4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 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.storage import bdev
58 from ganeti.storage import drbd
59 from ganeti.storage import filestorage
60 from ganeti import objects
61 from ganeti import ssconf
62 from ganeti import serializer
63 from ganeti import netutils
64 from ganeti import runtime
65 from ganeti import compat
66 from ganeti import pathutils
67 from ganeti import vcluster
69 from ganeti.storage.base import BlockDev
70 from ganeti.storage.drbd import DRBD8
71 from ganeti import hooksmaster
74 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
75 _ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
77 pathutils.JOB_QUEUE_ARCHIVE_DIR,
79 pathutils.CRYPTO_KEYS_DIR,
81 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
82 _X509_KEY_FILE = "key"
83 _X509_CERT_FILE = "cert"
84 _IES_STATUS_FILE = "status"
88 #: Valid LVS output line regex
89 _LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
91 # Actions for the master setup script
92 _MASTER_START = "start"
95 #: Maximum file permissions for restricted command directory and executables
96 _RCMD_MAX_MODE = (stat.S_IRWXU |
97 stat.S_IRGRP | stat.S_IXGRP |
98 stat.S_IROTH | stat.S_IXOTH)
100 #: Delay before returning an error for restricted commands
101 _RCMD_INVALID_DELAY = 10
103 #: How long to wait to acquire lock for restricted commands (shorter than
104 #: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
105 #: command requests arrive
106 _RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
109 class RPCFail(Exception):
110 """Class denoting RPC failure.
112 Its argument is the error message.
117 def _GetInstReasonFilename(instance_name):
118 """Path of the file containing the reason of the instance status change.
120 @type instance_name: string
121 @param instance_name: The name of the instance
123 @return: The path of the file
126 return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
129 def _StoreInstReasonTrail(instance_name, trail):
130 """Serialize a reason trail related to an instance change of state to file.
132 The exact location of the file depends on the name of the instance and on
133 the configuration of the Ganeti cluster defined at deploy time.
135 @type instance_name: string
136 @param instance_name: The name of the instance
140 json = serializer.DumpJson(trail)
141 filename = _GetInstReasonFilename(instance_name)
142 utils.WriteFile(filename, data=json)
145 def _Fail(msg, *args, **kwargs):
146 """Log an error and the raise an RPCFail exception.
148 This exception is then handled specially in the ganeti daemon and
149 turned into a 'failed' return type. As such, this function is a
150 useful shortcut for logging the error and returning it to the master
154 @param msg: the text of the exception
160 if "log" not in kwargs or kwargs["log"]: # if we should log this error
161 if "exc" in kwargs and kwargs["exc"]:
162 logging.exception(msg)
169 """Simple wrapper to return a SimpleStore.
171 @rtype: L{ssconf.SimpleStore}
172 @return: a SimpleStore instance
175 return ssconf.SimpleStore()
178 def _GetSshRunner(cluster_name):
179 """Simple wrapper to return an SshRunner.
181 @type cluster_name: str
182 @param cluster_name: the cluster name, which is needed
183 by the SshRunner constructor
184 @rtype: L{ssh.SshRunner}
185 @return: an SshRunner instance
188 return ssh.SshRunner(cluster_name)
191 def _Decompress(data):
192 """Unpacks data compressed by the RPC client.
194 @type data: list or tuple
195 @param data: Data sent by RPC client
197 @return: Decompressed data
200 assert isinstance(data, (list, tuple))
201 assert len(data) == 2
202 (encoding, content) = data
203 if encoding == constants.RPC_ENCODING_NONE:
205 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
206 return zlib.decompress(base64.b64decode(content))
208 raise AssertionError("Unknown data encoding")
211 def _CleanDirectory(path, exclude=None):
212 """Removes all regular files in a directory.
215 @param path: the directory to clean
217 @param exclude: list of files to be excluded, defaults
221 if path not in _ALLOWED_CLEAN_DIRS:
222 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
225 if not os.path.isdir(path):
230 # Normalize excluded paths
231 exclude = [os.path.normpath(i) for i in exclude]
233 for rel_name in utils.ListVisibleFiles(path):
234 full_name = utils.PathJoin(path, rel_name)
235 if full_name in exclude:
237 if os.path.isfile(full_name) and not os.path.islink(full_name):
238 utils.RemoveFile(full_name)
241 def _BuildUploadFileList():
242 """Build the list of allowed upload files.
244 This is abstracted so that it's built only once at module import time.
247 allowed_files = set([
248 pathutils.CLUSTER_CONF_FILE,
250 pathutils.SSH_KNOWN_HOSTS_FILE,
251 pathutils.VNC_PASSWORD_FILE,
252 pathutils.RAPI_CERT_FILE,
253 pathutils.SPICE_CERT_FILE,
254 pathutils.SPICE_CACERT_FILE,
255 pathutils.RAPI_USERS_FILE,
256 pathutils.CONFD_HMAC_KEY,
257 pathutils.CLUSTER_DOMAIN_SECRET_FILE,
260 for hv_name in constants.HYPER_TYPES:
261 hv_class = hypervisor.GetHypervisorClass(hv_name)
262 allowed_files.update(hv_class.GetAncillaryFiles()[0])
264 assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
265 "Allowed file storage paths should never be uploaded via RPC"
267 return frozenset(allowed_files)
270 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
274 """Removes job queue files and archived jobs.
280 _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
281 _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
285 """Returns master information.
287 This is an utility function to compute master information, either
288 for consumption here or from the node daemon.
291 @return: master_netdev, master_ip, master_name, primary_ip_family,
293 @raise RPCFail: in case of errors
298 master_netdev = cfg.GetMasterNetdev()
299 master_ip = cfg.GetMasterIP()
300 master_netmask = cfg.GetMasterNetmask()
301 master_node = cfg.GetMasterNode()
302 primary_ip_family = cfg.GetPrimaryIPFamily()
303 except errors.ConfigurationError, err:
304 _Fail("Cluster configuration incomplete: %s", err, exc=True)
305 return (master_netdev, master_ip, master_node, primary_ip_family,
309 def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
310 """Decorator that runs hooks before and after the decorated function.
312 @type hook_opcode: string
313 @param hook_opcode: opcode of the hook
314 @type hooks_path: string
315 @param hooks_path: path of the hooks
316 @type env_builder_fn: function
317 @param env_builder_fn: function that returns a dictionary containing the
318 environment variables for the hooks. Will get all the parameters of the
320 @raise RPCFail: in case of pre-hook failure
324 def wrapper(*args, **kwargs):
325 _, myself = ssconf.GetMasterAndMyself()
326 nodes = ([myself], [myself]) # these hooks run locally
328 env_fn = compat.partial(env_builder_fn, *args, **kwargs)
332 hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
333 hr.RunLocalHooks, None, env_fn,
334 logging.warning, cfg.GetClusterName(),
336 hm.RunPhase(constants.HOOKS_PHASE_PRE)
337 result = fn(*args, **kwargs)
338 hm.RunPhase(constants.HOOKS_PHASE_POST)
345 def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
346 """Builds environment variables for master IP hooks.
348 @type master_params: L{objects.MasterNetworkParameters}
349 @param master_params: network parameters of the master
350 @type use_external_mip_script: boolean
351 @param use_external_mip_script: whether to use an external master IP
352 address setup script (unused, but necessary per the implementation of the
353 _RunLocalHooks decorator)
356 # pylint: disable=W0613
357 ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
359 "MASTER_NETDEV": master_params.netdev,
360 "MASTER_IP": master_params.ip,
361 "MASTER_NETMASK": str(master_params.netmask),
362 "CLUSTER_IP_VERSION": str(ver),
368 def _RunMasterSetupScript(master_params, action, use_external_mip_script):
369 """Execute the master IP address setup script.
371 @type master_params: L{objects.MasterNetworkParameters}
372 @param master_params: network parameters of the master
374 @param action: action to pass to the script. Must be one of
375 L{backend._MASTER_START} or L{backend._MASTER_STOP}
376 @type use_external_mip_script: boolean
377 @param use_external_mip_script: whether to use an external master IP
379 @raise backend.RPCFail: if there are errors during the execution of the
383 env = _BuildMasterIpEnv(master_params)
385 if use_external_mip_script:
386 setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
388 setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
390 result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
393 _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
394 (action, result.exit_code, result.output), log=True)
397 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
399 def ActivateMasterIp(master_params, use_external_mip_script):
400 """Activate the IP address of the master daemon.
402 @type master_params: L{objects.MasterNetworkParameters}
403 @param master_params: network parameters of the master
404 @type use_external_mip_script: boolean
405 @param use_external_mip_script: whether to use an external master IP
407 @raise RPCFail: in case of errors during the IP startup
410 _RunMasterSetupScript(master_params, _MASTER_START,
411 use_external_mip_script)
414 def StartMasterDaemons(no_voting):
415 """Activate local node as master node.
417 The function will start the master daemons (ganeti-masterd and ganeti-rapi).
419 @type no_voting: boolean
420 @param no_voting: whether to start ganeti-masterd without a node vote
421 but still non-interactively
427 masterd_args = "--no-voting --yes-do-it"
432 "EXTRA_MASTERD_ARGS": masterd_args,
435 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
437 msg = "Can't start Ganeti master: %s" % result.output
442 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
444 def DeactivateMasterIp(master_params, use_external_mip_script):
445 """Deactivate the master IP on this node.
447 @type master_params: L{objects.MasterNetworkParameters}
448 @param master_params: network parameters of the master
449 @type use_external_mip_script: boolean
450 @param use_external_mip_script: whether to use an external master IP
452 @raise RPCFail: in case of errors during the IP turndown
455 _RunMasterSetupScript(master_params, _MASTER_STOP,
456 use_external_mip_script)
459 def StopMasterDaemons():
460 """Stop the master daemons on this node.
462 Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
467 # TODO: log and report back to the caller the error failures; we
468 # need to decide in which case we fail the RPC for this
470 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
472 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
474 result.cmd, result.exit_code, result.output)
477 def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
478 """Change the netmask of the master IP.
480 @param old_netmask: the old value of the netmask
481 @param netmask: the new value of the netmask
482 @param master_ip: the master IP
483 @param master_netdev: the master network device
486 if old_netmask == netmask:
489 if not netutils.IPAddress.Own(master_ip):
490 _Fail("The master IP address is not up, not attempting to change its"
493 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
494 "%s/%s" % (master_ip, netmask),
495 "dev", master_netdev, "label",
496 "%s:0" % master_netdev])
498 _Fail("Could not set the new netmask on the master IP address")
500 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
501 "%s/%s" % (master_ip, old_netmask),
502 "dev", master_netdev, "label",
503 "%s:0" % master_netdev])
505 _Fail("Could not bring down the master IP address with the old netmask")
508 def EtcHostsModify(mode, host, ip):
509 """Modify a host entry in /etc/hosts.
511 @param mode: The mode to operate. Either add or remove entry
512 @param host: The host to operate on
513 @param ip: The ip associated with the entry
516 if mode == constants.ETC_HOSTS_ADD:
518 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
520 utils.AddHostToEtcHosts(host, ip)
521 elif mode == constants.ETC_HOSTS_REMOVE:
523 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
524 " parameter is present")
525 utils.RemoveHostFromEtcHosts(host)
527 RPCFail("Mode not supported")
530 def LeaveCluster(modify_ssh_setup):
531 """Cleans up and remove the current node.
533 This function cleans up and prepares the current node to be removed
536 If processing is successful, then it raises an
537 L{errors.QuitGanetiException} which is used as a special case to
538 shutdown the node daemon.
540 @param modify_ssh_setup: boolean
543 _CleanDirectory(pathutils.DATA_DIR)
544 _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
549 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
551 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
553 utils.RemoveFile(priv_key)
554 utils.RemoveFile(pub_key)
555 except errors.OpExecError:
556 logging.exception("Error while processing ssh files")
559 utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
560 utils.RemoveFile(pathutils.RAPI_CERT_FILE)
561 utils.RemoveFile(pathutils.SPICE_CERT_FILE)
562 utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
563 utils.RemoveFile(pathutils.NODED_CERT_FILE)
564 except: # pylint: disable=W0702
565 logging.exception("Error while removing cluster secrets")
567 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
569 logging.error("Command %s failed with exitcode %s and error %s",
570 result.cmd, result.exit_code, result.output)
572 # Raise a custom exception (handled in ganeti-noded)
573 raise errors.QuitGanetiException(True, "Shutdown scheduled")
576 def _CheckStorageParams(params, num_params):
577 """Performs sanity checks for storage parameters.
580 @param params: list of storage parameters
581 @type num_params: int
582 @param num_params: expected number of parameters
586 raise errors.ProgrammerError("No storage parameters for storage"
587 " reporting is provided.")
588 if not isinstance(params, list):
589 raise errors.ProgrammerError("The storage parameters are not of type"
590 " list: '%s'" % params)
591 if not len(params) == num_params:
592 raise errors.ProgrammerError("Did not receive the expected number of"
593 "storage parameters: expected %s,"
594 " received '%s'" % (num_params, len(params)))
597 def _CheckLvmStorageParams(params):
598 """Performs sanity check for the 'exclusive storage' flag.
600 @see: C{_CheckStorageParams}
603 _CheckStorageParams(params, 1)
604 excl_stor = params[0]
605 if not isinstance(params[0], bool):
606 raise errors.ProgrammerError("Exclusive storage parameter is not"
607 " boolean: '%s'." % excl_stor)
611 def _GetLvmVgSpaceInfo(name, params):
612 """Wrapper around C{_GetVgInfo} which checks the storage parameters.
615 @param name: name of the volume group
617 @param params: list of storage parameters, which in this case should be
618 containing only one for exclusive storage
621 excl_stor = _CheckLvmStorageParams(params)
622 return _GetVgInfo(name, excl_stor)
626 name, excl_stor, info_fn=bdev.LogicalVolume.GetVGInfo):
627 """Retrieves information about a LVM volume group.
630 # TODO: GetVGInfo supports returning information for multiple VGs at once
631 vginfo = info_fn([name], excl_stor)
633 vg_free = int(round(vginfo[0][0], 0))
634 vg_size = int(round(vginfo[0][1], 0))
640 "type": constants.ST_LVM_VG,
642 "storage_free": vg_free,
643 "storage_size": vg_size,
647 def _GetLvmPvSpaceInfo(name, params):
648 """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
650 @see: C{_GetLvmVgSpaceInfo}
653 excl_stor = _CheckLvmStorageParams(params)
654 return _GetVgSpindlesInfo(name, excl_stor)
657 def _GetVgSpindlesInfo(
658 name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
659 """Retrieves information about spindles in an LVM volume group.
663 @type excl_stor: bool
664 @param excl_stor: exclusive storage
666 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
667 free spindles, total spindles respectively
671 (vg_free, vg_size) = info_fn(name)
676 "type": constants.ST_LVM_PV,
678 "storage_free": vg_free,
679 "storage_size": vg_size,
683 def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
684 """Retrieves node information from a hypervisor.
686 The information returned depends on the hypervisor. Common items:
688 - vg_size is the size of the configured volume group in MiB
689 - vg_free is the free size of the volume group in MiB
690 - memory_dom0 is the memory allocated for domain0 in MiB
691 - memory_free is the currently available (free) ram in MiB
692 - memory_total is the total number of ram in MiB
693 - hv_version: the hypervisor version, if available
695 @type hvparams: dict of string
696 @param hvparams: the hypervisor's hvparams
699 return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
702 def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
703 """Retrieves node information for all hypervisors.
705 See C{_GetHvInfo} for information on the output.
707 @type hv_specs: list of pairs (string, dict of strings)
708 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
715 for hvname, hvparams in hv_specs:
716 result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
720 def _GetNamedNodeInfo(names, fn):
721 """Calls C{fn} for all names in C{names} and returns a dictionary.
729 return map(fn, names)
732 def GetNodeInfo(storage_units, hv_specs):
733 """Gives back a hash with different information about the node.
735 @type storage_units: list of tuples (string, string, list)
736 @param storage_units: List of tuples (storage unit, identifier, parameters) to
737 ask for disk space information. In case of lvm-vg, the identifier is
738 the VG name. The parameters can contain additional, storage-type-specific
739 parameters, for example exclusive storage for lvm storage.
740 @type hv_specs: list of pairs (string, dict of strings)
741 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
742 @rtype: tuple; (string, None/dict, None/dict)
743 @return: Tuple containing boot ID, volume group information and hypervisor
747 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
748 storage_info = _GetNamedNodeInfo(
750 (lambda (storage_type, storage_key, storage_params):
751 _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
752 hv_info = _GetHvInfoAll(hv_specs)
753 return (bootid, storage_info, hv_info)
756 def _GetFileStorageSpaceInfo(path, params):
757 """Wrapper around filestorage.GetSpaceInfo.
759 The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
760 and ignore the *args parameter to not leak it into the filestorage
763 @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
767 _CheckStorageParams(params, 0)
768 return filestorage.GetFileStorageSpaceInfo(path)
771 # FIXME: implement storage reporting for all missing storage types.
772 _STORAGE_TYPE_INFO_FN = {
773 constants.ST_BLOCK: None,
774 constants.ST_DISKLESS: None,
775 constants.ST_EXT: None,
776 constants.ST_FILE: _GetFileStorageSpaceInfo,
777 constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
778 constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
779 constants.ST_RADOS: None,
783 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
784 """Looks up and applies the correct function to calculate free and total
785 storage for the given storage type.
787 @type storage_type: string
788 @param storage_type: the storage type for which the storage shall be reported.
789 @type storage_key: string
790 @param storage_key: identifier of a storage unit, e.g. the volume group name
791 of an LVM storage unit
793 @param args: various parameters that can be used for storage reporting. These
794 parameters and their semantics vary from storage type to storage type and
795 are just propagated in this function.
796 @return: the results of the application of the storage space function (see
797 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
799 @raises NotImplementedError: for storage types who don't support space
802 fn = _STORAGE_TYPE_INFO_FN[storage_type]
804 return fn(storage_key, *args)
806 raise NotImplementedError
809 def _CheckExclusivePvs(pvi_list):
810 """Check that PVs are not shared among LVs
812 @type pvi_list: list of L{objects.LvmPvInfo} objects
813 @param pvi_list: information about the PVs
815 @rtype: list of tuples (string, list of strings)
816 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
821 if len(pvi.lv_list) > 1:
822 res.append((pvi.name, pvi.lv_list))
826 def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
827 get_hv_fn=hypervisor.GetHypervisor):
828 """Verifies the hypervisor. Appends the results to the 'results' list.
831 @param what: a dictionary of things to check
832 @type vm_capable: boolean
833 @param vm_capable: whether or not this node is vm capable
835 @param result: dictionary of verification results; results of the
836 verifications in this function will be added here
837 @type all_hvparams: dict of dict of string
838 @param all_hvparams: dictionary mapping hypervisor names to hvparams
839 @type get_hv_fn: function
840 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
846 if constants.NV_HYPERVISOR in what:
847 result[constants.NV_HYPERVISOR] = {}
848 for hv_name in what[constants.NV_HYPERVISOR]:
849 hvparams = all_hvparams[hv_name]
851 val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
852 except errors.HypervisorError, err:
853 val = "Error while checking hypervisor: %s" % str(err)
854 result[constants.NV_HYPERVISOR][hv_name] = val
857 def _VerifyHvparams(what, vm_capable, result,
858 get_hv_fn=hypervisor.GetHypervisor):
859 """Verifies the hvparams. Appends the results to the 'results' list.
862 @param what: a dictionary of things to check
863 @type vm_capable: boolean
864 @param vm_capable: whether or not this node is vm capable
866 @param result: dictionary of verification results; results of the
867 verifications in this function will be added here
868 @type get_hv_fn: function
869 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
875 if constants.NV_HVPARAMS in what:
876 result[constants.NV_HVPARAMS] = []
877 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
879 logging.info("Validating hv %s, %s", hv_name, hvparms)
880 get_hv_fn(hv_name).ValidateParameters(hvparms)
881 except errors.HypervisorError, err:
882 result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
885 def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
886 """Verifies the instance list.
889 @param what: a dictionary of things to check
890 @type vm_capable: boolean
891 @param vm_capable: whether or not this node is vm capable
893 @param result: dictionary of verification results; results of the
894 verifications in this function will be added here
895 @type all_hvparams: dict of dict of string
896 @param all_hvparams: dictionary mapping hypervisor names to hvparams
899 if constants.NV_INSTANCELIST in what and vm_capable:
900 # GetInstanceList can fail
902 val = GetInstanceList(what[constants.NV_INSTANCELIST],
903 all_hvparams=all_hvparams)
906 result[constants.NV_INSTANCELIST] = val
909 def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
910 """Verifies the node info.
913 @param what: a dictionary of things to check
914 @type vm_capable: boolean
915 @param vm_capable: whether or not this node is vm capable
917 @param result: dictionary of verification results; results of the
918 verifications in this function will be added here
919 @type all_hvparams: dict of dict of string
920 @param all_hvparams: dictionary mapping hypervisor names to hvparams
923 if constants.NV_HVINFO in what and vm_capable:
924 hvname = what[constants.NV_HVINFO]
925 hyper = hypervisor.GetHypervisor(hvname)
926 hvparams = all_hvparams[hvname]
927 result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
930 def VerifyNode(what, cluster_name, all_hvparams):
931 """Verify the status of the local node.
933 Based on the input L{what} parameter, various checks are done on the
936 If the I{filelist} key is present, this list of
937 files is checksummed and the file/checksum pairs are returned.
939 If the I{nodelist} key is present, we check that we have
940 connectivity via ssh with the target nodes (and check the hostname
943 If the I{node-net-test} key is present, we check that we have
944 connectivity to the given nodes via both primary IP and, if
945 applicable, secondary IPs.
948 @param what: a dictionary of things to check:
949 - filelist: list of files for which to compute checksums
950 - nodelist: list of nodes we should check ssh communication with
951 - node-net-test: list of nodes we should check node daemon port
953 - hypervisor: list with hypervisors to run the verify for
954 @type cluster_name: string
955 @param cluster_name: the cluster's name
956 @type all_hvparams: dict of dict of strings
957 @param all_hvparams: a dictionary mapping hypervisor names to hvparams
959 @return: a dictionary with the same keys as the input dict, and
960 values representing the result of the checks
964 my_name = netutils.Hostname.GetSysName()
965 port = netutils.GetDaemonPort(constants.NODED)
966 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
968 _VerifyHypervisors(what, vm_capable, result, all_hvparams)
969 _VerifyHvparams(what, vm_capable, result)
971 if constants.NV_FILELIST in what:
972 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
973 what[constants.NV_FILELIST]))
974 result[constants.NV_FILELIST] = \
975 dict((vcluster.MakeVirtualPath(key), value)
976 for (key, value) in fingerprints.items())
978 if constants.NV_NODELIST in what:
979 (nodes, bynode) = what[constants.NV_NODELIST]
981 # Add nodes from other groups (different for each node)
983 nodes.extend(bynode[my_name])
988 random.shuffle(nodes)
990 # Try to contact all nodes
993 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
997 result[constants.NV_NODELIST] = val
999 if constants.NV_NODENETTEST in what:
1000 result[constants.NV_NODENETTEST] = tmp = {}
1001 my_pip = my_sip = None
1002 for name, pip, sip in what[constants.NV_NODENETTEST]:
1008 tmp[my_name] = ("Can't find my own primary/secondary IP"
1009 " in the node list")
1011 for name, pip, sip in what[constants.NV_NODENETTEST]:
1013 if not netutils.TcpPing(pip, port, source=my_pip):
1014 fail.append("primary")
1016 if not netutils.TcpPing(sip, port, source=my_sip):
1017 fail.append("secondary")
1019 tmp[name] = ("failure using the %s interface(s)" %
1022 if constants.NV_MASTERIP in what:
1023 # FIXME: add checks on incoming data structures (here and in the
1024 # rest of the function)
1025 master_name, master_ip = what[constants.NV_MASTERIP]
1026 if master_name == my_name:
1027 source = constants.IP4_ADDRESS_LOCALHOST
1030 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
1033 if constants.NV_USERSCRIPTS in what:
1034 result[constants.NV_USERSCRIPTS] = \
1035 [script for script in what[constants.NV_USERSCRIPTS]
1036 if not utils.IsExecutable(script)]
1038 if constants.NV_OOB_PATHS in what:
1039 result[constants.NV_OOB_PATHS] = tmp = []
1040 for path in what[constants.NV_OOB_PATHS]:
1043 except OSError, err:
1044 tmp.append("error stating out of band helper: %s" % err)
1046 if stat.S_ISREG(st.st_mode):
1047 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
1050 tmp.append("out of band helper %s is not executable" % path)
1052 tmp.append("out of band helper %s is not a file" % path)
1054 if constants.NV_LVLIST in what and vm_capable:
1056 val = GetVolumeList(utils.ListVolumeGroups().keys())
1057 except RPCFail, err:
1059 result[constants.NV_LVLIST] = val
1061 _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1063 if constants.NV_VGLIST in what and vm_capable:
1064 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1066 if constants.NV_PVLIST in what and vm_capable:
1067 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1068 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1069 filter_allocatable=False,
1070 include_lvs=check_exclusive_pvs)
1071 if check_exclusive_pvs:
1072 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1074 # Avoid sending useless data on the wire
1076 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1078 if constants.NV_VERSION in what:
1079 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1080 constants.RELEASE_VERSION)
1082 _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1084 if constants.NV_DRBDVERSION in what and vm_capable:
1086 drbd_version = DRBD8.GetProcInfo().GetVersionString()
1087 except errors.BlockDeviceError, err:
1088 logging.warning("Can't get DRBD version", exc_info=True)
1089 drbd_version = str(err)
1090 result[constants.NV_DRBDVERSION] = drbd_version
1092 if constants.NV_DRBDLIST in what and vm_capable:
1094 used_minors = drbd.DRBD8.GetUsedDevs()
1095 except errors.BlockDeviceError, err:
1096 logging.warning("Can't get used minors list", exc_info=True)
1097 used_minors = str(err)
1098 result[constants.NV_DRBDLIST] = used_minors
1100 if constants.NV_DRBDHELPER in what and vm_capable:
1103 payload = drbd.DRBD8.GetUsermodeHelper()
1104 except errors.BlockDeviceError, err:
1105 logging.error("Can't get DRBD usermode helper: %s", str(err))
1108 result[constants.NV_DRBDHELPER] = (status, payload)
1110 if constants.NV_NODESETUP in what:
1111 result[constants.NV_NODESETUP] = tmpr = []
1112 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1113 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1114 " under /sys, missing required directories /sys/block"
1115 " and /sys/class/net")
1116 if (not os.path.isdir("/proc/sys") or
1117 not os.path.isfile("/proc/sysrq-trigger")):
1118 tmpr.append("The procfs filesystem doesn't seem to be mounted"
1119 " under /proc, missing required directory /proc/sys and"
1120 " the file /proc/sysrq-trigger")
1122 if constants.NV_TIME in what:
1123 result[constants.NV_TIME] = utils.SplitTime(time.time())
1125 if constants.NV_OSLIST in what and vm_capable:
1126 result[constants.NV_OSLIST] = DiagnoseOS()
1128 if constants.NV_BRIDGES in what and vm_capable:
1129 result[constants.NV_BRIDGES] = [bridge
1130 for bridge in what[constants.NV_BRIDGES]
1131 if not utils.BridgeExists(bridge)]
1133 if what.get(constants.NV_ACCEPTED_STORAGE_PATHS) == my_name:
1134 result[constants.NV_ACCEPTED_STORAGE_PATHS] = \
1135 filestorage.ComputeWrongFileStoragePaths()
1137 if what.get(constants.NV_FILE_STORAGE_PATH):
1138 pathresult = filestorage.CheckFileStoragePath(
1139 what[constants.NV_FILE_STORAGE_PATH])
1141 result[constants.NV_FILE_STORAGE_PATH] = pathresult
1146 def GetBlockDevSizes(devices):
1147 """Return the size of the given block devices
1150 @param devices: list of block device nodes to query
1153 dictionary of all block devices under /dev (key). The value is their
1156 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1159 DEV_PREFIX = "/dev/"
1162 for devpath in devices:
1163 if not utils.IsBelowDir(DEV_PREFIX, devpath):
1167 st = os.stat(devpath)
1168 except EnvironmentError, err:
1169 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1172 if stat.S_ISBLK(st.st_mode):
1173 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1175 # We don't want to fail, just do not list this device as available
1176 logging.warning("Cannot get size for block device %s", devpath)
1179 size = int(result.stdout) / (1024 * 1024)
1180 blockdevs[devpath] = size
1184 def GetVolumeList(vg_names):
1185 """Compute list of logical volumes and their size.
1187 @type vg_names: list
1188 @param vg_names: the volume groups whose LVs we should list, or
1189 empty for all volume groups
1192 dictionary of all partions (key) with value being a tuple of
1193 their size (in MiB), inactive and online status::
1195 {'xenvg/test1': ('20.06', True, True)}
1197 in case of errors, a string is returned with the error
1205 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1206 "--separator=%s" % sep,
1207 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1209 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1211 for line in result.stdout.splitlines():
1213 match = _LVSLINE_REGEX.match(line)
1215 logging.error("Invalid line returned from lvs output: '%s'", line)
1217 vg_name, name, size, attr = match.groups()
1218 inactive = attr[4] == "-"
1219 online = attr[5] == "o"
1220 virtual = attr[0] == "v"
1222 # we don't want to report such volumes as existing, since they
1223 # don't really hold data
1225 lvs[vg_name + "/" + name] = (size, inactive, online)
1230 def ListVolumeGroups():
1231 """List the volume groups and their size.
1234 @return: dictionary with keys volume name and values the
1238 return utils.ListVolumeGroups()
1242 """List all volumes on this node.
1246 A list of dictionaries, each having four keys:
1247 - name: the logical volume name,
1248 - size: the size of the logical volume
1249 - dev: the physical device on which the LV lives
1250 - vg: the volume group to which it belongs
1252 In case of errors, we return an empty list and log the
1255 Note that since a logical volume can live on multiple physical
1256 volumes, the resulting list might include a logical volume
1260 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1262 "--options=lv_name,lv_size,devices,vg_name"])
1264 _Fail("Failed to list logical volumes, lvs output: %s",
1268 return dev.split("(")[0]
1270 def handle_dev(dev):
1271 return [parse_dev(x) for x in dev.split(",")]
1274 line = [v.strip() for v in line]
1275 return [{"name": line[0], "size": line[1],
1276 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1279 for line in result.stdout.splitlines():
1280 if line.count("|") >= 3:
1281 all_devs.extend(map_line(line.split("|")))
1283 logging.warning("Strange line in the output from lvs: '%s'", line)
1287 def BridgesExist(bridges_list):
1288 """Check if a list of bridges exist on the current node.
1291 @return: C{True} if all of them exist, C{False} otherwise
1295 for bridge in bridges_list:
1296 if not utils.BridgeExists(bridge):
1297 missing.append(bridge)
1300 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1303 def GetInstanceListForHypervisor(hname, hvparams=None,
1304 get_hv_fn=hypervisor.GetHypervisor):
1305 """Provides a list of instances of the given hypervisor.
1308 @param hname: name of the hypervisor
1309 @type hvparams: dict of strings
1310 @param hvparams: hypervisor parameters for the given hypervisor
1311 @type get_hv_fn: function
1312 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1313 name; optional parameter to increase testability
1316 @return: a list of all running instances on the current node
1317 - instance1.example.com
1318 - instance2.example.com
1323 hv = get_hv_fn(hname)
1324 names = hv.ListInstances(hvparams=hvparams)
1325 results.extend(names)
1326 except errors.HypervisorError, err:
1327 _Fail("Error enumerating instances (hypervisor %s): %s",
1328 hname, err, exc=True)
1332 def GetInstanceList(hypervisor_list, all_hvparams=None,
1333 get_hv_fn=hypervisor.GetHypervisor):
1334 """Provides a list of instances.
1336 @type hypervisor_list: list
1337 @param hypervisor_list: the list of hypervisors to query information
1338 @type all_hvparams: dict of dict of strings
1339 @param all_hvparams: a dictionary mapping hypervisor types to respective
1340 cluster-wide hypervisor parameters
1341 @type get_hv_fn: function
1342 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1343 name; optional parameter to increase testability
1346 @return: a list of all running instances on the current node
1347 - instance1.example.com
1348 - instance2.example.com
1352 for hname in hypervisor_list:
1353 hvparams = all_hvparams[hname]
1354 results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1355 get_hv_fn=get_hv_fn))
1359 def GetInstanceInfo(instance, hname, hvparams=None):
1360 """Gives back the information about an instance as a dictionary.
1362 @type instance: string
1363 @param instance: the instance name
1365 @param hname: the hypervisor type of the instance
1366 @type hvparams: dict of strings
1367 @param hvparams: the instance's hvparams
1370 @return: dictionary with the following keys:
1371 - memory: memory size of instance (int)
1372 - state: xen state of instance (string)
1373 - time: cpu time of instance (float)
1374 - vcpus: the number of vcpus (int)
1379 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1381 if iinfo is not None:
1382 output["memory"] = iinfo[2]
1383 output["vcpus"] = iinfo[3]
1384 output["state"] = iinfo[4]
1385 output["time"] = iinfo[5]
1390 def GetInstanceMigratable(instance):
1391 """Computes whether an instance can be migrated.
1393 @type instance: L{objects.Instance}
1394 @param instance: object representing the instance to be checked.
1397 @return: tuple of (result, description) where:
1398 - result: whether the instance can be migrated or not
1399 - description: a description of the issue, if relevant
1402 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1403 iname = instance.name
1404 if iname not in hyper.ListInstances(instance.hvparams):
1405 _Fail("Instance %s is not running", iname)
1407 for idx in range(len(instance.disks)):
1408 link_name = _GetBlockDevSymlinkPath(iname, idx)
1409 if not os.path.islink(link_name):
1410 logging.warning("Instance %s is missing symlink %s for disk %d",
1411 iname, link_name, idx)
1414 def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1415 """Gather data about all instances.
1417 This is the equivalent of L{GetInstanceInfo}, except that it
1418 computes data for all instances at once, thus being faster if one
1419 needs data about more than one instance.
1421 @type hypervisor_list: list
1422 @param hypervisor_list: list of hypervisors to query for instance data
1423 @type all_hvparams: dict of dict of strings
1424 @param all_hvparams: mapping of hypervisor names to hvparams
1427 @return: dictionary of instance: data, with data having the following keys:
1428 - memory: memory size of instance (int)
1429 - state: xen state of instance (string)
1430 - time: cpu time of instance (float)
1431 - vcpus: the number of vcpus
1436 for hname in hypervisor_list:
1437 hvparams = all_hvparams[hname]
1438 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1440 for name, _, memory, vcpus, state, times in iinfo:
1448 # we only check static parameters, like memory and vcpus,
1449 # and not state and time which can change between the
1450 # invocations of the different hypervisors
1451 for key in "memory", "vcpus":
1452 if value[key] != output[name][key]:
1453 _Fail("Instance %s is running twice"
1454 " with different parameters", name)
1455 output[name] = value
1460 def _InstanceLogName(kind, os_name, instance, component):
1461 """Compute the OS log filename for a given instance and operation.
1463 The instance name and os name are passed in as strings since not all
1464 operations have these as part of an instance object.
1467 @param kind: the operation type (e.g. add, import, etc.)
1468 @type os_name: string
1469 @param os_name: the os name
1470 @type instance: string
1471 @param instance: the name of the instance being imported/added/etc.
1472 @type component: string or None
1473 @param component: the name of the component of the instance being
1477 # TODO: Use tempfile.mkstemp to create unique filename
1479 assert "/" not in component
1480 c_msg = "-%s" % component
1483 base = ("%s-%s-%s%s-%s.log" %
1484 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1485 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1488 def InstanceOsAdd(instance, reinstall, debug):
1489 """Add an OS to an instance.
1491 @type instance: L{objects.Instance}
1492 @param instance: Instance whose OS is to be installed
1493 @type reinstall: boolean
1494 @param reinstall: whether this is an instance reinstall
1495 @type debug: integer
1496 @param debug: debug level, passed to the OS scripts
1500 inst_os = OSFromDisk(instance.os)
1502 create_env = OSEnvironment(instance, inst_os, debug)
1504 create_env["INSTANCE_REINSTALL"] = "1"
1506 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1508 result = utils.RunCmd([inst_os.create_script], env=create_env,
1509 cwd=inst_os.path, output=logfile, reset_env=True)
1511 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1512 " output: %s", result.cmd, result.fail_reason, logfile,
1514 lines = [utils.SafeEncode(val)
1515 for val in utils.TailFile(logfile, lines=20)]
1516 _Fail("OS create script failed (%s), last lines in the"
1517 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1520 def RunRenameInstance(instance, old_name, debug):
1521 """Run the OS rename script for an instance.
1523 @type instance: L{objects.Instance}
1524 @param instance: Instance whose OS is to be installed
1525 @type old_name: string
1526 @param old_name: previous instance name
1527 @type debug: integer
1528 @param debug: debug level, passed to the OS scripts
1530 @return: the success of the operation
1533 inst_os = OSFromDisk(instance.os)
1535 rename_env = OSEnvironment(instance, inst_os, debug)
1536 rename_env["OLD_INSTANCE_NAME"] = old_name
1538 logfile = _InstanceLogName("rename", instance.os,
1539 "%s-%s" % (old_name, instance.name), None)
1541 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1542 cwd=inst_os.path, output=logfile, reset_env=True)
1545 logging.error("os create command '%s' returned error: %s output: %s",
1546 result.cmd, result.fail_reason, result.output)
1547 lines = [utils.SafeEncode(val)
1548 for val in utils.TailFile(logfile, lines=20)]
1549 _Fail("OS rename script failed (%s), last lines in the"
1550 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1553 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1554 """Returns symlink path for block device.
1558 _dir = pathutils.DISK_LINKS_DIR
1560 return utils.PathJoin(_dir,
1562 (instance_name, constants.DISK_SEPARATOR, idx)))
1565 def _SymlinkBlockDev(instance_name, device_path, idx):
1566 """Set up symlinks to a instance's block device.
1568 This is an auxiliary function run when an instance is start (on the primary
1569 node) or when an instance is migrated (on the target node).
1572 @param instance_name: the name of the target instance
1573 @param device_path: path of the physical block device, on the node
1574 @param idx: the disk index
1575 @return: absolute path to the disk's symlink
1578 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1580 os.symlink(device_path, link_name)
1581 except OSError, err:
1582 if err.errno == errno.EEXIST:
1583 if (not os.path.islink(link_name) or
1584 os.readlink(link_name) != device_path):
1585 os.remove(link_name)
1586 os.symlink(device_path, link_name)
1593 def _RemoveBlockDevLinks(instance_name, disks):
1594 """Remove the block device symlinks belonging to the given instance.
1597 for idx, _ in enumerate(disks):
1598 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1599 if os.path.islink(link_name):
1601 os.remove(link_name)
1603 logging.exception("Can't remove symlink '%s'", link_name)
1606 def _GatherAndLinkBlockDevs(instance):
1607 """Set up an instance's block device(s).
1609 This is run on the primary node at instance startup. The block
1610 devices must be already assembled.
1612 @type instance: L{objects.Instance}
1613 @param instance: the instance whose disks we shoul assemble
1615 @return: list of (disk_object, device_path)
1619 for idx, disk in enumerate(instance.disks):
1620 device = _RecursiveFindBD(disk)
1622 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1626 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1628 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1631 block_devices.append((disk, link_name))
1633 return block_devices
1636 def StartInstance(instance, startup_paused, reason, store_reason=True):
1637 """Start an instance.
1639 @type instance: L{objects.Instance}
1640 @param instance: the instance object
1641 @type startup_paused: bool
1642 @param instance: pause instance at startup?
1643 @type reason: list of reasons
1644 @param reason: the reason trail for this startup
1645 @type store_reason: boolean
1646 @param store_reason: whether to store the shutdown reason trail on file
1650 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1653 if instance.name in running_instances:
1654 logging.info("Instance %s already running, not starting", instance.name)
1658 block_devices = _GatherAndLinkBlockDevs(instance)
1659 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1660 hyper.StartInstance(instance, block_devices, startup_paused)
1662 _StoreInstReasonTrail(instance.name, reason)
1663 except errors.BlockDeviceError, err:
1664 _Fail("Block device error: %s", err, exc=True)
1665 except errors.HypervisorError, err:
1666 _RemoveBlockDevLinks(instance.name, instance.disks)
1667 _Fail("Hypervisor error: %s", err, exc=True)
1670 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1671 """Shut an instance down.
1673 @note: this functions uses polling with a hardcoded timeout.
1675 @type instance: L{objects.Instance}
1676 @param instance: the instance object
1677 @type timeout: integer
1678 @param timeout: maximum timeout for soft shutdown
1679 @type reason: list of reasons
1680 @param reason: the reason trail for this shutdown
1681 @type store_reason: boolean
1682 @param store_reason: whether to store the shutdown reason trail on file
1686 hv_name = instance.hypervisor
1687 hyper = hypervisor.GetHypervisor(hv_name)
1688 iname = instance.name
1690 if instance.name not in hyper.ListInstances(instance.hvparams):
1691 logging.info("Instance %s not running, doing nothing", iname)
1696 self.tried_once = False
1699 if iname not in hyper.ListInstances(instance.hvparams):
1703 hyper.StopInstance(instance, retry=self.tried_once)
1705 _StoreInstReasonTrail(instance.name, reason)
1706 except errors.HypervisorError, err:
1707 if iname not in hyper.ListInstances(instance.hvparams):
1708 # if the instance is no longer existing, consider this a
1709 # success and go to cleanup
1712 _Fail("Failed to stop instance %s: %s", iname, err)
1714 self.tried_once = True
1716 raise utils.RetryAgain()
1719 utils.Retry(_TryShutdown(), 5, timeout)
1720 except utils.RetryTimeout:
1721 # the shutdown did not succeed
1722 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1725 hyper.StopInstance(instance, force=True)
1726 except errors.HypervisorError, err:
1727 if iname in hyper.ListInstances(instance.hvparams):
1728 # only raise an error if the instance still exists, otherwise
1729 # the error could simply be "instance ... unknown"!
1730 _Fail("Failed to force stop instance %s: %s", iname, err)
1734 if iname in hyper.ListInstances(instance.hvparams):
1735 _Fail("Could not shutdown instance %s even by destroy", iname)
1738 hyper.CleanupInstance(instance.name)
1739 except errors.HypervisorError, err:
1740 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1742 _RemoveBlockDevLinks(iname, instance.disks)
1745 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1746 """Reboot an instance.
1748 @type instance: L{objects.Instance}
1749 @param instance: the instance object to reboot
1750 @type reboot_type: str
1751 @param reboot_type: the type of reboot, one the following
1753 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1754 instance OS, do not recreate the VM
1755 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1756 restart the VM (at the hypervisor level)
1757 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1758 not accepted here, since that mode is handled differently, in
1759 cmdlib, and translates into full stop and start of the
1760 instance (instead of a call_instance_reboot RPC)
1761 @type shutdown_timeout: integer
1762 @param shutdown_timeout: maximum timeout for soft shutdown
1763 @type reason: list of reasons
1764 @param reason: the reason trail for this reboot
1768 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1771 if instance.name not in running_instances:
1772 _Fail("Cannot reboot instance %s that is not running", instance.name)
1774 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1775 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1777 hyper.RebootInstance(instance)
1778 except errors.HypervisorError, err:
1779 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1780 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1782 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1783 result = StartInstance(instance, False, reason, store_reason=False)
1784 _StoreInstReasonTrail(instance.name, reason)
1786 except errors.HypervisorError, err:
1787 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1789 _Fail("Invalid reboot_type received: %s", reboot_type)
1792 def InstanceBalloonMemory(instance, memory):
1793 """Resize an instance's memory.
1795 @type instance: L{objects.Instance}
1796 @param instance: the instance object
1798 @param memory: new memory amount in MB
1802 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1803 running = hyper.ListInstances(instance.hvparams)
1804 if instance.name not in running:
1805 logging.info("Instance %s is not running, cannot balloon", instance.name)
1808 hyper.BalloonInstanceMemory(instance, memory)
1809 except errors.HypervisorError, err:
1810 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1813 def MigrationInfo(instance):
1814 """Gather information about an instance to be migrated.
1816 @type instance: L{objects.Instance}
1817 @param instance: the instance definition
1820 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1822 info = hyper.MigrationInfo(instance)
1823 except errors.HypervisorError, err:
1824 _Fail("Failed to fetch migration information: %s", err, exc=True)
1828 def AcceptInstance(instance, info, target):
1829 """Prepare the node to accept an instance.
1831 @type instance: L{objects.Instance}
1832 @param instance: the instance definition
1833 @type info: string/data (opaque)
1834 @param info: migration information, from the source node
1835 @type target: string
1836 @param target: target host (usually ip), on this node
1839 # TODO: why is this required only for DTS_EXT_MIRROR?
1840 if instance.disk_template in constants.DTS_EXT_MIRROR:
1841 # Create the symlinks, as the disks are not active
1844 _GatherAndLinkBlockDevs(instance)
1845 except errors.BlockDeviceError, err:
1846 _Fail("Block device error: %s", err, exc=True)
1848 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1850 hyper.AcceptInstance(instance, info, target)
1851 except errors.HypervisorError, err:
1852 if instance.disk_template in constants.DTS_EXT_MIRROR:
1853 _RemoveBlockDevLinks(instance.name, instance.disks)
1854 _Fail("Failed to accept instance: %s", err, exc=True)
1857 def FinalizeMigrationDst(instance, info, success):
1858 """Finalize any preparation to accept an instance.
1860 @type instance: L{objects.Instance}
1861 @param instance: the instance definition
1862 @type info: string/data (opaque)
1863 @param info: migration information, from the source node
1864 @type success: boolean
1865 @param success: whether the migration was a success or a failure
1868 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1870 hyper.FinalizeMigrationDst(instance, info, success)
1871 except errors.HypervisorError, err:
1872 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1875 def MigrateInstance(cluster_name, instance, target, live):
1876 """Migrates an instance to another node.
1878 @type cluster_name: string
1879 @param cluster_name: name of the cluster
1880 @type instance: L{objects.Instance}
1881 @param instance: the instance definition
1882 @type target: string
1883 @param target: the target node name
1885 @param live: whether the migration should be done live or not (the
1886 interpretation of this parameter is left to the hypervisor)
1887 @raise RPCFail: if migration fails for some reason
1890 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1893 hyper.MigrateInstance(cluster_name, instance, target, live)
1894 except errors.HypervisorError, err:
1895 _Fail("Failed to migrate instance: %s", err, exc=True)
1898 def FinalizeMigrationSource(instance, success, live):
1899 """Finalize the instance migration on the source node.
1901 @type instance: L{objects.Instance}
1902 @param instance: the instance definition of the migrated instance
1904 @param success: whether the migration succeeded or not
1906 @param live: whether the user requested a live migration or not
1907 @raise RPCFail: If the execution fails for some reason
1910 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1913 hyper.FinalizeMigrationSource(instance, success, live)
1914 except Exception, err: # pylint: disable=W0703
1915 _Fail("Failed to finalize the migration on the source node: %s", err,
1919 def GetMigrationStatus(instance):
1920 """Get the migration status
1922 @type instance: L{objects.Instance}
1923 @param instance: the instance that is being migrated
1924 @rtype: L{objects.MigrationStatus}
1925 @return: the status of the current migration (one of
1926 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1927 progress info that can be retrieved from the hypervisor
1928 @raise RPCFail: If the migration status cannot be retrieved
1931 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1933 return hyper.GetMigrationStatus(instance)
1934 except Exception, err: # pylint: disable=W0703
1935 _Fail("Failed to get migration status: %s", err, exc=True)
1938 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1939 """Creates a block device for an instance.
1941 @type disk: L{objects.Disk}
1942 @param disk: the object describing the disk we should create
1944 @param size: the size of the physical underlying device, in MiB
1946 @param owner: the name of the instance for which disk is created,
1947 used for device cache data
1948 @type on_primary: boolean
1949 @param on_primary: indicates if it is the primary node or not
1951 @param info: string that will be sent to the physical device
1952 creation, used for example to set (LVM) tags on LVs
1953 @type excl_stor: boolean
1954 @param excl_stor: Whether exclusive_storage is active
1956 @return: the new unique_id of the device (this can sometime be
1957 computed only after creation), or None. On secondary nodes,
1958 it's not required to return anything.
1961 # TODO: remove the obsolete "size" argument
1962 # pylint: disable=W0613
1965 for child in disk.children:
1967 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1968 except errors.BlockDeviceError, err:
1969 _Fail("Can't assemble device %s: %s", child, err)
1970 if on_primary or disk.AssembleOnSecondary():
1971 # we need the children open in case the device itself has to
1974 # pylint: disable=E1103
1976 except errors.BlockDeviceError, err:
1977 _Fail("Can't make child '%s' read-write: %s", child, err)
1981 device = bdev.Create(disk, clist, excl_stor)
1982 except errors.BlockDeviceError, err:
1983 _Fail("Can't create block device: %s", err)
1985 if on_primary or disk.AssembleOnSecondary():
1988 except errors.BlockDeviceError, err:
1989 _Fail("Can't assemble device after creation, unusual event: %s", err)
1990 if on_primary or disk.OpenOnSecondary():
1992 device.Open(force=True)
1993 except errors.BlockDeviceError, err:
1994 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1995 DevCacheManager.UpdateCache(device.dev_path, owner,
1996 on_primary, disk.iv_name)
1998 device.SetInfo(info)
2000 return device.unique_id
2003 def _WipeDevice(path, offset, size):
2004 """This function actually wipes the device.
2006 @param path: The path to the device to wipe
2007 @param offset: The offset in MiB in the file
2008 @param size: The size in MiB to write
2011 # Internal sizes are always in Mebibytes; if the following "dd" command
2012 # should use a different block size the offset and size given to this
2013 # function must be adjusted accordingly before being passed to "dd".
2014 block_size = 1024 * 1024
2016 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2017 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2019 result = utils.RunCmd(cmd)
2022 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2023 result.fail_reason, result.output)
2026 def BlockdevWipe(disk, offset, size):
2027 """Wipes a block device.
2029 @type disk: L{objects.Disk}
2030 @param disk: the disk object we want to wipe
2032 @param offset: The offset in MiB in the file
2034 @param size: The size in MiB to write
2038 rdev = _RecursiveFindBD(disk)
2039 except errors.BlockDeviceError:
2043 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2045 # Do cross verify some of the parameters
2047 _Fail("Negative offset")
2049 _Fail("Negative size")
2050 if offset > rdev.size:
2051 _Fail("Offset is bigger than device size")
2052 if (offset + size) > rdev.size:
2053 _Fail("The provided offset and size to wipe is bigger than device size")
2055 _WipeDevice(rdev.dev_path, offset, size)
2058 def BlockdevPauseResumeSync(disks, pause):
2059 """Pause or resume the sync of the block device.
2061 @type disks: list of L{objects.Disk}
2062 @param disks: the disks object we want to pause/resume
2064 @param pause: Wheater to pause or resume
2070 rdev = _RecursiveFindBD(disk)
2071 except errors.BlockDeviceError:
2075 success.append((False, ("Cannot change sync for device %s:"
2076 " device not found" % disk.iv_name)))
2079 result = rdev.PauseResumeSync(pause)
2082 success.append((result, None))
2088 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2093 def BlockdevRemove(disk):
2094 """Remove a block device.
2096 @note: This is intended to be called recursively.
2098 @type disk: L{objects.Disk}
2099 @param disk: the disk object we should remove
2101 @return: the success of the operation
2106 rdev = _RecursiveFindBD(disk)
2107 except errors.BlockDeviceError, err:
2108 # probably can't attach
2109 logging.info("Can't attach to device %s in remove", disk)
2111 if rdev is not None:
2112 r_path = rdev.dev_path
2115 except errors.BlockDeviceError, err:
2116 msgs.append(str(err))
2118 DevCacheManager.RemoveCache(r_path)
2121 for child in disk.children:
2123 BlockdevRemove(child)
2124 except RPCFail, err:
2125 msgs.append(str(err))
2128 _Fail("; ".join(msgs))
2131 def _RecursiveAssembleBD(disk, owner, as_primary):
2132 """Activate a block device for an instance.
2134 This is run on the primary and secondary nodes for an instance.
2136 @note: this function is called recursively.
2138 @type disk: L{objects.Disk}
2139 @param disk: the disk we try to assemble
2141 @param owner: the name of the instance which owns the disk
2142 @type as_primary: boolean
2143 @param as_primary: if we should make the block device
2146 @return: the assembled device or None (in case no device
2148 @raise errors.BlockDeviceError: in case there is an error
2149 during the activation of the children or the device
2155 mcn = disk.ChildrenNeeded()
2157 mcn = 0 # max number of Nones allowed
2159 mcn = len(disk.children) - mcn # max number of Nones
2160 for chld_disk in disk.children:
2162 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2163 except errors.BlockDeviceError, err:
2164 if children.count(None) >= mcn:
2167 logging.error("Error in child activation (but continuing): %s",
2169 children.append(cdev)
2171 if as_primary or disk.AssembleOnSecondary():
2172 r_dev = bdev.Assemble(disk, children)
2174 if as_primary or disk.OpenOnSecondary():
2176 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2177 as_primary, disk.iv_name)
2184 def BlockdevAssemble(disk, owner, as_primary, idx):
2185 """Activate a block device for an instance.
2187 This is a wrapper over _RecursiveAssembleBD.
2189 @rtype: str or boolean
2190 @return: a C{/dev/...} path for primary nodes, and
2191 C{True} for secondary nodes
2195 result = _RecursiveAssembleBD(disk, owner, as_primary)
2196 if isinstance(result, BlockDev):
2197 # pylint: disable=E1103
2198 result = result.dev_path
2200 _SymlinkBlockDev(owner, result, idx)
2201 except errors.BlockDeviceError, err:
2202 _Fail("Error while assembling disk: %s", err, exc=True)
2203 except OSError, err:
2204 _Fail("Error while symlinking disk: %s", err, exc=True)
2209 def BlockdevShutdown(disk):
2210 """Shut down a block device.
2212 First, if the device is assembled (Attach() is successful), then
2213 the device is shutdown. Then the children of the device are
2216 This function is called recursively. Note that we don't cache the
2217 children or such, as oppossed to assemble, shutdown of different
2218 devices doesn't require that the upper device was active.
2220 @type disk: L{objects.Disk}
2221 @param disk: the description of the disk we should
2227 r_dev = _RecursiveFindBD(disk)
2228 if r_dev is not None:
2229 r_path = r_dev.dev_path
2232 DevCacheManager.RemoveCache(r_path)
2233 except errors.BlockDeviceError, err:
2234 msgs.append(str(err))
2237 for child in disk.children:
2239 BlockdevShutdown(child)
2240 except RPCFail, err:
2241 msgs.append(str(err))
2244 _Fail("; ".join(msgs))
2247 def BlockdevAddchildren(parent_cdev, new_cdevs):
2248 """Extend a mirrored block device.
2250 @type parent_cdev: L{objects.Disk}
2251 @param parent_cdev: the disk to which we should add children
2252 @type new_cdevs: list of L{objects.Disk}
2253 @param new_cdevs: the list of children which we should add
2257 parent_bdev = _RecursiveFindBD(parent_cdev)
2258 if parent_bdev is None:
2259 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2260 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2261 if new_bdevs.count(None) > 0:
2262 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2263 parent_bdev.AddChildren(new_bdevs)
2266 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2267 """Shrink a mirrored block device.
2269 @type parent_cdev: L{objects.Disk}
2270 @param parent_cdev: the disk from which we should remove children
2271 @type new_cdevs: list of L{objects.Disk}
2272 @param new_cdevs: the list of children which we should remove
2276 parent_bdev = _RecursiveFindBD(parent_cdev)
2277 if parent_bdev is None:
2278 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2280 for disk in new_cdevs:
2281 rpath = disk.StaticDevPath()
2283 bd = _RecursiveFindBD(disk)
2285 _Fail("Can't find device %s while removing children", disk)
2287 devs.append(bd.dev_path)
2289 if not utils.IsNormAbsPath(rpath):
2290 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2292 parent_bdev.RemoveChildren(devs)
2295 def BlockdevGetmirrorstatus(disks):
2296 """Get the mirroring status of a list of devices.
2298 @type disks: list of L{objects.Disk}
2299 @param disks: the list of disks which we should query
2301 @return: List of L{objects.BlockDevStatus}, one for each disk
2302 @raise errors.BlockDeviceError: if any of the disks cannot be
2308 rbd = _RecursiveFindBD(dsk)
2310 _Fail("Can't find device %s", dsk)
2312 stats.append(rbd.CombinedSyncStatus())
2317 def BlockdevGetmirrorstatusMulti(disks):
2318 """Get the mirroring status of a list of devices.
2320 @type disks: list of L{objects.Disk}
2321 @param disks: the list of disks which we should query
2323 @return: List of tuples, (bool, status), one for each disk; bool denotes
2324 success/failure, status is L{objects.BlockDevStatus} on success, string
2331 rbd = _RecursiveFindBD(disk)
2333 result.append((False, "Can't find device %s" % disk))
2336 status = rbd.CombinedSyncStatus()
2337 except errors.BlockDeviceError, err:
2338 logging.exception("Error while getting disk status")
2339 result.append((False, str(err)))
2341 result.append((True, status))
2343 assert len(disks) == len(result)
2348 def _RecursiveFindBD(disk):
2349 """Check if a device is activated.
2351 If so, return information about the real device.
2353 @type disk: L{objects.Disk}
2354 @param disk: the disk object we need to find
2356 @return: None if the device can't be found,
2357 otherwise the device instance
2362 for chdisk in disk.children:
2363 children.append(_RecursiveFindBD(chdisk))
2365 return bdev.FindDevice(disk, children)
2368 def _OpenRealBD(disk):
2369 """Opens the underlying block device of a disk.
2371 @type disk: L{objects.Disk}
2372 @param disk: the disk object we want to open
2375 real_disk = _RecursiveFindBD(disk)
2376 if real_disk is None:
2377 _Fail("Block device '%s' is not set up", disk)
2384 def BlockdevFind(disk):
2385 """Check if a device is activated.
2387 If it is, return information about the real device.
2389 @type disk: L{objects.Disk}
2390 @param disk: the disk to find
2391 @rtype: None or objects.BlockDevStatus
2392 @return: None if the disk cannot be found, otherwise a the current
2397 rbd = _RecursiveFindBD(disk)
2398 except errors.BlockDeviceError, err:
2399 _Fail("Failed to find device: %s", err, exc=True)
2404 return rbd.GetSyncStatus()
2407 def BlockdevGetdimensions(disks):
2408 """Computes the size of the given disks.
2410 If a disk is not found, returns None instead.
2412 @type disks: list of L{objects.Disk}
2413 @param disks: the list of disk to compute the size for
2415 @return: list with elements None if the disk cannot be found,
2416 otherwise the pair (size, spindles), where spindles is None if the
2417 device doesn't support that
2423 rbd = _RecursiveFindBD(cf)
2424 except errors.BlockDeviceError:
2430 result.append(rbd.GetActualDimensions())
2434 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2435 """Export a block device to a remote node.
2437 @type disk: L{objects.Disk}
2438 @param disk: the description of the disk to export
2439 @type dest_node: str
2440 @param dest_node: the destination node to export to
2441 @type dest_path: str
2442 @param dest_path: the destination path on the target node
2443 @type cluster_name: str
2444 @param cluster_name: the cluster name, needed for SSH hostalias
2448 real_disk = _OpenRealBD(disk)
2450 # the block size on the read dd is 1MiB to match our units
2451 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2452 "dd if=%s bs=1048576 count=%s",
2453 real_disk.dev_path, str(disk.size))
2455 # we set here a smaller block size as, due to ssh buffering, more
2456 # than 64-128k will mostly ignored; we use nocreat to fail if the
2457 # device is not already there or we pass a wrong path; we use
2458 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2459 # to not buffer too much memory; this means that at best, we flush
2460 # every 64k, which will not be very fast
2461 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2462 " oflag=dsync", dest_path)
2464 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2465 constants.SSH_LOGIN_USER,
2468 # all commands have been checked, so we're safe to combine them
2469 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2471 result = utils.RunCmd(["bash", "-c", command])
2474 _Fail("Disk copy command '%s' returned error: %s"
2475 " output: %s", command, result.fail_reason, result.output)
2478 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2479 """Write a file to the filesystem.
2481 This allows the master to overwrite(!) a file. It will only perform
2482 the operation if the file belongs to a list of configuration files.
2484 @type file_name: str
2485 @param file_name: the target file name
2487 @param data: the new contents of the file
2489 @param mode: the mode to give the file (can be None)
2491 @param uid: the owner of the file
2493 @param gid: the group of the file
2495 @param atime: the atime to set on the file (can be None)
2497 @param mtime: the mtime to set on the file (can be None)
2501 file_name = vcluster.LocalizeVirtualPath(file_name)
2503 if not os.path.isabs(file_name):
2504 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2506 if file_name not in _ALLOWED_UPLOAD_FILES:
2507 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2510 raw_data = _Decompress(data)
2512 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2513 _Fail("Invalid username/groupname type")
2515 getents = runtime.GetEnts()
2516 uid = getents.LookupUser(uid)
2517 gid = getents.LookupGroup(gid)
2519 utils.SafeWriteFile(file_name, None,
2520 data=raw_data, mode=mode, uid=uid, gid=gid,
2521 atime=atime, mtime=mtime)
2524 def RunOob(oob_program, command, node, timeout):
2525 """Executes oob_program with given command on given node.
2527 @param oob_program: The path to the executable oob_program
2528 @param command: The command to invoke on oob_program
2529 @param node: The node given as an argument to the program
2530 @param timeout: Timeout after which we kill the oob program
2533 @raise RPCFail: If execution fails for some reason
2536 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2539 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2540 result.fail_reason, result.output)
2542 return result.stdout
2545 def _OSOndiskAPIVersion(os_dir):
2546 """Compute and return the API version of a given OS.
2548 This function will try to read the API version of the OS residing in
2549 the 'os_dir' directory.
2552 @param os_dir: the directory in which we should look for the OS
2554 @return: tuple (status, data) with status denoting the validity and
2555 data holding either the vaid versions or an error message
2558 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2561 st = os.stat(api_file)
2562 except EnvironmentError, err:
2563 return False, ("Required file '%s' not found under path %s: %s" %
2564 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2566 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2567 return False, ("File '%s' in %s is not a regular file" %
2568 (constants.OS_API_FILE, os_dir))
2571 api_versions = utils.ReadFile(api_file).splitlines()
2572 except EnvironmentError, err:
2573 return False, ("Error while reading the API version file at %s: %s" %
2574 (api_file, utils.ErrnoOrStr(err)))
2577 api_versions = [int(version.strip()) for version in api_versions]
2578 except (TypeError, ValueError), err:
2579 return False, ("API version(s) can't be converted to integer: %s" %
2582 return True, api_versions
2585 def DiagnoseOS(top_dirs=None):
2586 """Compute the validity for all OSes.
2588 @type top_dirs: list
2589 @param top_dirs: the list of directories in which to
2590 search (if not given defaults to
2591 L{pathutils.OS_SEARCH_PATH})
2592 @rtype: list of L{objects.OS}
2593 @return: a list of tuples (name, path, status, diagnose, variants,
2594 parameters, api_version) for all (potential) OSes under all
2595 search paths, where:
2596 - name is the (potential) OS name
2597 - path is the full path to the OS
2598 - status True/False is the validity of the OS
2599 - diagnose is the error message for an invalid OS, otherwise empty
2600 - variants is a list of supported OS variants, if any
2601 - parameters is a list of (name, help) parameters, if any
2602 - api_version is a list of support OS API versions
2605 if top_dirs is None:
2606 top_dirs = pathutils.OS_SEARCH_PATH
2609 for dir_name in top_dirs:
2610 if os.path.isdir(dir_name):
2612 f_names = utils.ListVisibleFiles(dir_name)
2613 except EnvironmentError, err:
2614 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2616 for name in f_names:
2617 os_path = utils.PathJoin(dir_name, name)
2618 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2621 variants = os_inst.supported_variants
2622 parameters = os_inst.supported_parameters
2623 api_versions = os_inst.api_versions
2626 variants = parameters = api_versions = []
2627 result.append((name, os_path, status, diagnose, variants,
2628 parameters, api_versions))
2633 def _TryOSFromDisk(name, base_dir=None):
2634 """Create an OS instance from disk.
2636 This function will return an OS instance if the given name is a
2639 @type base_dir: string
2640 @keyword base_dir: Base directory containing OS installations.
2641 Defaults to a search in all the OS_SEARCH_PATH dirs.
2643 @return: success and either the OS instance if we find a valid one,
2647 if base_dir is None:
2648 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2650 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2653 return False, "Directory for OS %s not found in search path" % name
2655 status, api_versions = _OSOndiskAPIVersion(os_dir)
2658 return status, api_versions
2660 if not constants.OS_API_VERSIONS.intersection(api_versions):
2661 return False, ("API version mismatch for path '%s': found %s, want %s." %
2662 (os_dir, api_versions, constants.OS_API_VERSIONS))
2664 # OS Files dictionary, we will populate it with the absolute path
2665 # names; if the value is True, then it is a required file, otherwise
2667 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2669 if max(api_versions) >= constants.OS_API_V15:
2670 os_files[constants.OS_VARIANTS_FILE] = False
2672 if max(api_versions) >= constants.OS_API_V20:
2673 os_files[constants.OS_PARAMETERS_FILE] = True
2675 del os_files[constants.OS_SCRIPT_VERIFY]
2677 for (filename, required) in os_files.items():
2678 os_files[filename] = utils.PathJoin(os_dir, filename)
2681 st = os.stat(os_files[filename])
2682 except EnvironmentError, err:
2683 if err.errno == errno.ENOENT and not required:
2684 del os_files[filename]
2686 return False, ("File '%s' under path '%s' is missing (%s)" %
2687 (filename, os_dir, utils.ErrnoOrStr(err)))
2689 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2690 return False, ("File '%s' under path '%s' is not a regular file" %
2693 if filename in constants.OS_SCRIPTS:
2694 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2695 return False, ("File '%s' under path '%s' is not executable" %
2699 if constants.OS_VARIANTS_FILE in os_files:
2700 variants_file = os_files[constants.OS_VARIANTS_FILE]
2703 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2704 except EnvironmentError, err:
2705 # we accept missing files, but not other errors
2706 if err.errno != errno.ENOENT:
2707 return False, ("Error while reading the OS variants file at %s: %s" %
2708 (variants_file, utils.ErrnoOrStr(err)))
2711 if constants.OS_PARAMETERS_FILE in os_files:
2712 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2714 parameters = utils.ReadFile(parameters_file).splitlines()
2715 except EnvironmentError, err:
2716 return False, ("Error while reading the OS parameters file at %s: %s" %
2717 (parameters_file, utils.ErrnoOrStr(err)))
2718 parameters = [v.split(None, 1) for v in parameters]
2720 os_obj = objects.OS(name=name, path=os_dir,
2721 create_script=os_files[constants.OS_SCRIPT_CREATE],
2722 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2723 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2724 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2725 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2727 supported_variants=variants,
2728 supported_parameters=parameters,
2729 api_versions=api_versions)
2733 def OSFromDisk(name, base_dir=None):
2734 """Create an OS instance from disk.
2736 This function will return an OS instance if the given name is a
2737 valid OS name. Otherwise, it will raise an appropriate
2738 L{RPCFail} exception, detailing why this is not a valid OS.
2740 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2741 an exception but returns true/false status data.
2743 @type base_dir: string
2744 @keyword base_dir: Base directory containing OS installations.
2745 Defaults to a search in all the OS_SEARCH_PATH dirs.
2746 @rtype: L{objects.OS}
2747 @return: the OS instance if we find a valid one
2748 @raise RPCFail: if we don't find a valid OS
2751 name_only = objects.OS.GetName(name)
2752 status, payload = _TryOSFromDisk(name_only, base_dir)
2760 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2761 """Calculate the basic environment for an os script.
2764 @param os_name: full operating system name (including variant)
2765 @type inst_os: L{objects.OS}
2766 @param inst_os: operating system for which the environment is being built
2767 @type os_params: dict
2768 @param os_params: the OS parameters
2769 @type debug: integer
2770 @param debug: debug level (0 or 1, for OS Api 10)
2772 @return: dict of environment variables
2773 @raise errors.BlockDeviceError: if the block device
2779 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2780 result["OS_API_VERSION"] = "%d" % api_version
2781 result["OS_NAME"] = inst_os.name
2782 result["DEBUG_LEVEL"] = "%d" % debug
2785 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2786 variant = objects.OS.GetVariant(os_name)
2788 variant = inst_os.supported_variants[0]
2791 result["OS_VARIANT"] = variant
2794 for pname, pvalue in os_params.items():
2795 result["OSP_%s" % pname.upper()] = pvalue
2797 # Set a default path otherwise programs called by OS scripts (or
2798 # even hooks called from OS scripts) might break, and we don't want
2799 # to have each script require setting a PATH variable
2800 result["PATH"] = constants.HOOKS_PATH
2805 def OSEnvironment(instance, inst_os, debug=0):
2806 """Calculate the environment for an os script.
2808 @type instance: L{objects.Instance}
2809 @param instance: target instance for the os script run
2810 @type inst_os: L{objects.OS}
2811 @param inst_os: operating system for which the environment is being built
2812 @type debug: integer
2813 @param debug: debug level (0 or 1, for OS Api 10)
2815 @return: dict of environment variables
2816 @raise errors.BlockDeviceError: if the block device
2820 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2822 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2823 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2825 result["HYPERVISOR"] = instance.hypervisor
2826 result["DISK_COUNT"] = "%d" % len(instance.disks)
2827 result["NIC_COUNT"] = "%d" % len(instance.nics)
2828 result["INSTANCE_SECONDARY_NODES"] = \
2829 ("%s" % " ".join(instance.secondary_nodes))
2832 for idx, disk in enumerate(instance.disks):
2833 real_disk = _OpenRealBD(disk)
2834 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2835 result["DISK_%d_ACCESS" % idx] = disk.mode
2836 result["DISK_%d_UUID" % idx] = disk.uuid
2838 result["DISK_%d_NAME" % idx] = disk.name
2839 if constants.HV_DISK_TYPE in instance.hvparams:
2840 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2841 instance.hvparams[constants.HV_DISK_TYPE]
2842 if disk.dev_type in constants.LDS_BLOCK:
2843 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2844 elif disk.dev_type == constants.LD_FILE:
2845 result["DISK_%d_BACKEND_TYPE" % idx] = \
2846 "file:%s" % disk.physical_id[0]
2849 for idx, nic in enumerate(instance.nics):
2850 result["NIC_%d_MAC" % idx] = nic.mac
2851 result["NIC_%d_UUID" % idx] = nic.uuid
2853 result["NIC_%d_NAME" % idx] = nic.name
2855 result["NIC_%d_IP" % idx] = nic.ip
2856 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2857 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2858 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2859 if nic.nicparams[constants.NIC_LINK]:
2860 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2862 nobj = objects.Network.FromDict(nic.netinfo)
2863 result.update(nobj.HooksDict("NIC_%d_" % idx))
2864 if constants.HV_NIC_TYPE in instance.hvparams:
2865 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2866 instance.hvparams[constants.HV_NIC_TYPE]
2869 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2870 for key, value in source.items():
2871 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2876 def DiagnoseExtStorage(top_dirs=None):
2877 """Compute the validity for all ExtStorage Providers.
2879 @type top_dirs: list
2880 @param top_dirs: the list of directories in which to
2881 search (if not given defaults to
2882 L{pathutils.ES_SEARCH_PATH})
2883 @rtype: list of L{objects.ExtStorage}
2884 @return: a list of tuples (name, path, status, diagnose, parameters)
2885 for all (potential) ExtStorage Providers under all
2886 search paths, where:
2887 - name is the (potential) ExtStorage Provider
2888 - path is the full path to the ExtStorage Provider
2889 - status True/False is the validity of the ExtStorage Provider
2890 - diagnose is the error message for an invalid ExtStorage Provider,
2892 - parameters is a list of (name, help) parameters, if any
2895 if top_dirs is None:
2896 top_dirs = pathutils.ES_SEARCH_PATH
2899 for dir_name in top_dirs:
2900 if os.path.isdir(dir_name):
2902 f_names = utils.ListVisibleFiles(dir_name)
2903 except EnvironmentError, err:
2904 logging.exception("Can't list the ExtStorage directory %s: %s",
2907 for name in f_names:
2908 es_path = utils.PathJoin(dir_name, name)
2909 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2912 parameters = es_inst.supported_parameters
2916 result.append((name, es_path, status, diagnose, parameters))
2921 def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
2922 """Grow a stack of block devices.
2924 This function is called recursively, with the childrens being the
2925 first ones to resize.
2927 @type disk: L{objects.Disk}
2928 @param disk: the disk to be grown
2929 @type amount: integer
2930 @param amount: the amount (in mebibytes) to grow with
2931 @type dryrun: boolean
2932 @param dryrun: whether to execute the operation in simulation mode
2933 only, without actually increasing the size
2934 @param backingstore: whether to execute the operation on backing storage
2935 only, or on "logical" storage only; e.g. DRBD is logical storage,
2936 whereas LVM, file, RBD are backing storage
2937 @rtype: (status, result)
2938 @type excl_stor: boolean
2939 @param excl_stor: Whether exclusive_storage is active
2940 @return: a tuple with the status of the operation (True/False), and
2941 the errors message if status is False
2944 r_dev = _RecursiveFindBD(disk)
2946 _Fail("Cannot find block device %s", disk)
2949 r_dev.Grow(amount, dryrun, backingstore, excl_stor)
2950 except errors.BlockDeviceError, err:
2951 _Fail("Failed to grow block device: %s", err, exc=True)
2954 def BlockdevSnapshot(disk):
2955 """Create a snapshot copy of a block device.
2957 This function is called recursively, and the snapshot is actually created
2958 just for the leaf lvm backend device.
2960 @type disk: L{objects.Disk}
2961 @param disk: the disk to be snapshotted
2963 @return: snapshot disk ID as (vg, lv)
2966 if disk.dev_type == constants.LD_DRBD8:
2967 if not disk.children:
2968 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2970 return BlockdevSnapshot(disk.children[0])
2971 elif disk.dev_type == constants.LD_LV:
2972 r_dev = _RecursiveFindBD(disk)
2973 if r_dev is not None:
2974 # FIXME: choose a saner value for the snapshot size
2975 # let's stay on the safe side and ask for the full size, for now
2976 return r_dev.Snapshot(disk.size)
2978 _Fail("Cannot find block device %s", disk)
2980 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2981 disk.unique_id, disk.dev_type)
2984 def BlockdevSetInfo(disk, info):
2985 """Sets 'metadata' information on block devices.
2987 This function sets 'info' metadata on block devices. Initial
2988 information is set at device creation; this function should be used
2989 for example after renames.
2991 @type disk: L{objects.Disk}
2992 @param disk: the disk to be grown
2994 @param info: new 'info' metadata
2995 @rtype: (status, result)
2996 @return: a tuple with the status of the operation (True/False), and
2997 the errors message if status is False
3000 r_dev = _RecursiveFindBD(disk)
3002 _Fail("Cannot find block device %s", disk)
3006 except errors.BlockDeviceError, err:
3007 _Fail("Failed to set information on block device: %s", err, exc=True)
3010 def FinalizeExport(instance, snap_disks):
3011 """Write out the export configuration information.
3013 @type instance: L{objects.Instance}
3014 @param instance: the instance which we export, used for
3015 saving configuration
3016 @type snap_disks: list of L{objects.Disk}
3017 @param snap_disks: list of snapshot block devices, which
3018 will be used to get the actual name of the dump file
3023 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3024 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3026 config = objects.SerializableConfigParser()
3028 config.add_section(constants.INISECT_EXP)
3029 config.set(constants.INISECT_EXP, "version", "0")
3030 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3031 config.set(constants.INISECT_EXP, "source", instance.primary_node)
3032 config.set(constants.INISECT_EXP, "os", instance.os)
3033 config.set(constants.INISECT_EXP, "compression", "none")
3035 config.add_section(constants.INISECT_INS)
3036 config.set(constants.INISECT_INS, "name", instance.name)
3037 config.set(constants.INISECT_INS, "maxmem", "%d" %
3038 instance.beparams[constants.BE_MAXMEM])
3039 config.set(constants.INISECT_INS, "minmem", "%d" %
3040 instance.beparams[constants.BE_MINMEM])
3041 # "memory" is deprecated, but useful for exporting to old ganeti versions
3042 config.set(constants.INISECT_INS, "memory", "%d" %
3043 instance.beparams[constants.BE_MAXMEM])
3044 config.set(constants.INISECT_INS, "vcpus", "%d" %
3045 instance.beparams[constants.BE_VCPUS])
3046 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3047 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3048 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3051 for nic_count, nic in enumerate(instance.nics):
3053 config.set(constants.INISECT_INS, "nic%d_mac" %
3054 nic_count, "%s" % nic.mac)
3055 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3056 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3058 for param in constants.NICS_PARAMETER_TYPES:
3059 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3060 "%s" % nic.nicparams.get(param, None))
3061 # TODO: redundant: on load can read nics until it doesn't exist
3062 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3065 for disk_count, disk in enumerate(snap_disks):
3068 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3069 ("%s" % disk.iv_name))
3070 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3071 ("%s" % disk.physical_id[1]))
3072 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3075 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3077 # New-style hypervisor/backend parameters
3079 config.add_section(constants.INISECT_HYP)
3080 for name, value in instance.hvparams.items():
3081 if name not in constants.HVC_GLOBALS:
3082 config.set(constants.INISECT_HYP, name, str(value))
3084 config.add_section(constants.INISECT_BEP)
3085 for name, value in instance.beparams.items():
3086 config.set(constants.INISECT_BEP, name, str(value))
3088 config.add_section(constants.INISECT_OSP)
3089 for name, value in instance.osparams.items():
3090 config.set(constants.INISECT_OSP, name, str(value))
3092 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3093 data=config.Dumps())
3094 shutil.rmtree(finaldestdir, ignore_errors=True)
3095 shutil.move(destdir, finaldestdir)
3098 def ExportInfo(dest):
3099 """Get export configuration information.
3102 @param dest: directory containing the export
3104 @rtype: L{objects.SerializableConfigParser}
3105 @return: a serializable config file containing the
3109 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3111 config = objects.SerializableConfigParser()
3114 if (not config.has_section(constants.INISECT_EXP) or
3115 not config.has_section(constants.INISECT_INS)):
3116 _Fail("Export info file doesn't have the required fields")
3118 return config.Dumps()
3122 """Return a list of exports currently available on this machine.
3125 @return: list of the exports
3128 if os.path.isdir(pathutils.EXPORT_DIR):
3129 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3131 _Fail("No exports directory")
3134 def RemoveExport(export):
3135 """Remove an existing export from the node.
3138 @param export: the name of the export to remove
3142 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3145 shutil.rmtree(target)
3146 except EnvironmentError, err:
3147 _Fail("Error while removing the export: %s", err, exc=True)
3150 def BlockdevRename(devlist):
3151 """Rename a list of block devices.
3153 @type devlist: list of tuples
3154 @param devlist: list of tuples of the form (disk,
3155 new_logical_id, new_physical_id); disk is an
3156 L{objects.Disk} object describing the current disk,
3157 and new logical_id/physical_id is the name we
3160 @return: True if all renames succeeded, False otherwise
3165 for disk, unique_id in devlist:
3166 dev = _RecursiveFindBD(disk)
3168 msgs.append("Can't find device %s in rename" % str(disk))
3172 old_rpath = dev.dev_path
3173 dev.Rename(unique_id)
3174 new_rpath = dev.dev_path
3175 if old_rpath != new_rpath:
3176 DevCacheManager.RemoveCache(old_rpath)
3177 # FIXME: we should add the new cache information here, like:
3178 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3179 # but we don't have the owner here - maybe parse from existing
3180 # cache? for now, we only lose lvm data when we rename, which
3181 # is less critical than DRBD or MD
3182 except errors.BlockDeviceError, err:
3183 msgs.append("Can't rename device '%s' to '%s': %s" %
3184 (dev, unique_id, err))
3185 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3188 _Fail("; ".join(msgs))
3191 def _TransformFileStorageDir(fs_dir):
3192 """Checks whether given file_storage_dir is valid.
3194 Checks wheter the given fs_dir is within the cluster-wide default
3195 file_storage_dir or the shared_file_storage_dir, which are stored in
3196 SimpleStore. Only paths under those directories are allowed.
3199 @param fs_dir: the path to check
3201 @return: the normalized path if valid, None otherwise
3204 filestorage.CheckFileStoragePath(fs_dir)
3206 return os.path.normpath(fs_dir)
3209 def CreateFileStorageDir(file_storage_dir):
3210 """Create file storage directory.
3212 @type file_storage_dir: str
3213 @param file_storage_dir: directory to create
3216 @return: tuple with first element a boolean indicating wheter dir
3217 creation was successful or not
3220 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3221 if os.path.exists(file_storage_dir):
3222 if not os.path.isdir(file_storage_dir):
3223 _Fail("Specified storage dir '%s' is not a directory",
3227 os.makedirs(file_storage_dir, 0750)
3228 except OSError, err:
3229 _Fail("Cannot create file storage directory '%s': %s",
3230 file_storage_dir, err, exc=True)
3233 def RemoveFileStorageDir(file_storage_dir):
3234 """Remove file storage directory.
3236 Remove it only if it's empty. If not log an error and return.
3238 @type file_storage_dir: str
3239 @param file_storage_dir: the directory we should cleanup
3240 @rtype: tuple (success,)
3241 @return: tuple of one element, C{success}, denoting
3242 whether the operation was successful
3245 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3246 if os.path.exists(file_storage_dir):
3247 if not os.path.isdir(file_storage_dir):
3248 _Fail("Specified Storage directory '%s' is not a directory",
3250 # deletes dir only if empty, otherwise we want to fail the rpc call
3252 os.rmdir(file_storage_dir)
3253 except OSError, err:
3254 _Fail("Cannot remove file storage directory '%s': %s",
3255 file_storage_dir, err)
3258 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3259 """Rename the file storage directory.
3261 @type old_file_storage_dir: str
3262 @param old_file_storage_dir: the current path
3263 @type new_file_storage_dir: str
3264 @param new_file_storage_dir: the name we should rename to
3265 @rtype: tuple (success,)
3266 @return: tuple of one element, C{success}, denoting
3267 whether the operation was successful
3270 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3271 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3272 if not os.path.exists(new_file_storage_dir):
3273 if os.path.isdir(old_file_storage_dir):
3275 os.rename(old_file_storage_dir, new_file_storage_dir)
3276 except OSError, err:
3277 _Fail("Cannot rename '%s' to '%s': %s",
3278 old_file_storage_dir, new_file_storage_dir, err)
3280 _Fail("Specified storage dir '%s' is not a directory",
3281 old_file_storage_dir)
3283 if os.path.exists(old_file_storage_dir):
3284 _Fail("Cannot rename '%s' to '%s': both locations exist",
3285 old_file_storage_dir, new_file_storage_dir)
3288 def _EnsureJobQueueFile(file_name):
3289 """Checks whether the given filename is in the queue directory.
3291 @type file_name: str
3292 @param file_name: the file name we should check
3294 @raises RPCFail: if the file is not valid
3297 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3298 _Fail("Passed job queue file '%s' does not belong to"
3299 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3302 def JobQueueUpdate(file_name, content):
3303 """Updates a file in the queue directory.
3305 This is just a wrapper over L{utils.io.WriteFile}, with proper
3308 @type file_name: str
3309 @param file_name: the job file name
3311 @param content: the new job contents
3313 @return: the success of the operation
3316 file_name = vcluster.LocalizeVirtualPath(file_name)
3318 _EnsureJobQueueFile(file_name)
3319 getents = runtime.GetEnts()
3321 # Write and replace the file atomically
3322 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3323 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3326 def JobQueueRename(old, new):
3327 """Renames a job queue file.
3329 This is just a wrapper over os.rename with proper checking.
3332 @param old: the old (actual) file name
3334 @param new: the desired file name
3336 @return: the success of the operation and payload
3339 old = vcluster.LocalizeVirtualPath(old)
3340 new = vcluster.LocalizeVirtualPath(new)
3342 _EnsureJobQueueFile(old)
3343 _EnsureJobQueueFile(new)
3345 getents = runtime.GetEnts()
3347 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3348 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3351 def BlockdevClose(instance_name, disks):
3352 """Closes the given block devices.
3354 This means they will be switched to secondary mode (in case of
3357 @param instance_name: if the argument is not empty, the symlinks
3358 of this instance will be removed
3359 @type disks: list of L{objects.Disk}
3360 @param disks: the list of disks to be closed
3361 @rtype: tuple (success, message)
3362 @return: a tuple of success and message, where success
3363 indicates the succes of the operation, and message
3364 which will contain the error details in case we
3370 rd = _RecursiveFindBD(cf)
3372 _Fail("Can't find device %s", cf)
3379 except errors.BlockDeviceError, err:
3380 msg.append(str(err))
3382 _Fail("Can't make devices secondary: %s", ",".join(msg))
3385 _RemoveBlockDevLinks(instance_name, disks)
3388 def ValidateHVParams(hvname, hvparams):
3389 """Validates the given hypervisor parameters.
3391 @type hvname: string
3392 @param hvname: the hypervisor name
3393 @type hvparams: dict
3394 @param hvparams: the hypervisor parameters to be validated
3399 hv_type = hypervisor.GetHypervisor(hvname)
3400 hv_type.ValidateParameters(hvparams)
3401 except errors.HypervisorError, err:
3402 _Fail(str(err), log=False)
3405 def _CheckOSPList(os_obj, parameters):
3406 """Check whether a list of parameters is supported by the OS.
3408 @type os_obj: L{objects.OS}
3409 @param os_obj: OS object to check
3410 @type parameters: list
3411 @param parameters: the list of parameters to check
3414 supported = [v[0] for v in os_obj.supported_parameters]
3415 delta = frozenset(parameters).difference(supported)
3417 _Fail("The following parameters are not supported"
3418 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3421 def ValidateOS(required, osname, checks, osparams):
3422 """Validate the given OS' parameters.
3424 @type required: boolean
3425 @param required: whether absence of the OS should translate into
3427 @type osname: string
3428 @param osname: the OS to be validated
3430 @param checks: list of the checks to run (currently only 'parameters')
3431 @type osparams: dict
3432 @param osparams: dictionary with OS parameters
3434 @return: True if the validation passed, or False if the OS was not
3435 found and L{required} was false
3438 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3439 _Fail("Unknown checks required for OS %s: %s", osname,
3440 set(checks).difference(constants.OS_VALIDATE_CALLS))
3442 name_only = objects.OS.GetName(osname)
3443 status, tbv = _TryOSFromDisk(name_only, None)
3451 if max(tbv.api_versions) < constants.OS_API_V20:
3454 if constants.OS_VALIDATE_PARAMETERS in checks:
3455 _CheckOSPList(tbv, osparams.keys())
3457 validate_env = OSCoreEnv(osname, tbv, osparams)
3458 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3459 cwd=tbv.path, reset_env=True)
3461 logging.error("os validate command '%s' returned error: %s output: %s",
3462 result.cmd, result.fail_reason, result.output)
3463 _Fail("OS validation script failed (%s), output: %s",
3464 result.fail_reason, result.output, log=False)
3470 """Demotes the current node from master candidate role.
3473 # try to ensure we're not the master by mistake
3474 master, myself = ssconf.GetMasterAndMyself()
3475 if master == myself:
3476 _Fail("ssconf status shows I'm the master node, will not demote")
3478 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3479 if not result.failed:
3480 _Fail("The master daemon is running, will not demote")
3483 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3484 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3485 except EnvironmentError, err:
3486 if err.errno != errno.ENOENT:
3487 _Fail("Error while backing up cluster file: %s", err, exc=True)
3489 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3492 def _GetX509Filenames(cryptodir, name):
3493 """Returns the full paths for the private key and certificate.
3496 return (utils.PathJoin(cryptodir, name),
3497 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3498 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3501 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3502 """Creates a new X509 certificate for SSL/TLS.
3505 @param validity: Validity in seconds
3506 @rtype: tuple; (string, string)
3507 @return: Certificate name and public part
3510 (key_pem, cert_pem) = \
3511 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3512 min(validity, _MAX_SSL_CERT_VALIDITY))
3514 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3515 prefix="x509-%s-" % utils.TimestampForFilename())
3517 name = os.path.basename(cert_dir)
3518 assert len(name) > 5
3520 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3522 utils.WriteFile(key_file, mode=0400, data=key_pem)
3523 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3525 # Never return private key as it shouldn't leave the node
3526 return (name, cert_pem)
3528 shutil.rmtree(cert_dir, ignore_errors=True)
3532 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3533 """Removes a X509 certificate.
3536 @param name: Certificate name
3539 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3541 utils.RemoveFile(key_file)
3542 utils.RemoveFile(cert_file)
3546 except EnvironmentError, err:
3547 _Fail("Cannot remove certificate directory '%s': %s",
3551 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3552 """Returns the command for the requested input/output.
3554 @type instance: L{objects.Instance}
3555 @param instance: The instance object
3556 @param mode: Import/export mode
3557 @param ieio: Input/output type
3558 @param ieargs: Input/output arguments
3561 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3568 if ieio == constants.IEIO_FILE:
3569 (filename, ) = ieargs
3571 if not utils.IsNormAbsPath(filename):
3572 _Fail("Path '%s' is not normalized or absolute", filename)
3574 real_filename = os.path.realpath(filename)
3575 directory = os.path.dirname(real_filename)
3577 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3578 _Fail("File '%s' is not under exports directory '%s': %s",
3579 filename, pathutils.EXPORT_DIR, real_filename)
3582 utils.Makedirs(directory, mode=0750)
3584 quoted_filename = utils.ShellQuote(filename)
3586 if mode == constants.IEM_IMPORT:
3587 suffix = "> %s" % quoted_filename
3588 elif mode == constants.IEM_EXPORT:
3589 suffix = "< %s" % quoted_filename
3591 # Retrieve file size
3593 st = os.stat(filename)
3594 except EnvironmentError, err:
3595 logging.error("Can't stat(2) %s: %s", filename, err)
3597 exp_size = utils.BytesToMebibyte(st.st_size)
3599 elif ieio == constants.IEIO_RAW_DISK:
3602 real_disk = _OpenRealBD(disk)
3604 if mode == constants.IEM_IMPORT:
3605 # we set here a smaller block size as, due to transport buffering, more
3606 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3607 # is not already there or we pass a wrong path; we use notrunc to no
3608 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3609 # much memory; this means that at best, we flush every 64k, which will
3611 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3612 " bs=%s oflag=dsync"),
3616 elif mode == constants.IEM_EXPORT:
3617 # the block size on the read dd is 1MiB to match our units
3618 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3620 str(1024 * 1024), # 1 MB
3622 exp_size = disk.size
3624 elif ieio == constants.IEIO_SCRIPT:
3625 (disk, disk_index, ) = ieargs
3627 assert isinstance(disk_index, (int, long))
3629 real_disk = _OpenRealBD(disk)
3631 inst_os = OSFromDisk(instance.os)
3632 env = OSEnvironment(instance, inst_os)
3634 if mode == constants.IEM_IMPORT:
3635 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3636 env["IMPORT_INDEX"] = str(disk_index)
3637 script = inst_os.import_script
3639 elif mode == constants.IEM_EXPORT:
3640 env["EXPORT_DEVICE"] = real_disk.dev_path
3641 env["EXPORT_INDEX"] = str(disk_index)
3642 script = inst_os.export_script
3644 # TODO: Pass special environment only to script
3645 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3647 if mode == constants.IEM_IMPORT:
3648 suffix = "| %s" % script_cmd
3650 elif mode == constants.IEM_EXPORT:
3651 prefix = "%s |" % script_cmd
3653 # Let script predict size
3654 exp_size = constants.IE_CUSTOM_SIZE
3657 _Fail("Invalid %s I/O mode %r", mode, ieio)
3659 return (env, prefix, suffix, exp_size)
3662 def _CreateImportExportStatusDir(prefix):
3663 """Creates status directory for import/export.
3666 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3668 (prefix, utils.TimestampForFilename())))
3671 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3673 """Starts an import or export daemon.
3675 @param mode: Import/output mode
3676 @type opts: L{objects.ImportExportOptions}
3677 @param opts: Daemon options
3679 @param host: Remote host for export (None for import)
3681 @param port: Remote port for export (None for import)
3682 @type instance: L{objects.Instance}
3683 @param instance: Instance object
3684 @type component: string
3685 @param component: which part of the instance is transferred now,
3687 @param ieio: Input/output type
3688 @param ieioargs: Input/output arguments
3691 if mode == constants.IEM_IMPORT:
3694 if not (host is None and port is None):
3695 _Fail("Can not specify host or port on import")
3697 elif mode == constants.IEM_EXPORT:
3700 if host is None or port is None:
3701 _Fail("Host and port must be specified for an export")
3704 _Fail("Invalid mode %r", mode)
3706 if (opts.key_name is None) ^ (opts.ca_pem is None):
3707 _Fail("Cluster certificate can only be used for both key and CA")
3709 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3710 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3712 if opts.key_name is None:
3714 key_path = pathutils.NODED_CERT_FILE
3715 cert_path = pathutils.NODED_CERT_FILE
3716 assert opts.ca_pem is None
3718 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3720 assert opts.ca_pem is not None
3722 for i in [key_path, cert_path]:
3723 if not os.path.exists(i):
3724 _Fail("File '%s' does not exist" % i)
3726 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3728 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3729 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3730 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3732 if opts.ca_pem is None:
3734 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3739 utils.WriteFile(ca_file, data=ca, mode=0400)
3742 pathutils.IMPORT_EXPORT_DAEMON,
3744 "--key=%s" % key_path,
3745 "--cert=%s" % cert_path,
3746 "--ca=%s" % ca_file,
3750 cmd.append("--host=%s" % host)
3753 cmd.append("--port=%s" % port)
3756 cmd.append("--ipv6")
3758 cmd.append("--ipv4")
3761 cmd.append("--compress=%s" % opts.compress)
3764 cmd.append("--magic=%s" % opts.magic)
3766 if exp_size is not None:
3767 cmd.append("--expected-size=%s" % exp_size)
3770 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3773 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3775 if mode == constants.IEM_EXPORT:
3776 # Retry connection a few times when connecting to remote peer
3777 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3778 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3779 elif opts.connect_timeout is not None:
3780 assert mode == constants.IEM_IMPORT
3781 # Overall timeout for establishing connection while listening
3782 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3784 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3786 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3787 # support for receiving a file descriptor for output
3788 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3791 # The import/export name is simply the status directory name
3792 return os.path.basename(status_dir)
3795 shutil.rmtree(status_dir, ignore_errors=True)
3799 def GetImportExportStatus(names):
3800 """Returns import/export daemon status.
3802 @type names: sequence
3803 @param names: List of names
3804 @rtype: List of dicts
3805 @return: Returns a list of the state of each named import/export or None if a
3806 status couldn't be read
3812 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3816 data = utils.ReadFile(status_file)
3817 except EnvironmentError, err:
3818 if err.errno != errno.ENOENT:
3826 result.append(serializer.LoadJson(data))
3831 def AbortImportExport(name):
3832 """Sends SIGTERM to a running import/export daemon.
3835 logging.info("Abort import/export %s", name)
3837 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3838 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3841 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3843 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3846 def CleanupImportExport(name):
3847 """Cleanup after an import or export.
3849 If the import/export daemon is still running it's killed. Afterwards the
3850 whole status directory is removed.
3853 logging.info("Finalizing import/export %s", name)
3855 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3857 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3860 logging.info("Import/export %s is still running with PID %s",
3862 utils.KillProcess(pid, waitpid=False)
3864 shutil.rmtree(status_dir, ignore_errors=True)
3867 def _SetPhysicalId(target_node_uuid, nodes_ip, disks):
3868 """Sets the correct physical ID on all passed disks.
3872 cf.SetPhysicalID(target_node_uuid, nodes_ip)
3875 def _FindDisks(target_node_uuid, nodes_ip, disks):
3876 """Sets the physical ID on disks and returns the block devices.
3879 _SetPhysicalId(target_node_uuid, nodes_ip, disks)
3884 rd = _RecursiveFindBD(cf)
3886 _Fail("Can't find device %s", cf)
3891 def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
3892 """Disconnects the network on a list of drbd devices.
3895 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3901 except errors.BlockDeviceError, err:
3902 _Fail("Can't change network configuration to standalone mode: %s",
3906 def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
3908 """Attaches the network on a list of drbd devices.
3911 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3914 for idx, rd in enumerate(bdevs):
3916 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3917 except EnvironmentError, err:
3918 _Fail("Can't create symlink: %s", err)
3919 # reconnect disks, switch to new master configuration and if
3920 # needed primary mode
3923 rd.AttachNet(multimaster)
3924 except errors.BlockDeviceError, err:
3925 _Fail("Can't change network configuration: %s", err)
3927 # wait until the disks are connected; we need to retry the re-attach
3928 # if the device becomes standalone, as this might happen if the one
3929 # node disconnects and reconnects in a different mode before the
3930 # other node reconnects; in this case, one or both of the nodes will
3931 # decide it has wrong configuration and switch to standalone
3934 all_connected = True
3937 stats = rd.GetProcStatus()
3939 all_connected = (all_connected and
3940 (stats.is_connected or stats.is_in_resync))
3942 if stats.is_standalone:
3943 # peer had different config info and this node became
3944 # standalone, even though this should not happen with the
3945 # new staged way of changing disk configs
3947 rd.AttachNet(multimaster)
3948 except errors.BlockDeviceError, err:
3949 _Fail("Can't change network configuration: %s", err)
3951 if not all_connected:
3952 raise utils.RetryAgain()
3955 # Start with a delay of 100 miliseconds and go up to 5 seconds
3956 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3957 except utils.RetryTimeout:
3958 _Fail("Timeout in disk reconnecting")
3961 # change to primary mode
3965 except errors.BlockDeviceError, err:
3966 _Fail("Can't change to primary mode: %s", err)
3969 def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
3970 """Wait until DRBDs have synchronized.
3974 stats = rd.GetProcStatus()
3975 if not (stats.is_connected or stats.is_in_resync):
3976 raise utils.RetryAgain()
3979 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3985 # poll each second for 15 seconds
3986 stats = utils.Retry(_helper, 1, 15, args=[rd])
3987 except utils.RetryTimeout:
3988 stats = rd.GetProcStatus()
3990 if not (stats.is_connected or stats.is_in_resync):
3991 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3992 alldone = alldone and (not stats.is_in_resync)
3993 if stats.sync_percent is not None:
3994 min_resync = min(min_resync, stats.sync_percent)
3996 return (alldone, min_resync)
3999 def DrbdNeedsActivation(target_node_uuid, nodes_ip, disks):
4000 """Checks which of the passed disks needs activation and returns their UUIDs.
4003 _SetPhysicalId(target_node_uuid, nodes_ip, disks)
4007 rd = _RecursiveFindBD(disk)
4009 faulty_disks.append(disk)
4012 stats = rd.GetProcStatus()
4013 if stats.is_standalone or stats.is_diskless:
4014 faulty_disks.append(disk)
4016 return [disk.uuid for disk in faulty_disks]
4019 def GetDrbdUsermodeHelper():
4020 """Returns DRBD usermode helper currently configured.
4024 return drbd.DRBD8.GetUsermodeHelper()
4025 except errors.BlockDeviceError, err:
4029 def PowercycleNode(hypervisor_type, hvparams=None):
4030 """Hard-powercycle the node.
4032 Because we need to return first, and schedule the powercycle in the
4033 background, we won't be able to report failures nicely.
4036 hyper = hypervisor.GetHypervisor(hypervisor_type)
4040 # if we can't fork, we'll pretend that we're in the child process
4043 return "Reboot scheduled in 5 seconds"
4044 # ensure the child is running on ram
4047 except Exception: # pylint: disable=W0703
4050 hyper.PowercycleNode(hvparams=hvparams)
4053 def _VerifyRestrictedCmdName(cmd):
4054 """Verifies a restricted command name.
4057 @param cmd: Command name
4058 @rtype: tuple; (boolean, string or None)
4059 @return: The tuple's first element is the status; if C{False}, the second
4060 element is an error message string, otherwise it's C{None}
4064 return (False, "Missing command name")
4066 if os.path.basename(cmd) != cmd:
4067 return (False, "Invalid command name")
4069 if not constants.EXT_PLUGIN_MASK.match(cmd):
4070 return (False, "Command name contains forbidden characters")
4075 def _CommonRestrictedCmdCheck(path, owner):
4076 """Common checks for restricted command file system directories and files.
4079 @param path: Path to check
4080 @param owner: C{None} or tuple containing UID and GID
4081 @rtype: tuple; (boolean, string or C{os.stat} result)
4082 @return: The tuple's first element is the status; if C{False}, the second
4083 element is an error message string, otherwise it's the result of C{os.stat}
4087 # Default to root as owner
4092 except EnvironmentError, err:
4093 return (False, "Can't stat(2) '%s': %s" % (path, err))
4095 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4096 return (False, "Permissions on '%s' are too permissive" % path)
4098 if (st.st_uid, st.st_gid) != owner:
4099 (owner_uid, owner_gid) = owner
4100 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4105 def _VerifyRestrictedCmdDirectory(path, _owner=None):
4106 """Verifies restricted command directory.
4109 @param path: Path to check
4110 @rtype: tuple; (boolean, string or None)
4111 @return: The tuple's first element is the status; if C{False}, the second
4112 element is an error message string, otherwise it's C{None}
4115 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4118 return (False, value)
4120 if not stat.S_ISDIR(value.st_mode):
4121 return (False, "Path '%s' is not a directory" % path)
4126 def _VerifyRestrictedCmd(path, cmd, _owner=None):
4127 """Verifies a whole restricted command and returns its executable filename.
4130 @param path: Directory containing restricted commands
4132 @param cmd: Command name
4133 @rtype: tuple; (boolean, string)
4134 @return: The tuple's first element is the status; if C{False}, the second
4135 element is an error message string, otherwise the second element is the
4136 absolute path to the executable
4139 executable = utils.PathJoin(path, cmd)
4141 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4146 if not utils.IsExecutable(executable):
4147 return (False, "access(2) thinks '%s' can't be executed" % executable)
4149 return (True, executable)
4152 def _PrepareRestrictedCmd(path, cmd,
4153 _verify_dir=_VerifyRestrictedCmdDirectory,
4154 _verify_name=_VerifyRestrictedCmdName,
4155 _verify_cmd=_VerifyRestrictedCmd):
4156 """Performs a number of tests on a restricted command.
4159 @param path: Directory containing restricted commands
4161 @param cmd: Command name
4162 @return: Same as L{_VerifyRestrictedCmd}
4165 # Verify the directory first
4166 (status, msg) = _verify_dir(path)
4168 # Check command if everything was alright
4169 (status, msg) = _verify_name(cmd)
4174 # Check actual executable
4175 return _verify_cmd(path, cmd)
4178 def RunRestrictedCmd(cmd,
4179 _lock_timeout=_RCMD_LOCK_TIMEOUT,
4180 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4181 _path=pathutils.RESTRICTED_COMMANDS_DIR,
4182 _sleep_fn=time.sleep,
4183 _prepare_fn=_PrepareRestrictedCmd,
4184 _runcmd_fn=utils.RunCmd,
4185 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4186 """Executes a restricted command after performing strict tests.
4189 @param cmd: Command name
4191 @return: Command output
4192 @raise RPCFail: In case of an error
4195 logging.info("Preparing to run restricted command '%s'", cmd)
4198 _Fail("Restricted commands disabled at configure time")
4204 lock = utils.FileLock.Open(_lock_file)
4205 lock.Exclusive(blocking=True, timeout=_lock_timeout)
4207 (status, value) = _prepare_fn(_path, cmd)
4210 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4211 postfork_fn=lambda _: lock.Unlock())
4213 logging.error(value)
4214 except Exception: # pylint: disable=W0703
4215 # Keep original error in log
4216 logging.exception("Caught exception")
4218 if cmdresult is None:
4219 logging.info("Sleeping for %0.1f seconds before returning",
4220 _RCMD_INVALID_DELAY)
4221 _sleep_fn(_RCMD_INVALID_DELAY)
4223 # Do not include original error message in returned error
4224 _Fail("Executing command '%s' failed" % cmd)
4225 elif cmdresult.failed or cmdresult.fail_reason:
4226 _Fail("Restricted command '%s' failed: %s; output: %s",
4227 cmd, cmdresult.fail_reason, cmdresult.output)
4229 return cmdresult.output
4231 if lock is not None:
4232 # Release lock at last
4237 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4238 """Creates or removes the watcher pause file.
4240 @type until: None or number
4241 @param until: Unix timestamp saying until when the watcher shouldn't run
4245 logging.info("Received request to no longer pause watcher")
4246 utils.RemoveFile(_filename)
4248 logging.info("Received request to pause watcher until %s", until)
4250 if not ht.TNumber(until):
4251 _Fail("Duration must be numeric")
4253 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4256 class HooksRunner(object):
4259 This class is instantiated on the node side (ganeti-noded) and not
4263 def __init__(self, hooks_base_dir=None):
4264 """Constructor for hooks runner.
4266 @type hooks_base_dir: str or None
4267 @param hooks_base_dir: if not None, this overrides the
4268 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4271 if hooks_base_dir is None:
4272 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4273 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4275 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4277 def RunLocalHooks(self, node_list, hpath, phase, env):
4278 """Check that the hooks will be run only locally and then run them.
4281 assert len(node_list) == 1
4283 _, myself = ssconf.GetMasterAndMyself()
4284 assert node == myself
4286 results = self.RunHooks(hpath, phase, env)
4288 # Return values in the form expected by HooksMaster
4289 return {node: (None, False, results)}
4291 def RunHooks(self, hpath, phase, env):
4292 """Run the scripts in the hooks directory.
4295 @param hpath: the path to the hooks directory which
4298 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4299 L{constants.HOOKS_PHASE_POST}
4301 @param env: dictionary with the environment for the hook
4303 @return: list of 3-element tuples:
4305 - script result, either L{constants.HKR_SUCCESS} or
4306 L{constants.HKR_FAIL}
4307 - output of the script
4309 @raise errors.ProgrammerError: for invalid input
4313 if phase == constants.HOOKS_PHASE_PRE:
4315 elif phase == constants.HOOKS_PHASE_POST:
4318 _Fail("Unknown hooks phase '%s'", phase)
4320 subdir = "%s-%s.d" % (hpath, suffix)
4321 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4325 if not os.path.isdir(dir_name):
4326 # for non-existing/non-dirs, we simply exit instead of logging a
4327 # warning at every operation
4330 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4332 for (relname, relstatus, runresult) in runparts_results:
4333 if relstatus == constants.RUNPARTS_SKIP:
4334 rrval = constants.HKR_SKIP
4336 elif relstatus == constants.RUNPARTS_ERR:
4337 rrval = constants.HKR_FAIL
4338 output = "Hook script execution error: %s" % runresult
4339 elif relstatus == constants.RUNPARTS_RUN:
4340 if runresult.failed:
4341 rrval = constants.HKR_FAIL
4343 rrval = constants.HKR_SUCCESS
4344 output = utils.SafeEncode(runresult.output.strip())
4345 results.append(("%s/%s" % (subdir, relname), rrval, output))
4350 class IAllocatorRunner(object):
4351 """IAllocator runner.
4353 This class is instantiated on the node side (ganeti-noded) and not on
4358 def Run(name, idata):
4359 """Run an iallocator script.
4362 @param name: the iallocator script name
4364 @param idata: the allocator input data
4367 @return: two element tuple of:
4369 - either error message or stdout of allocator (for success)
4372 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4374 if alloc_script is None:
4375 _Fail("iallocator module '%s' not found in the search path", name)
4377 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4381 result = utils.RunCmd([alloc_script, fin_name])
4383 _Fail("iallocator module '%s' failed: %s, output '%s'",
4384 name, result.fail_reason, result.output)
4388 return result.stdout
4391 class DevCacheManager(object):
4392 """Simple class for managing a cache of block device information.
4395 _DEV_PREFIX = "/dev/"
4396 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4399 def _ConvertPath(cls, dev_path):
4400 """Converts a /dev/name path to the cache file name.
4402 This replaces slashes with underscores and strips the /dev
4403 prefix. It then returns the full path to the cache file.
4406 @param dev_path: the C{/dev/} path name
4408 @return: the converted path name
4411 if dev_path.startswith(cls._DEV_PREFIX):
4412 dev_path = dev_path[len(cls._DEV_PREFIX):]
4413 dev_path = dev_path.replace("/", "_")
4414 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4418 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4419 """Updates the cache information for a given device.
4422 @param dev_path: the pathname of the device
4424 @param owner: the owner (instance name) of the device
4425 @type on_primary: bool
4426 @param on_primary: whether this is the primary
4429 @param iv_name: the instance-visible name of the
4430 device, as in objects.Disk.iv_name
4435 if dev_path is None:
4436 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4438 fpath = cls._ConvertPath(dev_path)
4444 iv_name = "not_visible"
4445 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4447 utils.WriteFile(fpath, data=fdata)
4448 except EnvironmentError, err:
4449 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4452 def RemoveCache(cls, dev_path):
4453 """Remove data for a dev_path.
4455 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4456 path name and logging.
4459 @param dev_path: the pathname of the device
4464 if dev_path is None:
4465 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4467 fpath = cls._ConvertPath(dev_path)
4469 utils.RemoveFile(fpath)
4470 except EnvironmentError, err:
4471 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)