)
906 rename_env['OLD_INSTANCE_NAME'] = old_name
908 logfile = _InstanceLogName("rename", instance.os,
909 "%s-%s" % (old_name, instance.name))
911 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
912 cwd=inst_os.path, output=logfile)
915 logging.error("os create command '%s' returned error: %s output: %s",
916 result.cmd, result.fail_reason, result.output)
917 lines = [utils.SafeEncode(val)
918 for val in utils.TailFile(logfile, lines=20)]
919 _Fail("OS rename script failed (%s), last lines in the"
920 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
923 def _GetVGInfo(vg_name):
924 """Get information about the volume group.
927 @param vg_name: the volume group which we query
930 A dictionary with the following keys:
931 - C{vg_size} is the total size of the volume group in MiB
932 - C{vg_free} is the free size of the volume group in MiB
933 - C{pv_count} are the number of physical disks in that VG
935 If an error occurs during gathering of data, we return the same dict
936 with keys all set to None.
939 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
941 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
942 "--nosuffix", "--units=m", "--separator=:", vg_name])
945 logging.error("volume group %s not present", vg_name)
947 valarr = retval.stdout.strip().rstrip(':').split(':')
951 "vg_size": int(round(float(valarr[0]), 0)),
952 "vg_free": int(round(float(valarr[1]), 0)),
953 "pv_count": int(valarr[2]),
955 except (TypeError, ValueError), err:
956 logging.exception("Fail to parse vgs output: %s", err)
958 logging.error("vgs output has the wrong number of fields (expected"
959 " three): %s", str(valarr))
963 def _GetBlockDevSymlinkPath(instance_name, idx):
964 return utils.PathJoin(constants.DISK_LINKS_DIR,
965 "%s:%d" % (instance_name, idx))
968 def _SymlinkBlockDev(instance_name, device_path, idx):
969 """Set up symlinks to a instance's block device.
971 This is an auxiliary function run when an instance is start (on the primary
972 node) or when an instance is migrated (on the target node).
975 @param instance_name: the name of the target instance
976 @param device_path: path of the physical block device, on the node
977 @param idx: the disk index
978 @return: absolute path to the disk's symlink
981 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
983 os.symlink(device_path, link_name)
985 if err.errno == errno.EEXIST:
986 if (not os.path.islink(link_name) or
987 os.readlink(link_name) != device_path):
989 os.symlink(device_path, link_name)
996 def _RemoveBlockDevLinks(instance_name, disks):
997 """Remove the block device symlinks belonging to the given instance.
1000 for idx, _ in enumerate(disks):
1001 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1002 if os.path.islink(link_name):
1004 os.remove(link_name)
1006 logging.exception("Can't remove symlink '%s'", link_name)
1009 def _GatherAndLinkBlockDevs(instance):
1010 """Set up an instance's block device(s).
1012 This is run on the primary node at instance startup. The block
1013 devices must be already assembled.
1015 @type instance: L{objects.Instance}
1016 @param instance: the instance whose disks we shoul assemble
1018 @return: list of (disk_object, device_path)
1022 for idx, disk in enumerate(instance.disks):
1023 device = _RecursiveFindBD(disk)
1025 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1029 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1031 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1034 block_devices.append((disk, link_name))
1036 return block_devices
1039 def StartInstance(instance):
1040 """Start an instance.
1042 @type instance: L{objects.Instance}
1043 @param instance: the instance object
1047 running_instances = GetInstanceList([instance.hypervisor])
1049 if instance.name in running_instances:
1050 logging.info("Instance %s already running, not starting", instance.name)
1054 block_devices = _GatherAndLinkBlockDevs(instance)
1055 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1056 hyper.StartInstance(instance, block_devices)
1057 except errors.BlockDeviceError, err:
1058 _Fail("Block device error: %s", err, exc=True)
1059 except errors.HypervisorError, err:
1060 _RemoveBlockDevLinks(instance.name, instance.disks)
1061 _Fail("Hypervisor error: %s", err, exc=True)
1064 def InstanceShutdown(instance, timeout):
1065 """Shut an instance down.
1067 @note: this functions uses polling with a hardcoded timeout.
1069 @type instance: L{objects.Instance}
1070 @param instance: the instance object
1071 @type timeout: integer
1072 @param timeout: maximum timeout for soft shutdown
1076 hv_name = instance.hypervisor
1077 hyper = hypervisor.GetHypervisor(hv_name)
1078 iname = instance.name
1080 if instance.name not in hyper.ListInstances():
1081 logging.info("Instance %s not running, doing nothing", iname)
1086 self.tried_once = False
1089 if iname not in hyper.ListInstances():
1093 hyper.StopInstance(instance, retry=self.tried_once)
1094 except errors.HypervisorError, err:
1095 if iname not in hyper.ListInstances():
1096 # if the instance is no longer existing, consider this a
1097 # success and go to cleanup
1100 _Fail("Failed to stop instance %s: %s", iname, err)
1102 self.tried_once = True
1104 raise utils.RetryAgain()
1107 utils.Retry(_TryShutdown(), 5, timeout)
1108 except utils.RetryTimeout:
1109 # the shutdown did not succeed
1110 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1113 hyper.StopInstance(instance, force=True)
1114 except errors.HypervisorError, err:
1115 if iname in hyper.ListInstances():
1116 # only raise an error if the instance still exists, otherwise
1117 # the error could simply be "instance ... unknown"!
1118 _Fail("Failed to force stop instance %s: %s", iname, err)
1122 if iname in hyper.ListInstances():
1123 _Fail("Could not shutdown instance %s even by destroy", iname)
1126 hyper.CleanupInstance(instance.name)
1127 except errors.HypervisorError, err:
1128 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1130 _RemoveBlockDevLinks(iname, instance.disks)
1133 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1134 """Reboot an instance.
1136 @type instance: L{objects.Instance}
1137 @param instance: the instance object to reboot
1138 @type reboot_type: str
1139 @param reboot_type: the type of reboot, one the following
1141 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1142 instance OS, do not recreate the VM
1143 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1144 restart the VM (at the hypervisor level)
1145 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1146 not accepted here, since that mode is handled differently, in
1147 cmdlib, and translates into full stop and start of the
1148 instance (instead of a call_instance_reboot RPC)
1149 @type shutdown_timeout: integer
1150 @param shutdown_timeout: maximum timeout for soft shutdown
1154 running_instances = GetInstanceList([instance.hypervisor])
1156 if instance.name not in running_instances:
1157 _Fail("Cannot reboot instance %s that is not running", instance.name)
1159 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1160 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1162 hyper.RebootInstance(instance)
1163 except errors.HypervisorError, err:
1164 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1165 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1167 InstanceShutdown(instance, shutdown_timeout)
1168 return StartInstance(instance)
1169 except errors.HypervisorError, err:
1170 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1172 _Fail("Invalid reboot_type received: %s", reboot_type)
1175 def MigrationInfo(instance):
1176 """Gather information about an instance to be migrated.
1178 @type instance: L{objects.Instance}
1179 @param instance: the instance definition
1182 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1184 info = hyper.MigrationInfo(instance)
1185 except errors.HypervisorError, err:
1186 _Fail("Failed to fetch migration information: %s", err, exc=True)
1190 def AcceptInstance(instance, info, target):
1191 """Prepare the node to accept an instance.
1193 @type instance: L{objects.Instance}
1194 @param instance: the instance definition
1195 @type info: string/data (opaque)
1196 @param info: migration information, from the source node
1197 @type target: string
1198 @param target: target host (usually ip), on this node
1201 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1203 hyper.AcceptInstance(instance, info, target)
1204 except errors.HypervisorError, err:
1205 _Fail("Failed to accept instance: %s", err, exc=True)
1208 def FinalizeMigration(instance, info, success):
1209 """Finalize any preparation to accept an instance.
1211 @type instance: L{objects.Instance}
1212 @param instance: the instance definition
1213 @type info: string/data (opaque)
1214 @param info: migration information, from the source node
1215 @type success: boolean
1216 @param success: whether the migration was a success or a failure
1219 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1221 hyper.FinalizeMigration(instance, info, success)
1222 except errors.HypervisorError, err:
1223 _Fail("Failed to finalize migration: %s", err, exc=True)
1226 def MigrateInstance(instance, target, live):
1227 """Migrates an instance to another node.
1229 @type instance: L{objects.Instance}
1230 @param instance: the instance definition
1231 @type target: string
1232 @param target: the target node name
1234 @param live: whether the migration should be done live or not (the
1235 interpretation of this parameter is left to the hypervisor)
1237 @return: a tuple of (success, msg) where:
1238 - succes is a boolean denoting the success/failure of the operation
1239 - msg is a string with details in case of failure
1242 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1245 hyper.MigrateInstance(instance, target, live)
1246 except errors.HypervisorError, err:
1247 _Fail("Failed to migrate instance: %s", err, exc=True)
1250 def BlockdevCreate(disk, size, owner, on_primary, info):
1251 """Creates a block device for an instance.
1253 @type disk: L{objects.Disk}
1254 @param disk: the object describing the disk we should create
1256 @param size: the size of the physical underlying device, in MiB
1258 @param owner: the name of the instance for which disk is created,
1259 used for device cache data
1260 @type on_primary: boolean
1261 @param on_primary: indicates if it is the primary node or not
1263 @param info: string that will be sent to the physical device
1264 creation, used for example to set (LVM) tags on LVs
1266 @return: the new unique_id of the device (this can sometime be
1267 computed only after creation), or None. On secondary nodes,
1268 it's not required to return anything.
1271 # TODO: remove the obsolete 'size' argument
1272 # pylint: disable-msg=W0613
1275 for child in disk.children:
1277 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1278 except errors.BlockDeviceError, err:
1279 _Fail("Can't assemble device %s: %s", child, err)
1280 if on_primary or disk.AssembleOnSecondary():
1281 # we need the children open in case the device itself has to
1284 # pylint: disable-msg=E1103
1286 except errors.BlockDeviceError, err:
1287 _Fail("Can't make child '%s' read-write: %s", child, err)
1291 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1292 except errors.BlockDeviceError, err:
1293 _Fail("Can't create block device: %s", err)
1295 if on_primary or disk.AssembleOnSecondary():
1298 except errors.BlockDeviceError, err:
1299 _Fail("Can't assemble device after creation, unusual event: %s", err)
1300 device.SetSyncSpeed(constants.SYNC_SPEED)
1301 if on_primary or disk.OpenOnSecondary():
1303 device.Open(force=True)
1304 except errors.BlockDeviceError, err:
1305 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1306 DevCacheManager.UpdateCache(device.dev_path, owner,
1307 on_primary, disk.iv_name)
1309 device.SetInfo(info)
1311 return device.unique_id
1314 def BlockdevRemove(disk):
1315 """Remove a block device.
1317 @note: This is intended to be called recursively.
1319 @type disk: L{objects.Disk}
1320 @param disk: the disk object we should remove
1322 @return: the success of the operation
1327 rdev = _RecursiveFindBD(disk)
1328 except errors.BlockDeviceError, err:
1329 # probably can't attach
1330 logging.info("Can't attach to device %s in remove", disk)
1332 if rdev is not None:
1333 r_path = rdev.dev_path
1336 except errors.BlockDeviceError, err:
1337 msgs.append(str(err))
1339 DevCacheManager.RemoveCache(r_path)
1342 for child in disk.children:
1344 BlockdevRemove(child)
1345 except RPCFail, err:
1346 msgs.append(str(err))
1349 _Fail("; ".join(msgs))
1352 def _RecursiveAssembleBD(disk, owner, as_primary):
1353 """Activate a block device for an instance.
1355 This is run on the primary and secondary nodes for an instance.
1357 @note: this function is called recursively.
1359 @type disk: L{objects.Disk}
1360 @param disk: the disk we try to assemble
1362 @param owner: the name of the instance which owns the disk
1363 @type as_primary: boolean
1364 @param as_primary: if we should make the block device
1367 @return: the assembled device or None (in case no device
1369 @raise errors.BlockDeviceError: in case there is an error
1370 during the activation of the children or the device
1376 mcn = disk.ChildrenNeeded()
1378 mcn = 0 # max number of Nones allowed
1380 mcn = len(disk.children) - mcn # max number of Nones
1381 for chld_disk in disk.children:
1383 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1384 except errors.BlockDeviceError, err:
1385 if children.count(None) >= mcn:
1388 logging.error("Error in child activation (but continuing): %s",
1390 children.append(cdev)
1392 if as_primary or disk.AssembleOnSecondary():
1393 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1394 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1396 if as_primary or disk.OpenOnSecondary():
1398 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1399 as_primary, disk.iv_name)
1406 def BlockdevAssemble(disk, owner, as_primary):
1407 """Activate a block device for an instance.
1409 This is a wrapper over _RecursiveAssembleBD.
1411 @rtype: str or boolean
1412 @return: a C{/dev/...} path for primary nodes, and
1413 C{True} for secondary nodes
1417 result = _RecursiveAssembleBD(disk, owner, as_primary)
1418 if isinstance(result, bdev.BlockDev):
1419 # pylint: disable-msg=E1103
1420 result = result.dev_path
1421 except errors.BlockDeviceError, err:
1422 _Fail("Error while assembling disk: %s", err, exc=True)
1427 def BlockdevShutdown(disk):
1428 """Shut down a block device.
1430 First, if the device is assembled (Attach() is successful), then
1431 the device is shutdown. Then the children of the device are
1434 This function is called recursively. Note that we don't cache the
1435 children or such, as oppossed to assemble, shutdown of different
1436 devices doesn't require that the upper device was active.
1438 @type disk: L{objects.Disk}
1439 @param disk: the description of the disk we should
1445 r_dev = _RecursiveFindBD(disk)
1446 if r_dev is not None:
1447 r_path = r_dev.dev_path
1450 DevCacheManager.RemoveCache(r_path)
1451 except errors.BlockDeviceError, err:
1452 msgs.append(str(err))
1455 for child in disk.children:
1457 BlockdevShutdown(child)
1458 except RPCFail, err:
1459 msgs.append(str(err))
1462 _Fail("; ".join(msgs))
1465 def BlockdevAddchildren(parent_cdev, new_cdevs):
1466 """Extend a mirrored block device.
1468 @type parent_cdev: L{objects.Disk}
1469 @param parent_cdev: the disk to which we should add children
1470 @type new_cdevs: list of L{objects.Disk}
1471 @param new_cdevs: the list of children which we should add
1475 parent_bdev = _RecursiveFindBD(parent_cdev)
1476 if parent_bdev is None:
1477 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1478 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1479 if new_bdevs.count(None) > 0:
1480 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1481 parent_bdev.AddChildren(new_bdevs)
1484 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1485 """Shrink a mirrored block device.
1487 @type parent_cdev: L{objects.Disk}
1488 @param parent_cdev: the disk from which we should remove children
1489 @type new_cdevs: list of L{objects.Disk}
1490 @param new_cdevs: the list of children which we should remove
1494 parent_bdev = _RecursiveFindBD(parent_cdev)
1495 if parent_bdev is None:
1496 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1498 for disk in new_cdevs:
1499 rpath = disk.StaticDevPath()
1501 bd = _RecursiveFindBD(disk)
1503 _Fail("Can't find device %s while removing children", disk)
1505 devs.append(bd.dev_path)
1507 if not utils.IsNormAbsPath(rpath):
1508 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1510 parent_bdev.RemoveChildren(devs)
1513 def BlockdevGetmirrorstatus(disks):
1514 """Get the mirroring status of a list of devices.
1516 @type disks: list of L{objects.Disk}
1517 @param disks: the list of disks which we should query
1520 a list of (mirror_done, estimated_time) tuples, which
1521 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1522 @raise errors.BlockDeviceError: if any of the disks cannot be
1528 rbd = _RecursiveFindBD(dsk)
1530 _Fail("Can't find device %s", dsk)
1532 stats.append(rbd.CombinedSyncStatus())
1537 def _RecursiveFindBD(disk):
1538 """Check if a device is activated.
1540 If so, return information about the real device.
1542 @type disk: L{objects.Disk}
1543 @param disk: the disk object we need to find
1545 @return: None if the device can't be found,
1546 otherwise the device instance
1551 for chdisk in disk.children:
1552 children.append(_RecursiveFindBD(chdisk))
1554 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1557 def _OpenRealBD(disk):
1558 """Opens the underlying block device of a disk.
1560 @type disk: L{objects.Disk}
1561 @param disk: the disk object we want to open
1564 real_disk = _RecursiveFindBD(disk)
1565 if real_disk is None:
1566 _Fail("Block device '%s' is not set up", disk)
1573 def BlockdevFind(disk):
1574 """Check if a device is activated.
1576 If it is, return information about the real device.
1578 @type disk: L{objects.Disk}
1579 @param disk: the disk to find
1580 @rtype: None or objects.BlockDevStatus
1581 @return: None if the disk cannot be found, otherwise a the current
1586 rbd = _RecursiveFindBD(disk)
1587 except errors.BlockDeviceError, err:
1588 _Fail("Failed to find device: %s", err, exc=True)
1593 return rbd.GetSyncStatus()
1596 def BlockdevGetsize(disks):
1597 """Computes the size of the given disks.
1599 If a disk is not found, returns None instead.
1601 @type disks: list of L{objects.Disk}
1602 @param disks: the list of disk to compute the size for
1604 @return: list with elements None if the disk cannot be found,
1611 rbd = _RecursiveFindBD(cf)
1612 except errors.BlockDeviceError:
1618 result.append(rbd.GetActualSize())
1622 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1623 """Export a block device to a remote node.
1625 @type disk: L{objects.Disk}
1626 @param disk: the description of the disk to export
1627 @type dest_node: str
1628 @param dest_node: the destination node to export to
1629 @type dest_path: str
1630 @param dest_path: the destination path on the target node
1631 @type cluster_name: str
1632 @param cluster_name: the cluster name, needed for SSH hostalias
1636 real_disk = _OpenRealBD(disk)
1638 # the block size on the read dd is 1MiB to match our units
1639 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1640 "dd if=%s bs=1048576 count=%s",
1641 real_disk.dev_path, str(disk.size))
1643 # we set here a smaller block size as, due to ssh buffering, more
1644 # than 64-128k will mostly ignored; we use nocreat to fail if the
1645 # device is not already there or we pass a wrong path; we use
1646 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1647 # to not buffer too much memory; this means that at best, we flush
1648 # every 64k, which will not be very fast
1649 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1650 " oflag=dsync", dest_path)
1652 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1653 constants.GANETI_RUNAS,
1656 # all commands have been checked, so we're safe to combine them
1657 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1659 result = utils.RunCmd(["bash", "-c", command])
1662 _Fail("Disk copy command '%s' returned error: %s"
1663 " output: %s", command, result.fail_reason, result.output)
1666 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1667 """Write a file to the filesystem.
1669 This allows the master to overwrite(!) a file. It will only perform
1670 the operation if the file belongs to a list of configuration files.
1672 @type file_name: str
1673 @param file_name: the target file name
1675 @param data: the new contents of the file
1677 @param mode: the mode to give the file (can be None)
1679 @param uid: the owner of the file (can be -1 for default)
1681 @param gid: the group of the file (can be -1 for default)
1683 @param atime: the atime to set on the file (can be None)
1685 @param mtime: the mtime to set on the file (can be None)
1689 if not os.path.isabs(file_name):
1690 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1692 if file_name not in _ALLOWED_UPLOAD_FILES:
1693 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1696 raw_data = _Decompress(data)
1698 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1699 atime=atime, mtime=mtime)
1702 def WriteSsconfFiles(values):
1703 """Update all ssconf files.
1705 Wrapper around the SimpleStore.WriteFiles.
1708 ssconf.SimpleStore().WriteFiles(values)
1711 def _ErrnoOrStr(err):
1712 """Format an EnvironmentError exception.
1714 If the L{err} argument has an errno attribute, it will be looked up
1715 and converted into a textual C{E...} description. Otherwise the
1716 string representation of the error will be returned.
1718 @type err: L{EnvironmentError}
1719 @param err: the exception to format
1722 if hasattr(err, 'errno'):
1723 detail = errno.errorcode[err.errno]
1729 def _OSOndiskAPIVersion(os_dir):
1730 """Compute and return the API version of a given OS.
1732 This function will try to read the API version of the OS residing in
1733 the 'os_dir' directory.
1736 @param os_dir: the directory in which we should look for the OS
1738 @return: tuple (status, data) with status denoting the validity and
1739 data holding either the vaid versions or an error message
1742 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
1745 st = os.stat(api_file)
1746 except EnvironmentError, err:
1747 return False, ("Required file '%s' not found under path %s: %s" %
1748 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1750 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1751 return False, ("File '%s' in %s is not a regular file" %
1752 (constants.OS_API_FILE, os_dir))
1755 api_versions = utils.ReadFile(api_file).splitlines()
1756 except EnvironmentError, err:
1757 return False, ("Error while reading the API version file at %s: %s" %
1758 (api_file, _ErrnoOrStr(err)))
1761 api_versions = [int(version.strip()) for version in api_versions]
1762 except (TypeError, ValueError), err:
1763 return False, ("API version(s) can't be converted to integer: %s" %
1766 return True, api_versions
1769 def DiagnoseOS(top_dirs=None):
1770 """Compute the validity for all OSes.
1772 @type top_dirs: list
1773 @param top_dirs: the list of directories in which to
1774 search (if not given defaults to
1775 L{constants.OS_SEARCH_PATH})
1776 @rtype: list of L{objects.OS}
1777 @return: a list of tuples (name, path, status, diagnose, variants)
1778 for all (potential) OSes under all search paths, where:
1779 - name is the (potential) OS name
1780 - path is the full path to the OS
1781 - status True/False is the validity of the OS
1782 - diagnose is the error message for an invalid OS, otherwise empty
1783 - variants is a list of supported OS variants, if any
1786 if top_dirs is None:
1787 top_dirs = constants.OS_SEARCH_PATH
1790 for dir_name in top_dirs:
1791 if os.path.isdir(dir_name):
1793 f_names = utils.ListVisibleFiles(dir_name)
1794 except EnvironmentError, err:
1795 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1797 for name in f_names:
1798 os_path = utils.PathJoin(dir_name, name)
1799 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1802 variants = os_inst.supported_variants
1806 result.append((name, os_path, status, diagnose, variants))
1811 def _TryOSFromDisk(name, base_dir=None):
1812 """Create an OS instance from disk.
1814 This function will return an OS instance if the given name is a
1817 @type base_dir: string
1818 @keyword base_dir: Base directory containing OS installations.
1819 Defaults to a search in all the OS_SEARCH_PATH dirs.
1821 @return: success and either the OS instance if we find a valid one,
1825 if base_dir is None:
1826 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1828 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
1831 return False, "Directory for OS %s not found in search path" % name
1833 status, api_versions = _OSOndiskAPIVersion(os_dir)
1836 return status, api_versions
1838 if not constants.OS_API_VERSIONS.intersection(api_versions):
1839 return False, ("API version mismatch for path '%s': found %s, want %s." %
1840 (os_dir, api_versions, constants.OS_API_VERSIONS))
1842 # OS Files dictionary, we will populate it with the absolute path names
1843 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1845 if max(api_versions) >= constants.OS_API_V15:
1846 os_files[constants.OS_VARIANTS_FILE] = ''
1848 for filename in os_files:
1849 os_files[filename] = utils.PathJoin(os_dir, filename)
1852 st = os.stat(os_files[filename])
1853 except EnvironmentError, err:
1854 return False, ("File '%s' under path '%s' is missing (%s)" %
1855 (filename, os_dir, _ErrnoOrStr(err)))
1857 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1858 return False, ("File '%s' under path '%s' is not a regular file" %
1861 if filename in constants.OS_SCRIPTS:
1862 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1863 return False, ("File '%s' under path '%s' is not executable" %
1867 if constants.OS_VARIANTS_FILE in os_files:
1868 variants_file = os_files[constants.OS_VARIANTS_FILE]
1870 variants = utils.ReadFile(variants_file).splitlines()
1871 except EnvironmentError, err:
1872 return False, ("Error while reading the OS variants file at %s: %s" %
1873 (variants_file, _ErrnoOrStr(err)))
1875 return False, ("No supported os variant found")
1877 os_obj = objects.OS(name=name, path=os_dir,
1878 create_script=os_files[constants.OS_SCRIPT_CREATE],
1879 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1880 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1881 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1882 supported_variants=variants,
1883 api_versions=api_versions)
1887 def OSFromDisk(name, base_dir=None):
1888 """Create an OS instance from disk.
1890 This function will return an OS instance if the given name is a
1891 valid OS name. Otherwise, it will raise an appropriate
1892 L{RPCFail} exception, detailing why this is not a valid OS.
1894 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1895 an exception but returns true/false status data.
1897 @type base_dir: string
1898 @keyword base_dir: Base directory containing OS installations.
1899 Defaults to a search in all the OS_SEARCH_PATH dirs.
1900 @rtype: L{objects.OS}
1901 @return: the OS instance if we find a valid one
1902 @raise RPCFail: if we don't find a valid OS
1905 name_only = name.split("+", 1)[0]
1906 status, payload = _TryOSFromDisk(name_only, base_dir)
1914 def OSEnvironment(instance, inst_os, debug=0):
1915 """Calculate the environment for an os script.
1917 @type instance: L{objects.Instance}
1918 @param instance: target instance for the os script run
1919 @type inst_os: L{objects.OS}
1920 @param inst_os: operating system for which the environment is being built
1921 @type debug: integer
1922 @param debug: debug level (0 or 1, for OS Api 10)
1924 @return: dict of environment variables
1925 @raise errors.BlockDeviceError: if the block device
1931 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1932 result['OS_API_VERSION'] = '%d' % api_version
1933 result['INSTANCE_NAME'] = instance.name
1934 result['INSTANCE_OS'] = instance.os
1935 result['HYPERVISOR'] = instance.hypervisor
1936 result['DISK_COUNT'] = '%d' % len(instance.disks)
1937 result['NIC_COUNT'] = '%d' % len(instance.nics)
1938 result['DEBUG_LEVEL'] = '%d' % debug
1939 if api_version >= constants.OS_API_V15:
1941 variant = instance.os.split('+', 1)[1]
1943 variant = inst_os.supported_variants[0]
1944 result['OS_VARIANT'] = variant
1945 for idx, disk in enumerate(instance.disks):
1946 real_disk = _OpenRealBD(disk)
1947 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1948 result['DISK_%d_ACCESS' % idx] = disk.mode
1949 if constants.HV_DISK_TYPE in instance.hvparams:
1950 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1951 instance.hvparams[constants.HV_DISK_TYPE]
1952 if disk.dev_type in constants.LDS_BLOCK:
1953 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1954 elif disk.dev_type == constants.LD_FILE:
1955 result['DISK_%d_BACKEND_TYPE' % idx] = \
1956 'file:%s' % disk.physical_id[0]
1957 for idx, nic in enumerate(instance.nics):
1958 result['NIC_%d_MAC' % idx] = nic.mac
1960 result['NIC_%d_IP' % idx] = nic.ip
1961 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1962 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1963 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1964 if nic.nicparams[constants.NIC_LINK]:
1965 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1966 if constants.HV_NIC_TYPE in instance.hvparams:
1967 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1968 instance.hvparams[constants.HV_NIC_TYPE]
1970 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1971 for key, value in source.items():
1972 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1977 def BlockdevGrow(disk, amount):
1978 """Grow a stack of block devices.
1980 This function is called recursively, with the childrens being the
1981 first ones to resize.
1983 @type disk: L{objects.Disk}
1984 @param disk: the disk to be grown
1985 @rtype: (status, result)
1986 @return: a tuple with the status of the operation
1987 (True/False), and the errors message if status
1991 r_dev = _RecursiveFindBD(disk)
1993 _Fail("Cannot find block device %s", disk)
1997 except errors.BlockDeviceError, err:
1998 _Fail("Failed to grow block device: %s", err, exc=True)
2001 def BlockdevSnapshot(disk):
2002 """Create a snapshot copy of a block device.
2004 This function is called recursively, and the snapshot is actually created
2005 just for the leaf lvm backend device.
2007 @type disk: L{objects.Disk}
2008 @param disk: the disk to be snapshotted
2010 @return: snapshot disk path
2013 if disk.dev_type == constants.LD_DRBD8:
2014 if not disk.children:
2015 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2017 return BlockdevSnapshot(disk.children[0])
2018 elif disk.dev_type == constants.LD_LV:
2019 r_dev = _RecursiveFindBD(disk)
2020 if r_dev is not None:
2021 # FIXME: choose a saner value for the snapshot size
2022 # let's stay on the safe side and ask for the full size, for now
2023 return r_dev.Snapshot(disk.size)
2025 _Fail("Cannot find block device %s", disk)
2027 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2028 disk.unique_id, disk.dev_type)
2031 def FinalizeExport(instance, snap_disks):
2032 """Write out the export configuration information.
2034 @type instance: L{objects.Instance}
2035 @param instance: the instance which we export, used for
2036 saving configuration
2037 @type snap_disks: list of L{objects.Disk}
2038 @param snap_disks: list of snapshot block devices, which
2039 will be used to get the actual name of the dump file
2044 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2045 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2047 config = objects.SerializableConfigParser()
2049 config.add_section(constants.INISECT_EXP)
2050 config.set(constants.INISECT_EXP, 'version', '0')
2051 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2052 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2053 config.set(constants.INISECT_EXP, 'os', instance.os)
2054 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2056 config.add_section(constants.INISECT_INS)
2057 config.set(constants.INISECT_INS, 'name', instance.name)
2058 config.set(constants.INISECT_INS, 'memory', '%d' %
2059 instance.beparams[constants.BE_MEMORY])
2060 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2061 instance.beparams[constants.BE_VCPUS])
2062 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2063 config.set(constants.INISECT_INS, 'hypervisor', instance.hypervisor)
2066 for nic_count, nic in enumerate(instance.nics):
2068 config.set(constants.INISECT_INS, 'nic%d_mac' %
2069 nic_count, '%s' % nic.mac)
2070 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2071 for param in constants.NICS_PARAMETER_TYPES:
2072 config.set(constants.INISECT_INS, 'nic%d_%s' % (nic_count, param),
2073 '%s' % nic.nicparams.get(param, None))
2074 # TODO: redundant: on load can read nics until it doesn't exist
2075 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2078 for disk_count, disk in enumerate(snap_disks):
2081 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2082 ('%s' % disk.iv_name))
2083 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2084 ('%s' % disk.physical_id[1]))
2085 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2088 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2090 # New-style hypervisor/backend parameters
2092 config.add_section(constants.INISECT_HYP)
2093 for name, value in instance.hvparams.items():
2094 if name not in constants.HVC_GLOBALS:
2095 config.set(constants.INISECT_HYP, name, str(value))
2097 config.add_section(constants.INISECT_BEP)
2098 for name, value in instance.beparams.items():
2099 config.set(constants.INISECT_BEP, name, str(value))
2101 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2102 data=config.Dumps())
2103 shutil.rmtree(finaldestdir, ignore_errors=True)
2104 shutil.move(destdir, finaldestdir)
2107 def ExportInfo(dest):
2108 """Get export configuration information.
2111 @param dest: directory containing the export
2113 @rtype: L{objects.SerializableConfigParser}
2114 @return: a serializable config file containing the
2118 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2120 config = objects.SerializableConfigParser()
2123 if (not config.has_section(constants.INISECT_EXP) or
2124 not config.has_section(constants.INISECT_INS)):
2125 _Fail("Export info file doesn't have the required fields")
2127 return config.Dumps()
2131 """Return a list of exports currently available on this machine.
2134 @return: list of the exports
2137 if os.path.isdir(constants.EXPORT_DIR):
2138 return utils.ListVisibleFiles(constants.EXPORT_DIR)
2140 _Fail("No exports directory")
2143 def RemoveExport(export):
2144 """Remove an existing export from the node.
2147 @param export: the name of the export to remove
2151 target = utils.PathJoin(constants.EXPORT_DIR, export)
2154 shutil.rmtree(target)
2155 except EnvironmentError, err:
2156 _Fail("Error while removing the export: %s", err, exc=True)
2159 def BlockdevRename(devlist):
2160 """Rename a list of block devices.
2162 @type devlist: list of tuples
2163 @param devlist: list of tuples of the form (disk,
2164 new_logical_id, new_physical_id); disk is an
2165 L{objects.Disk} object describing the current disk,
2166 and new logical_id/physical_id is the name we
2169 @return: True if all renames succeeded, False otherwise
2174 for disk, unique_id in devlist:
2175 dev = _RecursiveFindBD(disk)
2177 msgs.append("Can't find device %s in rename" % str(disk))
2181 old_rpath = dev.dev_path
2182 dev.Rename(unique_id)
2183 new_rpath = dev.dev_path
2184 if old_rpath != new_rpath:
2185 DevCacheManager.RemoveCache(old_rpath)
2186 # FIXME: we should add the new cache information here, like:
2187 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2188 # but we don't have the owner here - maybe parse from existing
2189 # cache? for now, we only lose lvm data when we rename, which
2190 # is less critical than DRBD or MD
2191 except errors.BlockDeviceError, err:
2192 msgs.append("Can't rename device '%s' to '%s': %s" %
2193 (dev, unique_id, err))
2194 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2197 _Fail("; ".join(msgs))
2200 def _TransformFileStorageDir(file_storage_dir):
2201 """Checks whether given file_storage_dir is valid.
2203 Checks wheter the given file_storage_dir is within the cluster-wide
2204 default file_storage_dir stored in SimpleStore. Only paths under that
2205 directory are allowed.
2207 @type file_storage_dir: str
2208 @param file_storage_dir: the path to check
2210 @return: the normalized path if valid, None otherwise
2213 if not constants.ENABLE_FILE_STORAGE:
2214 _Fail("File storage disabled at configure time")
2216 file_storage_dir = os.path.normpath(file_storage_dir)
2217 base_file_storage_dir = cfg.GetFileStorageDir()
2218 if (os.path.commonprefix([file_storage_dir, base_file_storage_dir]) !=
2219 base_file_storage_dir):
2220 _Fail("File storage directory '%s' is not under base file"
2221 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2222 return file_storage_dir
2225 def CreateFileStorageDir(file_storage_dir):
2226 """Create file storage directory.
2228 @type file_storage_dir: str
2229 @param file_storage_dir: directory to create
2232 @return: tuple with first element a boolean indicating wheter dir
2233 creation was successful or not
2236 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2237 if os.path.exists(file_storage_dir):
2238 if not os.path.isdir(file_storage_dir):
2239 _Fail("Specified storage dir '%s' is not a directory",
2243 os.makedirs(file_storage_dir, 0750)
2244 except OSError, err:
2245 _Fail("Cannot create file storage directory '%s': %s",
2246 file_storage_dir, err, exc=True)
2249 def RemoveFileStorageDir(file_storage_dir):
2250 """Remove file storage directory.
2252 Remove it only if it's empty. If not log an error and return.
2254 @type file_storage_dir: str
2255 @param file_storage_dir: the directory we should cleanup
2256 @rtype: tuple (success,)
2257 @return: tuple of one element, C{success}, denoting
2258 whether the operation was successful
2261 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2262 if os.path.exists(file_storage_dir):
2263 if not os.path.isdir(file_storage_dir):
2264 _Fail("Specified Storage directory '%s' is not a directory",
2266 # deletes dir only if empty, otherwise we want to fail the rpc call
2268 os.rmdir(file_storage_dir)
2269 except OSError, err:
2270 _Fail("Cannot remove file storage directory '%s': %s",
2271 file_storage_dir, err)
2274 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2275 """Rename the file storage directory.
2277 @type old_file_storage_dir: str
2278 @param old_file_storage_dir: the current path
2279 @type new_file_storage_dir: str
2280 @param new_file_storage_dir: the name we should rename to
2281 @rtype: tuple (success,)
2282 @return: tuple of one element, C{success}, denoting
2283 whether the operation was successful
2286 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2287 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2288 if not os.path.exists(new_file_storage_dir):
2289 if os.path.isdir(old_file_storage_dir):
2291 os.rename(old_file_storage_dir, new_file_storage_dir)
2292 except OSError, err:
2293 _Fail("Cannot rename '%s' to '%s': %s",
2294 old_file_storage_dir, new_file_storage_dir, err)
2296 _Fail("Specified storage dir '%s' is not a directory",
2297 old_file_storage_dir)
2299 if os.path.exists(old_file_storage_dir):
2300 _Fail("Cannot rename '%s' to '%s': both locations exist",
2301 old_file_storage_dir, new_file_storage_dir)
2304 def _EnsureJobQueueFile(file_name):
2305 """Checks whether the given filename is in the queue directory.
2307 @type file_name: str
2308 @param file_name: the file name we should check
2310 @raises RPCFail: if the file is not valid
2313 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2314 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2317 _Fail("Passed job queue file '%s' does not belong to"
2318 " the queue directory '%s'", file_name, queue_dir)
2321 def JobQueueUpdate(file_name, content):
2322 """Updates a file in the queue directory.
2324 This is just a wrapper over L{utils.WriteFile}, with proper
2327 @type file_name: str
2328 @param file_name: the job file name
2330 @param content: the new job contents
2332 @return: the success of the operation
2335 _EnsureJobQueueFile(file_name)
2337 # Write and replace the file atomically
2338 utils.WriteFile(file_name, data=_Decompress(content))
2341 def JobQueueRename(old, new):
2342 """Renames a job queue file.
2344 This is just a wrapper over os.rename with proper checking.
2347 @param old: the old (actual) file name
2349 @param new: the desired file name
2351 @return: the success of the operation and payload
2354 _EnsureJobQueueFile(old)
2355 _EnsureJobQueueFile(new)
2357 utils.RenameFile(old, new, mkdir=True)
2360 def JobQueueSetDrainFlag(drain_flag):
2361 """Set the drain flag for the queue.
2363 This will set or unset the queue drain flag.
2365 @type drain_flag: boolean
2366 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2368 @return: always True, None
2369 @warning: the function always returns True
2373 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2375 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2378 def BlockdevClose(instance_name, disks):
2379 """Closes the given block devices.
2381 This means they will be switched to secondary mode (in case of
2384 @param instance_name: if the argument is not empty, the symlinks
2385 of this instance will be removed
2386 @type disks: list of L{objects.Disk}
2387 @param disks: the list of disks to be closed
2388 @rtype: tuple (success, message)
2389 @return: a tuple of success and message, where success
2390 indicates the succes of the operation, and message
2391 which will contain the error details in case we
2397 rd = _RecursiveFindBD(cf)
2399 _Fail("Can't find device %s", cf)
2406 except errors.BlockDeviceError, err:
2407 msg.append(str(err))
2409 _Fail("Can't make devices secondary: %s", ",".join(msg))
2412 _RemoveBlockDevLinks(instance_name, disks)
2415 def ValidateHVParams(hvname, hvparams):
2416 """Validates the given hypervisor parameters.
2418 @type hvname: string
2419 @param hvname: the hypervisor name
2420 @type hvparams: dict
2421 @param hvparams: the hypervisor parameters to be validated
2426 hv_type = hypervisor.GetHypervisor(hvname)
2427 hv_type.ValidateParameters(hvparams)
2428 except errors.HypervisorError, err:
2429 _Fail(str(err), log=False)
2433 """Demotes the current node from master candidate role.
2436 # try to ensure we're not the master by mistake
2437 master, myself = ssconf.GetMasterAndMyself()
2438 if master == myself:
2439 _Fail("ssconf status shows I'm the master node, will not demote")
2441 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2442 if not result.failed:
2443 _Fail("The master daemon is running, will not demote")
2446 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2447 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2448 except EnvironmentError, err:
2449 if err.errno != errno.ENOENT:
2450 _Fail("Error while backing up cluster file: %s", err, exc=True)
2452 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2455 def _GetX509Filenames(cryptodir, name):
2456 """Returns the full paths for the private key and certificate.
2459 return (utils.PathJoin(cryptodir, name),
2460 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2461 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2464 def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2465 """Creates a new X509 certificate for SSL/TLS.
2468 @param validity: Validity in seconds
2469 @rtype: tuple; (string, string)
2470 @return: Certificate name and public part
2473 (key_pem, cert_pem) = \
2474 utils.GenerateSelfSignedX509Cert(utils.HostInfo.SysName(),
2475 min(validity, _MAX_SSL_CERT_VALIDITY))
2477 cert_dir = tempfile.mkdtemp(dir=cryptodir,
2478 prefix="x509-%s-" % utils.TimestampForFilename())
2480 name = os.path.basename(cert_dir)
2481 assert len(name) > 5
2483 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2485 utils.WriteFile(key_file, mode=0400, data=key_pem)
2486 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
2488 # Never return private key as it shouldn't leave the node
2489 return (name, cert_pem)
2491 shutil.rmtree(cert_dir, ignore_errors=True)
2495 def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
2496 """Removes a X509 certificate.
2499 @param name: Certificate name
2502 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
2504 utils.RemoveFile(key_file)
2505 utils.RemoveFile(cert_file)
2509 except EnvironmentError, err:
2510 _Fail("Cannot remove certificate directory '%s': %s",
2514 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
2515 """Returns the command for the requested input/output.
2517 @type instance: L{objects.Instance}
2518 @param instance: The instance object
2519 @param mode: Import/export mode
2520 @param ieio: Input/output type
2521 @param ieargs: Input/output arguments
2524 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
2530 if ieio == constants.IEIO_FILE:
2531 (filename, ) = ieargs
2533 if not utils.IsNormAbsPath(filename):
2534 _Fail("Path '%s' is not normalized or absolute", filename)
2536 directory = os.path.normpath(os.path.dirname(filename))
2538 if (os.path.commonprefix([constants.EXPORT_DIR, directory]) !=
2539 constants.EXPORT_DIR):
2540 _Fail("File '%s' is not under exports directory '%s'",
2541 filename, constants.EXPORT_DIR)
2544 utils.Makedirs(directory, mode=0750)
2546 quoted_filename = utils.ShellQuote(filename)
2548 if mode == constants.IEM_IMPORT:
2549 suffix = "> %s" % quoted_filename
2550 elif mode == constants.IEM_EXPORT:
2551 suffix = "< %s" % quoted_filename
2553 elif ieio == constants.IEIO_RAW_DISK:
2556 real_disk = _OpenRealBD(disk)
2558 if mode == constants.IEM_IMPORT:
2559 # we set here a smaller block size as, due to transport buffering, more
2560 # than 64-128k will mostly ignored; we use nocreat to fail if the device
2561 # is not already there or we pass a wrong path; we use notrunc to no
2562 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
2563 # much memory; this means that at best, we flush every 64k, which will
2565 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
2566 " bs=%s oflag=dsync"),
2570 elif mode == constants.IEM_EXPORT:
2571 # the block size on the read dd is 1MiB to match our units
2572 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
2574 str(1024 * 1024), # 1 MB
2577 elif ieio == constants.IEIO_SCRIPT:
2578 (disk, disk_index, ) = ieargs
2580 assert isinstance(disk_index, (int, long))
2582 real_disk = _OpenRealBD(disk)
2584 inst_os = OSFromDisk(instance.os)
2585 env = OSEnvironment(instance, inst_os)
2587 if mode == constants.IEM_IMPORT:
2588 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
2589 env["IMPORT_INDEX"] = str(disk_index)
2590 script = inst_os.import_script
2592 elif mode == constants.IEM_EXPORT:
2593 env["EXPORT_DEVICE"] = real_disk.dev_path
2594 env["EXPORT_INDEX"] = str(disk_index)
2595 script = inst_os.export_script
2597 # TODO: Pass special environment only to script
2598 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
2600 if mode == constants.IEM_IMPORT:
2601 suffix = "| %s" % script_cmd
2603 elif mode == constants.IEM_EXPORT:
2604 prefix = "%s |" % script_cmd
2607 _Fail("Invalid %s I/O mode %r", mode, ieio)
2609 return (env, prefix, suffix)
2612 def _CreateImportExportStatusDir(prefix):
2613 """Creates status directory for import/export.
2616 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
2618 (prefix, utils.TimestampForFilename())))
2621 def StartImportExportDaemon(mode, opts, host, port, instance, ieio, ieioargs):
2622 """Starts an import or export daemon.
2624 @param mode: Import/output mode
2625 @type opts: L{objects.ImportExportOptions}
2626 @param opts: Daemon options
2628 @param host: Remote host for export (None for import)
2630 @param port: Remote port for export (None for import)
2631 @type instance: L{objects.Instance}
2632 @param instance: Instance object
2633 @param ieio: Input/output type
2634 @param ieioargs: Input/output arguments
2637 if mode == constants.IEM_IMPORT:
2640 if not (host is None and port is None):
2641 _Fail("Can not specify host or port on import")
2643 elif mode == constants.IEM_EXPORT:
2646 if host is None or port is None:
2647 _Fail("Host and port must be specified for an export")
2650 _Fail("Invalid mode %r", mode)
2652 if (opts.key_name is None) ^ (opts.ca_pem is None):
2653 _Fail("Cluster certificate can only be used for both key and CA")
2655 (cmd_env, cmd_prefix, cmd_suffix) = \
2656 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
2658 if opts.key_name is None:
2660 key_path = constants.NODED_CERT_FILE
2661 cert_path = constants.NODED_CERT_FILE
2662 assert opts.ca_pem is None
2664 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
2666 assert opts.ca_pem is not None
2668 for i in [key_path, cert_path]:
2669 if not os.path.exists(i):
2670 _Fail("File '%s' does not exist" % i)
2672 status_dir = _CreateImportExportStatusDir(prefix)
2674 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
2675 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
2676 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
2678 if opts.ca_pem is None:
2680 ca = utils.ReadFile(constants.NODED_CERT_FILE)
2685 utils.WriteFile(ca_file, data=ca, mode=0400)
2688 constants.IMPORT_EXPORT_DAEMON,
2690 "--key=%s" % key_path,
2691 "--cert=%s" % cert_path,
2692 "--ca=%s" % ca_file,
2696 cmd.append("--host=%s" % host)
2699 cmd.append("--port=%s" % port)
2702 cmd.append("--compress=%s" % opts.compress)
2705 cmd.append("--cmd-prefix=%s" % cmd_prefix)
2708 cmd.append("--cmd-suffix=%s" % cmd_suffix)
2710 logfile = _InstanceLogName(prefix, instance.os, instance.name)
2712 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
2713 # support for receiving a file descriptor for output
2714 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
2717 # The import/export name is simply the status directory name
2718 return os.path.basename(status_dir)
2721 shutil.rmtree(status_dir, ignore_errors=True)
2725 def GetImportExportStatus(names):
2726 """Returns import/export daemon status.
2728 @type names: sequence
2729 @param names: List of names
2730 @rtype: List of dicts
2731 @return: Returns a list of the state of each named import/export or None if a
2732 status couldn't be read
2738 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
2742 data = utils.ReadFile(status_file)
2743 except EnvironmentError, err:
2744 if err.errno != errno.ENOENT:
2752 result.append(serializer.LoadJson(data))
2757 def AbortImportExport(name):
2758 """Sends SIGTERM to a running import/export daemon.
2761 logging.info("Abort import/export %s", name)
2763 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
2764 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
2767 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
2769 os.kill(pid, signal.SIGTERM)
2772 def CleanupImportExport(name):
2773 """Cleanup after an import or export.
2775 If the import/export daemon is still running it's killed. Afterwards the
2776 whole status directory is removed.
2779 logging.info("Finalizing import/export %s", name)
2781 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
2783 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
2786 logging.info("Import/export %s is still running with PID %s",
2788 utils.KillProcess(pid, waitpid=False)
2790 shutil.rmtree(status_dir, ignore_errors=True)
2793 def _FindDisks(nodes_ip, disks):
2794 """Sets the physical ID on disks and returns the block devices.
2797 # set the correct physical ID
2798 my_name = utils.HostInfo().name
2800 cf.SetPhysicalID(my_name, nodes_ip)
2805 rd = _RecursiveFindBD(cf)
2807 _Fail("Can't find device %s", cf)
2812 def DrbdDisconnectNet(nodes_ip, disks):
2813 """Disconnects the network on a list of drbd devices.
2816 bdevs = _FindDisks(nodes_ip, disks)
2822 except errors.BlockDeviceError, err:
2823 _Fail("Can't change network configuration to standalone mode: %s",
2827 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2828 """Attaches the network on a list of drbd devices.
2831 bdevs = _FindDisks(nodes_ip, disks)
2834 for idx, rd in enumerate(bdevs):
2836 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2837 except EnvironmentError, err:
2838 _Fail("Can't create symlink: %s", err)
2839 # reconnect disks, switch to new master configuration and if
2840 # needed primary mode
2843 rd.AttachNet(multimaster)
2844 except errors.BlockDeviceError, err:
2845 _Fail("Can't change network configuration: %s", err)
2847 # wait until the disks are connected; we need to retry the re-attach
2848 # if the device becomes standalone, as this might happen if the one
2849 # node disconnects and reconnects in a different mode before the
2850 # other node reconnects; in this case, one or both of the nodes will
2851 # decide it has wrong configuration and switch to standalone
2854 all_connected = True
2857 stats = rd.GetProcStatus()
2859 all_connected = (all_connected and
2860 (stats.is_connected or stats.is_in_resync))
2862 if stats.is_standalone:
2863 # peer had different config info and this node became
2864 # standalone, even though this should not happen with the
2865 # new staged way of changing disk configs
2867 rd.AttachNet(multimaster)
2868 except errors.BlockDeviceError, err:
2869 _Fail("Can't change network configuration: %s", err)
2871 if not all_connected:
2872 raise utils.RetryAgain()
2875 # Start with a delay of 100 miliseconds and go up to 5 seconds
2876 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2877 except utils.RetryTimeout:
2878 _Fail("Timeout in disk reconnecting")
2881 # change to primary mode
2885 except errors.BlockDeviceError, err:
2886 _Fail("Can't change to primary mode: %s", err)
2889 def DrbdWaitSync(nodes_ip, disks):
2890 """Wait until DRBDs have synchronized.
2894 stats = rd.GetProcStatus()
2895 if not (stats.is_connected or stats.is_in_resync):
2896 raise utils.RetryAgain()
2899 bdevs = _FindDisks(nodes_ip, disks)
2905 # poll each second for 15 seconds
2906 stats = utils.Retry(_helper, 1, 15, args=[rd])
2907 except utils.RetryTimeout:
2908 stats = rd.GetProcStatus()
2910 if not (stats.is_connected or stats.is_in_resync):
2911 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2912 alldone = alldone and (not stats.is_in_resync)
2913 if stats.sync_percent is not None:
2914 min_resync = min(min_resync, stats.sync_percent)
2916 return (alldone, min_resync)
2919 def PowercycleNode(hypervisor_type):
2920 """Hard-powercycle the node.
2922 Because we need to return first, and schedule the powercycle in the
2923 background, we won't be able to report failures nicely.
2926 hyper = hypervisor.GetHypervisor(hypervisor_type)
2930 # if we can't fork, we'll pretend that we're in the child process
2933 return "Reboot scheduled in 5 seconds"
2934 # ensure the child is running on ram
2937 except Exception: # pylint: disable-msg=W0703
2940 hyper.PowercycleNode()
2943 class HooksRunner(object):
2946 This class is instantiated on the node side (ganeti-noded) and not
2950 def __init__(self, hooks_base_dir=None):
2951 """Constructor for hooks runner.
2953 @type hooks_base_dir: str or None
2954 @param hooks_base_dir: if not None, this overrides the
2955 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2958 if hooks_base_dir is None:
2959 hooks_base_dir = constants.HOOKS_BASE_DIR
2960 # yeah, _BASE_DIR is not valid for attributes, we use it like a
2962 self._BASE_DIR = hooks_base_dir # pylint: disable-msg=C0103
2964 def RunHooks(self, hpath, phase, env):
2965 """Run the scripts in the hooks directory.
2968 @param hpath: the path to the hooks directory which
2971 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2972 L{constants.HOOKS_PHASE_POST}
2974 @param env: dictionary with the environment for the hook
2976 @return: list of 3-element tuples:
2978 - script result, either L{constants.HKR_SUCCESS} or
2979 L{constants.HKR_FAIL}
2980 - output of the script
2982 @raise errors.ProgrammerError: for invalid input
2986 if phase == constants.HOOKS_PHASE_PRE:
2988 elif phase == constants.HOOKS_PHASE_POST:
2991 _Fail("Unknown hooks phase '%s'", phase)
2994 subdir = "%s-%s.d" % (hpath, suffix)
2995 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
2999 if not os.path.isdir(dir_name):
3000 # for non-existing/non-dirs, we simply exit instead of logging a
3001 # warning at every operation
3004 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3006 for (relname, relstatus, runresult) in runparts_results:
3007 if relstatus == constants.RUNPARTS_SKIP:
3008 rrval = constants.HKR_SKIP
3010 elif relstatus == constants.RUNPARTS_ERR:
3011 rrval = constants.HKR_FAIL
3012 output = "Hook script execution error: %s" % runresult
3013 elif relstatus == constants.RUNPARTS_RUN:
3014 if runresult.failed:
3015 rrval = constants.HKR_FAIL
3017 rrval = constants.HKR_SUCCESS
3018 output = utils.SafeEncode(runresult.output.strip())
3019 results.append(("%s/%s" % (subdir, relname), rrval, output))
3024 class IAllocatorRunner(object):
3025 """IAllocator runner.
3027 This class is instantiated on the node side (ganeti-noded) and not on
3032 def Run(name, idata):
3033 """Run an iallocator script.
3036 @param name: the iallocator script name
3038 @param idata: the allocator input data
3041 @return: two element tuple of:
3043 - either error message or stdout of allocator (for success)
3046 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3048 if alloc_script is None:
3049 _Fail("iallocator module '%s' not found in the search path", name)
3051 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3055 result = utils.RunCmd([alloc_script, fin_name])
3057 _Fail("iallocator module '%s' failed: %s, output '%s'",
3058 name, result.fail_reason, result.output)
3062 return result.stdout
3065 class DevCacheManager(object):
3066 """Simple class for managing a cache of block device information.
3069 _DEV_PREFIX = "/dev/"
3070 _ROOT_DIR = constants.BDEV_CACHE_DIR
3073 def _ConvertPath(cls, dev_path):
3074 """Converts a /dev/name path to the cache file name.
3076 This replaces slashes with underscores and strips the /dev
3077 prefix. It then returns the full path to the cache file.
3080 @param dev_path: the C{/dev/} path name
3082 @return: the converted path name
3085 if dev_path.startswith(cls._DEV_PREFIX):
3086 dev_path = dev_path[len(cls._DEV_PREFIX):]
3087 dev_path = dev_path.replace("/", "_")
3088 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3092 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3093 """Updates the cache information for a given device.
3096 @param dev_path: the pathname of the device
3098 @param owner: the owner (instance name) of the device
3099 @type on_primary: bool
3100 @param on_primary: whether this is the primary
3103 @param iv_name: the instance-visible name of the
3104 device, as in objects.Disk.iv_name
3109 if dev_path is None:
3110 logging.error("DevCacheManager.UpdateCache got a None dev_path")
3112 fpath = cls._ConvertPath(dev_path)
3118 iv_name = "not_visible"
3119 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3121 utils.WriteFile(fpath, data=fdata)
3122 except EnvironmentError, err:
3123 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3126 def RemoveCache(cls, dev_path):
3127 """Remove data for a dev_path.
3129 This is just a wrapper over L{utils.RemoveFile} with a converted
3130 path name and logging.
3133 @param dev_path: the pathname of the device
3138 if dev_path is None:
3139 logging.error("DevCacheManager.RemoveCache got a None dev_path")
3141 fpath = cls._ConvertPath(dev_path)
3143 utils.RemoveFile(fpath)
3144 except EnvironmentError, err:
3145 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)