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
44 from ganeti import errors
45 from ganeti import utils
46 from ganeti import ssh
47 from ganeti import hypervisor
48 from ganeti import constants
49 from ganeti import bdev
50 from ganeti import objects
51 from ganeti import ssconf
54 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
57 class RPCFail(Exception):
58 """Class denoting RPC failure.
60 Its argument is the error message.
65 def _Fail(msg, *args, **kwargs):
66 """Log an error and the raise an RPCFail exception.
68 This exception is then handled specially in the ganeti daemon and
69 turned into a 'failed' return type. As such, this function is a
70 useful shortcut for logging the error and returning it to the master
74 @param msg: the text of the exception
80 if "log" not in kwargs or kwargs["log"]: # if we should log this error
81 if "exc" in kwargs and kwargs["exc"]:
82 logging.exception(msg)
89 """Simple wrapper to return a SimpleStore.
91 @rtype: L{ssconf.SimpleStore}
92 @return: a SimpleStore instance
95 return ssconf.SimpleStore()
98 def _GetSshRunner(cluster_name):
99 """Simple wrapper to return an SshRunner.
101 @type cluster_name: str
102 @param cluster_name: the cluster name, which is needed
103 by the SshRunner constructor
104 @rtype: L{ssh.SshRunner}
105 @return: an SshRunner instance
108 return ssh.SshRunner(cluster_name)
111 def _Decompress(data):
112 """Unpacks data compressed by the RPC client.
114 @type data: list or tuple
115 @param data: Data sent by RPC client
117 @return: Decompressed data
120 assert isinstance(data, (list, tuple))
121 assert len(data) == 2
122 (encoding, content) = data
123 if encoding == constants.RPC_ENCODING_NONE:
125 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
126 return zlib.decompress(base64.b64decode(content))
128 raise AssertionError("Unknown data encoding")
131 def _CleanDirectory(path, exclude=None):
132 """Removes all regular files in a directory.
135 @param path: the directory to clean
137 @param exclude: list of files to be excluded, defaults
141 if not os.path.isdir(path):
146 # Normalize excluded paths
147 exclude = [os.path.normpath(i) for i in exclude]
149 for rel_name in utils.ListVisibleFiles(path):
150 full_name = os.path.normpath(os.path.join(path, rel_name))
151 if full_name in exclude:
153 if os.path.isfile(full_name) and not os.path.islink(full_name):
154 utils.RemoveFile(full_name)
157 def _BuildUploadFileList():
158 """Build the list of allowed upload files.
160 This is abstracted so that it's built only once at module import time.
163 allowed_files = set([
164 constants.CLUSTER_CONF_FILE,
166 constants.SSH_KNOWN_HOSTS_FILE,
167 constants.VNC_PASSWORD_FILE,
168 constants.RAPI_CERT_FILE,
169 constants.RAPI_USERS_FILE,
170 constants.HMAC_CLUSTER_KEY,
173 for hv_name in constants.HYPER_TYPES:
174 hv_class = hypervisor.GetHypervisorClass(hv_name)
175 allowed_files.update(hv_class.GetAncillaryFiles())
177 return frozenset(allowed_files)
180 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
184 """Removes job queue files and archived jobs.
190 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
191 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
195 """Returns master information.
197 This is an utility function to compute master information, either
198 for consumption here or from the node daemon.
201 @return: master_netdev, master_ip, master_name
202 @raise RPCFail: in case of errors
207 master_netdev = cfg.GetMasterNetdev()
208 master_ip = cfg.GetMasterIP()
209 master_node = cfg.GetMasterNode()
210 except errors.ConfigurationError, err:
211 _Fail("Cluster configuration incomplete: %s", err, exc=True)
212 return (master_netdev, master_ip, master_node)
215 def StartMaster(start_daemons, no_voting):
216 """Activate local node as master node.
218 The function will always try activate the IP address of the master
219 (unless someone else has it). It will also start the master daemons,
220 based on the start_daemons parameter.
222 @type start_daemons: boolean
223 @param start_daemons: whether to also start the master
224 daemons (ganeti-masterd and ganeti-rapi)
225 @type no_voting: boolean
226 @param no_voting: whether to start ganeti-masterd without a node vote
227 (if start_daemons is True), but still non-interactively
231 # GetMasterInfo will raise an exception if not able to return data
232 master_netdev, master_ip, _ = GetMasterInfo()
235 if utils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
236 if utils.OwnIpAddress(master_ip):
237 # we already have the ip:
238 logging.debug("Master IP already configured, doing nothing")
240 msg = "Someone else has the master ip, not activating"
244 result = utils.RunCmd(["ip", "address", "add", "%s/32" % master_ip,
245 "dev", master_netdev, "label",
246 "%s:0" % master_netdev])
248 msg = "Can't activate master IP: %s" % result.output
252 result = utils.RunCmd(["arping", "-q", "-U", "-c 3", "-I", master_netdev,
253 "-s", master_ip, master_ip])
254 # we'll ignore the exit code of arping
256 # and now start the master and rapi daemons
259 'ganeti-masterd': [],
263 daemons_params['ganeti-masterd'].append('--no-voting')
264 daemons_params['ganeti-masterd'].append('--yes-do-it')
265 for daemon in daemons_params:
267 cmd.extend(daemons_params[daemon])
268 result = utils.RunCmd(cmd)
270 msg = "Can't start daemon %s: %s" % (daemon, result.output)
275 _Fail("; ".join(err_msgs))
278 def StopMaster(stop_daemons):
279 """Deactivate this node as master.
281 The function will always try to deactivate the IP address of the
282 master. It will also stop the master daemons depending on the
283 stop_daemons parameter.
285 @type stop_daemons: boolean
286 @param stop_daemons: whether to also stop the master daemons
287 (ganeti-masterd and ganeti-rapi)
291 # TODO: log and report back to the caller the error failures; we
292 # need to decide in which case we fail the RPC for this
294 # GetMasterInfo will raise an exception if not able to return data
295 master_netdev, master_ip, _ = GetMasterInfo()
297 result = utils.RunCmd(["ip", "address", "del", "%s/32" % master_ip,
298 "dev", master_netdev])
300 logging.error("Can't remove the master IP, error: %s", result.output)
301 # but otherwise ignore the failure
304 # stop/kill the rapi and the master daemon
305 for daemon in constants.RAPI, constants.MASTERD:
306 utils.KillProcess(utils.ReadPidFile(utils.DaemonPidFileName(daemon)))
309 def AddNode(dsa, dsapub, rsa, rsapub, sshkey, sshpub):
310 """Joins this node to the cluster.
312 This does the following:
313 - updates the hostkeys of the machine (rsa and dsa)
314 - adds the ssh private key to the user
315 - adds the ssh public key to the users' authorized_keys file
318 @param dsa: the DSA private key to write
320 @param dsapub: the DSA public key to write
322 @param rsa: the RSA private key to write
324 @param rsapub: the RSA public key to write
326 @param sshkey: the SSH private key to write
328 @param sshpub: the SSH public key to write
330 @return: the success of the operation
333 sshd_keys = [(constants.SSH_HOST_RSA_PRIV, rsa, 0600),
334 (constants.SSH_HOST_RSA_PUB, rsapub, 0644),
335 (constants.SSH_HOST_DSA_PRIV, dsa, 0600),
336 (constants.SSH_HOST_DSA_PUB, dsapub, 0644)]
337 for name, content, mode in sshd_keys:
338 utils.WriteFile(name, data=content, mode=mode)
341 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS,
343 except errors.OpExecError, err:
344 _Fail("Error while processing user ssh files: %s", err, exc=True)
346 for name, content in [(priv_key, sshkey), (pub_key, sshpub)]:
347 utils.WriteFile(name, data=content, mode=0600)
349 utils.AddAuthorizedKey(auth_keys, sshpub)
351 utils.RunCmd([constants.SSH_INITD_SCRIPT, "restart"])
354 def LeaveCluster(modify_ssh_setup):
355 """Cleans up and remove the current node.
357 This function cleans up and prepares the current node to be removed
360 If processing is successful, then it raises an
361 L{errors.QuitGanetiException} which is used as a special case to
362 shutdown the node daemon.
364 @param modify_ssh_setup: boolean
367 _CleanDirectory(constants.DATA_DIR)
372 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
374 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
376 utils.RemoveFile(priv_key)
377 utils.RemoveFile(pub_key)
378 except errors.OpExecError:
379 logging.exception("Error while processing ssh files")
382 utils.RemoveFile(constants.HMAC_CLUSTER_KEY)
383 utils.RemoveFile(constants.RAPI_CERT_FILE)
384 utils.RemoveFile(constants.SSL_CERT_FILE)
386 logging.exception("Error while removing cluster secrets")
388 confd_pid = utils.ReadPidFile(utils.DaemonPidFileName(constants.CONFD))
391 utils.KillProcess(confd_pid, timeout=2)
393 # Raise a custom exception (handled in ganeti-noded)
394 raise errors.QuitGanetiException(True, 'Shutdown scheduled')
397 def GetNodeInfo(vgname, hypervisor_type):
398 """Gives back a hash with different information about the node.
400 @type vgname: C{string}
401 @param vgname: the name of the volume group to ask for disk space information
402 @type hypervisor_type: C{str}
403 @param hypervisor_type: the name of the hypervisor to ask for
406 @return: dictionary with the following keys:
407 - vg_size is the size of the configured volume group in MiB
408 - vg_free is the free size of the volume group in MiB
409 - memory_dom0 is the memory allocated for domain0 in MiB
410 - memory_free is the currently available (free) ram in MiB
411 - memory_total is the total number of ram in MiB
415 vginfo = _GetVGInfo(vgname)
416 outputarray['vg_size'] = vginfo['vg_size']
417 outputarray['vg_free'] = vginfo['vg_free']
419 hyper = hypervisor.GetHypervisor(hypervisor_type)
420 hyp_info = hyper.GetNodeInfo()
421 if hyp_info is not None:
422 outputarray.update(hyp_info)
424 outputarray["bootid"] = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
429 def VerifyNode(what, cluster_name):
430 """Verify the status of the local node.
432 Based on the input L{what} parameter, various checks are done on the
435 If the I{filelist} key is present, this list of
436 files is checksummed and the file/checksum pairs are returned.
438 If the I{nodelist} key is present, we check that we have
439 connectivity via ssh with the target nodes (and check the hostname
442 If the I{node-net-test} key is present, we check that we have
443 connectivity to the given nodes via both primary IP and, if
444 applicable, secondary IPs.
447 @param what: a dictionary of things to check:
448 - filelist: list of files for which to compute checksums
449 - nodelist: list of nodes we should check ssh communication with
450 - node-net-test: list of nodes we should check node daemon port
452 - hypervisor: list with hypervisors to run the verify for
454 @return: a dictionary with the same keys as the input dict, and
455 values representing the result of the checks
460 if constants.NV_HYPERVISOR in what:
461 result[constants.NV_HYPERVISOR] = tmp = {}
462 for hv_name in what[constants.NV_HYPERVISOR]:
463 tmp[hv_name] = hypervisor.GetHypervisor(hv_name).Verify()
465 if constants.NV_FILELIST in what:
466 result[constants.NV_FILELIST] = utils.FingerprintFiles(
467 what[constants.NV_FILELIST])
469 if constants.NV_NODELIST in what:
470 result[constants.NV_NODELIST] = tmp = {}
471 random.shuffle(what[constants.NV_NODELIST])
472 for node in what[constants.NV_NODELIST]:
473 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
477 if constants.NV_NODENETTEST in what:
478 result[constants.NV_NODENETTEST] = tmp = {}
479 my_name = utils.HostInfo().name
480 my_pip = my_sip = None
481 for name, pip, sip in what[constants.NV_NODENETTEST]:
487 tmp[my_name] = ("Can't find my own primary/secondary IP"
490 port = utils.GetDaemonPort(constants.NODED)
491 for name, pip, sip in what[constants.NV_NODENETTEST]:
493 if not utils.TcpPing(pip, port, source=my_pip):
494 fail.append("primary")
496 if not utils.TcpPing(sip, port, source=my_sip):
497 fail.append("secondary")
499 tmp[name] = ("failure using the %s interface(s)" %
502 if constants.NV_LVLIST in what:
503 result[constants.NV_LVLIST] = GetVolumeList(what[constants.NV_LVLIST])
505 if constants.NV_INSTANCELIST in what:
506 result[constants.NV_INSTANCELIST] = GetInstanceList(
507 what[constants.NV_INSTANCELIST])
509 if constants.NV_VGLIST in what:
510 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
512 if constants.NV_PVLIST in what:
513 result[constants.NV_PVLIST] = \
514 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
515 filter_allocatable=False)
517 if constants.NV_VERSION in what:
518 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
519 constants.RELEASE_VERSION)
521 if constants.NV_HVINFO in what:
522 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
523 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
525 if constants.NV_DRBDLIST in what:
527 used_minors = bdev.DRBD8.GetUsedDevs().keys()
528 except errors.BlockDeviceError, err:
529 logging.warning("Can't get used minors list", exc_info=True)
530 used_minors = str(err)
531 result[constants.NV_DRBDLIST] = used_minors
533 if constants.NV_NODESETUP in what:
534 result[constants.NV_NODESETUP] = tmpr = []
535 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
536 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
537 " under /sys, missing required directories /sys/block"
538 " and /sys/class/net")
539 if (not os.path.isdir("/proc/sys") or
540 not os.path.isfile("/proc/sysrq-trigger")):
541 tmpr.append("The procfs filesystem doesn't seem to be mounted"
542 " under /proc, missing required directory /proc/sys and"
543 " the file /proc/sysrq-trigger")
547 def GetVolumeList(vg_name):
548 """Compute list of logical volumes and their size.
551 @param vg_name: the volume group whose LVs we should list
554 dictionary of all partions (key) with value being a tuple of
555 their size (in MiB), inactive and online status::
557 {'test1': ('20.06', True, True)}
559 in case of errors, a string is returned with the error
565 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
566 "--separator=%s" % sep,
567 "-olv_name,lv_size,lv_attr", vg_name])
569 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
571 valid_line_re = re.compile("^ *([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
572 for line in result.stdout.splitlines():
574 match = valid_line_re.match(line)
576 logging.error("Invalid line returned from lvs output: '%s'", line)
578 name, size, attr = match.groups()
579 inactive = attr[4] == '-'
580 online = attr[5] == 'o'
581 virtual = attr[0] == 'v'
583 # we don't want to report such volumes as existing, since they
584 # don't really hold data
586 lvs[name] = (size, inactive, online)
591 def ListVolumeGroups():
592 """List the volume groups and their size.
595 @return: dictionary with keys volume name and values the
599 return utils.ListVolumeGroups()
603 """List all volumes on this node.
607 A list of dictionaries, each having four keys:
608 - name: the logical volume name,
609 - size: the size of the logical volume
610 - dev: the physical device on which the LV lives
611 - vg: the volume group to which it belongs
613 In case of errors, we return an empty list and log the
616 Note that since a logical volume can live on multiple physical
617 volumes, the resulting list might include a logical volume
621 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
623 "--options=lv_name,lv_size,devices,vg_name"])
625 _Fail("Failed to list logical volumes, lvs output: %s",
630 return dev.split('(')[0]
636 'name': line[0].strip(),
637 'size': line[1].strip(),
638 'dev': parse_dev(line[2].strip()),
639 'vg': line[3].strip(),
642 return [map_line(line.split('|')) for line in result.stdout.splitlines()
643 if line.count('|') >= 3]
646 def BridgesExist(bridges_list):
647 """Check if a list of bridges exist on the current node.
650 @return: C{True} if all of them exist, C{False} otherwise
654 for bridge in bridges_list:
655 if not utils.BridgeExists(bridge):
656 missing.append(bridge)
659 _Fail("Missing bridges %s", ", ".join(missing))
662 def GetInstanceList(hypervisor_list):
663 """Provides a list of instances.
665 @type hypervisor_list: list
666 @param hypervisor_list: the list of hypervisors to query information
669 @return: a list of all running instances on the current node
670 - instance1.example.com
671 - instance2.example.com
675 for hname in hypervisor_list:
677 names = hypervisor.GetHypervisor(hname).ListInstances()
678 results.extend(names)
679 except errors.HypervisorError, err:
680 _Fail("Error enumerating instances (hypervisor %s): %s",
681 hname, err, exc=True)
686 def GetInstanceInfo(instance, hname):
687 """Gives back the information about an instance as a dictionary.
689 @type instance: string
690 @param instance: the instance name
692 @param hname: the hypervisor type of the instance
695 @return: dictionary with the following keys:
696 - memory: memory size of instance (int)
697 - state: xen state of instance (string)
698 - time: cpu time of instance (float)
703 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
704 if iinfo is not None:
705 output['memory'] = iinfo[2]
706 output['state'] = iinfo[4]
707 output['time'] = iinfo[5]
712 def GetInstanceMigratable(instance):
713 """Gives whether an instance can be migrated.
715 @type instance: L{objects.Instance}
716 @param instance: object representing the instance to be checked.
719 @return: tuple of (result, description) where:
720 - result: whether the instance can be migrated or not
721 - description: a description of the issue, if relevant
724 hyper = hypervisor.GetHypervisor(instance.hypervisor)
725 iname = instance.name
726 if iname not in hyper.ListInstances():
727 _Fail("Instance %s is not running", iname)
729 for idx in range(len(instance.disks)):
730 link_name = _GetBlockDevSymlinkPath(iname, idx)
731 if not os.path.islink(link_name):
732 _Fail("Instance %s was not restarted since ganeti 1.2.5", iname)
735 def GetAllInstancesInfo(hypervisor_list):
736 """Gather data about all instances.
738 This is the equivalent of L{GetInstanceInfo}, except that it
739 computes data for all instances at once, thus being faster if one
740 needs data about more than one instance.
742 @type hypervisor_list: list
743 @param hypervisor_list: list of hypervisors to query for instance data
746 @return: dictionary of instance: data, with data having the following keys:
747 - memory: memory size of instance (int)
748 - state: xen state of instance (string)
749 - time: cpu time of instance (float)
750 - vcpus: the number of vcpus
755 for hname in hypervisor_list:
756 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
758 for name, _, memory, vcpus, state, times in iinfo:
766 # we only check static parameters, like memory and vcpus,
767 # and not state and time which can change between the
768 # invocations of the different hypervisors
769 for key in 'memory', 'vcpus':
770 if value[key] != output[name][key]:
771 _Fail("Instance %s is running twice"
772 " with different parameters", name)
778 def InstanceOsAdd(instance, reinstall):
779 """Add an OS to an instance.
781 @type instance: L{objects.Instance}
782 @param instance: Instance whose OS is to be installed
783 @type reinstall: boolean
784 @param reinstall: whether this is an instance reinstall
788 inst_os = OSFromDisk(instance.os)
790 create_env = OSEnvironment(instance, inst_os)
792 create_env['INSTANCE_REINSTALL'] = "1"
794 logfile = "%s/add-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
795 instance.name, int(time.time()))
797 result = utils.RunCmd([inst_os.create_script], env=create_env,
798 cwd=inst_os.path, output=logfile,)
800 logging.error("os create command '%s' returned error: %s, logfile: %s,"
801 " output: %s", result.cmd, result.fail_reason, logfile,
803 lines = [utils.SafeEncode(val)
804 for val in utils.TailFile(logfile, lines=20)]
805 _Fail("OS create script failed (%s), last lines in the"
806 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
809 def RunRenameInstance(instance, old_name):
810 """Run the OS rename script for an instance.
812 @type instance: L{objects.Instance}
813 @param instance: Instance whose OS is to be installed
814 @type old_name: string
815 @param old_name: previous instance name
817 @return: the success of the operation
820 inst_os = OSFromDisk(instance.os)
822 rename_env = OSEnvironment(instance, inst_os)
823 rename_env['OLD_INSTANCE_NAME'] = old_name
825 logfile = "%s/rename-%s-%s-%s-%d.log" % (constants.LOG_OS_DIR, instance.os,
827 instance.name, int(time.time()))
829 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
830 cwd=inst_os.path, output=logfile)
833 logging.error("os create command '%s' returned error: %s output: %s",
834 result.cmd, result.fail_reason, result.output)
835 lines = [utils.SafeEncode(val)
836 for val in utils.TailFile(logfile, lines=20)]
837 _Fail("OS rename script failed (%s), last lines in the"
838 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
841 def _GetVGInfo(vg_name):
842 """Get information about the volume group.
845 @param vg_name: the volume group which we query
848 A dictionary with the following keys:
849 - C{vg_size} is the total size of the volume group in MiB
850 - C{vg_free} is the free size of the volume group in MiB
851 - C{pv_count} are the number of physical disks in that VG
853 If an error occurs during gathering of data, we return the same dict
854 with keys all set to None.
857 retdic = dict.fromkeys(["vg_size", "vg_free", "pv_count"])
859 retval = utils.RunCmd(["vgs", "-ovg_size,vg_free,pv_count", "--noheadings",
860 "--nosuffix", "--units=m", "--separator=:", vg_name])
863 logging.error("volume group %s not present", vg_name)
865 valarr = retval.stdout.strip().rstrip(':').split(':')
869 "vg_size": int(round(float(valarr[0]), 0)),
870 "vg_free": int(round(float(valarr[1]), 0)),
871 "pv_count": int(valarr[2]),
873 except ValueError, err:
874 logging.exception("Fail to parse vgs output: %s", err)
876 logging.error("vgs output has the wrong number of fields (expected"
877 " three): %s", str(valarr))
881 def _GetBlockDevSymlinkPath(instance_name, idx):
882 return os.path.join(constants.DISK_LINKS_DIR,
883 "%s:%d" % (instance_name, idx))
886 def _SymlinkBlockDev(instance_name, device_path, idx):
887 """Set up symlinks to a instance's block device.
889 This is an auxiliary function run when an instance is start (on the primary
890 node) or when an instance is migrated (on the target node).
893 @param instance_name: the name of the target instance
894 @param device_path: path of the physical block device, on the node
895 @param idx: the disk index
896 @return: absolute path to the disk's symlink
899 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
901 os.symlink(device_path, link_name)
903 if err.errno == errno.EEXIST:
904 if (not os.path.islink(link_name) or
905 os.readlink(link_name) != device_path):
907 os.symlink(device_path, link_name)
914 def _RemoveBlockDevLinks(instance_name, disks):
915 """Remove the block device symlinks belonging to the given instance.
918 for idx, _ in enumerate(disks):
919 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
920 if os.path.islink(link_name):
924 logging.exception("Can't remove symlink '%s'", link_name)
927 def _GatherAndLinkBlockDevs(instance):
928 """Set up an instance's block device(s).
930 This is run on the primary node at instance startup. The block
931 devices must be already assembled.
933 @type instance: L{objects.Instance}
934 @param instance: the instance whose disks we shoul assemble
936 @return: list of (disk_object, device_path)
940 for idx, disk in enumerate(instance.disks):
941 device = _RecursiveFindBD(disk)
943 raise errors.BlockDeviceError("Block device '%s' is not set up." %
947 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
949 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
952 block_devices.append((disk, link_name))
957 def StartInstance(instance):
958 """Start an instance.
960 @type instance: L{objects.Instance}
961 @param instance: the instance object
965 running_instances = GetInstanceList([instance.hypervisor])
967 if instance.name in running_instances:
968 logging.info("Instance %s already running, not starting", instance.name)
972 block_devices = _GatherAndLinkBlockDevs(instance)
973 hyper = hypervisor.GetHypervisor(instance.hypervisor)
974 hyper.StartInstance(instance, block_devices)
975 except errors.BlockDeviceError, err:
976 _Fail("Block device error: %s", err, exc=True)
977 except errors.HypervisorError, err:
978 _RemoveBlockDevLinks(instance.name, instance.disks)
979 _Fail("Hypervisor error: %s", err, exc=True)
982 def InstanceShutdown(instance, timeout):
983 """Shut an instance down.
985 @note: this functions uses polling with a hardcoded timeout.
987 @type instance: L{objects.Instance}
988 @param instance: the instance object
989 @type timeout: integer
990 @param timeout: maximum timeout for soft shutdown
994 hv_name = instance.hypervisor
995 hyper = hypervisor.GetHypervisor(hv_name)
996 iname = instance.name
998 if instance.name not in hyper.ListInstances():
999 logging.info("Instance %s not running, doing nothing", iname)
1004 self.tried_once = False
1007 if iname not in hyper.ListInstances():
1011 hyper.StopInstance(instance, retry=self.tried_once)
1012 except errors.HypervisorError, err:
1013 if iname not in hyper.ListInstances():
1014 # if the instance is no longer existing, consider this a
1015 # success and go to cleanup
1018 _Fail("Failed to stop instance %s: %s", iname, err)
1020 self.tried_once = True
1022 raise utils.RetryAgain()
1025 utils.Retry(_TryShutdown(), 5, timeout)
1026 except utils.RetryTimeout:
1027 # the shutdown did not succeed
1028 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1031 hyper.StopInstance(instance, force=True)
1032 except errors.HypervisorError, err:
1033 if iname in hyper.ListInstances():
1034 # only raise an error if the instance still exists, otherwise
1035 # the error could simply be "instance ... unknown"!
1036 _Fail("Failed to force stop instance %s: %s", iname, err)
1040 if iname in hyper.ListInstances():
1041 _Fail("Could not shutdown instance %s even by destroy", iname)
1043 _RemoveBlockDevLinks(iname, instance.disks)
1046 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1047 """Reboot an instance.
1049 @type instance: L{objects.Instance}
1050 @param instance: the instance object to reboot
1051 @type reboot_type: str
1052 @param reboot_type: the type of reboot, one the following
1054 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1055 instance OS, do not recreate the VM
1056 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1057 restart the VM (at the hypervisor level)
1058 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1059 not accepted here, since that mode is handled differently, in
1060 cmdlib, and translates into full stop and start of the
1061 instance (instead of a call_instance_reboot RPC)
1062 @type shutdown_timeout: integer
1063 @param shutdown_timeout: maximum timeout for soft shutdown
1067 running_instances = GetInstanceList([instance.hypervisor])
1069 if instance.name not in running_instances:
1070 _Fail("Cannot reboot instance %s that is not running", instance.name)
1072 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1073 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1075 hyper.RebootInstance(instance)
1076 except errors.HypervisorError, err:
1077 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1078 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1080 InstanceShutdown(instance, shutdown_timeout)
1081 return StartInstance(instance)
1082 except errors.HypervisorError, err:
1083 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1085 _Fail("Invalid reboot_type received: %s", reboot_type)
1088 def MigrationInfo(instance):
1089 """Gather information about an instance to be migrated.
1091 @type instance: L{objects.Instance}
1092 @param instance: the instance definition
1095 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1097 info = hyper.MigrationInfo(instance)
1098 except errors.HypervisorError, err:
1099 _Fail("Failed to fetch migration information: %s", err, exc=True)
1103 def AcceptInstance(instance, info, target):
1104 """Prepare the node to accept an instance.
1106 @type instance: L{objects.Instance}
1107 @param instance: the instance definition
1108 @type info: string/data (opaque)
1109 @param info: migration information, from the source node
1110 @type target: string
1111 @param target: target host (usually ip), on this node
1114 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1116 hyper.AcceptInstance(instance, info, target)
1117 except errors.HypervisorError, err:
1118 _Fail("Failed to accept instance: %s", err, exc=True)
1121 def FinalizeMigration(instance, info, success):
1122 """Finalize any preparation to accept an instance.
1124 @type instance: L{objects.Instance}
1125 @param instance: the instance definition
1126 @type info: string/data (opaque)
1127 @param info: migration information, from the source node
1128 @type success: boolean
1129 @param success: whether the migration was a success or a failure
1132 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1134 hyper.FinalizeMigration(instance, info, success)
1135 except errors.HypervisorError, err:
1136 _Fail("Failed to finalize migration: %s", err, exc=True)
1139 def MigrateInstance(instance, target, live):
1140 """Migrates an instance to another node.
1142 @type instance: L{objects.Instance}
1143 @param instance: the instance definition
1144 @type target: string
1145 @param target: the target node name
1147 @param live: whether the migration should be done live or not (the
1148 interpretation of this parameter is left to the hypervisor)
1150 @return: a tuple of (success, msg) where:
1151 - succes is a boolean denoting the success/failure of the operation
1152 - msg is a string with details in case of failure
1155 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1158 hyper.MigrateInstance(instance, target, live)
1159 except errors.HypervisorError, err:
1160 _Fail("Failed to migrate instance: %s", err, exc=True)
1163 def BlockdevCreate(disk, size, owner, on_primary, info):
1164 """Creates a block device for an instance.
1166 @type disk: L{objects.Disk}
1167 @param disk: the object describing the disk we should create
1169 @param size: the size of the physical underlying device, in MiB
1171 @param owner: the name of the instance for which disk is created,
1172 used for device cache data
1173 @type on_primary: boolean
1174 @param on_primary: indicates if it is the primary node or not
1176 @param info: string that will be sent to the physical device
1177 creation, used for example to set (LVM) tags on LVs
1179 @return: the new unique_id of the device (this can sometime be
1180 computed only after creation), or None. On secondary nodes,
1181 it's not required to return anything.
1186 for child in disk.children:
1188 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1189 except errors.BlockDeviceError, err:
1190 _Fail("Can't assemble device %s: %s", child, err)
1191 if on_primary or disk.AssembleOnSecondary():
1192 # we need the children open in case the device itself has to
1196 except errors.BlockDeviceError, err:
1197 _Fail("Can't make child '%s' read-write: %s", child, err)
1201 device = bdev.Create(disk.dev_type, disk.physical_id, clist, disk.size)
1202 except errors.BlockDeviceError, err:
1203 _Fail("Can't create block device: %s", err)
1205 if on_primary or disk.AssembleOnSecondary():
1208 except errors.BlockDeviceError, err:
1209 _Fail("Can't assemble device after creation, unusual event: %s", err)
1210 device.SetSyncSpeed(constants.SYNC_SPEED)
1211 if on_primary or disk.OpenOnSecondary():
1213 device.Open(force=True)
1214 except errors.BlockDeviceError, err:
1215 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1216 DevCacheManager.UpdateCache(device.dev_path, owner,
1217 on_primary, disk.iv_name)
1219 device.SetInfo(info)
1221 return device.unique_id
1224 def BlockdevRemove(disk):
1225 """Remove a block device.
1227 @note: This is intended to be called recursively.
1229 @type disk: L{objects.Disk}
1230 @param disk: the disk object we should remove
1232 @return: the success of the operation
1237 rdev = _RecursiveFindBD(disk)
1238 except errors.BlockDeviceError, err:
1239 # probably can't attach
1240 logging.info("Can't attach to device %s in remove", disk)
1242 if rdev is not None:
1243 r_path = rdev.dev_path
1246 except errors.BlockDeviceError, err:
1247 msgs.append(str(err))
1249 DevCacheManager.RemoveCache(r_path)
1252 for child in disk.children:
1254 BlockdevRemove(child)
1255 except RPCFail, err:
1256 msgs.append(str(err))
1259 _Fail("; ".join(msgs))
1262 def _RecursiveAssembleBD(disk, owner, as_primary):
1263 """Activate a block device for an instance.
1265 This is run on the primary and secondary nodes for an instance.
1267 @note: this function is called recursively.
1269 @type disk: L{objects.Disk}
1270 @param disk: the disk we try to assemble
1272 @param owner: the name of the instance which owns the disk
1273 @type as_primary: boolean
1274 @param as_primary: if we should make the block device
1277 @return: the assembled device or None (in case no device
1279 @raise errors.BlockDeviceError: in case there is an error
1280 during the activation of the children or the device
1286 mcn = disk.ChildrenNeeded()
1288 mcn = 0 # max number of Nones allowed
1290 mcn = len(disk.children) - mcn # max number of Nones
1291 for chld_disk in disk.children:
1293 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1294 except errors.BlockDeviceError, err:
1295 if children.count(None) >= mcn:
1298 logging.error("Error in child activation (but continuing): %s",
1300 children.append(cdev)
1302 if as_primary or disk.AssembleOnSecondary():
1303 r_dev = bdev.Assemble(disk.dev_type, disk.physical_id, children, disk.size)
1304 r_dev.SetSyncSpeed(constants.SYNC_SPEED)
1306 if as_primary or disk.OpenOnSecondary():
1308 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1309 as_primary, disk.iv_name)
1316 def BlockdevAssemble(disk, owner, as_primary):
1317 """Activate a block device for an instance.
1319 This is a wrapper over _RecursiveAssembleBD.
1321 @rtype: str or boolean
1322 @return: a C{/dev/...} path for primary nodes, and
1323 C{True} for secondary nodes
1327 result = _RecursiveAssembleBD(disk, owner, as_primary)
1328 if isinstance(result, bdev.BlockDev):
1329 result = result.dev_path
1330 except errors.BlockDeviceError, err:
1331 _Fail("Error while assembling disk: %s", err, exc=True)
1336 def BlockdevShutdown(disk):
1337 """Shut down a block device.
1339 First, if the device is assembled (Attach() is successful), then
1340 the device is shutdown. Then the children of the device are
1343 This function is called recursively. Note that we don't cache the
1344 children or such, as oppossed to assemble, shutdown of different
1345 devices doesn't require that the upper device was active.
1347 @type disk: L{objects.Disk}
1348 @param disk: the description of the disk we should
1354 r_dev = _RecursiveFindBD(disk)
1355 if r_dev is not None:
1356 r_path = r_dev.dev_path
1359 DevCacheManager.RemoveCache(r_path)
1360 except errors.BlockDeviceError, err:
1361 msgs.append(str(err))
1364 for child in disk.children:
1366 BlockdevShutdown(child)
1367 except RPCFail, err:
1368 msgs.append(str(err))
1371 _Fail("; ".join(msgs))
1374 def BlockdevAddchildren(parent_cdev, new_cdevs):
1375 """Extend a mirrored block device.
1377 @type parent_cdev: L{objects.Disk}
1378 @param parent_cdev: the disk to which we should add children
1379 @type new_cdevs: list of L{objects.Disk}
1380 @param new_cdevs: the list of children which we should add
1384 parent_bdev = _RecursiveFindBD(parent_cdev)
1385 if parent_bdev is None:
1386 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1387 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1388 if new_bdevs.count(None) > 0:
1389 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1390 parent_bdev.AddChildren(new_bdevs)
1393 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1394 """Shrink a mirrored block device.
1396 @type parent_cdev: L{objects.Disk}
1397 @param parent_cdev: the disk from which we should remove children
1398 @type new_cdevs: list of L{objects.Disk}
1399 @param new_cdevs: the list of children which we should remove
1403 parent_bdev = _RecursiveFindBD(parent_cdev)
1404 if parent_bdev is None:
1405 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1407 for disk in new_cdevs:
1408 rpath = disk.StaticDevPath()
1410 bd = _RecursiveFindBD(disk)
1412 _Fail("Can't find device %s while removing children", disk)
1414 devs.append(bd.dev_path)
1417 parent_bdev.RemoveChildren(devs)
1420 def BlockdevGetmirrorstatus(disks):
1421 """Get the mirroring status of a list of devices.
1423 @type disks: list of L{objects.Disk}
1424 @param disks: the list of disks which we should query
1427 a list of (mirror_done, estimated_time) tuples, which
1428 are the result of L{bdev.BlockDev.CombinedSyncStatus}
1429 @raise errors.BlockDeviceError: if any of the disks cannot be
1435 rbd = _RecursiveFindBD(dsk)
1437 _Fail("Can't find device %s", dsk)
1439 stats.append(rbd.CombinedSyncStatus())
1444 def _RecursiveFindBD(disk):
1445 """Check if a device is activated.
1447 If so, return information about the real device.
1449 @type disk: L{objects.Disk}
1450 @param disk: the disk object we need to find
1452 @return: None if the device can't be found,
1453 otherwise the device instance
1458 for chdisk in disk.children:
1459 children.append(_RecursiveFindBD(chdisk))
1461 return bdev.FindDevice(disk.dev_type, disk.physical_id, children, disk.size)
1464 def BlockdevFind(disk):
1465 """Check if a device is activated.
1467 If it is, return information about the real device.
1469 @type disk: L{objects.Disk}
1470 @param disk: the disk to find
1471 @rtype: None or objects.BlockDevStatus
1472 @return: None if the disk cannot be found, otherwise a the current
1477 rbd = _RecursiveFindBD(disk)
1478 except errors.BlockDeviceError, err:
1479 _Fail("Failed to find device: %s", err, exc=True)
1484 return rbd.GetSyncStatus()
1487 def BlockdevGetsize(disks):
1488 """Computes the size of the given disks.
1490 If a disk is not found, returns None instead.
1492 @type disks: list of L{objects.Disk}
1493 @param disks: the list of disk to compute the size for
1495 @return: list with elements None if the disk cannot be found,
1502 rbd = _RecursiveFindBD(cf)
1503 except errors.BlockDeviceError, err:
1509 result.append(rbd.GetActualSize())
1513 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
1514 """Export a block device to a remote node.
1516 @type disk: L{objects.Disk}
1517 @param disk: the description of the disk to export
1518 @type dest_node: str
1519 @param dest_node: the destination node to export to
1520 @type dest_path: str
1521 @param dest_path: the destination path on the target node
1522 @type cluster_name: str
1523 @param cluster_name: the cluster name, needed for SSH hostalias
1527 real_disk = _RecursiveFindBD(disk)
1528 if real_disk is None:
1529 _Fail("Block device '%s' is not set up", disk)
1533 # the block size on the read dd is 1MiB to match our units
1534 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
1535 "dd if=%s bs=1048576 count=%s",
1536 real_disk.dev_path, str(disk.size))
1538 # we set here a smaller block size as, due to ssh buffering, more
1539 # than 64-128k will mostly ignored; we use nocreat to fail if the
1540 # device is not already there or we pass a wrong path; we use
1541 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
1542 # to not buffer too much memory; this means that at best, we flush
1543 # every 64k, which will not be very fast
1544 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
1545 " oflag=dsync", dest_path)
1547 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1548 constants.GANETI_RUNAS,
1551 # all commands have been checked, so we're safe to combine them
1552 command = '|'.join([expcmd, utils.ShellQuoteArgs(remotecmd)])
1554 result = utils.RunCmd(["bash", "-c", command])
1557 _Fail("Disk copy command '%s' returned error: %s"
1558 " output: %s", command, result.fail_reason, result.output)
1561 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
1562 """Write a file to the filesystem.
1564 This allows the master to overwrite(!) a file. It will only perform
1565 the operation if the file belongs to a list of configuration files.
1567 @type file_name: str
1568 @param file_name: the target file name
1570 @param data: the new contents of the file
1572 @param mode: the mode to give the file (can be None)
1574 @param uid: the owner of the file (can be -1 for default)
1576 @param gid: the group of the file (can be -1 for default)
1578 @param atime: the atime to set on the file (can be None)
1580 @param mtime: the mtime to set on the file (can be None)
1584 if not os.path.isabs(file_name):
1585 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
1587 if file_name not in _ALLOWED_UPLOAD_FILES:
1588 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
1591 raw_data = _Decompress(data)
1593 utils.WriteFile(file_name, data=raw_data, mode=mode, uid=uid, gid=gid,
1594 atime=atime, mtime=mtime)
1597 def WriteSsconfFiles(values):
1598 """Update all ssconf files.
1600 Wrapper around the SimpleStore.WriteFiles.
1603 ssconf.SimpleStore().WriteFiles(values)
1606 def _ErrnoOrStr(err):
1607 """Format an EnvironmentError exception.
1609 If the L{err} argument has an errno attribute, it will be looked up
1610 and converted into a textual C{E...} description. Otherwise the
1611 string representation of the error will be returned.
1613 @type err: L{EnvironmentError}
1614 @param err: the exception to format
1617 if hasattr(err, 'errno'):
1618 detail = errno.errorcode[err.errno]
1624 def _OSOndiskAPIVersion(name, os_dir):
1625 """Compute and return the API version of a given OS.
1627 This function will try to read the API version of the OS given by
1628 the 'name' parameter and residing in the 'os_dir' directory.
1631 @param name: the OS name we should look for
1633 @param os_dir: the directory inwhich we should look for the OS
1635 @return: tuple (status, data) with status denoting the validity and
1636 data holding either the vaid versions or an error message
1639 api_file = os.path.sep.join([os_dir, constants.OS_API_FILE])
1642 st = os.stat(api_file)
1643 except EnvironmentError, err:
1644 return False, ("Required file '%s' not found under path %s: %s" %
1645 (constants.OS_API_FILE, os_dir, _ErrnoOrStr(err)))
1647 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1648 return False, ("File '%s' in %s is not a regular file" %
1649 (constants.OS_API_FILE, os_dir))
1652 api_versions = utils.ReadFile(api_file).splitlines()
1653 except EnvironmentError, err:
1654 return False, ("Error while reading the API version file at %s: %s" %
1655 (api_file, _ErrnoOrStr(err)))
1658 api_versions = [int(version.strip()) for version in api_versions]
1659 except (TypeError, ValueError), err:
1660 return False, ("API version(s) can't be converted to integer: %s" %
1663 return True, api_versions
1666 def DiagnoseOS(top_dirs=None):
1667 """Compute the validity for all OSes.
1669 @type top_dirs: list
1670 @param top_dirs: the list of directories in which to
1671 search (if not given defaults to
1672 L{constants.OS_SEARCH_PATH})
1673 @rtype: list of L{objects.OS}
1674 @return: a list of tuples (name, path, status, diagnose, variants)
1675 for all (potential) OSes under all search paths, where:
1676 - name is the (potential) OS name
1677 - path is the full path to the OS
1678 - status True/False is the validity of the OS
1679 - diagnose is the error message for an invalid OS, otherwise empty
1680 - variants is a list of supported OS variants, if any
1683 if top_dirs is None:
1684 top_dirs = constants.OS_SEARCH_PATH
1687 for dir_name in top_dirs:
1688 if os.path.isdir(dir_name):
1690 f_names = utils.ListVisibleFiles(dir_name)
1691 except EnvironmentError, err:
1692 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
1694 for name in f_names:
1695 os_path = os.path.sep.join([dir_name, name])
1696 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
1699 variants = os_inst.supported_variants
1703 result.append((name, os_path, status, diagnose, variants))
1708 def _TryOSFromDisk(name, base_dir=None):
1709 """Create an OS instance from disk.
1711 This function will return an OS instance if the given name is a
1714 @type base_dir: string
1715 @keyword base_dir: Base directory containing OS installations.
1716 Defaults to a search in all the OS_SEARCH_PATH dirs.
1718 @return: success and either the OS instance if we find a valid one,
1722 if base_dir is None:
1723 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
1725 return False, "Directory for OS %s not found in search path" % name
1727 os_dir = os.path.sep.join([base_dir, name])
1729 status, api_versions = _OSOndiskAPIVersion(name, os_dir)
1732 return status, api_versions
1734 if not constants.OS_API_VERSIONS.intersection(api_versions):
1735 return False, ("API version mismatch for path '%s': found %s, want %s." %
1736 (os_dir, api_versions, constants.OS_API_VERSIONS))
1738 # OS Files dictionary, we will populate it with the absolute path names
1739 os_files = dict.fromkeys(constants.OS_SCRIPTS)
1741 if max(api_versions) >= constants.OS_API_V15:
1742 os_files[constants.OS_VARIANTS_FILE] = ''
1744 for filename in os_files:
1745 os_files[filename] = os.path.sep.join([os_dir, filename])
1748 st = os.stat(os_files[filename])
1749 except EnvironmentError, err:
1750 return False, ("File '%s' under path '%s' is missing (%s)" %
1751 (filename, os_dir, _ErrnoOrStr(err)))
1753 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
1754 return False, ("File '%s' under path '%s' is not a regular file" %
1757 if filename in constants.OS_SCRIPTS:
1758 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
1759 return False, ("File '%s' under path '%s' is not executable" %
1763 if constants.OS_VARIANTS_FILE in os_files:
1764 variants_file = os_files[constants.OS_VARIANTS_FILE]
1766 variants = utils.ReadFile(variants_file).splitlines()
1767 except EnvironmentError, err:
1768 return False, ("Error while reading the OS variants file at %s: %s" %
1769 (variants_file, _ErrnoOrStr(err)))
1771 return False, ("No supported os variant found")
1773 os_obj = objects.OS(name=name, path=os_dir,
1774 create_script=os_files[constants.OS_SCRIPT_CREATE],
1775 export_script=os_files[constants.OS_SCRIPT_EXPORT],
1776 import_script=os_files[constants.OS_SCRIPT_IMPORT],
1777 rename_script=os_files[constants.OS_SCRIPT_RENAME],
1778 supported_variants=variants,
1779 api_versions=api_versions)
1783 def OSFromDisk(name, base_dir=None):
1784 """Create an OS instance from disk.
1786 This function will return an OS instance if the given name is a
1787 valid OS name. Otherwise, it will raise an appropriate
1788 L{RPCFail} exception, detailing why this is not a valid OS.
1790 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
1791 an exception but returns true/false status data.
1793 @type base_dir: string
1794 @keyword base_dir: Base directory containing OS installations.
1795 Defaults to a search in all the OS_SEARCH_PATH dirs.
1796 @rtype: L{objects.OS}
1797 @return: the OS instance if we find a valid one
1798 @raise RPCFail: if we don't find a valid OS
1801 name_only = name.split("+", 1)[0]
1802 status, payload = _TryOSFromDisk(name_only, base_dir)
1810 def OSEnvironment(instance, inst_os, debug=0):
1811 """Calculate the environment for an os script.
1813 @type instance: L{objects.Instance}
1814 @param instance: target instance for the os script run
1815 @type inst_os: L{objects.OS}
1816 @param inst_os: operating system for which the environment is being built
1817 @type debug: integer
1818 @param debug: debug level (0 or 1, for OS Api 10)
1820 @return: dict of environment variables
1821 @raise errors.BlockDeviceError: if the block device
1827 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
1828 result['OS_API_VERSION'] = '%d' % api_version
1829 result['INSTANCE_NAME'] = instance.name
1830 result['INSTANCE_OS'] = instance.os
1831 result['HYPERVISOR'] = instance.hypervisor
1832 result['DISK_COUNT'] = '%d' % len(instance.disks)
1833 result['NIC_COUNT'] = '%d' % len(instance.nics)
1834 result['DEBUG_LEVEL'] = '%d' % debug
1835 if api_version >= constants.OS_API_V15:
1837 variant = instance.os.split('+', 1)[1]
1839 variant = inst_os.supported_variants[0]
1840 result['OS_VARIANT'] = variant
1841 for idx, disk in enumerate(instance.disks):
1842 real_disk = _RecursiveFindBD(disk)
1843 if real_disk is None:
1844 raise errors.BlockDeviceError("Block device '%s' is not set up" %
1847 result['DISK_%d_PATH' % idx] = real_disk.dev_path
1848 result['DISK_%d_ACCESS' % idx] = disk.mode
1849 if constants.HV_DISK_TYPE in instance.hvparams:
1850 result['DISK_%d_FRONTEND_TYPE' % idx] = \
1851 instance.hvparams[constants.HV_DISK_TYPE]
1852 if disk.dev_type in constants.LDS_BLOCK:
1853 result['DISK_%d_BACKEND_TYPE' % idx] = 'block'
1854 elif disk.dev_type == constants.LD_FILE:
1855 result['DISK_%d_BACKEND_TYPE' % idx] = \
1856 'file:%s' % disk.physical_id[0]
1857 for idx, nic in enumerate(instance.nics):
1858 result['NIC_%d_MAC' % idx] = nic.mac
1860 result['NIC_%d_IP' % idx] = nic.ip
1861 result['NIC_%d_MODE' % idx] = nic.nicparams[constants.NIC_MODE]
1862 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
1863 result['NIC_%d_BRIDGE' % idx] = nic.nicparams[constants.NIC_LINK]
1864 if nic.nicparams[constants.NIC_LINK]:
1865 result['NIC_%d_LINK' % idx] = nic.nicparams[constants.NIC_LINK]
1866 if constants.HV_NIC_TYPE in instance.hvparams:
1867 result['NIC_%d_FRONTEND_TYPE' % idx] = \
1868 instance.hvparams[constants.HV_NIC_TYPE]
1870 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
1871 for key, value in source.items():
1872 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
1876 def BlockdevGrow(disk, amount):
1877 """Grow a stack of block devices.
1879 This function is called recursively, with the childrens being the
1880 first ones to resize.
1882 @type disk: L{objects.Disk}
1883 @param disk: the disk to be grown
1884 @rtype: (status, result)
1885 @return: a tuple with the status of the operation
1886 (True/False), and the errors message if status
1890 r_dev = _RecursiveFindBD(disk)
1892 _Fail("Cannot find block device %s", disk)
1896 except errors.BlockDeviceError, err:
1897 _Fail("Failed to grow block device: %s", err, exc=True)
1900 def BlockdevSnapshot(disk):
1901 """Create a snapshot copy of a block device.
1903 This function is called recursively, and the snapshot is actually created
1904 just for the leaf lvm backend device.
1906 @type disk: L{objects.Disk}
1907 @param disk: the disk to be snapshotted
1909 @return: snapshot disk path
1913 if len(disk.children) == 1:
1914 # only one child, let's recurse on it
1915 return BlockdevSnapshot(disk.children[0])
1917 # more than one child, choose one that matches
1918 for child in disk.children:
1919 if child.size == disk.size:
1920 # return implies breaking the loop
1921 return BlockdevSnapshot(child)
1922 elif disk.dev_type == constants.LD_LV:
1923 r_dev = _RecursiveFindBD(disk)
1924 if r_dev is not None:
1925 # let's stay on the safe side and ask for the full size, for now
1926 return r_dev.Snapshot(disk.size)
1928 _Fail("Cannot find block device %s", disk)
1930 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
1931 disk.unique_id, disk.dev_type)
1934 def ExportSnapshot(disk, dest_node, instance, cluster_name, idx):
1935 """Export a block device snapshot to a remote node.
1937 @type disk: L{objects.Disk}
1938 @param disk: the description of the disk to export
1939 @type dest_node: str
1940 @param dest_node: the destination node to export to
1941 @type instance: L{objects.Instance}
1942 @param instance: the instance object to whom the disk belongs
1943 @type cluster_name: str
1944 @param cluster_name: the cluster name, needed for SSH hostalias
1946 @param idx: the index of the disk in the instance's disk list,
1947 used to export to the OS scripts environment
1951 inst_os = OSFromDisk(instance.os)
1952 export_env = OSEnvironment(instance, inst_os)
1954 export_script = inst_os.export_script
1956 logfile = "%s/exp-%s-%s-%s.log" % (constants.LOG_OS_DIR, inst_os.name,
1957 instance.name, int(time.time()))
1958 if not os.path.exists(constants.LOG_OS_DIR):
1959 os.mkdir(constants.LOG_OS_DIR, 0750)
1960 real_disk = _RecursiveFindBD(disk)
1961 if real_disk is None:
1962 _Fail("Block device '%s' is not set up", disk)
1966 export_env['EXPORT_DEVICE'] = real_disk.dev_path
1967 export_env['EXPORT_INDEX'] = str(idx)
1969 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
1970 destfile = disk.physical_id[1]
1972 # the target command is built out of three individual commands,
1973 # which are joined by pipes; we check each individual command for
1975 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; cd %s; %s 2>%s",
1976 inst_os.path, export_script, logfile)
1980 destcmd = utils.BuildShellCmd("mkdir -p %s && cat > %s/%s",
1981 destdir, destdir, destfile)
1982 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
1983 constants.GANETI_RUNAS,
1986 # all commands have been checked, so we're safe to combine them
1987 command = '|'.join([expcmd, comprcmd, utils.ShellQuoteArgs(remotecmd)])
1989 result = utils.RunCmd(["bash", "-c", command], env=export_env)
1992 _Fail("OS snapshot export command '%s' returned error: %s"
1993 " output: %s", command, result.fail_reason, result.output)
1996 def FinalizeExport(instance, snap_disks):
1997 """Write out the export configuration information.
1999 @type instance: L{objects.Instance}
2000 @param instance: the instance which we export, used for
2001 saving configuration
2002 @type snap_disks: list of L{objects.Disk}
2003 @param snap_disks: list of snapshot block devices, which
2004 will be used to get the actual name of the dump file
2009 destdir = os.path.join(constants.EXPORT_DIR, instance.name + ".new")
2010 finaldestdir = os.path.join(constants.EXPORT_DIR, instance.name)
2012 config = objects.SerializableConfigParser()
2014 config.add_section(constants.INISECT_EXP)
2015 config.set(constants.INISECT_EXP, 'version', '0')
2016 config.set(constants.INISECT_EXP, 'timestamp', '%d' % int(time.time()))
2017 config.set(constants.INISECT_EXP, 'source', instance.primary_node)
2018 config.set(constants.INISECT_EXP, 'os', instance.os)
2019 config.set(constants.INISECT_EXP, 'compression', 'gzip')
2021 config.add_section(constants.INISECT_INS)
2022 config.set(constants.INISECT_INS, 'name', instance.name)
2023 config.set(constants.INISECT_INS, 'memory', '%d' %
2024 instance.beparams[constants.BE_MEMORY])
2025 config.set(constants.INISECT_INS, 'vcpus', '%d' %
2026 instance.beparams[constants.BE_VCPUS])
2027 config.set(constants.INISECT_INS, 'disk_template', instance.disk_template)
2030 for nic_count, nic in enumerate(instance.nics):
2032 config.set(constants.INISECT_INS, 'nic%d_mac' %
2033 nic_count, '%s' % nic.mac)
2034 config.set(constants.INISECT_INS, 'nic%d_ip' % nic_count, '%s' % nic.ip)
2035 config.set(constants.INISECT_INS, 'nic%d_bridge' % nic_count,
2037 # TODO: redundant: on load can read nics until it doesn't exist
2038 config.set(constants.INISECT_INS, 'nic_count' , '%d' % nic_total)
2041 for disk_count, disk in enumerate(snap_disks):
2044 config.set(constants.INISECT_INS, 'disk%d_ivname' % disk_count,
2045 ('%s' % disk.iv_name))
2046 config.set(constants.INISECT_INS, 'disk%d_dump' % disk_count,
2047 ('%s' % disk.physical_id[1]))
2048 config.set(constants.INISECT_INS, 'disk%d_size' % disk_count,
2051 config.set(constants.INISECT_INS, 'disk_count' , '%d' % disk_total)
2053 utils.WriteFile(os.path.join(destdir, constants.EXPORT_CONF_FILE),
2054 data=config.Dumps())
2055 shutil.rmtree(finaldestdir, True)
2056 shutil.move(destdir, finaldestdir)
2059 def ExportInfo(dest):
2060 """Get export configuration information.
2063 @param dest: directory containing the export
2065 @rtype: L{objects.SerializableConfigParser}
2066 @return: a serializable config file containing the
2070 cff = os.path.join(dest, constants.EXPORT_CONF_FILE)
2072 config = objects.SerializableConfigParser()
2075 if (not config.has_section(constants.INISECT_EXP) or
2076 not config.has_section(constants.INISECT_INS)):
2077 _Fail("Export info file doesn't have the required fields")
2079 return config.Dumps()
2082 def ImportOSIntoInstance(instance, src_node, src_images, cluster_name):
2083 """Import an os image into an instance.
2085 @type instance: L{objects.Instance}
2086 @param instance: instance to import the disks into
2087 @type src_node: string
2088 @param src_node: source node for the disk images
2089 @type src_images: list of string
2090 @param src_images: absolute paths of the disk images
2091 @rtype: list of boolean
2092 @return: each boolean represent the success of importing the n-th disk
2095 inst_os = OSFromDisk(instance.os)
2096 import_env = OSEnvironment(instance, inst_os)
2097 import_script = inst_os.import_script
2099 logfile = "%s/import-%s-%s-%s.log" % (constants.LOG_OS_DIR, instance.os,
2100 instance.name, int(time.time()))
2101 if not os.path.exists(constants.LOG_OS_DIR):
2102 os.mkdir(constants.LOG_OS_DIR, 0750)
2105 impcmd = utils.BuildShellCmd("(cd %s; %s >%s 2>&1)", inst_os.path,
2106 import_script, logfile)
2109 for idx, image in enumerate(src_images):
2111 destcmd = utils.BuildShellCmd('cat %s', image)
2112 remotecmd = _GetSshRunner(cluster_name).BuildCmd(src_node,
2113 constants.GANETI_RUNAS,
2115 command = '|'.join([utils.ShellQuoteArgs(remotecmd), comprcmd, impcmd])
2116 import_env['IMPORT_DEVICE'] = import_env['DISK_%d_PATH' % idx]
2117 import_env['IMPORT_INDEX'] = str(idx)
2118 result = utils.RunCmd(command, env=import_env)
2120 logging.error("Disk import command '%s' returned error: %s"
2121 " output: %s", command, result.fail_reason,
2123 final_result.append("error importing disk %d: %s, %s" %
2124 (idx, result.fail_reason, result.output[-100]))
2127 _Fail("; ".join(final_result), log=False)
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 = os.path.join(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
2214 file_storage_dir = os.path.normpath(file_storage_dir)
2215 base_file_storage_dir = cfg.GetFileStorageDir()
2216 if (not os.path.commonprefix([file_storage_dir, base_file_storage_dir]) ==
2217 base_file_storage_dir):
2218 _Fail("File storage directory '%s' is not under base file"
2219 " storage directory '%s'", file_storage_dir, base_file_storage_dir)
2220 return file_storage_dir
2223 def CreateFileStorageDir(file_storage_dir):
2224 """Create file storage directory.
2226 @type file_storage_dir: str
2227 @param file_storage_dir: directory to create
2230 @return: tuple with first element a boolean indicating wheter dir
2231 creation was successful or not
2234 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2235 if os.path.exists(file_storage_dir):
2236 if not os.path.isdir(file_storage_dir):
2237 _Fail("Specified storage dir '%s' is not a directory",
2241 os.makedirs(file_storage_dir, 0750)
2242 except OSError, err:
2243 _Fail("Cannot create file storage directory '%s': %s",
2244 file_storage_dir, err, exc=True)
2247 def RemoveFileStorageDir(file_storage_dir):
2248 """Remove file storage directory.
2250 Remove it only if it's empty. If not log an error and return.
2252 @type file_storage_dir: str
2253 @param file_storage_dir: the directory we should cleanup
2254 @rtype: tuple (success,)
2255 @return: tuple of one element, C{success}, denoting
2256 whether the operation was successful
2259 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2260 if os.path.exists(file_storage_dir):
2261 if not os.path.isdir(file_storage_dir):
2262 _Fail("Specified Storage directory '%s' is not a directory",
2264 # deletes dir only if empty, otherwise we want to fail the rpc call
2266 os.rmdir(file_storage_dir)
2267 except OSError, err:
2268 _Fail("Cannot remove file storage directory '%s': %s",
2269 file_storage_dir, err)
2272 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2273 """Rename the file storage directory.
2275 @type old_file_storage_dir: str
2276 @param old_file_storage_dir: the current path
2277 @type new_file_storage_dir: str
2278 @param new_file_storage_dir: the name we should rename to
2279 @rtype: tuple (success,)
2280 @return: tuple of one element, C{success}, denoting
2281 whether the operation was successful
2284 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2285 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2286 if not os.path.exists(new_file_storage_dir):
2287 if os.path.isdir(old_file_storage_dir):
2289 os.rename(old_file_storage_dir, new_file_storage_dir)
2290 except OSError, err:
2291 _Fail("Cannot rename '%s' to '%s': %s",
2292 old_file_storage_dir, new_file_storage_dir, err)
2294 _Fail("Specified storage dir '%s' is not a directory",
2295 old_file_storage_dir)
2297 if os.path.exists(old_file_storage_dir):
2298 _Fail("Cannot rename '%s' to '%s': both locations exist",
2299 old_file_storage_dir, new_file_storage_dir)
2302 def _EnsureJobQueueFile(file_name):
2303 """Checks whether the given filename is in the queue directory.
2305 @type file_name: str
2306 @param file_name: the file name we should check
2308 @raises RPCFail: if the file is not valid
2311 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2312 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2315 _Fail("Passed job queue file '%s' does not belong to"
2316 " the queue directory '%s'", file_name, queue_dir)
2319 def JobQueueUpdate(file_name, content):
2320 """Updates a file in the queue directory.
2322 This is just a wrapper over L{utils.WriteFile}, with proper
2325 @type file_name: str
2326 @param file_name: the job file name
2328 @param content: the new job contents
2330 @return: the success of the operation
2333 _EnsureJobQueueFile(file_name)
2335 # Write and replace the file atomically
2336 utils.WriteFile(file_name, data=_Decompress(content))
2339 def JobQueueRename(old, new):
2340 """Renames a job queue file.
2342 This is just a wrapper over os.rename with proper checking.
2345 @param old: the old (actual) file name
2347 @param new: the desired file name
2349 @return: the success of the operation and payload
2352 _EnsureJobQueueFile(old)
2353 _EnsureJobQueueFile(new)
2355 utils.RenameFile(old, new, mkdir=True)
2358 def JobQueueSetDrainFlag(drain_flag):
2359 """Set the drain flag for the queue.
2361 This will set or unset the queue drain flag.
2363 @type drain_flag: boolean
2364 @param drain_flag: if True, will set the drain flag, otherwise reset it.
2366 @return: always True, None
2367 @warning: the function always returns True
2371 utils.WriteFile(constants.JOB_QUEUE_DRAIN_FILE, data="", close=True)
2373 utils.RemoveFile(constants.JOB_QUEUE_DRAIN_FILE)
2376 def BlockdevClose(instance_name, disks):
2377 """Closes the given block devices.
2379 This means they will be switched to secondary mode (in case of
2382 @param instance_name: if the argument is not empty, the symlinks
2383 of this instance will be removed
2384 @type disks: list of L{objects.Disk}
2385 @param disks: the list of disks to be closed
2386 @rtype: tuple (success, message)
2387 @return: a tuple of success and message, where success
2388 indicates the succes of the operation, and message
2389 which will contain the error details in case we
2395 rd = _RecursiveFindBD(cf)
2397 _Fail("Can't find device %s", cf)
2404 except errors.BlockDeviceError, err:
2405 msg.append(str(err))
2407 _Fail("Can't make devices secondary: %s", ",".join(msg))
2410 _RemoveBlockDevLinks(instance_name, disks)
2413 def ValidateHVParams(hvname, hvparams):
2414 """Validates the given hypervisor parameters.
2416 @type hvname: string
2417 @param hvname: the hypervisor name
2418 @type hvparams: dict
2419 @param hvparams: the hypervisor parameters to be validated
2424 hv_type = hypervisor.GetHypervisor(hvname)
2425 hv_type.ValidateParameters(hvparams)
2426 except errors.HypervisorError, err:
2427 _Fail(str(err), log=False)
2431 """Demotes the current node from master candidate role.
2434 # try to ensure we're not the master by mistake
2435 master, myself = ssconf.GetMasterAndMyself()
2436 if master == myself:
2437 _Fail("ssconf status shows I'm the master node, will not demote")
2438 pid_file = utils.DaemonPidFileName(constants.MASTERD)
2439 if utils.IsProcessAlive(utils.ReadPidFile(pid_file)):
2440 _Fail("The master daemon is running, will not demote")
2442 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2443 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2444 except EnvironmentError, err:
2445 if err.errno != errno.ENOENT:
2446 _Fail("Error while backing up cluster file: %s", err, exc=True)
2447 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2450 def _FindDisks(nodes_ip, disks):
2451 """Sets the physical ID on disks and returns the block devices.
2454 # set the correct physical ID
2455 my_name = utils.HostInfo().name
2457 cf.SetPhysicalID(my_name, nodes_ip)
2462 rd = _RecursiveFindBD(cf)
2464 _Fail("Can't find device %s", cf)
2469 def DrbdDisconnectNet(nodes_ip, disks):
2470 """Disconnects the network on a list of drbd devices.
2473 bdevs = _FindDisks(nodes_ip, disks)
2479 except errors.BlockDeviceError, err:
2480 _Fail("Can't change network configuration to standalone mode: %s",
2484 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
2485 """Attaches the network on a list of drbd devices.
2488 bdevs = _FindDisks(nodes_ip, disks)
2491 for idx, rd in enumerate(bdevs):
2493 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
2494 except EnvironmentError, err:
2495 _Fail("Can't create symlink: %s", err)
2496 # reconnect disks, switch to new master configuration and if
2497 # needed primary mode
2500 rd.AttachNet(multimaster)
2501 except errors.BlockDeviceError, err:
2502 _Fail("Can't change network configuration: %s", err)
2504 # wait until the disks are connected; we need to retry the re-attach
2505 # if the device becomes standalone, as this might happen if the one
2506 # node disconnects and reconnects in a different mode before the
2507 # other node reconnects; in this case, one or both of the nodes will
2508 # decide it has wrong configuration and switch to standalone
2511 all_connected = True
2514 stats = rd.GetProcStatus()
2516 all_connected = (all_connected and
2517 (stats.is_connected or stats.is_in_resync))
2519 if stats.is_standalone:
2520 # peer had different config info and this node became
2521 # standalone, even though this should not happen with the
2522 # new staged way of changing disk configs
2524 rd.AttachNet(multimaster)
2525 except errors.BlockDeviceError, err:
2526 _Fail("Can't change network configuration: %s", err)
2528 if not all_connected:
2529 raise utils.RetryAgain()
2532 # Start with a delay of 100 miliseconds and go up to 5 seconds
2533 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
2534 except utils.RetryTimeout:
2535 _Fail("Timeout in disk reconnecting")
2538 # change to primary mode
2542 except errors.BlockDeviceError, err:
2543 _Fail("Can't change to primary mode: %s", err)
2546 def DrbdWaitSync(nodes_ip, disks):
2547 """Wait until DRBDs have synchronized.
2551 stats = rd.GetProcStatus()
2552 if not (stats.is_connected or stats.is_in_resync):
2553 raise utils.RetryAgain()
2556 bdevs = _FindDisks(nodes_ip, disks)
2562 # poll each second for 15 seconds
2563 stats = utils.Retry(_helper, 1, 15, args=[rd])
2564 except utils.RetryTimeout:
2565 stats = rd.GetProcStatus()
2567 if not (stats.is_connected or stats.is_in_resync):
2568 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
2569 alldone = alldone and (not stats.is_in_resync)
2570 if stats.sync_percent is not None:
2571 min_resync = min(min_resync, stats.sync_percent)
2573 return (alldone, min_resync)
2576 def PowercycleNode(hypervisor_type):
2577 """Hard-powercycle the node.
2579 Because we need to return first, and schedule the powercycle in the
2580 background, we won't be able to report failures nicely.
2583 hyper = hypervisor.GetHypervisor(hypervisor_type)
2587 # if we can't fork, we'll pretend that we're in the child process
2590 return "Reboot scheduled in 5 seconds"
2592 hyper.PowercycleNode()
2595 class HooksRunner(object):
2598 This class is instantiated on the node side (ganeti-noded) and not
2602 RE_MASK = re.compile("^[a-zA-Z0-9_-]+$")
2604 def __init__(self, hooks_base_dir=None):
2605 """Constructor for hooks runner.
2607 @type hooks_base_dir: str or None
2608 @param hooks_base_dir: if not None, this overrides the
2609 L{constants.HOOKS_BASE_DIR} (useful for unittests)
2612 if hooks_base_dir is None:
2613 hooks_base_dir = constants.HOOKS_BASE_DIR
2614 self._BASE_DIR = hooks_base_dir
2617 def ExecHook(script, env):
2618 """Exec one hook script.
2621 @param script: the full path to the script
2623 @param env: the environment with which to exec the script
2624 @rtype: tuple (success, message)
2625 @return: a tuple of success and message, where success
2626 indicates the succes of the operation, and message
2627 which will contain the error details in case we
2631 # exec the process using subprocess and log the output
2634 fdstdin = open("/dev/null", "r")
2635 child = subprocess.Popen([script], stdin=fdstdin, stdout=subprocess.PIPE,
2636 stderr=subprocess.STDOUT, close_fds=True,
2637 shell=False, cwd="/", env=env)
2640 output = child.stdout.read(4096)
2641 child.stdout.close()
2642 except EnvironmentError, err:
2643 output += "Hook script error: %s" % str(err)
2647 result = child.wait()
2649 except EnvironmentError, err:
2650 if err.errno == errno.EINTR:
2654 # try not to leak fds
2655 for fd in (fdstdin, ):
2659 except EnvironmentError, err:
2660 # just log the error
2661 #logging.exception("Error while closing fd %s", fd)
2664 return result == 0, utils.SafeEncode(output.strip())
2666 def RunHooks(self, hpath, phase, env):
2667 """Run the scripts in the hooks directory.
2670 @param hpath: the path to the hooks directory which
2673 @param phase: either L{constants.HOOKS_PHASE_PRE} or
2674 L{constants.HOOKS_PHASE_POST}
2676 @param env: dictionary with the environment for the hook
2678 @return: list of 3-element tuples:
2680 - script result, either L{constants.HKR_SUCCESS} or
2681 L{constants.HKR_FAIL}
2682 - output of the script
2684 @raise errors.ProgrammerError: for invalid input
2688 if phase == constants.HOOKS_PHASE_PRE:
2690 elif phase == constants.HOOKS_PHASE_POST:
2693 _Fail("Unknown hooks phase '%s'", phase)
2697 subdir = "%s-%s.d" % (hpath, suffix)
2698 dir_name = "%s/%s" % (self._BASE_DIR, subdir)
2700 dir_contents = utils.ListVisibleFiles(dir_name)
2702 # FIXME: must log output in case of failures
2705 # we use the standard python sort order,
2706 # so 00name is the recommended naming scheme
2708 for relname in dir_contents:
2709 fname = os.path.join(dir_name, relname)
2710 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
2711 self.RE_MASK.match(relname) is not None):
2712 rrval = constants.HKR_SKIP
2715 result, output = self.ExecHook(fname, env)
2717 rrval = constants.HKR_FAIL
2719 rrval = constants.HKR_SUCCESS
2720 rr.append(("%s/%s" % (subdir, relname), rrval, output))
2725 class IAllocatorRunner(object):
2726 """IAllocator runner.
2728 This class is instantiated on the node side (ganeti-noded) and not on
2732 def Run(self, name, idata):
2733 """Run an iallocator script.
2736 @param name: the iallocator script name
2738 @param idata: the allocator input data
2741 @return: two element tuple of:
2743 - either error message or stdout of allocator (for success)
2746 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
2748 if alloc_script is None:
2749 _Fail("iallocator module '%s' not found in the search path", name)
2751 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
2755 result = utils.RunCmd([alloc_script, fin_name])
2757 _Fail("iallocator module '%s' failed: %s, output '%s'",
2758 name, result.fail_reason, result.output)
2762 return result.stdout
2765 class DevCacheManager(object):
2766 """Simple class for managing a cache of block device information.
2769 _DEV_PREFIX = "/dev/"
2770 _ROOT_DIR = constants.BDEV_CACHE_DIR
2773 def _ConvertPath(cls, dev_path):
2774 """Converts a /dev/name path to the cache file name.
2776 This replaces slashes with underscores and strips the /dev
2777 prefix. It then returns the full path to the cache file.
2780 @param dev_path: the C{/dev/} path name
2782 @return: the converted path name
2785 if dev_path.startswith(cls._DEV_PREFIX):
2786 dev_path = dev_path[len(cls._DEV_PREFIX):]
2787 dev_path = dev_path.replace("/", "_")
2788 fpath = "%s/bdev_%s" % (cls._ROOT_DIR, dev_path)
2792 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
2793 """Updates the cache information for a given device.
2796 @param dev_path: the pathname of the device
2798 @param owner: the owner (instance name) of the device
2799 @type on_primary: bool
2800 @param on_primary: whether this is the primary
2803 @param iv_name: the instance-visible name of the
2804 device, as in objects.Disk.iv_name
2809 if dev_path is None:
2810 logging.error("DevCacheManager.UpdateCache got a None dev_path")
2812 fpath = cls._ConvertPath(dev_path)
2818 iv_name = "not_visible"
2819 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
2821 utils.WriteFile(fpath, data=fdata)
2822 except EnvironmentError, err:
2823 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
2826 def RemoveCache(cls, dev_path):
2827 """Remove data for a dev_path.
2829 This is just a wrapper over L{utils.RemoveFile} with a converted
2830 path name and logging.
2833 @param dev_path: the pathname of the device
2838 if dev_path is None:
2839 logging.error("DevCacheManager.RemoveCache got a None dev_path")
2841 fpath = cls._ConvertPath(dev_path)
2843 utils.RemoveFile(fpath)
2844 except EnvironmentError, err:
2845 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)