4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Functions used by the node daemon
24 @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
25 the L{UploadFile} function
26 @var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted
27 in the L{_CleanDirectory} function
31 # pylint: disable=E1103
33 # E1103: %s %r has no %r member (but some types could not be
34 # inferred), because the _TryOSFromDisk returns either (True, os_obj)
35 # or (False, "string") which confuses pylint
52 from ganeti import errors
53 from ganeti import utils
54 from ganeti import ssh
55 from ganeti import hypervisor
56 from ganeti import constants
57 from ganeti import bdev
58 from ganeti import objects
59 from ganeti import ssconf
60 from ganeti import serializer
61 from ganeti import netutils
62 from ganeti import runtime
63 from ganeti import mcpu
64 from ganeti import compat
67 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
68 _ALLOWED_CLEAN_DIRS = frozenset([
70 constants.JOB_QUEUE_ARCHIVE_DIR,
72 constants.CRYPTO_KEYS_DIR,
74 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
75 _X509_KEY_FILE = "key"
76 _X509_CERT_FILE = "cert"
77 _IES_STATUS_FILE = "status"
81 #: Valid LVS output line regex
82 _LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6})\|?$")
84 # Actions for the master setup script
85 _MASTER_START = "start"
89 class RPCFail(Exception):
90 """Class denoting RPC failure.
92 Its argument is the error message.
97 def _Fail(msg, *args, **kwargs):
98 """Log an error and the raise an RPCFail exception.
100 This exception is then handled specially in the ganeti daemon and
101 turned into a 'failed' return type. As such, this function is a
102 useful shortcut for logging the error and returning it to the master
106 @param msg: the text of the exception
112 if "log" not in kwargs or kwargs["log"]: # if we should log this error
113 if "exc" in kwargs and kwargs["exc"]:
114 logging.exception(msg)
121 """Simple wrapper to return a SimpleStore.
123 @rtype: L{ssconf.SimpleStore}
124 @return: a SimpleStore instance
127 return ssconf.SimpleStore()
130 def _GetSshRunner(cluster_name):
131 """Simple wrapper to return an SshRunner.
133 @type cluster_name: str
134 @param cluster_name: the cluster name, which is needed
135 by the SshRunner constructor
136 @rtype: L{ssh.SshRunner}
137 @return: an SshRunner instance
140 return ssh.SshRunner(cluster_name)
143 def _Decompress(data):
144 """Unpacks data compressed by the RPC client.
146 @type data: list or tuple
147 @param data: Data sent by RPC client
149 @return: Decompressed data
152 assert isinstance(data, (list, tuple))
153 assert len(data) == 2
154 (encoding, content) = data
155 if encoding == constants.RPC_ENCODING_NONE:
157 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
158 return zlib.decompress(base64.b64decode(content))
160 raise AssertionError("Unknown data encoding")
163 def _CleanDirectory(path, exclude=None):
164 """Removes all regular files in a directory.
167 @param path: the directory to clean
169 @param exclude: list of files to be excluded, defaults
173 if path not in _ALLOWED_CLEAN_DIRS:
174 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
177 if not os.path.isdir(path):
182 # Normalize excluded paths
183 exclude = [os.path.normpath(i) for i in exclude]
185 for rel_name in utils.ListVisibleFiles(path):
186 full_name = utils.PathJoin(path, rel_name)
187 if full_name in exclude:
189 if os.path.isfile(full_name) and not os.path.islink(full_name):
190 utils.RemoveFile(full_name)
193 def _BuildUploadFileList():
194 """Build the list of allowed upload files.
196 This is abstracted so that it's built only once at module import time.
199 allowed_files = set([
200 constants.CLUSTER_CONF_FILE,
202 constants.SSH_KNOWN_HOSTS_FILE,
203 constants.VNC_PASSWORD_FILE,
204 constants.RAPI_CERT_FILE,
205 constants.SPICE_CERT_FILE,
206 constants.SPICE_CACERT_FILE,
207 constants.RAPI_USERS_FILE,
208 constants.CONFD_HMAC_KEY,
209 constants.CLUSTER_DOMAIN_SECRET_FILE,
212 for hv_name in constants.HYPER_TYPES:
213 hv_class = hypervisor.GetHypervisorClass(hv_name)
214 allowed_files.update(hv_class.GetAncillaryFiles()[0])
216 return frozenset(allowed_files)
219 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
223 """Removes job queue files and archived jobs.
229 _CleanDirectory(constants.QUEUE_DIR, exclude=[constants.JOB_QUEUE_LOCK_FILE])
230 _CleanDirectory(constants.JOB_QUEUE_ARCHIVE_DIR)
234 """Returns master information.
236 This is an utility function to compute master information, either
237 for consumption here or from the node daemon.
240 @return: master_netdev, master_ip, master_name, primary_ip_family,
242 @raise RPCFail: in case of errors
247 master_netdev = cfg.GetMasterNetdev()
248 master_ip = cfg.GetMasterIP()
249 master_netmask = cfg.GetMasterNetmask()
250 master_node = cfg.GetMasterNode()
251 primary_ip_family = cfg.GetPrimaryIPFamily()
252 except errors.ConfigurationError, err:
253 _Fail("Cluster configuration incomplete: %s", err, exc=True)
254 return (master_netdev, master_ip, master_node, primary_ip_family,
258 def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
259 """Decorator that runs hooks before and after the decorated function.
261 @type hook_opcode: string
262 @param hook_opcode: opcode of the hook
263 @type hooks_path: string
264 @param hooks_path: path of the hooks
265 @type env_builder_fn: function
266 @param env_builder_fn: function that returns a dictionary containing the
267 environment variables for the hooks. Will get all the parameters of the
269 @raise RPCFail: in case of pre-hook failure
273 def wrapper(*args, **kwargs):
274 _, myself = ssconf.GetMasterAndMyself()
275 nodes = ([myself], [myself]) # these hooks run locally
277 env_fn = compat.partial(env_builder_fn, *args, **kwargs)
281 hm = mcpu.HooksMaster(hook_opcode, hooks_path, nodes, hr.RunLocalHooks,
282 None, env_fn, logging.warning, cfg.GetClusterName(),
285 hm.RunPhase(constants.HOOKS_PHASE_PRE)
286 result = fn(*args, **kwargs)
287 hm.RunPhase(constants.HOOKS_PHASE_POST)
294 def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
295 """Builds environment variables for master IP hooks.
297 @type master_params: L{objects.MasterNetworkParameters}
298 @param master_params: network parameters of the master
299 @type use_external_mip_script: boolean
300 @param use_external_mip_script: whether to use an external master IP
301 address setup script (unused, but necessary per the implementation of the
302 _RunLocalHooks decorator)
305 # pylint: disable=W0613
306 ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
308 "MASTER_NETDEV": master_params.netdev,
309 "MASTER_IP": master_params.ip,
310 "MASTER_NETMASK": str(master_params.netmask),
311 "CLUSTER_IP_VERSION": str(ver),
317 def _RunMasterSetupScript(master_params, action, use_external_mip_script):
318 """Execute the master IP address setup script.
320 @type master_params: L{objects.MasterNetworkParameters}
321 @param master_params: network parameters of the master
323 @param action: action to pass to the script. Must be one of
324 L{backend._MASTER_START} or L{backend._MASTER_STOP}
325 @type use_external_mip_script: boolean
326 @param use_external_mip_script: whether to use an external master IP
328 @raise backend.RPCFail: if there are errors during the execution of the
332 env = _BuildMasterIpEnv(master_params)
334 if use_external_mip_script:
335 setup_script = constants.EXTERNAL_MASTER_SETUP_SCRIPT
337 setup_script = constants.DEFAULT_MASTER_SETUP_SCRIPT
339 result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
342 _Fail("Failed to %s the master IP. Script return value: %s" %
343 (action, result.exit_code), log=True)
346 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
348 def ActivateMasterIp(master_params, use_external_mip_script):
349 """Activate the IP address of the master daemon.
351 @type master_params: L{objects.MasterNetworkParameters}
352 @param master_params: network parameters of the master
353 @type use_external_mip_script: boolean
354 @param use_external_mip_script: whether to use an external master IP
356 @raise RPCFail: in case of errors during the IP startup
359 _RunMasterSetupScript(master_params, _MASTER_START,
360 use_external_mip_script)
363 def StartMasterDaemons(no_voting):
364 """Activate local node as master node.
366 The function will start the master daemons (ganeti-masterd and ganeti-rapi).
368 @type no_voting: boolean
369 @param no_voting: whether to start ganeti-masterd without a node vote
370 but still non-interactively
376 masterd_args = "--no-voting --yes-do-it"
381 "EXTRA_MASTERD_ARGS": masterd_args,
384 result = utils.RunCmd([constants.DAEMON_UTIL, "start-master"], env=env)
386 msg = "Can't start Ganeti master: %s" % result.output
391 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
393 def DeactivateMasterIp(master_params, use_external_mip_script):
394 """Deactivate the master IP on this node.
396 @type master_params: L{objects.MasterNetworkParameters}
397 @param master_params: network parameters of the master
398 @type use_external_mip_script: boolean
399 @param use_external_mip_script: whether to use an external master IP
401 @raise RPCFail: in case of errors during the IP turndown
404 _RunMasterSetupScript(master_params, _MASTER_STOP,
405 use_external_mip_script)
408 def StopMasterDaemons():
409 """Stop the master daemons on this node.
411 Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
416 # TODO: log and report back to the caller the error failures; we
417 # need to decide in which case we fail the RPC for this
419 result = utils.RunCmd([constants.DAEMON_UTIL, "stop-master"])
421 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
423 result.cmd, result.exit_code, result.output)
426 def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
427 """Change the netmask of the master IP.
429 @param old_netmask: the old value of the netmask
430 @param netmask: the new value of the netmask
431 @param master_ip: the master IP
432 @param master_netdev: the master network device
435 if old_netmask == netmask:
438 if not netutils.IPAddress.Own(master_ip):
439 _Fail("The master IP address is not up, not attempting to change its"
442 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
443 "%s/%s" % (master_ip, netmask),
444 "dev", master_netdev, "label",
445 "%s:0" % master_netdev])
447 _Fail("Could not set the new netmask on the master IP address")
449 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
450 "%s/%s" % (master_ip, old_netmask),
451 "dev", master_netdev, "label",
452 "%s:0" % master_netdev])
454 _Fail("Could not bring down the master IP address with the old netmask")
457 def EtcHostsModify(mode, host, ip):
458 """Modify a host entry in /etc/hosts.
460 @param mode: The mode to operate. Either add or remove entry
461 @param host: The host to operate on
462 @param ip: The ip associated with the entry
465 if mode == constants.ETC_HOSTS_ADD:
467 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
469 utils.AddHostToEtcHosts(host, ip)
470 elif mode == constants.ETC_HOSTS_REMOVE:
472 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
473 " parameter is present")
474 utils.RemoveHostFromEtcHosts(host)
476 RPCFail("Mode not supported")
479 def LeaveCluster(modify_ssh_setup):
480 """Cleans up and remove the current node.
482 This function cleans up and prepares the current node to be removed
485 If processing is successful, then it raises an
486 L{errors.QuitGanetiException} which is used as a special case to
487 shutdown the node daemon.
489 @param modify_ssh_setup: boolean
492 _CleanDirectory(constants.DATA_DIR)
493 _CleanDirectory(constants.CRYPTO_KEYS_DIR)
498 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
500 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
502 utils.RemoveFile(priv_key)
503 utils.RemoveFile(pub_key)
504 except errors.OpExecError:
505 logging.exception("Error while processing ssh files")
508 utils.RemoveFile(constants.CONFD_HMAC_KEY)
509 utils.RemoveFile(constants.RAPI_CERT_FILE)
510 utils.RemoveFile(constants.SPICE_CERT_FILE)
511 utils.RemoveFile(constants.SPICE_CACERT_FILE)
512 utils.RemoveFile(constants.NODED_CERT_FILE)
513 except: # pylint: disable=W0702
514 logging.exception("Error while removing cluster secrets")
516 result = utils.RunCmd([constants.DAEMON_UTIL, "stop", constants.CONFD])
518 logging.error("Command %s failed with exitcode %s and error %s",
519 result.cmd, result.exit_code, result.output)
521 # Raise a custom exception (handled in ganeti-noded)
522 raise errors.QuitGanetiException(True, "Shutdown scheduled")
525 def _GetVgInfo(name):
526 """Retrieves information about a LVM volume group.
529 # TODO: GetVGInfo supports returning information for multiple VGs at once
530 vginfo = bdev.LogicalVolume.GetVGInfo([name])
532 vg_free = int(round(vginfo[0][0], 0))
533 vg_size = int(round(vginfo[0][1], 0))
545 def _GetHvInfo(name):
546 """Retrieves node information from a hypervisor.
548 The information returned depends on the hypervisor. Common items:
550 - vg_size is the size of the configured volume group in MiB
551 - vg_free is the free size of the volume group in MiB
552 - memory_dom0 is the memory allocated for domain0 in MiB
553 - memory_free is the currently available (free) ram in MiB
554 - memory_total is the total number of ram in MiB
555 - hv_version: the hypervisor version, if available
558 return hypervisor.GetHypervisor(name).GetNodeInfo()
561 def _GetNamedNodeInfo(names, fn):
562 """Calls C{fn} for all names in C{names} and returns a dictionary.
570 return map(fn, names)
573 def GetNodeInfo(vg_names, hv_names):
574 """Gives back a hash with different information about the node.
576 @type vg_names: list of string
577 @param vg_names: Names of the volume groups to ask for disk space information
578 @type hv_names: list of string
579 @param hv_names: Names of the hypervisors to ask for node information
580 @rtype: tuple; (string, None/dict, None/dict)
581 @return: Tuple containing boot ID, volume group information and hypervisor
585 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
586 vg_info = _GetNamedNodeInfo(vg_names, _GetVgInfo)
587 hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo)
589 return (bootid, vg_info, hv_info)
592 def VerifyNode(what, cluster_name):
593 """Verify the status of the local node.
595 Based on the input L{what} parameter, various checks are done on the
598 If the I{filelist} key is present, this list of
599 files is checksummed and the file/checksum pairs are returned.
601 If the I{nodelist} key is present, we check that we have
602 connectivity via ssh with the target nodes (and check the hostname
605 If the I{node-net-test} key is present, we check that we have
606 connectivity to the given nodes via both primary IP and, if
607 applicable, secondary IPs.
610 @param what: a dictionary of things to check:
611 - filelist: list of files for which to compute checksums
612 - nodelist: list of nodes we should check ssh communication with
613 - node-net-test: list of nodes we should check node daemon port
615 - hypervisor: list with hypervisors to run the verify for
617 @return: a dictionary with the same keys as the input dict, and
618 values representing the result of the checks
622 my_name = netutils.Hostname.GetSysName()
623 port = netutils.GetDaemonPort(constants.NODED)
624 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
626 if constants.NV_HYPERVISOR in what and vm_capable:
627 result[constants.NV_HYPERVISOR] = tmp = {}
628 for hv_name in what[constants.NV_HYPERVISOR]:
630 val = hypervisor.GetHypervisor(hv_name).Verify()
631 except errors.HypervisorError, err:
632 val = "Error while checking hypervisor: %s" % str(err)
635 if constants.NV_HVPARAMS in what and vm_capable:
636 result[constants.NV_HVPARAMS] = tmp = []
637 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
639 logging.info("Validating hv %s, %s", hv_name, hvparms)
640 hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
641 except errors.HypervisorError, err:
642 tmp.append((source, hv_name, str(err)))
644 if constants.NV_FILELIST in what:
645 result[constants.NV_FILELIST] = utils.FingerprintFiles(
646 what[constants.NV_FILELIST])
648 if constants.NV_NODELIST in what:
649 (nodes, bynode) = what[constants.NV_NODELIST]
651 # Add nodes from other groups (different for each node)
653 nodes.extend(bynode[my_name])
658 random.shuffle(nodes)
660 # Try to contact all nodes
663 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
667 result[constants.NV_NODELIST] = val
669 if constants.NV_NODENETTEST in what:
670 result[constants.NV_NODENETTEST] = tmp = {}
671 my_pip = my_sip = None
672 for name, pip, sip in what[constants.NV_NODENETTEST]:
678 tmp[my_name] = ("Can't find my own primary/secondary IP"
681 for name, pip, sip in what[constants.NV_NODENETTEST]:
683 if not netutils.TcpPing(pip, port, source=my_pip):
684 fail.append("primary")
686 if not netutils.TcpPing(sip, port, source=my_sip):
687 fail.append("secondary")
689 tmp[name] = ("failure using the %s interface(s)" %
692 if constants.NV_MASTERIP in what:
693 # FIXME: add checks on incoming data structures (here and in the
694 # rest of the function)
695 master_name, master_ip = what[constants.NV_MASTERIP]
696 if master_name == my_name:
697 source = constants.IP4_ADDRESS_LOCALHOST
700 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
703 if constants.NV_USERSCRIPTS in what:
704 result[constants.NV_USERSCRIPTS] = \
705 [script for script in what[constants.NV_USERSCRIPTS]
706 if not (os.path.exists(script) and os.access(script, os.X_OK))]
708 if constants.NV_OOB_PATHS in what:
709 result[constants.NV_OOB_PATHS] = tmp = []
710 for path in what[constants.NV_OOB_PATHS]:
714 tmp.append("error stating out of band helper: %s" % err)
716 if stat.S_ISREG(st.st_mode):
717 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
720 tmp.append("out of band helper %s is not executable" % path)
722 tmp.append("out of band helper %s is not a file" % path)
724 if constants.NV_LVLIST in what and vm_capable:
726 val = GetVolumeList(utils.ListVolumeGroups().keys())
729 result[constants.NV_LVLIST] = val
731 if constants.NV_INSTANCELIST in what and vm_capable:
732 # GetInstanceList can fail
734 val = GetInstanceList(what[constants.NV_INSTANCELIST])
737 result[constants.NV_INSTANCELIST] = val
739 if constants.NV_VGLIST in what and vm_capable:
740 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
742 if constants.NV_PVLIST in what and vm_capable:
743 result[constants.NV_PVLIST] = \
744 bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
745 filter_allocatable=False)
747 if constants.NV_VERSION in what:
748 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
749 constants.RELEASE_VERSION)
751 if constants.NV_HVINFO in what and vm_capable:
752 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
753 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
755 if constants.NV_DRBDLIST in what and vm_capable:
757 used_minors = bdev.DRBD8.GetUsedDevs().keys()
758 except errors.BlockDeviceError, err:
759 logging.warning("Can't get used minors list", exc_info=True)
760 used_minors = str(err)
761 result[constants.NV_DRBDLIST] = used_minors
763 if constants.NV_DRBDHELPER in what and vm_capable:
766 payload = bdev.BaseDRBD.GetUsermodeHelper()
767 except errors.BlockDeviceError, err:
768 logging.error("Can't get DRBD usermode helper: %s", str(err))
771 result[constants.NV_DRBDHELPER] = (status, payload)
773 if constants.NV_NODESETUP in what:
774 result[constants.NV_NODESETUP] = tmpr = []
775 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
776 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
777 " under /sys, missing required directories /sys/block"
778 " and /sys/class/net")
779 if (not os.path.isdir("/proc/sys") or
780 not os.path.isfile("/proc/sysrq-trigger")):
781 tmpr.append("The procfs filesystem doesn't seem to be mounted"
782 " under /proc, missing required directory /proc/sys and"
783 " the file /proc/sysrq-trigger")
785 if constants.NV_TIME in what:
786 result[constants.NV_TIME] = utils.SplitTime(time.time())
788 if constants.NV_OSLIST in what and vm_capable:
789 result[constants.NV_OSLIST] = DiagnoseOS()
791 if constants.NV_BRIDGES in what and vm_capable:
792 result[constants.NV_BRIDGES] = [bridge
793 for bridge in what[constants.NV_BRIDGES]
794 if not utils.BridgeExists(bridge)]
798 def GetBlockDevSizes(devices):
799 """Return the size of the given block devices
802 @param devices: list of block device nodes to query
805 dictionary of all block devices under /dev (key). The value is their
808 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
814 for devpath in devices:
815 if not utils.IsBelowDir(DEV_PREFIX, devpath):
819 st = os.stat(devpath)
820 except EnvironmentError, err:
821 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
824 if stat.S_ISBLK(st.st_mode):
825 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
827 # We don't want to fail, just do not list this device as available
828 logging.warning("Cannot get size for block device %s", devpath)
831 size = int(result.stdout) / (1024 * 1024)
832 blockdevs[devpath] = size
836 def GetVolumeList(vg_names):
837 """Compute list of logical volumes and their size.
840 @param vg_names: the volume groups whose LVs we should list, or
841 empty for all volume groups
844 dictionary of all partions (key) with value being a tuple of
845 their size (in MiB), inactive and online status::
847 {'xenvg/test1': ('20.06', True, True)}
849 in case of errors, a string is returned with the error
857 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
858 "--separator=%s" % sep,
859 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
861 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
863 for line in result.stdout.splitlines():
865 match = _LVSLINE_REGEX.match(line)
867 logging.error("Invalid line returned from lvs output: '%s'", line)
869 vg_name, name, size, attr = match.groups()
870 inactive = attr[4] == "-"
871 online = attr[5] == "o"
872 virtual = attr[0] == "v"
874 # we don't want to report such volumes as existing, since they
875 # don't really hold data
877 lvs[vg_name + "/" + name] = (size, inactive, online)
882 def ListVolumeGroups():
883 """List the volume groups and their size.
886 @return: dictionary with keys volume name and values the
890 return utils.ListVolumeGroups()
894 """List all volumes on this node.
898 A list of dictionaries, each having four keys:
899 - name: the logical volume name,
900 - size: the size of the logical volume
901 - dev: the physical device on which the LV lives
902 - vg: the volume group to which it belongs
904 In case of errors, we return an empty list and log the
907 Note that since a logical volume can live on multiple physical
908 volumes, the resulting list might include a logical volume
912 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
914 "--options=lv_name,lv_size,devices,vg_name"])
916 _Fail("Failed to list logical volumes, lvs output: %s",
920 return dev.split("(")[0]
923 return [parse_dev(x) for x in dev.split(",")]
926 line = [v.strip() for v in line]
927 return [{"name": line[0], "size": line[1],
928 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
931 for line in result.stdout.splitlines():
932 if line.count("|") >= 3:
933 all_devs.extend(map_line(line.split("|")))
935 logging.warning("Strange line in the output from lvs: '%s'", line)
939 def BridgesExist(bridges_list):
940 """Check if a list of bridges exist on the current node.
943 @return: C{True} if all of them exist, C{False} otherwise
947 for bridge in bridges_list:
948 if not utils.BridgeExists(bridge):
949 missing.append(bridge)
952 _Fail("Missing bridges %s", utils.CommaJoin(missing))
955 def GetInstanceList(hypervisor_list):
956 """Provides a list of instances.
958 @type hypervisor_list: list
959 @param hypervisor_list: the list of hypervisors to query information
962 @return: a list of all running instances on the current node
963 - instance1.example.com
964 - instance2.example.com
968 for hname in hypervisor_list:
970 names = hypervisor.GetHypervisor(hname).ListInstances()
971 results.extend(names)
972 except errors.HypervisorError, err:
973 _Fail("Error enumerating instances (hypervisor %s): %s",
974 hname, err, exc=True)
979 def GetInstanceInfo(instance, hname):
980 """Gives back the information about an instance as a dictionary.
982 @type instance: string
983 @param instance: the instance name
985 @param hname: the hypervisor type of the instance
988 @return: dictionary with the following keys:
989 - memory: memory size of instance (int)
990 - state: xen state of instance (string)
991 - time: cpu time of instance (float)
996 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
997 if iinfo is not None:
998 output["memory"] = iinfo[2]
999 output["state"] = iinfo[4]
1000 output["time"] = iinfo[5]
1005 def GetInstanceMigratable(instance):
1006 """Gives whether an instance can be migrated.
1008 @type instance: L{objects.Instance}
1009 @param instance: object representing the instance to be checked.
1012 @return: tuple of (result, description) where:
1013 - result: whether the instance can be migrated or not
1014 - description: a description of the issue, if relevant
1017 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1018 iname = instance.name
1019 if iname not in hyper.ListInstances():
1020 _Fail("Instance %s is not running", iname)
1022 for idx in range(len(instance.disks)):
1023 link_name = _GetBlockDevSymlinkPath(iname, idx)
1024 if not os.path.islink(link_name):
1025 logging.warning("Instance %s is missing symlink %s for disk %d",
1026 iname, link_name, idx)
1029 def GetAllInstancesInfo(hypervisor_list):
1030 """Gather data about all instances.
1032 This is the equivalent of L{GetInstanceInfo}, except that it
1033 computes data for all instances at once, thus being faster if one
1034 needs data about more than one instance.
1036 @type hypervisor_list: list
1037 @param hypervisor_list: list of hypervisors to query for instance data
1040 @return: dictionary of instance: data, with data having the following keys:
1041 - memory: memory size of instance (int)
1042 - state: xen state of instance (string)
1043 - time: cpu time of instance (float)
1044 - vcpus: the number of vcpus
1049 for hname in hypervisor_list:
1050 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
1052 for name, _, memory, vcpus, state, times in iinfo:
1060 # we only check static parameters, like memory and vcpus,
1061 # and not state and time which can change between the
1062 # invocations of the different hypervisors
1063 for key in "memory", "vcpus":
1064 if value[key] != output[name][key]:
1065 _Fail("Instance %s is running twice"
1066 " with different parameters", name)
1067 output[name] = value
1072 def _InstanceLogName(kind, os_name, instance, component):
1073 """Compute the OS log filename for a given instance and operation.
1075 The instance name and os name are passed in as strings since not all
1076 operations have these as part of an instance object.
1079 @param kind: the operation type (e.g. add, import, etc.)
1080 @type os_name: string
1081 @param os_name: the os name
1082 @type instance: string
1083 @param instance: the name of the instance being imported/added/etc.
1084 @type component: string or None
1085 @param component: the name of the component of the instance being
1089 # TODO: Use tempfile.mkstemp to create unique filename
1091 assert "/" not in component
1092 c_msg = "-%s" % component
1095 base = ("%s-%s-%s%s-%s.log" %
1096 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1097 return utils.PathJoin(constants.LOG_OS_DIR, base)
1100 def InstanceOsAdd(instance, reinstall, debug):
1101 """Add an OS to an instance.
1103 @type instance: L{objects.Instance}
1104 @param instance: Instance whose OS is to be installed
1105 @type reinstall: boolean
1106 @param reinstall: whether this is an instance reinstall
1107 @type debug: integer
1108 @param debug: debug level, passed to the OS scripts
1112 inst_os = OSFromDisk(instance.os)
1114 create_env = OSEnvironment(instance, inst_os, debug)
1116 create_env["INSTANCE_REINSTALL"] = "1"
1118 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1120 result = utils.RunCmd([inst_os.create_script], env=create_env,
1121 cwd=inst_os.path, output=logfile, reset_env=True)
1123 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1124 " output: %s", result.cmd, result.fail_reason, logfile,
1126 lines = [utils.SafeEncode(val)
1127 for val in utils.TailFile(logfile, lines=20)]
1128 _Fail("OS create script failed (%s), last lines in the"
1129 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1132 def RunRenameInstance(instance, old_name, debug):
1133 """Run the OS rename script for an instance.
1135 @type instance: L{objects.Instance}
1136 @param instance: Instance whose OS is to be installed
1137 @type old_name: string
1138 @param old_name: previous instance name
1139 @type debug: integer
1140 @param debug: debug level, passed to the OS scripts
1142 @return: the success of the operation
1145 inst_os = OSFromDisk(instance.os)
1147 rename_env = OSEnvironment(instance, inst_os, debug)
1148 rename_env["OLD_INSTANCE_NAME"] = old_name
1150 logfile = _InstanceLogName("rename", instance.os,
1151 "%s-%s" % (old_name, instance.name), None)
1153 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1154 cwd=inst_os.path, output=logfile, reset_env=True)
1157 logging.error("os create command '%s' returned error: %s output: %s",
1158 result.cmd, result.fail_reason, result.output)
1159 lines = [utils.SafeEncode(val)
1160 for val in utils.TailFile(logfile, lines=20)]
1161 _Fail("OS rename script failed (%s), last lines in the"
1162 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1165 def _GetBlockDevSymlinkPath(instance_name, idx):
1166 return utils.PathJoin(constants.DISK_LINKS_DIR, "%s%s%d" %
1167 (instance_name, constants.DISK_SEPARATOR, idx))
1170 def _SymlinkBlockDev(instance_name, device_path, idx):
1171 """Set up symlinks to a instance's block device.
1173 This is an auxiliary function run when an instance is start (on the primary
1174 node) or when an instance is migrated (on the target node).
1177 @param instance_name: the name of the target instance
1178 @param device_path: path of the physical block device, on the node
1179 @param idx: the disk index
1180 @return: absolute path to the disk's symlink
1183 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1185 os.symlink(device_path, link_name)
1186 except OSError, err:
1187 if err.errno == errno.EEXIST:
1188 if (not os.path.islink(link_name) or
1189 os.readlink(link_name) != device_path):
1190 os.remove(link_name)
1191 os.symlink(device_path, link_name)
1198 def _RemoveBlockDevLinks(instance_name, disks):
1199 """Remove the block device symlinks belonging to the given instance.
1202 for idx, _ in enumerate(disks):
1203 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1204 if os.path.islink(link_name):
1206 os.remove(link_name)
1208 logging.exception("Can't remove symlink '%s'", link_name)
1211 def _GatherAndLinkBlockDevs(instance):
1212 """Set up an instance's block device(s).
1214 This is run on the primary node at instance startup. The block
1215 devices must be already assembled.
1217 @type instance: L{objects.Instance}
1218 @param instance: the instance whose disks we shoul assemble
1220 @return: list of (disk_object, device_path)
1224 for idx, disk in enumerate(instance.disks):
1225 device = _RecursiveFindBD(disk)
1227 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1231 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1233 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1236 block_devices.append((disk, link_name))
1238 return block_devices
1241 def StartInstance(instance, startup_paused):
1242 """Start an instance.
1244 @type instance: L{objects.Instance}
1245 @param instance: the instance object
1246 @type startup_paused: bool
1247 @param instance: pause instance at startup?
1251 running_instances = GetInstanceList([instance.hypervisor])
1253 if instance.name in running_instances:
1254 logging.info("Instance %s already running, not starting", instance.name)
1258 block_devices = _GatherAndLinkBlockDevs(instance)
1259 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1260 hyper.StartInstance(instance, block_devices, startup_paused)
1261 except errors.BlockDeviceError, err:
1262 _Fail("Block device error: %s", err, exc=True)
1263 except errors.HypervisorError, err:
1264 _RemoveBlockDevLinks(instance.name, instance.disks)
1265 _Fail("Hypervisor error: %s", err, exc=True)
1268 def InstanceShutdown(instance, timeout):
1269 """Shut an instance down.
1271 @note: this functions uses polling with a hardcoded timeout.
1273 @type instance: L{objects.Instance}
1274 @param instance: the instance object
1275 @type timeout: integer
1276 @param timeout: maximum timeout for soft shutdown
1280 hv_name = instance.hypervisor
1281 hyper = hypervisor.GetHypervisor(hv_name)
1282 iname = instance.name
1284 if instance.name not in hyper.ListInstances():
1285 logging.info("Instance %s not running, doing nothing", iname)
1290 self.tried_once = False
1293 if iname not in hyper.ListInstances():
1297 hyper.StopInstance(instance, retry=self.tried_once)
1298 except errors.HypervisorError, err:
1299 if iname not in hyper.ListInstances():
1300 # if the instance is no longer existing, consider this a
1301 # success and go to cleanup
1304 _Fail("Failed to stop instance %s: %s", iname, err)
1306 self.tried_once = True
1308 raise utils.RetryAgain()
1311 utils.Retry(_TryShutdown(), 5, timeout)
1312 except utils.RetryTimeout:
1313 # the shutdown did not succeed
1314 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1317 hyper.StopInstance(instance, force=True)
1318 except errors.HypervisorError, err:
1319 if iname in hyper.ListInstances():
1320 # only raise an error if the instance still exists, otherwise
1321 # the error could simply be "instance ... unknown"!
1322 _Fail("Failed to force stop instance %s: %s", iname, err)
1326 if iname in hyper.ListInstances():
1327 _Fail("Could not shutdown instance %s even by destroy", iname)
1330 hyper.CleanupInstance(instance.name)
1331 except errors.HypervisorError, err:
1332 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1334 _RemoveBlockDevLinks(iname, instance.disks)
1337 def InstanceReboot(instance, reboot_type, shutdown_timeout):
1338 """Reboot an instance.
1340 @type instance: L{objects.Instance}
1341 @param instance: the instance object to reboot
1342 @type reboot_type: str
1343 @param reboot_type: the type of reboot, one the following
1345 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1346 instance OS, do not recreate the VM
1347 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1348 restart the VM (at the hypervisor level)
1349 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1350 not accepted here, since that mode is handled differently, in
1351 cmdlib, and translates into full stop and start of the
1352 instance (instead of a call_instance_reboot RPC)
1353 @type shutdown_timeout: integer
1354 @param shutdown_timeout: maximum timeout for soft shutdown
1358 running_instances = GetInstanceList([instance.hypervisor])
1360 if instance.name not in running_instances:
1361 _Fail("Cannot reboot instance %s that is not running", instance.name)
1363 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1364 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1366 hyper.RebootInstance(instance)
1367 except errors.HypervisorError, err:
1368 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1369 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1371 InstanceShutdown(instance, shutdown_timeout)
1372 return StartInstance(instance, False)
1373 except errors.HypervisorError, err:
1374 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1376 _Fail("Invalid reboot_type received: %s", reboot_type)
1379 def InstanceBalloonMemory(instance, memory):
1380 """Resize an instance's memory.
1382 @type instance: L{objects.Instance}
1383 @param instance: the instance object
1385 @param memory: new memory amount in MB
1389 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1390 running = hyper.ListInstances()
1391 if instance.name not in running:
1392 logging.info("Instance %s is not running, cannot balloon", instance.name)
1395 hyper.BalloonInstanceMemory(instance, memory)
1396 except errors.HypervisorError, err:
1397 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1400 def MigrationInfo(instance):
1401 """Gather information about an instance to be migrated.
1403 @type instance: L{objects.Instance}
1404 @param instance: the instance definition
1407 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1409 info = hyper.MigrationInfo(instance)
1410 except errors.HypervisorError, err:
1411 _Fail("Failed to fetch migration information: %s", err, exc=True)
1415 def AcceptInstance(instance, info, target):
1416 """Prepare the node to accept an instance.
1418 @type instance: L{objects.Instance}
1419 @param instance: the instance definition
1420 @type info: string/data (opaque)
1421 @param info: migration information, from the source node
1422 @type target: string
1423 @param target: target host (usually ip), on this node
1426 # TODO: why is this required only for DTS_EXT_MIRROR?
1427 if instance.disk_template in constants.DTS_EXT_MIRROR:
1428 # Create the symlinks, as the disks are not active
1431 _GatherAndLinkBlockDevs(instance)
1432 except errors.BlockDeviceError, err:
1433 _Fail("Block device error: %s", err, exc=True)
1435 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1437 hyper.AcceptInstance(instance, info, target)
1438 except errors.HypervisorError, err:
1439 if instance.disk_template in constants.DTS_EXT_MIRROR:
1440 _RemoveBlockDevLinks(instance.name, instance.disks)
1441 _Fail("Failed to accept instance: %s", err, exc=True)
1444 def FinalizeMigrationDst(instance, info, success):
1445 """Finalize any preparation to accept an instance.
1447 @type instance: L{objects.Instance}
1448 @param instance: the instance definition
1449 @type info: string/data (opaque)
1450 @param info: migration information, from the source node
1451 @type success: boolean
1452 @param success: whether the migration was a success or a failure
1455 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1457 hyper.FinalizeMigrationDst(instance, info, success)
1458 except errors.HypervisorError, err:
1459 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1462 def MigrateInstance(instance, target, live):
1463 """Migrates an instance to another node.
1465 @type instance: L{objects.Instance}
1466 @param instance: the instance definition
1467 @type target: string
1468 @param target: the target node name
1470 @param live: whether the migration should be done live or not (the
1471 interpretation of this parameter is left to the hypervisor)
1472 @raise RPCFail: if migration fails for some reason
1475 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1478 hyper.MigrateInstance(instance, target, live)
1479 except errors.HypervisorError, err:
1480 _Fail("Failed to migrate instance: %s", err, exc=True)
1483 def FinalizeMigrationSource(instance, success, live):
1484 """Finalize the instance migration on the source node.
1486 @type instance: L{objects.Instance}
1487 @param instance: the instance definition of the migrated instance
1489 @param success: whether the migration succeeded or not
1491 @param live: whether the user requested a live migration or not
1492 @raise RPCFail: If the execution fails for some reason
1495 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1498 hyper.FinalizeMigrationSource(instance, success, live)
1499 except Exception, err: # pylint: disable=W0703
1500 _Fail("Failed to finalize the migration on the source node: %s", err,
1504 def GetMigrationStatus(instance):
1505 """Get the migration status
1507 @type instance: L{objects.Instance}
1508 @param instance: the instance that is being migrated
1509 @rtype: L{objects.MigrationStatus}
1510 @return: the status of the current migration (one of
1511 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1512 progress info that can be retrieved from the hypervisor
1513 @raise RPCFail: If the migration status cannot be retrieved
1516 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1518 return hyper.GetMigrationStatus(instance)
1519 except Exception, err: # pylint: disable=W0703
1520 _Fail("Failed to get migration status: %s", err, exc=True)
1523 def BlockdevCreate(disk, size, owner, on_primary, info):
1524 """Creates a block device for an instance.
1526 @type disk: L{objects.Disk}
1527 @param disk: the object describing the disk we should create
1529 @param size: the size of the physical underlying device, in MiB
1531 @param owner: the name of the instance for which disk is created,
1532 used for device cache data
1533 @type on_primary: boolean
1534 @param on_primary: indicates if it is the primary node or not
1536 @param info: string that will be sent to the physical device
1537 creation, used for example to set (LVM) tags on LVs
1539 @return: the new unique_id of the device (this can sometime be
1540 computed only after creation), or None. On secondary nodes,
1541 it's not required to return anything.
1544 # TODO: remove the obsolete "size" argument
1545 # pylint: disable=W0613
1548 for child in disk.children:
1550 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1551 except errors.BlockDeviceError, err:
1552 _Fail("Can't assemble device %s: %s", child, err)
1553 if on_primary or disk.AssembleOnSecondary():
1554 # we need the children open in case the device itself has to
1557 # pylint: disable=E1103
1559 except errors.BlockDeviceError, err:
1560 _Fail("Can't make child '%s' read-write: %s", child, err)
1564 device = bdev.Create(disk, clist)
1565 except errors.BlockDeviceError, err:
1566 _Fail("Can't create block device: %s", err)
1568 if on_primary or disk.AssembleOnSecondary():
1571 except errors.BlockDeviceError, err:
1572 _Fail("Can't assemble device after creation, unusual event: %s", err)
1573 if on_primary or disk.OpenOnSecondary():
1575 device.Open(force=True)
1576 except errors.BlockDeviceError, err:
1577 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1578 DevCacheManager.UpdateCache(device.dev_path, owner,
1579 on_primary, disk.iv_name)
1581 device.SetInfo(info)
1583 return device.unique_id
1586 def _WipeDevice(path, offset, size):
1587 """This function actually wipes the device.
1589 @param path: The path to the device to wipe
1590 @param offset: The offset in MiB in the file
1591 @param size: The size in MiB to write
1594 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1595 "bs=%d" % constants.WIPE_BLOCK_SIZE, "oflag=direct", "of=%s" % path,
1597 result = utils.RunCmd(cmd)
1600 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1601 result.fail_reason, result.output)
1604 def BlockdevWipe(disk, offset, size):
1605 """Wipes a block device.
1607 @type disk: L{objects.Disk}
1608 @param disk: the disk object we want to wipe
1610 @param offset: The offset in MiB in the file
1612 @param size: The size in MiB to write
1616 rdev = _RecursiveFindBD(disk)
1617 except errors.BlockDeviceError:
1621 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1623 # Do cross verify some of the parameters
1624 if offset > rdev.size:
1625 _Fail("Offset is bigger than device size")
1626 if (offset + size) > rdev.size:
1627 _Fail("The provided offset and size to wipe is bigger than device size")
1629 _WipeDevice(rdev.dev_path, offset, size)
1632 def BlockdevPauseResumeSync(disks, pause):
1633 """Pause or resume the sync of the block device.
1635 @type disks: list of L{objects.Disk}
1636 @param disks: the disks object we want to pause/resume
1638 @param pause: Wheater to pause or resume
1644 rdev = _RecursiveFindBD(disk)
1645 except errors.BlockDeviceError:
1649 success.append((False, ("Cannot change sync for device %s:"
1650 " device not found" % disk.iv_name)))
1653 result = rdev.PauseResumeSync(pause)
1656 success.append((result, None))
1662 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1667 def BlockdevRemove(disk):
1668 """Remove a block device.
1670 @note: This is intended to be called recursively.
1672 @type disk: L{objects.Disk}
1673 @param disk: the disk object we should remove
1675 @return: the success of the operation
1680 rdev = _RecursiveFindBD(disk)
1681 except errors.BlockDeviceError, err:
1682 # probably can't attach
1683 logging.info("Can't attach to device %s in remove", disk)
1685 if rdev is not None:
1686 r_path = rdev.dev_path
1689 except errors.BlockDeviceError, err:
1690 msgs.append(str(err))
1692 DevCacheManager.RemoveCache(r_path)
1695 for child in disk.children:
1697 BlockdevRemove(child)
1698 except RPCFail, err:
1699 msgs.append(str(err))
1702 _Fail("; ".join(msgs))
1705 def _RecursiveAssembleBD(disk, owner, as_primary):
1706 """Activate a block device for an instance.
1708 This is run on the primary and secondary nodes for an instance.
1710 @note: this function is called recursively.
1712 @type disk: L{objects.Disk}
1713 @param disk: the disk we try to assemble
1715 @param owner: the name of the instance which owns the disk
1716 @type as_primary: boolean
1717 @param as_primary: if we should make the block device
1720 @return: the assembled device or None (in case no device
1722 @raise errors.BlockDeviceError: in case there is an error
1723 during the activation of the children or the device
1729 mcn = disk.ChildrenNeeded()
1731 mcn = 0 # max number of Nones allowed
1733 mcn = len(disk.children) - mcn # max number of Nones
1734 for chld_disk in disk.children:
1736 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1737 except errors.BlockDeviceError, err:
1738 if children.count(None) >= mcn:
1741 logging.error("Error in child activation (but continuing): %s",
1743 children.append(cdev)
1745 if as_primary or disk.AssembleOnSecondary():
1746 r_dev = bdev.Assemble(disk, children)
1748 if as_primary or disk.OpenOnSecondary():
1750 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1751 as_primary, disk.iv_name)
1758 def BlockdevAssemble(disk, owner, as_primary, idx):
1759 """Activate a block device for an instance.
1761 This is a wrapper over _RecursiveAssembleBD.
1763 @rtype: str or boolean
1764 @return: a C{/dev/...} path for primary nodes, and
1765 C{True} for secondary nodes
1769 result = _RecursiveAssembleBD(disk, owner, as_primary)
1770 if isinstance(result, bdev.BlockDev):
1771 # pylint: disable=E1103
1772 result = result.dev_path
1774 _SymlinkBlockDev(owner, result, idx)
1775 except errors.BlockDeviceError, err:
1776 _Fail("Error while assembling disk: %s", err, exc=True)
1777 except OSError, err:
1778 _Fail("Error while symlinking disk: %s", err, exc=True)
1783 def BlockdevShutdown(disk):
1784 """Shut down a block device.
1786 First, if the device is assembled (Attach() is successful), then
1787 the device is shutdown. Then the children of the device are
1790 This function is called recursively. Note that we don't cache the
1791 children or such, as oppossed to assemble, shutdown of different
1792 devices doesn't require that the upper device was active.
1794 @type disk: L{objects.Disk}
1795 @param disk: the description of the disk we should
1801 r_dev = _RecursiveFindBD(disk)
1802 if r_dev is not None:
1803 r_path = r_dev.dev_path
1806 DevCacheManager.RemoveCache(r_path)
1807 except errors.BlockDeviceError, err:
1808 msgs.append(str(err))
1811 for child in disk.children:
1813 BlockdevShutdown(child)
1814 except RPCFail, err:
1815 msgs.append(str(err))
1818 _Fail("; ".join(msgs))
1821 def BlockdevAddchildren(parent_cdev, new_cdevs):
1822 """Extend a mirrored block device.
1824 @type parent_cdev: L{objects.Disk}
1825 @param parent_cdev: the disk to which we should add children
1826 @type new_cdevs: list of L{objects.Disk}
1827 @param new_cdevs: the list of children which we should add
1831 parent_bdev = _RecursiveFindBD(parent_cdev)
1832 if parent_bdev is None:
1833 _Fail("Can't find parent device '%s' in add children", parent_cdev)
1834 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1835 if new_bdevs.count(None) > 0:
1836 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1837 parent_bdev.AddChildren(new_bdevs)
1840 def BlockdevRemovechildren(parent_cdev, new_cdevs):
1841 """Shrink a mirrored block device.
1843 @type parent_cdev: L{objects.Disk}
1844 @param parent_cdev: the disk from which we should remove children
1845 @type new_cdevs: list of L{objects.Disk}
1846 @param new_cdevs: the list of children which we should remove
1850 parent_bdev = _RecursiveFindBD(parent_cdev)
1851 if parent_bdev is None:
1852 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1854 for disk in new_cdevs:
1855 rpath = disk.StaticDevPath()
1857 bd = _RecursiveFindBD(disk)
1859 _Fail("Can't find device %s while removing children", disk)
1861 devs.append(bd.dev_path)
1863 if not utils.IsNormAbsPath(rpath):
1864 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1866 parent_bdev.RemoveChildren(devs)
1869 def BlockdevGetmirrorstatus(disks):
1870 """Get the mirroring status of a list of devices.
1872 @type disks: list of L{objects.Disk}
1873 @param disks: the list of disks which we should query
1875 @return: List of L{objects.BlockDevStatus}, one for each disk
1876 @raise errors.BlockDeviceError: if any of the disks cannot be
1882 rbd = _RecursiveFindBD(dsk)
1884 _Fail("Can't find device %s", dsk)
1886 stats.append(rbd.CombinedSyncStatus())
1891 def BlockdevGetmirrorstatusMulti(disks):
1892 """Get the mirroring status of a list of devices.
1894 @type disks: list of L{objects.Disk}
1895 @param disks: the list of disks which we should query
1897 @return: List of tuples, (bool, status), one for each disk; bool denotes
1898 success/failure, status is L{objects.BlockDevStatus} on success, string
1905 rbd = _RecursiveFindBD(disk)
1907 result.append((False, "Can't find device %s" % disk))
1910 status = rbd.CombinedSyncStatus()
1911 except errors.BlockDeviceError, err:
1912 logging.exception("Error while getting disk status")
1913 result.append((False, str(err)))
1915 result.append((True, status))
1917 assert len(disks) == len(result)
1922 def _RecursiveFindBD(disk):
1923 """Check if a device is activated.
1925 If so, return information about the real device.
1927 @type disk: L{objects.Disk}
1928 @param disk: the disk object we need to find
1930 @return: None if the device can't be found,
1931 otherwise the device instance
1936 for chdisk in disk.children:
1937 children.append(_RecursiveFindBD(chdisk))
1939 return bdev.FindDevice(disk, children)
1942 def _OpenRealBD(disk):
1943 """Opens the underlying block device of a disk.
1945 @type disk: L{objects.Disk}
1946 @param disk: the disk object we want to open
1949 real_disk = _RecursiveFindBD(disk)
1950 if real_disk is None:
1951 _Fail("Block device '%s' is not set up", disk)
1958 def BlockdevFind(disk):
1959 """Check if a device is activated.
1961 If it is, return information about the real device.
1963 @type disk: L{objects.Disk}
1964 @param disk: the disk to find
1965 @rtype: None or objects.BlockDevStatus
1966 @return: None if the disk cannot be found, otherwise a the current
1971 rbd = _RecursiveFindBD(disk)
1972 except errors.BlockDeviceError, err:
1973 _Fail("Failed to find device: %s", err, exc=True)
1978 return rbd.GetSyncStatus()
1981 def BlockdevGetsize(disks):
1982 """Computes the size of the given disks.
1984 If a disk is not found, returns None instead.
1986 @type disks: list of L{objects.Disk}
1987 @param disks: the list of disk to compute the size for
1989 @return: list with elements None if the disk cannot be found,
1996 rbd = _RecursiveFindBD(cf)
1997 except errors.BlockDeviceError:
2003 result.append(rbd.GetActualSize())
2007 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2008 """Export a block device to a remote node.
2010 @type disk: L{objects.Disk}
2011 @param disk: the description of the disk to export
2012 @type dest_node: str
2013 @param dest_node: the destination node to export to
2014 @type dest_path: str
2015 @param dest_path: the destination path on the target node
2016 @type cluster_name: str
2017 @param cluster_name: the cluster name, needed for SSH hostalias
2021 real_disk = _OpenRealBD(disk)
2023 # the block size on the read dd is 1MiB to match our units
2024 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2025 "dd if=%s bs=1048576 count=%s",
2026 real_disk.dev_path, str(disk.size))
2028 # we set here a smaller block size as, due to ssh buffering, more
2029 # than 64-128k will mostly ignored; we use nocreat to fail if the
2030 # device is not already there or we pass a wrong path; we use
2031 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2032 # to not buffer too much memory; this means that at best, we flush
2033 # every 64k, which will not be very fast
2034 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2035 " oflag=dsync", dest_path)
2037 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2038 constants.GANETI_RUNAS,
2041 # all commands have been checked, so we're safe to combine them
2042 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2044 result = utils.RunCmd(["bash", "-c", command])
2047 _Fail("Disk copy command '%s' returned error: %s"
2048 " output: %s", command, result.fail_reason, result.output)
2051 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2052 """Write a file to the filesystem.
2054 This allows the master to overwrite(!) a file. It will only perform
2055 the operation if the file belongs to a list of configuration files.
2057 @type file_name: str
2058 @param file_name: the target file name
2060 @param data: the new contents of the file
2062 @param mode: the mode to give the file (can be None)
2064 @param uid: the owner of the file
2066 @param gid: the group of the file
2068 @param atime: the atime to set on the file (can be None)
2070 @param mtime: the mtime to set on the file (can be None)
2074 if not os.path.isabs(file_name):
2075 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2077 if file_name not in _ALLOWED_UPLOAD_FILES:
2078 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2081 raw_data = _Decompress(data)
2083 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2084 _Fail("Invalid username/groupname type")
2086 getents = runtime.GetEnts()
2087 uid = getents.LookupUser(uid)
2088 gid = getents.LookupGroup(gid)
2090 utils.SafeWriteFile(file_name, None,
2091 data=raw_data, mode=mode, uid=uid, gid=gid,
2092 atime=atime, mtime=mtime)
2095 def RunOob(oob_program, command, node, timeout):
2096 """Executes oob_program with given command on given node.
2098 @param oob_program: The path to the executable oob_program
2099 @param command: The command to invoke on oob_program
2100 @param node: The node given as an argument to the program
2101 @param timeout: Timeout after which we kill the oob program
2104 @raise RPCFail: If execution fails for some reason
2107 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2110 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2111 result.fail_reason, result.output)
2113 return result.stdout
2116 def WriteSsconfFiles(values):
2117 """Update all ssconf files.
2119 Wrapper around the SimpleStore.WriteFiles.
2122 ssconf.SimpleStore().WriteFiles(values)
2125 def _OSOndiskAPIVersion(os_dir):
2126 """Compute and return the API version of a given OS.
2128 This function will try to read the API version of the OS residing in
2129 the 'os_dir' directory.
2132 @param os_dir: the directory in which we should look for the OS
2134 @return: tuple (status, data) with status denoting the validity and
2135 data holding either the vaid versions or an error message
2138 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2141 st = os.stat(api_file)
2142 except EnvironmentError, err:
2143 return False, ("Required file '%s' not found under path %s: %s" %
2144 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2146 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2147 return False, ("File '%s' in %s is not a regular file" %
2148 (constants.OS_API_FILE, os_dir))
2151 api_versions = utils.ReadFile(api_file).splitlines()
2152 except EnvironmentError, err:
2153 return False, ("Error while reading the API version file at %s: %s" %
2154 (api_file, utils.ErrnoOrStr(err)))
2157 api_versions = [int(version.strip()) for version in api_versions]
2158 except (TypeError, ValueError), err:
2159 return False, ("API version(s) can't be converted to integer: %s" %
2162 return True, api_versions
2165 def DiagnoseOS(top_dirs=None):
2166 """Compute the validity for all OSes.
2168 @type top_dirs: list
2169 @param top_dirs: the list of directories in which to
2170 search (if not given defaults to
2171 L{constants.OS_SEARCH_PATH})
2172 @rtype: list of L{objects.OS}
2173 @return: a list of tuples (name, path, status, diagnose, variants,
2174 parameters, api_version) for all (potential) OSes under all
2175 search paths, where:
2176 - name is the (potential) OS name
2177 - path is the full path to the OS
2178 - status True/False is the validity of the OS
2179 - diagnose is the error message for an invalid OS, otherwise empty
2180 - variants is a list of supported OS variants, if any
2181 - parameters is a list of (name, help) parameters, if any
2182 - api_version is a list of support OS API versions
2185 if top_dirs is None:
2186 top_dirs = constants.OS_SEARCH_PATH
2189 for dir_name in top_dirs:
2190 if os.path.isdir(dir_name):
2192 f_names = utils.ListVisibleFiles(dir_name)
2193 except EnvironmentError, err:
2194 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2196 for name in f_names:
2197 os_path = utils.PathJoin(dir_name, name)
2198 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2201 variants = os_inst.supported_variants
2202 parameters = os_inst.supported_parameters
2203 api_versions = os_inst.api_versions
2206 variants = parameters = api_versions = []
2207 result.append((name, os_path, status, diagnose, variants,
2208 parameters, api_versions))
2213 def _TryOSFromDisk(name, base_dir=None):
2214 """Create an OS instance from disk.
2216 This function will return an OS instance if the given name is a
2219 @type base_dir: string
2220 @keyword base_dir: Base directory containing OS installations.
2221 Defaults to a search in all the OS_SEARCH_PATH dirs.
2223 @return: success and either the OS instance if we find a valid one,
2227 if base_dir is None:
2228 os_dir = utils.FindFile(name, constants.OS_SEARCH_PATH, os.path.isdir)
2230 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2233 return False, "Directory for OS %s not found in search path" % name
2235 status, api_versions = _OSOndiskAPIVersion(os_dir)
2238 return status, api_versions
2240 if not constants.OS_API_VERSIONS.intersection(api_versions):
2241 return False, ("API version mismatch for path '%s': found %s, want %s." %
2242 (os_dir, api_versions, constants.OS_API_VERSIONS))
2244 # OS Files dictionary, we will populate it with the absolute path
2245 # names; if the value is True, then it is a required file, otherwise
2247 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2249 if max(api_versions) >= constants.OS_API_V15:
2250 os_files[constants.OS_VARIANTS_FILE] = False
2252 if max(api_versions) >= constants.OS_API_V20:
2253 os_files[constants.OS_PARAMETERS_FILE] = True
2255 del os_files[constants.OS_SCRIPT_VERIFY]
2257 for (filename, required) in os_files.items():
2258 os_files[filename] = utils.PathJoin(os_dir, filename)
2261 st = os.stat(os_files[filename])
2262 except EnvironmentError, err:
2263 if err.errno == errno.ENOENT and not required:
2264 del os_files[filename]
2266 return False, ("File '%s' under path '%s' is missing (%s)" %
2267 (filename, os_dir, utils.ErrnoOrStr(err)))
2269 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2270 return False, ("File '%s' under path '%s' is not a regular file" %
2273 if filename in constants.OS_SCRIPTS:
2274 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2275 return False, ("File '%s' under path '%s' is not executable" %
2279 if constants.OS_VARIANTS_FILE in os_files:
2280 variants_file = os_files[constants.OS_VARIANTS_FILE]
2282 variants = utils.ReadFile(variants_file).splitlines()
2283 except EnvironmentError, err:
2284 # we accept missing files, but not other errors
2285 if err.errno != errno.ENOENT:
2286 return False, ("Error while reading the OS variants file at %s: %s" %
2287 (variants_file, utils.ErrnoOrStr(err)))
2290 if constants.OS_PARAMETERS_FILE in os_files:
2291 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2293 parameters = utils.ReadFile(parameters_file).splitlines()
2294 except EnvironmentError, err:
2295 return False, ("Error while reading the OS parameters file at %s: %s" %
2296 (parameters_file, utils.ErrnoOrStr(err)))
2297 parameters = [v.split(None, 1) for v in parameters]
2299 os_obj = objects.OS(name=name, path=os_dir,
2300 create_script=os_files[constants.OS_SCRIPT_CREATE],
2301 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2302 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2303 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2304 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2306 supported_variants=variants,
2307 supported_parameters=parameters,
2308 api_versions=api_versions)
2312 def OSFromDisk(name, base_dir=None):
2313 """Create an OS instance from disk.
2315 This function will return an OS instance if the given name is a
2316 valid OS name. Otherwise, it will raise an appropriate
2317 L{RPCFail} exception, detailing why this is not a valid OS.
2319 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2320 an exception but returns true/false status data.
2322 @type base_dir: string
2323 @keyword base_dir: Base directory containing OS installations.
2324 Defaults to a search in all the OS_SEARCH_PATH dirs.
2325 @rtype: L{objects.OS}
2326 @return: the OS instance if we find a valid one
2327 @raise RPCFail: if we don't find a valid OS
2330 name_only = objects.OS.GetName(name)
2331 status, payload = _TryOSFromDisk(name_only, base_dir)
2339 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2340 """Calculate the basic environment for an os script.
2343 @param os_name: full operating system name (including variant)
2344 @type inst_os: L{objects.OS}
2345 @param inst_os: operating system for which the environment is being built
2346 @type os_params: dict
2347 @param os_params: the OS parameters
2348 @type debug: integer
2349 @param debug: debug level (0 or 1, for OS Api 10)
2351 @return: dict of environment variables
2352 @raise errors.BlockDeviceError: if the block device
2358 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2359 result["OS_API_VERSION"] = "%d" % api_version
2360 result["OS_NAME"] = inst_os.name
2361 result["DEBUG_LEVEL"] = "%d" % debug
2364 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2365 variant = objects.OS.GetVariant(os_name)
2367 variant = inst_os.supported_variants[0]
2370 result["OS_VARIANT"] = variant
2373 for pname, pvalue in os_params.items():
2374 result["OSP_%s" % pname.upper()] = pvalue
2379 def OSEnvironment(instance, inst_os, debug=0):
2380 """Calculate the environment for an os script.
2382 @type instance: L{objects.Instance}
2383 @param instance: target instance for the os script run
2384 @type inst_os: L{objects.OS}
2385 @param inst_os: operating system for which the environment is being built
2386 @type debug: integer
2387 @param debug: debug level (0 or 1, for OS Api 10)
2389 @return: dict of environment variables
2390 @raise errors.BlockDeviceError: if the block device
2394 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2396 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2397 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2399 result["HYPERVISOR"] = instance.hypervisor
2400 result["DISK_COUNT"] = "%d" % len(instance.disks)
2401 result["NIC_COUNT"] = "%d" % len(instance.nics)
2402 result["INSTANCE_SECONDARY_NODES"] = \
2403 ("%s" % " ".join(instance.secondary_nodes))
2406 for idx, disk in enumerate(instance.disks):
2407 real_disk = _OpenRealBD(disk)
2408 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2409 result["DISK_%d_ACCESS" % idx] = disk.mode
2410 if constants.HV_DISK_TYPE in instance.hvparams:
2411 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2412 instance.hvparams[constants.HV_DISK_TYPE]
2413 if disk.dev_type in constants.LDS_BLOCK:
2414 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2415 elif disk.dev_type == constants.LD_FILE:
2416 result["DISK_%d_BACKEND_TYPE" % idx] = \
2417 "file:%s" % disk.physical_id[0]
2420 for idx, nic in enumerate(instance.nics):
2421 result["NIC_%d_MAC" % idx] = nic.mac
2423 result["NIC_%d_IP" % idx] = nic.ip
2424 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2425 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2426 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2427 if nic.nicparams[constants.NIC_LINK]:
2428 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2429 if constants.HV_NIC_TYPE in instance.hvparams:
2430 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2431 instance.hvparams[constants.HV_NIC_TYPE]
2434 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2435 for key, value in source.items():
2436 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2441 def BlockdevGrow(disk, amount, dryrun):
2442 """Grow a stack of block devices.
2444 This function is called recursively, with the childrens being the
2445 first ones to resize.
2447 @type disk: L{objects.Disk}
2448 @param disk: the disk to be grown
2449 @type amount: integer
2450 @param amount: the amount (in mebibytes) to grow with
2451 @type dryrun: boolean
2452 @param dryrun: whether to execute the operation in simulation mode
2453 only, without actually increasing the size
2454 @rtype: (status, result)
2455 @return: a tuple with the status of the operation (True/False), and
2456 the errors message if status is False
2459 r_dev = _RecursiveFindBD(disk)
2461 _Fail("Cannot find block device %s", disk)
2464 r_dev.Grow(amount, dryrun)
2465 except errors.BlockDeviceError, err:
2466 _Fail("Failed to grow block device: %s", err, exc=True)
2469 def BlockdevSnapshot(disk):
2470 """Create a snapshot copy of a block device.
2472 This function is called recursively, and the snapshot is actually created
2473 just for the leaf lvm backend device.
2475 @type disk: L{objects.Disk}
2476 @param disk: the disk to be snapshotted
2478 @return: snapshot disk ID as (vg, lv)
2481 if disk.dev_type == constants.LD_DRBD8:
2482 if not disk.children:
2483 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2485 return BlockdevSnapshot(disk.children[0])
2486 elif disk.dev_type == constants.LD_LV:
2487 r_dev = _RecursiveFindBD(disk)
2488 if r_dev is not None:
2489 # FIXME: choose a saner value for the snapshot size
2490 # let's stay on the safe side and ask for the full size, for now
2491 return r_dev.Snapshot(disk.size)
2493 _Fail("Cannot find block device %s", disk)
2495 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2496 disk.unique_id, disk.dev_type)
2499 def FinalizeExport(instance, snap_disks):
2500 """Write out the export configuration information.
2502 @type instance: L{objects.Instance}
2503 @param instance: the instance which we export, used for
2504 saving configuration
2505 @type snap_disks: list of L{objects.Disk}
2506 @param snap_disks: list of snapshot block devices, which
2507 will be used to get the actual name of the dump file
2512 destdir = utils.PathJoin(constants.EXPORT_DIR, instance.name + ".new")
2513 finaldestdir = utils.PathJoin(constants.EXPORT_DIR, instance.name)
2515 config = objects.SerializableConfigParser()
2517 config.add_section(constants.INISECT_EXP)
2518 config.set(constants.INISECT_EXP, "version", "0")
2519 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2520 config.set(constants.INISECT_EXP, "source", instance.primary_node)
2521 config.set(constants.INISECT_EXP, "os", instance.os)
2522 config.set(constants.INISECT_EXP, "compression", "none")
2524 config.add_section(constants.INISECT_INS)
2525 config.set(constants.INISECT_INS, "name", instance.name)
2526 config.set(constants.INISECT_INS, "maxmem", "%d" %
2527 instance.beparams[constants.BE_MAXMEM])
2528 config.set(constants.INISECT_INS, "minmem", "%d" %
2529 instance.beparams[constants.BE_MINMEM])
2530 # "memory" is deprecated, but useful for exporting to old ganeti versions
2531 config.set(constants.INISECT_INS, "memory", "%d" %
2532 instance.beparams[constants.BE_MAXMEM])
2533 config.set(constants.INISECT_INS, "vcpus", "%d" %
2534 instance.beparams[constants.BE_VCPUS])
2535 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2536 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2537 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2540 for nic_count, nic in enumerate(instance.nics):
2542 config.set(constants.INISECT_INS, "nic%d_mac" %
2543 nic_count, "%s" % nic.mac)
2544 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2545 for param in constants.NICS_PARAMETER_TYPES:
2546 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2547 "%s" % nic.nicparams.get(param, None))
2548 # TODO: redundant: on load can read nics until it doesn't exist
2549 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2552 for disk_count, disk in enumerate(snap_disks):
2555 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2556 ("%s" % disk.iv_name))
2557 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2558 ("%s" % disk.physical_id[1]))
2559 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2562 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2564 # New-style hypervisor/backend parameters
2566 config.add_section(constants.INISECT_HYP)
2567 for name, value in instance.hvparams.items():
2568 if name not in constants.HVC_GLOBALS:
2569 config.set(constants.INISECT_HYP, name, str(value))
2571 config.add_section(constants.INISECT_BEP)
2572 for name, value in instance.beparams.items():
2573 config.set(constants.INISECT_BEP, name, str(value))
2575 config.add_section(constants.INISECT_OSP)
2576 for name, value in instance.osparams.items():
2577 config.set(constants.INISECT_OSP, name, str(value))
2579 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2580 data=config.Dumps())
2581 shutil.rmtree(finaldestdir, ignore_errors=True)
2582 shutil.move(destdir, finaldestdir)
2585 def ExportInfo(dest):
2586 """Get export configuration information.
2589 @param dest: directory containing the export
2591 @rtype: L{objects.SerializableConfigParser}
2592 @return: a serializable config file containing the
2596 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2598 config = objects.SerializableConfigParser()
2601 if (not config.has_section(constants.INISECT_EXP) or
2602 not config.has_section(constants.INISECT_INS)):
2603 _Fail("Export info file doesn't have the required fields")
2605 return config.Dumps()
2609 """Return a list of exports currently available on this machine.
2612 @return: list of the exports
2615 if os.path.isdir(constants.EXPORT_DIR):
2616 return sorted(utils.ListVisibleFiles(constants.EXPORT_DIR))
2618 _Fail("No exports directory")
2621 def RemoveExport(export):
2622 """Remove an existing export from the node.
2625 @param export: the name of the export to remove
2629 target = utils.PathJoin(constants.EXPORT_DIR, export)
2632 shutil.rmtree(target)
2633 except EnvironmentError, err:
2634 _Fail("Error while removing the export: %s", err, exc=True)
2637 def BlockdevRename(devlist):
2638 """Rename a list of block devices.
2640 @type devlist: list of tuples
2641 @param devlist: list of tuples of the form (disk,
2642 new_logical_id, new_physical_id); disk is an
2643 L{objects.Disk} object describing the current disk,
2644 and new logical_id/physical_id is the name we
2647 @return: True if all renames succeeded, False otherwise
2652 for disk, unique_id in devlist:
2653 dev = _RecursiveFindBD(disk)
2655 msgs.append("Can't find device %s in rename" % str(disk))
2659 old_rpath = dev.dev_path
2660 dev.Rename(unique_id)
2661 new_rpath = dev.dev_path
2662 if old_rpath != new_rpath:
2663 DevCacheManager.RemoveCache(old_rpath)
2664 # FIXME: we should add the new cache information here, like:
2665 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2666 # but we don't have the owner here - maybe parse from existing
2667 # cache? for now, we only lose lvm data when we rename, which
2668 # is less critical than DRBD or MD
2669 except errors.BlockDeviceError, err:
2670 msgs.append("Can't rename device '%s' to '%s': %s" %
2671 (dev, unique_id, err))
2672 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2675 _Fail("; ".join(msgs))
2678 def _TransformFileStorageDir(fs_dir):
2679 """Checks whether given file_storage_dir is valid.
2681 Checks wheter the given fs_dir is within the cluster-wide default
2682 file_storage_dir or the shared_file_storage_dir, which are stored in
2683 SimpleStore. Only paths under those directories are allowed.
2686 @param fs_dir: the path to check
2688 @return: the normalized path if valid, None otherwise
2691 if not constants.ENABLE_FILE_STORAGE:
2692 _Fail("File storage disabled at configure time")
2694 fs_dir = os.path.normpath(fs_dir)
2695 base_fstore = cfg.GetFileStorageDir()
2696 base_shared = cfg.GetSharedFileStorageDir()
2697 if not (utils.IsBelowDir(base_fstore, fs_dir) or
2698 utils.IsBelowDir(base_shared, fs_dir)):
2699 _Fail("File storage directory '%s' is not under base file"
2700 " storage directory '%s' or shared storage directory '%s'",
2701 fs_dir, base_fstore, base_shared)
2705 def CreateFileStorageDir(file_storage_dir):
2706 """Create file storage directory.
2708 @type file_storage_dir: str
2709 @param file_storage_dir: directory to create
2712 @return: tuple with first element a boolean indicating wheter dir
2713 creation was successful or not
2716 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2717 if os.path.exists(file_storage_dir):
2718 if not os.path.isdir(file_storage_dir):
2719 _Fail("Specified storage dir '%s' is not a directory",
2723 os.makedirs(file_storage_dir, 0750)
2724 except OSError, err:
2725 _Fail("Cannot create file storage directory '%s': %s",
2726 file_storage_dir, err, exc=True)
2729 def RemoveFileStorageDir(file_storage_dir):
2730 """Remove file storage directory.
2732 Remove it only if it's empty. If not log an error and return.
2734 @type file_storage_dir: str
2735 @param file_storage_dir: the directory we should cleanup
2736 @rtype: tuple (success,)
2737 @return: tuple of one element, C{success}, denoting
2738 whether the operation was successful
2741 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2742 if os.path.exists(file_storage_dir):
2743 if not os.path.isdir(file_storage_dir):
2744 _Fail("Specified Storage directory '%s' is not a directory",
2746 # deletes dir only if empty, otherwise we want to fail the rpc call
2748 os.rmdir(file_storage_dir)
2749 except OSError, err:
2750 _Fail("Cannot remove file storage directory '%s': %s",
2751 file_storage_dir, err)
2754 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2755 """Rename the file storage directory.
2757 @type old_file_storage_dir: str
2758 @param old_file_storage_dir: the current path
2759 @type new_file_storage_dir: str
2760 @param new_file_storage_dir: the name we should rename to
2761 @rtype: tuple (success,)
2762 @return: tuple of one element, C{success}, denoting
2763 whether the operation was successful
2766 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2767 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2768 if not os.path.exists(new_file_storage_dir):
2769 if os.path.isdir(old_file_storage_dir):
2771 os.rename(old_file_storage_dir, new_file_storage_dir)
2772 except OSError, err:
2773 _Fail("Cannot rename '%s' to '%s': %s",
2774 old_file_storage_dir, new_file_storage_dir, err)
2776 _Fail("Specified storage dir '%s' is not a directory",
2777 old_file_storage_dir)
2779 if os.path.exists(old_file_storage_dir):
2780 _Fail("Cannot rename '%s' to '%s': both locations exist",
2781 old_file_storage_dir, new_file_storage_dir)
2784 def _EnsureJobQueueFile(file_name):
2785 """Checks whether the given filename is in the queue directory.
2787 @type file_name: str
2788 @param file_name: the file name we should check
2790 @raises RPCFail: if the file is not valid
2793 queue_dir = os.path.normpath(constants.QUEUE_DIR)
2794 result = (os.path.commonprefix([queue_dir, file_name]) == queue_dir)
2797 _Fail("Passed job queue file '%s' does not belong to"
2798 " the queue directory '%s'", file_name, queue_dir)
2801 def JobQueueUpdate(file_name, content):
2802 """Updates a file in the queue directory.
2804 This is just a wrapper over L{utils.io.WriteFile}, with proper
2807 @type file_name: str
2808 @param file_name: the job file name
2810 @param content: the new job contents
2812 @return: the success of the operation
2815 _EnsureJobQueueFile(file_name)
2816 getents = runtime.GetEnts()
2818 # Write and replace the file atomically
2819 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2820 gid=getents.masterd_gid)
2823 def JobQueueRename(old, new):
2824 """Renames a job queue file.
2826 This is just a wrapper over os.rename with proper checking.
2829 @param old: the old (actual) file name
2831 @param new: the desired file name
2833 @return: the success of the operation and payload
2836 _EnsureJobQueueFile(old)
2837 _EnsureJobQueueFile(new)
2839 getents = runtime.GetEnts()
2841 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0700,
2842 dir_uid=getents.masterd_uid, dir_gid=getents.masterd_gid)
2845 def BlockdevClose(instance_name, disks):
2846 """Closes the given block devices.
2848 This means they will be switched to secondary mode (in case of
2851 @param instance_name: if the argument is not empty, the symlinks
2852 of this instance will be removed
2853 @type disks: list of L{objects.Disk}
2854 @param disks: the list of disks to be closed
2855 @rtype: tuple (success, message)
2856 @return: a tuple of success and message, where success
2857 indicates the succes of the operation, and message
2858 which will contain the error details in case we
2864 rd = _RecursiveFindBD(cf)
2866 _Fail("Can't find device %s", cf)
2873 except errors.BlockDeviceError, err:
2874 msg.append(str(err))
2876 _Fail("Can't make devices secondary: %s", ",".join(msg))
2879 _RemoveBlockDevLinks(instance_name, disks)
2882 def ValidateHVParams(hvname, hvparams):
2883 """Validates the given hypervisor parameters.
2885 @type hvname: string
2886 @param hvname: the hypervisor name
2887 @type hvparams: dict
2888 @param hvparams: the hypervisor parameters to be validated
2893 hv_type = hypervisor.GetHypervisor(hvname)
2894 hv_type.ValidateParameters(hvparams)
2895 except errors.HypervisorError, err:
2896 _Fail(str(err), log=False)
2899 def _CheckOSPList(os_obj, parameters):
2900 """Check whether a list of parameters is supported by the OS.
2902 @type os_obj: L{objects.OS}
2903 @param os_obj: OS object to check
2904 @type parameters: list
2905 @param parameters: the list of parameters to check
2908 supported = [v[0] for v in os_obj.supported_parameters]
2909 delta = frozenset(parameters).difference(supported)
2911 _Fail("The following parameters are not supported"
2912 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
2915 def ValidateOS(required, osname, checks, osparams):
2916 """Validate the given OS' parameters.
2918 @type required: boolean
2919 @param required: whether absence of the OS should translate into
2921 @type osname: string
2922 @param osname: the OS to be validated
2924 @param checks: list of the checks to run (currently only 'parameters')
2925 @type osparams: dict
2926 @param osparams: dictionary with OS parameters
2928 @return: True if the validation passed, or False if the OS was not
2929 found and L{required} was false
2932 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
2933 _Fail("Unknown checks required for OS %s: %s", osname,
2934 set(checks).difference(constants.OS_VALIDATE_CALLS))
2936 name_only = objects.OS.GetName(osname)
2937 status, tbv = _TryOSFromDisk(name_only, None)
2945 if max(tbv.api_versions) < constants.OS_API_V20:
2948 if constants.OS_VALIDATE_PARAMETERS in checks:
2949 _CheckOSPList(tbv, osparams.keys())
2951 validate_env = OSCoreEnv(osname, tbv, osparams)
2952 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
2953 cwd=tbv.path, reset_env=True)
2955 logging.error("os validate command '%s' returned error: %s output: %s",
2956 result.cmd, result.fail_reason, result.output)
2957 _Fail("OS validation script failed (%s), output: %s",
2958 result.fail_reason, result.output, log=False)
2964 """Demotes the current node from master candidate role.
2967 # try to ensure we're not the master by mistake
2968 master, myself = ssconf.GetMasterAndMyself()
2969 if master == myself:
2970 _Fail("ssconf status shows I'm the master node, will not demote")
2972 result = utils.RunCmd([constants.DAEMON_UTIL, "check", constants.MASTERD])
2973 if not result.failed:
2974 _Fail("The master daemon is running, will not demote")
2977 if os.path.isfile(constants.CLUSTER_CONF_FILE):
2978 utils.CreateBackup(constants.CLUSTER_CONF_FILE)
2979 except EnvironmentError, err:
2980 if err.errno != errno.ENOENT:
2981 _Fail("Error while backing up cluster file: %s", err, exc=True)
2983 utils.RemoveFile(constants.CLUSTER_CONF_FILE)
2986 def _GetX509Filenames(cryptodir, name):
2987 """Returns the full paths for the private key and certificate.
2990 return (utils.PathJoin(cryptodir, name),
2991 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
2992 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
2995 def CreateX509Certificate(validity, cryptodir=constants.CRYPTO_KEYS_DIR):
2996 """Creates a new X509 certificate for SSL/TLS.
2999 @param validity: Validity in seconds
3000 @rtype: tuple; (string, string)
3001 @return: Certificate name and public part
3004 (key_pem, cert_pem) = \
3005 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3006 min(validity, _MAX_SSL_CERT_VALIDITY))
3008 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3009 prefix="x509-%s-" % utils.TimestampForFilename())
3011 name = os.path.basename(cert_dir)
3012 assert len(name) > 5
3014 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3016 utils.WriteFile(key_file, mode=0400, data=key_pem)
3017 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3019 # Never return private key as it shouldn't leave the node
3020 return (name, cert_pem)
3022 shutil.rmtree(cert_dir, ignore_errors=True)
3026 def RemoveX509Certificate(name, cryptodir=constants.CRYPTO_KEYS_DIR):
3027 """Removes a X509 certificate.
3030 @param name: Certificate name
3033 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3035 utils.RemoveFile(key_file)
3036 utils.RemoveFile(cert_file)
3040 except EnvironmentError, err:
3041 _Fail("Cannot remove certificate directory '%s': %s",
3045 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3046 """Returns the command for the requested input/output.
3048 @type instance: L{objects.Instance}
3049 @param instance: The instance object
3050 @param mode: Import/export mode
3051 @param ieio: Input/output type
3052 @param ieargs: Input/output arguments
3055 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3062 if ieio == constants.IEIO_FILE:
3063 (filename, ) = ieargs
3065 if not utils.IsNormAbsPath(filename):
3066 _Fail("Path '%s' is not normalized or absolute", filename)
3068 real_filename = os.path.realpath(filename)
3069 directory = os.path.dirname(real_filename)
3071 if not utils.IsBelowDir(constants.EXPORT_DIR, real_filename):
3072 _Fail("File '%s' is not under exports directory '%s': %s",
3073 filename, constants.EXPORT_DIR, real_filename)
3076 utils.Makedirs(directory, mode=0750)
3078 quoted_filename = utils.ShellQuote(filename)
3080 if mode == constants.IEM_IMPORT:
3081 suffix = "> %s" % quoted_filename
3082 elif mode == constants.IEM_EXPORT:
3083 suffix = "< %s" % quoted_filename
3085 # Retrieve file size
3087 st = os.stat(filename)
3088 except EnvironmentError, err:
3089 logging.error("Can't stat(2) %s: %s", filename, err)
3091 exp_size = utils.BytesToMebibyte(st.st_size)
3093 elif ieio == constants.IEIO_RAW_DISK:
3096 real_disk = _OpenRealBD(disk)
3098 if mode == constants.IEM_IMPORT:
3099 # we set here a smaller block size as, due to transport buffering, more
3100 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3101 # is not already there or we pass a wrong path; we use notrunc to no
3102 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3103 # much memory; this means that at best, we flush every 64k, which will
3105 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3106 " bs=%s oflag=dsync"),
3110 elif mode == constants.IEM_EXPORT:
3111 # the block size on the read dd is 1MiB to match our units
3112 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3114 str(1024 * 1024), # 1 MB
3116 exp_size = disk.size
3118 elif ieio == constants.IEIO_SCRIPT:
3119 (disk, disk_index, ) = ieargs
3121 assert isinstance(disk_index, (int, long))
3123 real_disk = _OpenRealBD(disk)
3125 inst_os = OSFromDisk(instance.os)
3126 env = OSEnvironment(instance, inst_os)
3128 if mode == constants.IEM_IMPORT:
3129 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3130 env["IMPORT_INDEX"] = str(disk_index)
3131 script = inst_os.import_script
3133 elif mode == constants.IEM_EXPORT:
3134 env["EXPORT_DEVICE"] = real_disk.dev_path
3135 env["EXPORT_INDEX"] = str(disk_index)
3136 script = inst_os.export_script
3138 # TODO: Pass special environment only to script
3139 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3141 if mode == constants.IEM_IMPORT:
3142 suffix = "| %s" % script_cmd
3144 elif mode == constants.IEM_EXPORT:
3145 prefix = "%s |" % script_cmd
3147 # Let script predict size
3148 exp_size = constants.IE_CUSTOM_SIZE
3151 _Fail("Invalid %s I/O mode %r", mode, ieio)
3153 return (env, prefix, suffix, exp_size)
3156 def _CreateImportExportStatusDir(prefix):
3157 """Creates status directory for import/export.
3160 return tempfile.mkdtemp(dir=constants.IMPORT_EXPORT_DIR,
3162 (prefix, utils.TimestampForFilename())))
3165 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3167 """Starts an import or export daemon.
3169 @param mode: Import/output mode
3170 @type opts: L{objects.ImportExportOptions}
3171 @param opts: Daemon options
3173 @param host: Remote host for export (None for import)
3175 @param port: Remote port for export (None for import)
3176 @type instance: L{objects.Instance}
3177 @param instance: Instance object
3178 @type component: string
3179 @param component: which part of the instance is transferred now,
3181 @param ieio: Input/output type
3182 @param ieioargs: Input/output arguments
3185 if mode == constants.IEM_IMPORT:
3188 if not (host is None and port is None):
3189 _Fail("Can not specify host or port on import")
3191 elif mode == constants.IEM_EXPORT:
3194 if host is None or port is None:
3195 _Fail("Host and port must be specified for an export")
3198 _Fail("Invalid mode %r", mode)
3200 if (opts.key_name is None) ^ (opts.ca_pem is None):
3201 _Fail("Cluster certificate can only be used for both key and CA")
3203 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3204 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3206 if opts.key_name is None:
3208 key_path = constants.NODED_CERT_FILE
3209 cert_path = constants.NODED_CERT_FILE
3210 assert opts.ca_pem is None
3212 (_, key_path, cert_path) = _GetX509Filenames(constants.CRYPTO_KEYS_DIR,
3214 assert opts.ca_pem is not None
3216 for i in [key_path, cert_path]:
3217 if not os.path.exists(i):
3218 _Fail("File '%s' does not exist" % i)
3220 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3222 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3223 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3224 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3226 if opts.ca_pem is None:
3228 ca = utils.ReadFile(constants.NODED_CERT_FILE)
3233 utils.WriteFile(ca_file, data=ca, mode=0400)
3236 constants.IMPORT_EXPORT_DAEMON,
3238 "--key=%s" % key_path,
3239 "--cert=%s" % cert_path,
3240 "--ca=%s" % ca_file,
3244 cmd.append("--host=%s" % host)
3247 cmd.append("--port=%s" % port)
3250 cmd.append("--ipv6")
3252 cmd.append("--ipv4")
3255 cmd.append("--compress=%s" % opts.compress)
3258 cmd.append("--magic=%s" % opts.magic)
3260 if exp_size is not None:
3261 cmd.append("--expected-size=%s" % exp_size)
3264 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3267 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3269 if mode == constants.IEM_EXPORT:
3270 # Retry connection a few times when connecting to remote peer
3271 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3272 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3273 elif opts.connect_timeout is not None:
3274 assert mode == constants.IEM_IMPORT
3275 # Overall timeout for establishing connection while listening
3276 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3278 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3280 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3281 # support for receiving a file descriptor for output
3282 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3285 # The import/export name is simply the status directory name
3286 return os.path.basename(status_dir)
3289 shutil.rmtree(status_dir, ignore_errors=True)
3293 def GetImportExportStatus(names):
3294 """Returns import/export daemon status.
3296 @type names: sequence
3297 @param names: List of names
3298 @rtype: List of dicts
3299 @return: Returns a list of the state of each named import/export or None if a
3300 status couldn't be read
3306 status_file = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name,
3310 data = utils.ReadFile(status_file)
3311 except EnvironmentError, err:
3312 if err.errno != errno.ENOENT:
3320 result.append(serializer.LoadJson(data))
3325 def AbortImportExport(name):
3326 """Sends SIGTERM to a running import/export daemon.
3329 logging.info("Abort import/export %s", name)
3331 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3332 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3335 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3337 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3340 def CleanupImportExport(name):
3341 """Cleanup after an import or export.
3343 If the import/export daemon is still running it's killed. Afterwards the
3344 whole status directory is removed.
3347 logging.info("Finalizing import/export %s", name)
3349 status_dir = utils.PathJoin(constants.IMPORT_EXPORT_DIR, name)
3351 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3354 logging.info("Import/export %s is still running with PID %s",
3356 utils.KillProcess(pid, waitpid=False)
3358 shutil.rmtree(status_dir, ignore_errors=True)
3361 def _FindDisks(nodes_ip, disks):
3362 """Sets the physical ID on disks and returns the block devices.
3365 # set the correct physical ID
3366 my_name = netutils.Hostname.GetSysName()
3368 cf.SetPhysicalID(my_name, nodes_ip)
3373 rd = _RecursiveFindBD(cf)
3375 _Fail("Can't find device %s", cf)
3380 def DrbdDisconnectNet(nodes_ip, disks):
3381 """Disconnects the network on a list of drbd devices.
3384 bdevs = _FindDisks(nodes_ip, disks)
3390 except errors.BlockDeviceError, err:
3391 _Fail("Can't change network configuration to standalone mode: %s",
3395 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3396 """Attaches the network on a list of drbd devices.
3399 bdevs = _FindDisks(nodes_ip, disks)
3402 for idx, rd in enumerate(bdevs):
3404 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3405 except EnvironmentError, err:
3406 _Fail("Can't create symlink: %s", err)
3407 # reconnect disks, switch to new master configuration and if
3408 # needed primary mode
3411 rd.AttachNet(multimaster)
3412 except errors.BlockDeviceError, err:
3413 _Fail("Can't change network configuration: %s", err)
3415 # wait until the disks are connected; we need to retry the re-attach
3416 # if the device becomes standalone, as this might happen if the one
3417 # node disconnects and reconnects in a different mode before the
3418 # other node reconnects; in this case, one or both of the nodes will
3419 # decide it has wrong configuration and switch to standalone
3422 all_connected = True
3425 stats = rd.GetProcStatus()
3427 all_connected = (all_connected and
3428 (stats.is_connected or stats.is_in_resync))
3430 if stats.is_standalone:
3431 # peer had different config info and this node became
3432 # standalone, even though this should not happen with the
3433 # new staged way of changing disk configs
3435 rd.AttachNet(multimaster)
3436 except errors.BlockDeviceError, err:
3437 _Fail("Can't change network configuration: %s", err)
3439 if not all_connected:
3440 raise utils.RetryAgain()
3443 # Start with a delay of 100 miliseconds and go up to 5 seconds
3444 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3445 except utils.RetryTimeout:
3446 _Fail("Timeout in disk reconnecting")
3449 # change to primary mode
3453 except errors.BlockDeviceError, err:
3454 _Fail("Can't change to primary mode: %s", err)
3457 def DrbdWaitSync(nodes_ip, disks):
3458 """Wait until DRBDs have synchronized.
3462 stats = rd.GetProcStatus()
3463 if not (stats.is_connected or stats.is_in_resync):
3464 raise utils.RetryAgain()
3467 bdevs = _FindDisks(nodes_ip, disks)
3473 # poll each second for 15 seconds
3474 stats = utils.Retry(_helper, 1, 15, args=[rd])
3475 except utils.RetryTimeout:
3476 stats = rd.GetProcStatus()
3478 if not (stats.is_connected or stats.is_in_resync):
3479 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3480 alldone = alldone and (not stats.is_in_resync)
3481 if stats.sync_percent is not None:
3482 min_resync = min(min_resync, stats.sync_percent)
3484 return (alldone, min_resync)
3487 def GetDrbdUsermodeHelper():
3488 """Returns DRBD usermode helper currently configured.
3492 return bdev.BaseDRBD.GetUsermodeHelper()
3493 except errors.BlockDeviceError, err:
3497 def PowercycleNode(hypervisor_type):
3498 """Hard-powercycle the node.
3500 Because we need to return first, and schedule the powercycle in the
3501 background, we won't be able to report failures nicely.
3504 hyper = hypervisor.GetHypervisor(hypervisor_type)
3508 # if we can't fork, we'll pretend that we're in the child process
3511 return "Reboot scheduled in 5 seconds"
3512 # ensure the child is running on ram
3515 except Exception: # pylint: disable=W0703
3518 hyper.PowercycleNode()
3521 class HooksRunner(object):
3524 This class is instantiated on the node side (ganeti-noded) and not
3528 def __init__(self, hooks_base_dir=None):
3529 """Constructor for hooks runner.
3531 @type hooks_base_dir: str or None
3532 @param hooks_base_dir: if not None, this overrides the
3533 L{constants.HOOKS_BASE_DIR} (useful for unittests)
3536 if hooks_base_dir is None:
3537 hooks_base_dir = constants.HOOKS_BASE_DIR
3538 # yeah, _BASE_DIR is not valid for attributes, we use it like a
3540 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
3542 def RunLocalHooks(self, node_list, hpath, phase, env):
3543 """Check that the hooks will be run only locally and then run them.
3546 assert len(node_list) == 1
3548 _, myself = ssconf.GetMasterAndMyself()
3549 assert node == myself
3551 results = self.RunHooks(hpath, phase, env)
3553 # Return values in the form expected by HooksMaster
3554 return {node: (None, False, results)}
3556 def RunHooks(self, hpath, phase, env):
3557 """Run the scripts in the hooks directory.
3560 @param hpath: the path to the hooks directory which
3563 @param phase: either L{constants.HOOKS_PHASE_PRE} or
3564 L{constants.HOOKS_PHASE_POST}
3566 @param env: dictionary with the environment for the hook
3568 @return: list of 3-element tuples:
3570 - script result, either L{constants.HKR_SUCCESS} or
3571 L{constants.HKR_FAIL}
3572 - output of the script
3574 @raise errors.ProgrammerError: for invalid input
3578 if phase == constants.HOOKS_PHASE_PRE:
3580 elif phase == constants.HOOKS_PHASE_POST:
3583 _Fail("Unknown hooks phase '%s'", phase)
3585 subdir = "%s-%s.d" % (hpath, suffix)
3586 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3590 if not os.path.isdir(dir_name):
3591 # for non-existing/non-dirs, we simply exit instead of logging a
3592 # warning at every operation
3595 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3597 for (relname, relstatus, runresult) in runparts_results:
3598 if relstatus == constants.RUNPARTS_SKIP:
3599 rrval = constants.HKR_SKIP
3601 elif relstatus == constants.RUNPARTS_ERR:
3602 rrval = constants.HKR_FAIL
3603 output = "Hook script execution error: %s" % runresult
3604 elif relstatus == constants.RUNPARTS_RUN:
3605 if runresult.failed:
3606 rrval = constants.HKR_FAIL
3608 rrval = constants.HKR_SUCCESS
3609 output = utils.SafeEncode(runresult.output.strip())
3610 results.append(("%s/%s" % (subdir, relname), rrval, output))
3615 class IAllocatorRunner(object):
3616 """IAllocator runner.
3618 This class is instantiated on the node side (ganeti-noded) and not on
3623 def Run(name, idata):
3624 """Run an iallocator script.
3627 @param name: the iallocator script name
3629 @param idata: the allocator input data
3632 @return: two element tuple of:
3634 - either error message or stdout of allocator (for success)
3637 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3639 if alloc_script is None:
3640 _Fail("iallocator module '%s' not found in the search path", name)
3642 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3646 result = utils.RunCmd([alloc_script, fin_name])
3648 _Fail("iallocator module '%s' failed: %s, output '%s'",
3649 name, result.fail_reason, result.output)
3653 return result.stdout
3656 class DevCacheManager(object):
3657 """Simple class for managing a cache of block device information.
3660 _DEV_PREFIX = "/dev/"
3661 _ROOT_DIR = constants.BDEV_CACHE_DIR
3664 def _ConvertPath(cls, dev_path):
3665 """Converts a /dev/name path to the cache file name.
3667 This replaces slashes with underscores and strips the /dev
3668 prefix. It then returns the full path to the cache file.
3671 @param dev_path: the C{/dev/} path name
3673 @return: the converted path name
3676 if dev_path.startswith(cls._DEV_PREFIX):
3677 dev_path = dev_path[len(cls._DEV_PREFIX):]
3678 dev_path = dev_path.replace("/", "_")
3679 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3683 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
3684 """Updates the cache information for a given device.
3687 @param dev_path: the pathname of the device
3689 @param owner: the owner (instance name) of the device
3690 @type on_primary: bool
3691 @param on_primary: whether this is the primary
3694 @param iv_name: the instance-visible name of the
3695 device, as in objects.Disk.iv_name
3700 if dev_path is None:
3701 logging.error("DevCacheManager.UpdateCache got a None dev_path")
3703 fpath = cls._ConvertPath(dev_path)
3709 iv_name = "not_visible"
3710 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
3712 utils.WriteFile(fpath, data=fdata)
3713 except EnvironmentError, err:
3714 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
3717 def RemoveCache(cls, dev_path):
3718 """Remove data for a dev_path.
3720 This is just a wrapper over L{utils.io.RemoveFile} with a converted
3721 path name and logging.
3724 @param dev_path: the pathname of the device
3729 if dev_path is None:
3730 logging.error("DevCacheManager.RemoveCache got a None dev_path")
3732 fpath = cls._ConvertPath(dev_path)
3734 utils.RemoveFile(fpath)
3735 except EnvironmentError, err:
3736 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)