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)
625 def _GetVgInfo(name, excl_stor):
626 """Retrieves information about a LVM volume group.
629 # TODO: GetVGInfo supports returning information for multiple VGs at once
630 vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
632 vg_free = int(round(vginfo[0][0], 0))
633 vg_size = int(round(vginfo[0][1], 0))
639 "type": constants.ST_LVM_VG,
641 "storage_free": vg_free,
642 "storage_size": vg_size,
646 def _GetLvmPvSpaceInfo(name, params):
647 """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
649 @see: C{_GetLvmVgSpaceInfo}
652 excl_stor = _CheckLvmStorageParams(params)
653 return _GetVgSpindlesInfo(name, excl_stor)
656 def _GetVgSpindlesInfo(
657 name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
658 """Retrieves information about spindles in an LVM volume group.
662 @type excl_stor: bool
663 @param excl_stor: exclusive storage
665 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
666 free spindles, total spindles respectively
670 (vg_free, vg_size) = info_fn(name)
675 "type": constants.ST_LVM_PV,
677 "storage_free": vg_free,
678 "storage_size": vg_size,
682 def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
683 """Retrieves node information from a hypervisor.
685 The information returned depends on the hypervisor. Common items:
687 - vg_size is the size of the configured volume group in MiB
688 - vg_free is the free size of the volume group in MiB
689 - memory_dom0 is the memory allocated for domain0 in MiB
690 - memory_free is the currently available (free) ram in MiB
691 - memory_total is the total number of ram in MiB
692 - hv_version: the hypervisor version, if available
694 @type hvparams: dict of string
695 @param hvparams: the hypervisor's hvparams
698 return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
701 def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
702 """Retrieves node information for all hypervisors.
704 See C{_GetHvInfo} for information on the output.
706 @type hv_specs: list of pairs (string, dict of strings)
707 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
714 for hvname, hvparams in hv_specs:
715 result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
719 def _GetNamedNodeInfo(names, fn):
720 """Calls C{fn} for all names in C{names} and returns a dictionary.
728 return map(fn, names)
731 def GetNodeInfo(storage_units, hv_specs):
732 """Gives back a hash with different information about the node.
734 @type storage_units: list of tuples (string, string, list)
735 @param storage_units: List of tuples (storage unit, identifier, parameters) to
736 ask for disk space information. In case of lvm-vg, the identifier is
737 the VG name. The parameters can contain additional, storage-type-specific
738 parameters, for example exclusive storage for lvm storage.
739 @type hv_specs: list of pairs (string, dict of strings)
740 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
741 @rtype: tuple; (string, None/dict, None/dict)
742 @return: Tuple containing boot ID, volume group information and hypervisor
746 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
747 storage_info = _GetNamedNodeInfo(
749 (lambda (storage_type, storage_key, storage_params):
750 _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
751 hv_info = _GetHvInfoAll(hv_specs)
752 return (bootid, storage_info, hv_info)
755 def _GetFileStorageSpaceInfo(path, params):
756 """Wrapper around filestorage.GetSpaceInfo.
758 The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
759 and ignore the *args parameter to not leak it into the filestorage
762 @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
766 _CheckStorageParams(params, 0)
767 return filestorage.GetFileStorageSpaceInfo(path)
770 # FIXME: implement storage reporting for all missing storage types.
771 _STORAGE_TYPE_INFO_FN = {
772 constants.ST_BLOCK: None,
773 constants.ST_DISKLESS: None,
774 constants.ST_EXT: None,
775 constants.ST_FILE: _GetFileStorageSpaceInfo,
776 constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
777 constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
778 constants.ST_RADOS: None,
782 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
783 """Looks up and applies the correct function to calculate free and total
784 storage for the given storage type.
786 @type storage_type: string
787 @param storage_type: the storage type for which the storage shall be reported.
788 @type storage_key: string
789 @param storage_key: identifier of a storage unit, e.g. the volume group name
790 of an LVM storage unit
792 @param args: various parameters that can be used for storage reporting. These
793 parameters and their semantics vary from storage type to storage type and
794 are just propagated in this function.
795 @return: the results of the application of the storage space function (see
796 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
798 @raises NotImplementedError: for storage types who don't support space
801 fn = _STORAGE_TYPE_INFO_FN[storage_type]
803 return fn(storage_key, *args)
805 raise NotImplementedError
808 def _CheckExclusivePvs(pvi_list):
809 """Check that PVs are not shared among LVs
811 @type pvi_list: list of L{objects.LvmPvInfo} objects
812 @param pvi_list: information about the PVs
814 @rtype: list of tuples (string, list of strings)
815 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
820 if len(pvi.lv_list) > 1:
821 res.append((pvi.name, pvi.lv_list))
825 def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
826 get_hv_fn=hypervisor.GetHypervisor):
827 """Verifies the hypervisor. Appends the results to the 'results' list.
830 @param what: a dictionary of things to check
831 @type vm_capable: boolean
832 @param vm_capable: whether or not this node is vm capable
834 @param result: dictionary of verification results; results of the
835 verifications in this function will be added here
836 @type all_hvparams: dict of dict of string
837 @param all_hvparams: dictionary mapping hypervisor names to hvparams
838 @type get_hv_fn: function
839 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
845 if constants.NV_HYPERVISOR in what:
846 result[constants.NV_HYPERVISOR] = {}
847 for hv_name in what[constants.NV_HYPERVISOR]:
848 hvparams = all_hvparams[hv_name]
850 val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
851 except errors.HypervisorError, err:
852 val = "Error while checking hypervisor: %s" % str(err)
853 result[constants.NV_HYPERVISOR][hv_name] = val
856 def _VerifyHvparams(what, vm_capable, result,
857 get_hv_fn=hypervisor.GetHypervisor):
858 """Verifies the hvparams. Appends the results to the 'results' list.
861 @param what: a dictionary of things to check
862 @type vm_capable: boolean
863 @param vm_capable: whether or not this node is vm capable
865 @param result: dictionary of verification results; results of the
866 verifications in this function will be added here
867 @type get_hv_fn: function
868 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
874 if constants.NV_HVPARAMS in what:
875 result[constants.NV_HVPARAMS] = []
876 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
878 logging.info("Validating hv %s, %s", hv_name, hvparms)
879 get_hv_fn(hv_name).ValidateParameters(hvparms)
880 except errors.HypervisorError, err:
881 result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
884 def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
885 """Verifies the instance list.
888 @param what: a dictionary of things to check
889 @type vm_capable: boolean
890 @param vm_capable: whether or not this node is vm capable
892 @param result: dictionary of verification results; results of the
893 verifications in this function will be added here
894 @type all_hvparams: dict of dict of string
895 @param all_hvparams: dictionary mapping hypervisor names to hvparams
898 if constants.NV_INSTANCELIST in what and vm_capable:
899 # GetInstanceList can fail
901 val = GetInstanceList(what[constants.NV_INSTANCELIST],
902 all_hvparams=all_hvparams)
905 result[constants.NV_INSTANCELIST] = val
908 def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
909 """Verifies the node info.
912 @param what: a dictionary of things to check
913 @type vm_capable: boolean
914 @param vm_capable: whether or not this node is vm capable
916 @param result: dictionary of verification results; results of the
917 verifications in this function will be added here
918 @type all_hvparams: dict of dict of string
919 @param all_hvparams: dictionary mapping hypervisor names to hvparams
922 if constants.NV_HVINFO in what and vm_capable:
923 hvname = what[constants.NV_HVINFO]
924 hyper = hypervisor.GetHypervisor(hvname)
925 hvparams = all_hvparams[hvname]
926 result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
929 def VerifyNode(what, cluster_name, all_hvparams):
930 """Verify the status of the local node.
932 Based on the input L{what} parameter, various checks are done on the
935 If the I{filelist} key is present, this list of
936 files is checksummed and the file/checksum pairs are returned.
938 If the I{nodelist} key is present, we check that we have
939 connectivity via ssh with the target nodes (and check the hostname
942 If the I{node-net-test} key is present, we check that we have
943 connectivity to the given nodes via both primary IP and, if
944 applicable, secondary IPs.
947 @param what: a dictionary of things to check:
948 - filelist: list of files for which to compute checksums
949 - nodelist: list of nodes we should check ssh communication with
950 - node-net-test: list of nodes we should check node daemon port
952 - hypervisor: list with hypervisors to run the verify for
953 @type cluster_name: string
954 @param cluster_name: the cluster's name
955 @type all_hvparams: dict of dict of strings
956 @param all_hvparams: a dictionary mapping hypervisor names to hvparams
958 @return: a dictionary with the same keys as the input dict, and
959 values representing the result of the checks
963 my_name = netutils.Hostname.GetSysName()
964 port = netutils.GetDaemonPort(constants.NODED)
965 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
967 _VerifyHypervisors(what, vm_capable, result, all_hvparams)
968 _VerifyHvparams(what, vm_capable, result)
970 if constants.NV_FILELIST in what:
971 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
972 what[constants.NV_FILELIST]))
973 result[constants.NV_FILELIST] = \
974 dict((vcluster.MakeVirtualPath(key), value)
975 for (key, value) in fingerprints.items())
977 if constants.NV_NODELIST in what:
978 (nodes, bynode) = what[constants.NV_NODELIST]
980 # Add nodes from other groups (different for each node)
982 nodes.extend(bynode[my_name])
987 random.shuffle(nodes)
989 # Try to contact all nodes
992 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
996 result[constants.NV_NODELIST] = val
998 if constants.NV_NODENETTEST in what:
999 result[constants.NV_NODENETTEST] = tmp = {}
1000 my_pip = my_sip = None
1001 for name, pip, sip in what[constants.NV_NODENETTEST]:
1007 tmp[my_name] = ("Can't find my own primary/secondary IP"
1008 " in the node list")
1010 for name, pip, sip in what[constants.NV_NODENETTEST]:
1012 if not netutils.TcpPing(pip, port, source=my_pip):
1013 fail.append("primary")
1015 if not netutils.TcpPing(sip, port, source=my_sip):
1016 fail.append("secondary")
1018 tmp[name] = ("failure using the %s interface(s)" %
1021 if constants.NV_MASTERIP in what:
1022 # FIXME: add checks on incoming data structures (here and in the
1023 # rest of the function)
1024 master_name, master_ip = what[constants.NV_MASTERIP]
1025 if master_name == my_name:
1026 source = constants.IP4_ADDRESS_LOCALHOST
1029 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
1032 if constants.NV_USERSCRIPTS in what:
1033 result[constants.NV_USERSCRIPTS] = \
1034 [script for script in what[constants.NV_USERSCRIPTS]
1035 if not utils.IsExecutable(script)]
1037 if constants.NV_OOB_PATHS in what:
1038 result[constants.NV_OOB_PATHS] = tmp = []
1039 for path in what[constants.NV_OOB_PATHS]:
1042 except OSError, err:
1043 tmp.append("error stating out of band helper: %s" % err)
1045 if stat.S_ISREG(st.st_mode):
1046 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
1049 tmp.append("out of band helper %s is not executable" % path)
1051 tmp.append("out of band helper %s is not a file" % path)
1053 if constants.NV_LVLIST in what and vm_capable:
1055 val = GetVolumeList(utils.ListVolumeGroups().keys())
1056 except RPCFail, err:
1058 result[constants.NV_LVLIST] = val
1060 _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1062 if constants.NV_VGLIST in what and vm_capable:
1063 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1065 if constants.NV_PVLIST in what and vm_capable:
1066 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1067 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1068 filter_allocatable=False,
1069 include_lvs=check_exclusive_pvs)
1070 if check_exclusive_pvs:
1071 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1073 # Avoid sending useless data on the wire
1075 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1077 if constants.NV_VERSION in what:
1078 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1079 constants.RELEASE_VERSION)
1081 _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1083 if constants.NV_DRBDVERSION in what and vm_capable:
1085 drbd_version = DRBD8.GetProcInfo().GetVersionString()
1086 except errors.BlockDeviceError, err:
1087 logging.warning("Can't get DRBD version", exc_info=True)
1088 drbd_version = str(err)
1089 result[constants.NV_DRBDVERSION] = drbd_version
1091 if constants.NV_DRBDLIST in what and vm_capable:
1093 used_minors = drbd.DRBD8.GetUsedDevs()
1094 except errors.BlockDeviceError, err:
1095 logging.warning("Can't get used minors list", exc_info=True)
1096 used_minors = str(err)
1097 result[constants.NV_DRBDLIST] = used_minors
1099 if constants.NV_DRBDHELPER in what and vm_capable:
1102 payload = drbd.DRBD8.GetUsermodeHelper()
1103 except errors.BlockDeviceError, err:
1104 logging.error("Can't get DRBD usermode helper: %s", str(err))
1107 result[constants.NV_DRBDHELPER] = (status, payload)
1109 if constants.NV_NODESETUP in what:
1110 result[constants.NV_NODESETUP] = tmpr = []
1111 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1112 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1113 " under /sys, missing required directories /sys/block"
1114 " and /sys/class/net")
1115 if (not os.path.isdir("/proc/sys") or
1116 not os.path.isfile("/proc/sysrq-trigger")):
1117 tmpr.append("The procfs filesystem doesn't seem to be mounted"
1118 " under /proc, missing required directory /proc/sys and"
1119 " the file /proc/sysrq-trigger")
1121 if constants.NV_TIME in what:
1122 result[constants.NV_TIME] = utils.SplitTime(time.time())
1124 if constants.NV_OSLIST in what and vm_capable:
1125 result[constants.NV_OSLIST] = DiagnoseOS()
1127 if constants.NV_BRIDGES in what and vm_capable:
1128 result[constants.NV_BRIDGES] = [bridge
1129 for bridge in what[constants.NV_BRIDGES]
1130 if not utils.BridgeExists(bridge)]
1132 if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
1133 result[constants.NV_FILE_STORAGE_PATHS] = \
1134 bdev.ComputeWrongFileStoragePaths()
1139 def GetBlockDevSizes(devices):
1140 """Return the size of the given block devices
1143 @param devices: list of block device nodes to query
1146 dictionary of all block devices under /dev (key). The value is their
1149 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1152 DEV_PREFIX = "/dev/"
1155 for devpath in devices:
1156 if not utils.IsBelowDir(DEV_PREFIX, devpath):
1160 st = os.stat(devpath)
1161 except EnvironmentError, err:
1162 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1165 if stat.S_ISBLK(st.st_mode):
1166 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1168 # We don't want to fail, just do not list this device as available
1169 logging.warning("Cannot get size for block device %s", devpath)
1172 size = int(result.stdout) / (1024 * 1024)
1173 blockdevs[devpath] = size
1177 def GetVolumeList(vg_names):
1178 """Compute list of logical volumes and their size.
1180 @type vg_names: list
1181 @param vg_names: the volume groups whose LVs we should list, or
1182 empty for all volume groups
1185 dictionary of all partions (key) with value being a tuple of
1186 their size (in MiB), inactive and online status::
1188 {'xenvg/test1': ('20.06', True, True)}
1190 in case of errors, a string is returned with the error
1198 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1199 "--separator=%s" % sep,
1200 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1202 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1204 for line in result.stdout.splitlines():
1206 match = _LVSLINE_REGEX.match(line)
1208 logging.error("Invalid line returned from lvs output: '%s'", line)
1210 vg_name, name, size, attr = match.groups()
1211 inactive = attr[4] == "-"
1212 online = attr[5] == "o"
1213 virtual = attr[0] == "v"
1215 # we don't want to report such volumes as existing, since they
1216 # don't really hold data
1218 lvs[vg_name + "/" + name] = (size, inactive, online)
1223 def ListVolumeGroups():
1224 """List the volume groups and their size.
1227 @return: dictionary with keys volume name and values the
1231 return utils.ListVolumeGroups()
1235 """List all volumes on this node.
1239 A list of dictionaries, each having four keys:
1240 - name: the logical volume name,
1241 - size: the size of the logical volume
1242 - dev: the physical device on which the LV lives
1243 - vg: the volume group to which it belongs
1245 In case of errors, we return an empty list and log the
1248 Note that since a logical volume can live on multiple physical
1249 volumes, the resulting list might include a logical volume
1253 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1255 "--options=lv_name,lv_size,devices,vg_name"])
1257 _Fail("Failed to list logical volumes, lvs output: %s",
1261 return dev.split("(")[0]
1263 def handle_dev(dev):
1264 return [parse_dev(x) for x in dev.split(",")]
1267 line = [v.strip() for v in line]
1268 return [{"name": line[0], "size": line[1],
1269 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1272 for line in result.stdout.splitlines():
1273 if line.count("|") >= 3:
1274 all_devs.extend(map_line(line.split("|")))
1276 logging.warning("Strange line in the output from lvs: '%s'", line)
1280 def BridgesExist(bridges_list):
1281 """Check if a list of bridges exist on the current node.
1284 @return: C{True} if all of them exist, C{False} otherwise
1288 for bridge in bridges_list:
1289 if not utils.BridgeExists(bridge):
1290 missing.append(bridge)
1293 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1296 def GetInstanceListForHypervisor(hname, hvparams=None,
1297 get_hv_fn=hypervisor.GetHypervisor):
1298 """Provides a list of instances of the given hypervisor.
1301 @param hname: name of the hypervisor
1302 @type hvparams: dict of strings
1303 @param hvparams: hypervisor parameters for the given hypervisor
1304 @type get_hv_fn: function
1305 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1306 name; optional parameter to increase testability
1309 @return: a list of all running instances on the current node
1310 - instance1.example.com
1311 - instance2.example.com
1316 hv = get_hv_fn(hname)
1317 names = hv.ListInstances(hvparams=hvparams)
1318 results.extend(names)
1319 except errors.HypervisorError, err:
1320 _Fail("Error enumerating instances (hypervisor %s): %s",
1321 hname, err, exc=True)
1325 def GetInstanceList(hypervisor_list, all_hvparams=None,
1326 get_hv_fn=hypervisor.GetHypervisor):
1327 """Provides a list of instances.
1329 @type hypervisor_list: list
1330 @param hypervisor_list: the list of hypervisors to query information
1331 @type all_hvparams: dict of dict of strings
1332 @param all_hvparams: a dictionary mapping hypervisor types to respective
1333 cluster-wide hypervisor parameters
1334 @type get_hv_fn: function
1335 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1336 name; optional parameter to increase testability
1339 @return: a list of all running instances on the current node
1340 - instance1.example.com
1341 - instance2.example.com
1345 for hname in hypervisor_list:
1346 hvparams = all_hvparams[hname]
1347 results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1348 get_hv_fn=get_hv_fn))
1352 def GetInstanceInfo(instance, hname, hvparams=None):
1353 """Gives back the information about an instance as a dictionary.
1355 @type instance: string
1356 @param instance: the instance name
1358 @param hname: the hypervisor type of the instance
1359 @type hvparams: dict of strings
1360 @param hvparams: the instance's hvparams
1363 @return: dictionary with the following keys:
1364 - memory: memory size of instance (int)
1365 - state: xen state of instance (string)
1366 - time: cpu time of instance (float)
1367 - vcpus: the number of vcpus (int)
1372 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1374 if iinfo is not None:
1375 output["memory"] = iinfo[2]
1376 output["vcpus"] = iinfo[3]
1377 output["state"] = iinfo[4]
1378 output["time"] = iinfo[5]
1383 def GetInstanceMigratable(instance):
1384 """Computes whether an instance can be migrated.
1386 @type instance: L{objects.Instance}
1387 @param instance: object representing the instance to be checked.
1390 @return: tuple of (result, description) where:
1391 - result: whether the instance can be migrated or not
1392 - description: a description of the issue, if relevant
1395 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1396 iname = instance.name
1397 if iname not in hyper.ListInstances(instance.hvparams):
1398 _Fail("Instance %s is not running", iname)
1400 for idx in range(len(instance.disks)):
1401 link_name = _GetBlockDevSymlinkPath(iname, idx)
1402 if not os.path.islink(link_name):
1403 logging.warning("Instance %s is missing symlink %s for disk %d",
1404 iname, link_name, idx)
1407 def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1408 """Gather data about all instances.
1410 This is the equivalent of L{GetInstanceInfo}, except that it
1411 computes data for all instances at once, thus being faster if one
1412 needs data about more than one instance.
1414 @type hypervisor_list: list
1415 @param hypervisor_list: list of hypervisors to query for instance data
1416 @type all_hvparams: dict of dict of strings
1417 @param all_hvparams: mapping of hypervisor names to hvparams
1420 @return: dictionary of instance: data, with data having the following keys:
1421 - memory: memory size of instance (int)
1422 - state: xen state of instance (string)
1423 - time: cpu time of instance (float)
1424 - vcpus: the number of vcpus
1429 for hname in hypervisor_list:
1430 hvparams = all_hvparams[hname]
1431 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1433 for name, _, memory, vcpus, state, times in iinfo:
1441 # we only check static parameters, like memory and vcpus,
1442 # and not state and time which can change between the
1443 # invocations of the different hypervisors
1444 for key in "memory", "vcpus":
1445 if value[key] != output[name][key]:
1446 _Fail("Instance %s is running twice"
1447 " with different parameters", name)
1448 output[name] = value
1453 def _InstanceLogName(kind, os_name, instance, component):
1454 """Compute the OS log filename for a given instance and operation.
1456 The instance name and os name are passed in as strings since not all
1457 operations have these as part of an instance object.
1460 @param kind: the operation type (e.g. add, import, etc.)
1461 @type os_name: string
1462 @param os_name: the os name
1463 @type instance: string
1464 @param instance: the name of the instance being imported/added/etc.
1465 @type component: string or None
1466 @param component: the name of the component of the instance being
1470 # TODO: Use tempfile.mkstemp to create unique filename
1472 assert "/" not in component
1473 c_msg = "-%s" % component
1476 base = ("%s-%s-%s%s-%s.log" %
1477 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1478 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1481 def InstanceOsAdd(instance, reinstall, debug):
1482 """Add an OS to an instance.
1484 @type instance: L{objects.Instance}
1485 @param instance: Instance whose OS is to be installed
1486 @type reinstall: boolean
1487 @param reinstall: whether this is an instance reinstall
1488 @type debug: integer
1489 @param debug: debug level, passed to the OS scripts
1493 inst_os = OSFromDisk(instance.os)
1495 create_env = OSEnvironment(instance, inst_os, debug)
1497 create_env["INSTANCE_REINSTALL"] = "1"
1499 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1501 result = utils.RunCmd([inst_os.create_script], env=create_env,
1502 cwd=inst_os.path, output=logfile, reset_env=True)
1504 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1505 " output: %s", result.cmd, result.fail_reason, logfile,
1507 lines = [utils.SafeEncode(val)
1508 for val in utils.TailFile(logfile, lines=20)]
1509 _Fail("OS create script failed (%s), last lines in the"
1510 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1513 def RunRenameInstance(instance, old_name, debug):
1514 """Run the OS rename script for an instance.
1516 @type instance: L{objects.Instance}
1517 @param instance: Instance whose OS is to be installed
1518 @type old_name: string
1519 @param old_name: previous instance name
1520 @type debug: integer
1521 @param debug: debug level, passed to the OS scripts
1523 @return: the success of the operation
1526 inst_os = OSFromDisk(instance.os)
1528 rename_env = OSEnvironment(instance, inst_os, debug)
1529 rename_env["OLD_INSTANCE_NAME"] = old_name
1531 logfile = _InstanceLogName("rename", instance.os,
1532 "%s-%s" % (old_name, instance.name), None)
1534 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1535 cwd=inst_os.path, output=logfile, reset_env=True)
1538 logging.error("os create command '%s' returned error: %s output: %s",
1539 result.cmd, result.fail_reason, result.output)
1540 lines = [utils.SafeEncode(val)
1541 for val in utils.TailFile(logfile, lines=20)]
1542 _Fail("OS rename script failed (%s), last lines in the"
1543 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1546 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1547 """Returns symlink path for block device.
1551 _dir = pathutils.DISK_LINKS_DIR
1553 return utils.PathJoin(_dir,
1555 (instance_name, constants.DISK_SEPARATOR, idx)))
1558 def _SymlinkBlockDev(instance_name, device_path, idx):
1559 """Set up symlinks to a instance's block device.
1561 This is an auxiliary function run when an instance is start (on the primary
1562 node) or when an instance is migrated (on the target node).
1565 @param instance_name: the name of the target instance
1566 @param device_path: path of the physical block device, on the node
1567 @param idx: the disk index
1568 @return: absolute path to the disk's symlink
1571 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1573 os.symlink(device_path, link_name)
1574 except OSError, err:
1575 if err.errno == errno.EEXIST:
1576 if (not os.path.islink(link_name) or
1577 os.readlink(link_name) != device_path):
1578 os.remove(link_name)
1579 os.symlink(device_path, link_name)
1586 def _RemoveBlockDevLinks(instance_name, disks):
1587 """Remove the block device symlinks belonging to the given instance.
1590 for idx, _ in enumerate(disks):
1591 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1592 if os.path.islink(link_name):
1594 os.remove(link_name)
1596 logging.exception("Can't remove symlink '%s'", link_name)
1599 def _GatherAndLinkBlockDevs(instance):
1600 """Set up an instance's block device(s).
1602 This is run on the primary node at instance startup. The block
1603 devices must be already assembled.
1605 @type instance: L{objects.Instance}
1606 @param instance: the instance whose disks we shoul assemble
1608 @return: list of (disk_object, device_path)
1612 for idx, disk in enumerate(instance.disks):
1613 device = _RecursiveFindBD(disk)
1615 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1619 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1621 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1624 block_devices.append((disk, link_name))
1626 return block_devices
1629 def StartInstance(instance, startup_paused, reason, store_reason=True):
1630 """Start an instance.
1632 @type instance: L{objects.Instance}
1633 @param instance: the instance object
1634 @type startup_paused: bool
1635 @param instance: pause instance at startup?
1636 @type reason: list of reasons
1637 @param reason: the reason trail for this startup
1638 @type store_reason: boolean
1639 @param store_reason: whether to store the shutdown reason trail on file
1643 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1646 if instance.name in running_instances:
1647 logging.info("Instance %s already running, not starting", instance.name)
1651 block_devices = _GatherAndLinkBlockDevs(instance)
1652 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1653 hyper.StartInstance(instance, block_devices, startup_paused)
1655 _StoreInstReasonTrail(instance.name, reason)
1656 except errors.BlockDeviceError, err:
1657 _Fail("Block device error: %s", err, exc=True)
1658 except errors.HypervisorError, err:
1659 _RemoveBlockDevLinks(instance.name, instance.disks)
1660 _Fail("Hypervisor error: %s", err, exc=True)
1663 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1664 """Shut an instance down.
1666 @note: this functions uses polling with a hardcoded timeout.
1668 @type instance: L{objects.Instance}
1669 @param instance: the instance object
1670 @type timeout: integer
1671 @param timeout: maximum timeout for soft shutdown
1672 @type reason: list of reasons
1673 @param reason: the reason trail for this shutdown
1674 @type store_reason: boolean
1675 @param store_reason: whether to store the shutdown reason trail on file
1679 hv_name = instance.hypervisor
1680 hyper = hypervisor.GetHypervisor(hv_name)
1681 iname = instance.name
1683 if instance.name not in hyper.ListInstances(instance.hvparams):
1684 logging.info("Instance %s not running, doing nothing", iname)
1689 self.tried_once = False
1692 if iname not in hyper.ListInstances(instance.hvparams):
1696 hyper.StopInstance(instance, retry=self.tried_once)
1698 _StoreInstReasonTrail(instance.name, reason)
1699 except errors.HypervisorError, err:
1700 if iname not in hyper.ListInstances(instance.hvparams):
1701 # if the instance is no longer existing, consider this a
1702 # success and go to cleanup
1705 _Fail("Failed to stop instance %s: %s", iname, err)
1707 self.tried_once = True
1709 raise utils.RetryAgain()
1712 utils.Retry(_TryShutdown(), 5, timeout)
1713 except utils.RetryTimeout:
1714 # the shutdown did not succeed
1715 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1718 hyper.StopInstance(instance, force=True)
1719 except errors.HypervisorError, err:
1720 if iname in hyper.ListInstances(instance.hvparams):
1721 # only raise an error if the instance still exists, otherwise
1722 # the error could simply be "instance ... unknown"!
1723 _Fail("Failed to force stop instance %s: %s", iname, err)
1727 if iname in hyper.ListInstances(instance.hvparams):
1728 _Fail("Could not shutdown instance %s even by destroy", iname)
1731 hyper.CleanupInstance(instance.name)
1732 except errors.HypervisorError, err:
1733 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1735 _RemoveBlockDevLinks(iname, instance.disks)
1738 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1739 """Reboot an instance.
1741 @type instance: L{objects.Instance}
1742 @param instance: the instance object to reboot
1743 @type reboot_type: str
1744 @param reboot_type: the type of reboot, one the following
1746 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1747 instance OS, do not recreate the VM
1748 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1749 restart the VM (at the hypervisor level)
1750 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1751 not accepted here, since that mode is handled differently, in
1752 cmdlib, and translates into full stop and start of the
1753 instance (instead of a call_instance_reboot RPC)
1754 @type shutdown_timeout: integer
1755 @param shutdown_timeout: maximum timeout for soft shutdown
1756 @type reason: list of reasons
1757 @param reason: the reason trail for this reboot
1761 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1764 if instance.name not in running_instances:
1765 _Fail("Cannot reboot instance %s that is not running", instance.name)
1767 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1768 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1770 hyper.RebootInstance(instance)
1771 except errors.HypervisorError, err:
1772 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1773 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1775 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1776 result = StartInstance(instance, False, reason, store_reason=False)
1777 _StoreInstReasonTrail(instance.name, reason)
1779 except errors.HypervisorError, err:
1780 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1782 _Fail("Invalid reboot_type received: %s", reboot_type)
1785 def InstanceBalloonMemory(instance, memory):
1786 """Resize an instance's memory.
1788 @type instance: L{objects.Instance}
1789 @param instance: the instance object
1791 @param memory: new memory amount in MB
1795 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1796 running = hyper.ListInstances(instance.hvparams)
1797 if instance.name not in running:
1798 logging.info("Instance %s is not running, cannot balloon", instance.name)
1801 hyper.BalloonInstanceMemory(instance, memory)
1802 except errors.HypervisorError, err:
1803 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1806 def MigrationInfo(instance):
1807 """Gather information about an instance to be migrated.
1809 @type instance: L{objects.Instance}
1810 @param instance: the instance definition
1813 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1815 info = hyper.MigrationInfo(instance)
1816 except errors.HypervisorError, err:
1817 _Fail("Failed to fetch migration information: %s", err, exc=True)
1821 def AcceptInstance(instance, info, target):
1822 """Prepare the node to accept an instance.
1824 @type instance: L{objects.Instance}
1825 @param instance: the instance definition
1826 @type info: string/data (opaque)
1827 @param info: migration information, from the source node
1828 @type target: string
1829 @param target: target host (usually ip), on this node
1832 # TODO: why is this required only for DTS_EXT_MIRROR?
1833 if instance.disk_template in constants.DTS_EXT_MIRROR:
1834 # Create the symlinks, as the disks are not active
1837 _GatherAndLinkBlockDevs(instance)
1838 except errors.BlockDeviceError, err:
1839 _Fail("Block device error: %s", err, exc=True)
1841 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1843 hyper.AcceptInstance(instance, info, target)
1844 except errors.HypervisorError, err:
1845 if instance.disk_template in constants.DTS_EXT_MIRROR:
1846 _RemoveBlockDevLinks(instance.name, instance.disks)
1847 _Fail("Failed to accept instance: %s", err, exc=True)
1850 def FinalizeMigrationDst(instance, info, success):
1851 """Finalize any preparation to accept an instance.
1853 @type instance: L{objects.Instance}
1854 @param instance: the instance definition
1855 @type info: string/data (opaque)
1856 @param info: migration information, from the source node
1857 @type success: boolean
1858 @param success: whether the migration was a success or a failure
1861 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1863 hyper.FinalizeMigrationDst(instance, info, success)
1864 except errors.HypervisorError, err:
1865 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1868 def MigrateInstance(cluster_name, instance, target, live):
1869 """Migrates an instance to another node.
1871 @type cluster_name: string
1872 @param cluster_name: name of the cluster
1873 @type instance: L{objects.Instance}
1874 @param instance: the instance definition
1875 @type target: string
1876 @param target: the target node name
1878 @param live: whether the migration should be done live or not (the
1879 interpretation of this parameter is left to the hypervisor)
1880 @raise RPCFail: if migration fails for some reason
1883 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1886 hyper.MigrateInstance(cluster_name, instance, target, live)
1887 except errors.HypervisorError, err:
1888 _Fail("Failed to migrate instance: %s", err, exc=True)
1891 def FinalizeMigrationSource(instance, success, live):
1892 """Finalize the instance migration on the source node.
1894 @type instance: L{objects.Instance}
1895 @param instance: the instance definition of the migrated instance
1897 @param success: whether the migration succeeded or not
1899 @param live: whether the user requested a live migration or not
1900 @raise RPCFail: If the execution fails for some reason
1903 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1906 hyper.FinalizeMigrationSource(instance, success, live)
1907 except Exception, err: # pylint: disable=W0703
1908 _Fail("Failed to finalize the migration on the source node: %s", err,
1912 def GetMigrationStatus(instance):
1913 """Get the migration status
1915 @type instance: L{objects.Instance}
1916 @param instance: the instance that is being migrated
1917 @rtype: L{objects.MigrationStatus}
1918 @return: the status of the current migration (one of
1919 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1920 progress info that can be retrieved from the hypervisor
1921 @raise RPCFail: If the migration status cannot be retrieved
1924 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1926 return hyper.GetMigrationStatus(instance)
1927 except Exception, err: # pylint: disable=W0703
1928 _Fail("Failed to get migration status: %s", err, exc=True)
1931 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1932 """Creates a block device for an instance.
1934 @type disk: L{objects.Disk}
1935 @param disk: the object describing the disk we should create
1937 @param size: the size of the physical underlying device, in MiB
1939 @param owner: the name of the instance for which disk is created,
1940 used for device cache data
1941 @type on_primary: boolean
1942 @param on_primary: indicates if it is the primary node or not
1944 @param info: string that will be sent to the physical device
1945 creation, used for example to set (LVM) tags on LVs
1946 @type excl_stor: boolean
1947 @param excl_stor: Whether exclusive_storage is active
1949 @return: the new unique_id of the device (this can sometime be
1950 computed only after creation), or None. On secondary nodes,
1951 it's not required to return anything.
1954 # TODO: remove the obsolete "size" argument
1955 # pylint: disable=W0613
1958 for child in disk.children:
1960 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1961 except errors.BlockDeviceError, err:
1962 _Fail("Can't assemble device %s: %s", child, err)
1963 if on_primary or disk.AssembleOnSecondary():
1964 # we need the children open in case the device itself has to
1967 # pylint: disable=E1103
1969 except errors.BlockDeviceError, err:
1970 _Fail("Can't make child '%s' read-write: %s", child, err)
1974 device = bdev.Create(disk, clist, excl_stor)
1975 except errors.BlockDeviceError, err:
1976 _Fail("Can't create block device: %s", err)
1978 if on_primary or disk.AssembleOnSecondary():
1981 except errors.BlockDeviceError, err:
1982 _Fail("Can't assemble device after creation, unusual event: %s", err)
1983 if on_primary or disk.OpenOnSecondary():
1985 device.Open(force=True)
1986 except errors.BlockDeviceError, err:
1987 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1988 DevCacheManager.UpdateCache(device.dev_path, owner,
1989 on_primary, disk.iv_name)
1991 device.SetInfo(info)
1993 return device.unique_id
1996 def _WipeDevice(path, offset, size):
1997 """This function actually wipes the device.
1999 @param path: The path to the device to wipe
2000 @param offset: The offset in MiB in the file
2001 @param size: The size in MiB to write
2004 # Internal sizes are always in Mebibytes; if the following "dd" command
2005 # should use a different block size the offset and size given to this
2006 # function must be adjusted accordingly before being passed to "dd".
2007 block_size = 1024 * 1024
2009 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2010 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2012 result = utils.RunCmd(cmd)
2015 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2016 result.fail_reason, result.output)
2019 def BlockdevWipe(disk, offset, size):
2020 """Wipes a block device.
2022 @type disk: L{objects.Disk}
2023 @param disk: the disk object we want to wipe
2025 @param offset: The offset in MiB in the file
2027 @param size: The size in MiB to write
2031 rdev = _RecursiveFindBD(disk)
2032 except errors.BlockDeviceError:
2036 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2038 # Do cross verify some of the parameters
2040 _Fail("Negative offset")
2042 _Fail("Negative size")
2043 if offset > rdev.size:
2044 _Fail("Offset is bigger than device size")
2045 if (offset + size) > rdev.size:
2046 _Fail("The provided offset and size to wipe is bigger than device size")
2048 _WipeDevice(rdev.dev_path, offset, size)
2051 def BlockdevPauseResumeSync(disks, pause):
2052 """Pause or resume the sync of the block device.
2054 @type disks: list of L{objects.Disk}
2055 @param disks: the disks object we want to pause/resume
2057 @param pause: Wheater to pause or resume
2063 rdev = _RecursiveFindBD(disk)
2064 except errors.BlockDeviceError:
2068 success.append((False, ("Cannot change sync for device %s:"
2069 " device not found" % disk.iv_name)))
2072 result = rdev.PauseResumeSync(pause)
2075 success.append((result, None))
2081 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2086 def BlockdevRemove(disk):
2087 """Remove a block device.
2089 @note: This is intended to be called recursively.
2091 @type disk: L{objects.Disk}
2092 @param disk: the disk object we should remove
2094 @return: the success of the operation
2099 rdev = _RecursiveFindBD(disk)
2100 except errors.BlockDeviceError, err:
2101 # probably can't attach
2102 logging.info("Can't attach to device %s in remove", disk)
2104 if rdev is not None:
2105 r_path = rdev.dev_path
2108 except errors.BlockDeviceError, err:
2109 msgs.append(str(err))
2111 DevCacheManager.RemoveCache(r_path)
2114 for child in disk.children:
2116 BlockdevRemove(child)
2117 except RPCFail, err:
2118 msgs.append(str(err))
2121 _Fail("; ".join(msgs))
2124 def _RecursiveAssembleBD(disk, owner, as_primary):
2125 """Activate a block device for an instance.
2127 This is run on the primary and secondary nodes for an instance.
2129 @note: this function is called recursively.
2131 @type disk: L{objects.Disk}
2132 @param disk: the disk we try to assemble
2134 @param owner: the name of the instance which owns the disk
2135 @type as_primary: boolean
2136 @param as_primary: if we should make the block device
2139 @return: the assembled device or None (in case no device
2141 @raise errors.BlockDeviceError: in case there is an error
2142 during the activation of the children or the device
2148 mcn = disk.ChildrenNeeded()
2150 mcn = 0 # max number of Nones allowed
2152 mcn = len(disk.children) - mcn # max number of Nones
2153 for chld_disk in disk.children:
2155 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2156 except errors.BlockDeviceError, err:
2157 if children.count(None) >= mcn:
2160 logging.error("Error in child activation (but continuing): %s",
2162 children.append(cdev)
2164 if as_primary or disk.AssembleOnSecondary():
2165 r_dev = bdev.Assemble(disk, children)
2167 if as_primary or disk.OpenOnSecondary():
2169 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2170 as_primary, disk.iv_name)
2177 def BlockdevAssemble(disk, owner, as_primary, idx):
2178 """Activate a block device for an instance.
2180 This is a wrapper over _RecursiveAssembleBD.
2182 @rtype: str or boolean
2183 @return: a C{/dev/...} path for primary nodes, and
2184 C{True} for secondary nodes
2188 result = _RecursiveAssembleBD(disk, owner, as_primary)
2189 if isinstance(result, BlockDev):
2190 # pylint: disable=E1103
2191 result = result.dev_path
2193 _SymlinkBlockDev(owner, result, idx)
2194 except errors.BlockDeviceError, err:
2195 _Fail("Error while assembling disk: %s", err, exc=True)
2196 except OSError, err:
2197 _Fail("Error while symlinking disk: %s", err, exc=True)
2202 def BlockdevShutdown(disk):
2203 """Shut down a block device.
2205 First, if the device is assembled (Attach() is successful), then
2206 the device is shutdown. Then the children of the device are
2209 This function is called recursively. Note that we don't cache the
2210 children or such, as oppossed to assemble, shutdown of different
2211 devices doesn't require that the upper device was active.
2213 @type disk: L{objects.Disk}
2214 @param disk: the description of the disk we should
2220 r_dev = _RecursiveFindBD(disk)
2221 if r_dev is not None:
2222 r_path = r_dev.dev_path
2225 DevCacheManager.RemoveCache(r_path)
2226 except errors.BlockDeviceError, err:
2227 msgs.append(str(err))
2230 for child in disk.children:
2232 BlockdevShutdown(child)
2233 except RPCFail, err:
2234 msgs.append(str(err))
2237 _Fail("; ".join(msgs))
2240 def BlockdevAddchildren(parent_cdev, new_cdevs):
2241 """Extend a mirrored block device.
2243 @type parent_cdev: L{objects.Disk}
2244 @param parent_cdev: the disk to which we should add children
2245 @type new_cdevs: list of L{objects.Disk}
2246 @param new_cdevs: the list of children which we should add
2250 parent_bdev = _RecursiveFindBD(parent_cdev)
2251 if parent_bdev is None:
2252 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2253 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2254 if new_bdevs.count(None) > 0:
2255 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2256 parent_bdev.AddChildren(new_bdevs)
2259 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2260 """Shrink a mirrored block device.
2262 @type parent_cdev: L{objects.Disk}
2263 @param parent_cdev: the disk from which we should remove children
2264 @type new_cdevs: list of L{objects.Disk}
2265 @param new_cdevs: the list of children which we should remove
2269 parent_bdev = _RecursiveFindBD(parent_cdev)
2270 if parent_bdev is None:
2271 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2273 for disk in new_cdevs:
2274 rpath = disk.StaticDevPath()
2276 bd = _RecursiveFindBD(disk)
2278 _Fail("Can't find device %s while removing children", disk)
2280 devs.append(bd.dev_path)
2282 if not utils.IsNormAbsPath(rpath):
2283 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2285 parent_bdev.RemoveChildren(devs)
2288 def BlockdevGetmirrorstatus(disks):
2289 """Get the mirroring status of a list of devices.
2291 @type disks: list of L{objects.Disk}
2292 @param disks: the list of disks which we should query
2294 @return: List of L{objects.BlockDevStatus}, one for each disk
2295 @raise errors.BlockDeviceError: if any of the disks cannot be
2301 rbd = _RecursiveFindBD(dsk)
2303 _Fail("Can't find device %s", dsk)
2305 stats.append(rbd.CombinedSyncStatus())
2310 def BlockdevGetmirrorstatusMulti(disks):
2311 """Get the mirroring status of a list of devices.
2313 @type disks: list of L{objects.Disk}
2314 @param disks: the list of disks which we should query
2316 @return: List of tuples, (bool, status), one for each disk; bool denotes
2317 success/failure, status is L{objects.BlockDevStatus} on success, string
2324 rbd = _RecursiveFindBD(disk)
2326 result.append((False, "Can't find device %s" % disk))
2329 status = rbd.CombinedSyncStatus()
2330 except errors.BlockDeviceError, err:
2331 logging.exception("Error while getting disk status")
2332 result.append((False, str(err)))
2334 result.append((True, status))
2336 assert len(disks) == len(result)
2341 def _RecursiveFindBD(disk):
2342 """Check if a device is activated.
2344 If so, return information about the real device.
2346 @type disk: L{objects.Disk}
2347 @param disk: the disk object we need to find
2349 @return: None if the device can't be found,
2350 otherwise the device instance
2355 for chdisk in disk.children:
2356 children.append(_RecursiveFindBD(chdisk))
2358 return bdev.FindDevice(disk, children)
2361 def _OpenRealBD(disk):
2362 """Opens the underlying block device of a disk.
2364 @type disk: L{objects.Disk}
2365 @param disk: the disk object we want to open
2368 real_disk = _RecursiveFindBD(disk)
2369 if real_disk is None:
2370 _Fail("Block device '%s' is not set up", disk)
2377 def BlockdevFind(disk):
2378 """Check if a device is activated.
2380 If it is, return information about the real device.
2382 @type disk: L{objects.Disk}
2383 @param disk: the disk to find
2384 @rtype: None or objects.BlockDevStatus
2385 @return: None if the disk cannot be found, otherwise a the current
2390 rbd = _RecursiveFindBD(disk)
2391 except errors.BlockDeviceError, err:
2392 _Fail("Failed to find device: %s", err, exc=True)
2397 return rbd.GetSyncStatus()
2400 def BlockdevGetdimensions(disks):
2401 """Computes the size of the given disks.
2403 If a disk is not found, returns None instead.
2405 @type disks: list of L{objects.Disk}
2406 @param disks: the list of disk to compute the size for
2408 @return: list with elements None if the disk cannot be found,
2409 otherwise the pair (size, spindles), where spindles is None if the
2410 device doesn't support that
2416 rbd = _RecursiveFindBD(cf)
2417 except errors.BlockDeviceError:
2423 result.append(rbd.GetActualDimensions())
2427 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2428 """Export a block device to a remote node.
2430 @type disk: L{objects.Disk}
2431 @param disk: the description of the disk to export
2432 @type dest_node: str
2433 @param dest_node: the destination node to export to
2434 @type dest_path: str
2435 @param dest_path: the destination path on the target node
2436 @type cluster_name: str
2437 @param cluster_name: the cluster name, needed for SSH hostalias
2441 real_disk = _OpenRealBD(disk)
2443 # the block size on the read dd is 1MiB to match our units
2444 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2445 "dd if=%s bs=1048576 count=%s",
2446 real_disk.dev_path, str(disk.size))
2448 # we set here a smaller block size as, due to ssh buffering, more
2449 # than 64-128k will mostly ignored; we use nocreat to fail if the
2450 # device is not already there or we pass a wrong path; we use
2451 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2452 # to not buffer too much memory; this means that at best, we flush
2453 # every 64k, which will not be very fast
2454 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2455 " oflag=dsync", dest_path)
2457 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2458 constants.SSH_LOGIN_USER,
2461 # all commands have been checked, so we're safe to combine them
2462 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2464 result = utils.RunCmd(["bash", "-c", command])
2467 _Fail("Disk copy command '%s' returned error: %s"
2468 " output: %s", command, result.fail_reason, result.output)
2471 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2472 """Write a file to the filesystem.
2474 This allows the master to overwrite(!) a file. It will only perform
2475 the operation if the file belongs to a list of configuration files.
2477 @type file_name: str
2478 @param file_name: the target file name
2480 @param data: the new contents of the file
2482 @param mode: the mode to give the file (can be None)
2484 @param uid: the owner of the file
2486 @param gid: the group of the file
2488 @param atime: the atime to set on the file (can be None)
2490 @param mtime: the mtime to set on the file (can be None)
2494 file_name = vcluster.LocalizeVirtualPath(file_name)
2496 if not os.path.isabs(file_name):
2497 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2499 if file_name not in _ALLOWED_UPLOAD_FILES:
2500 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2503 raw_data = _Decompress(data)
2505 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2506 _Fail("Invalid username/groupname type")
2508 getents = runtime.GetEnts()
2509 uid = getents.LookupUser(uid)
2510 gid = getents.LookupGroup(gid)
2512 utils.SafeWriteFile(file_name, None,
2513 data=raw_data, mode=mode, uid=uid, gid=gid,
2514 atime=atime, mtime=mtime)
2517 def RunOob(oob_program, command, node, timeout):
2518 """Executes oob_program with given command on given node.
2520 @param oob_program: The path to the executable oob_program
2521 @param command: The command to invoke on oob_program
2522 @param node: The node given as an argument to the program
2523 @param timeout: Timeout after which we kill the oob program
2526 @raise RPCFail: If execution fails for some reason
2529 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2532 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2533 result.fail_reason, result.output)
2535 return result.stdout
2538 def _OSOndiskAPIVersion(os_dir):
2539 """Compute and return the API version of a given OS.
2541 This function will try to read the API version of the OS residing in
2542 the 'os_dir' directory.
2545 @param os_dir: the directory in which we should look for the OS
2547 @return: tuple (status, data) with status denoting the validity and
2548 data holding either the vaid versions or an error message
2551 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2554 st = os.stat(api_file)
2555 except EnvironmentError, err:
2556 return False, ("Required file '%s' not found under path %s: %s" %
2557 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2559 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2560 return False, ("File '%s' in %s is not a regular file" %
2561 (constants.OS_API_FILE, os_dir))
2564 api_versions = utils.ReadFile(api_file).splitlines()
2565 except EnvironmentError, err:
2566 return False, ("Error while reading the API version file at %s: %s" %
2567 (api_file, utils.ErrnoOrStr(err)))
2570 api_versions = [int(version.strip()) for version in api_versions]
2571 except (TypeError, ValueError), err:
2572 return False, ("API version(s) can't be converted to integer: %s" %
2575 return True, api_versions
2578 def DiagnoseOS(top_dirs=None):
2579 """Compute the validity for all OSes.
2581 @type top_dirs: list
2582 @param top_dirs: the list of directories in which to
2583 search (if not given defaults to
2584 L{pathutils.OS_SEARCH_PATH})
2585 @rtype: list of L{objects.OS}
2586 @return: a list of tuples (name, path, status, diagnose, variants,
2587 parameters, api_version) for all (potential) OSes under all
2588 search paths, where:
2589 - name is the (potential) OS name
2590 - path is the full path to the OS
2591 - status True/False is the validity of the OS
2592 - diagnose is the error message for an invalid OS, otherwise empty
2593 - variants is a list of supported OS variants, if any
2594 - parameters is a list of (name, help) parameters, if any
2595 - api_version is a list of support OS API versions
2598 if top_dirs is None:
2599 top_dirs = pathutils.OS_SEARCH_PATH
2602 for dir_name in top_dirs:
2603 if os.path.isdir(dir_name):
2605 f_names = utils.ListVisibleFiles(dir_name)
2606 except EnvironmentError, err:
2607 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2609 for name in f_names:
2610 os_path = utils.PathJoin(dir_name, name)
2611 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2614 variants = os_inst.supported_variants
2615 parameters = os_inst.supported_parameters
2616 api_versions = os_inst.api_versions
2619 variants = parameters = api_versions = []
2620 result.append((name, os_path, status, diagnose, variants,
2621 parameters, api_versions))
2626 def _TryOSFromDisk(name, base_dir=None):
2627 """Create an OS instance from disk.
2629 This function will return an OS instance if the given name is a
2632 @type base_dir: string
2633 @keyword base_dir: Base directory containing OS installations.
2634 Defaults to a search in all the OS_SEARCH_PATH dirs.
2636 @return: success and either the OS instance if we find a valid one,
2640 if base_dir is None:
2641 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2643 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2646 return False, "Directory for OS %s not found in search path" % name
2648 status, api_versions = _OSOndiskAPIVersion(os_dir)
2651 return status, api_versions
2653 if not constants.OS_API_VERSIONS.intersection(api_versions):
2654 return False, ("API version mismatch for path '%s': found %s, want %s." %
2655 (os_dir, api_versions, constants.OS_API_VERSIONS))
2657 # OS Files dictionary, we will populate it with the absolute path
2658 # names; if the value is True, then it is a required file, otherwise
2660 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2662 if max(api_versions) >= constants.OS_API_V15:
2663 os_files[constants.OS_VARIANTS_FILE] = False
2665 if max(api_versions) >= constants.OS_API_V20:
2666 os_files[constants.OS_PARAMETERS_FILE] = True
2668 del os_files[constants.OS_SCRIPT_VERIFY]
2670 for (filename, required) in os_files.items():
2671 os_files[filename] = utils.PathJoin(os_dir, filename)
2674 st = os.stat(os_files[filename])
2675 except EnvironmentError, err:
2676 if err.errno == errno.ENOENT and not required:
2677 del os_files[filename]
2679 return False, ("File '%s' under path '%s' is missing (%s)" %
2680 (filename, os_dir, utils.ErrnoOrStr(err)))
2682 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2683 return False, ("File '%s' under path '%s' is not a regular file" %
2686 if filename in constants.OS_SCRIPTS:
2687 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2688 return False, ("File '%s' under path '%s' is not executable" %
2692 if constants.OS_VARIANTS_FILE in os_files:
2693 variants_file = os_files[constants.OS_VARIANTS_FILE]
2696 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2697 except EnvironmentError, err:
2698 # we accept missing files, but not other errors
2699 if err.errno != errno.ENOENT:
2700 return False, ("Error while reading the OS variants file at %s: %s" %
2701 (variants_file, utils.ErrnoOrStr(err)))
2704 if constants.OS_PARAMETERS_FILE in os_files:
2705 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2707 parameters = utils.ReadFile(parameters_file).splitlines()
2708 except EnvironmentError, err:
2709 return False, ("Error while reading the OS parameters file at %s: %s" %
2710 (parameters_file, utils.ErrnoOrStr(err)))
2711 parameters = [v.split(None, 1) for v in parameters]
2713 os_obj = objects.OS(name=name, path=os_dir,
2714 create_script=os_files[constants.OS_SCRIPT_CREATE],
2715 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2716 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2717 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2718 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2720 supported_variants=variants,
2721 supported_parameters=parameters,
2722 api_versions=api_versions)
2726 def OSFromDisk(name, base_dir=None):
2727 """Create an OS instance from disk.
2729 This function will return an OS instance if the given name is a
2730 valid OS name. Otherwise, it will raise an appropriate
2731 L{RPCFail} exception, detailing why this is not a valid OS.
2733 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2734 an exception but returns true/false status data.
2736 @type base_dir: string
2737 @keyword base_dir: Base directory containing OS installations.
2738 Defaults to a search in all the OS_SEARCH_PATH dirs.
2739 @rtype: L{objects.OS}
2740 @return: the OS instance if we find a valid one
2741 @raise RPCFail: if we don't find a valid OS
2744 name_only = objects.OS.GetName(name)
2745 status, payload = _TryOSFromDisk(name_only, base_dir)
2753 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2754 """Calculate the basic environment for an os script.
2757 @param os_name: full operating system name (including variant)
2758 @type inst_os: L{objects.OS}
2759 @param inst_os: operating system for which the environment is being built
2760 @type os_params: dict
2761 @param os_params: the OS parameters
2762 @type debug: integer
2763 @param debug: debug level (0 or 1, for OS Api 10)
2765 @return: dict of environment variables
2766 @raise errors.BlockDeviceError: if the block device
2772 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2773 result["OS_API_VERSION"] = "%d" % api_version
2774 result["OS_NAME"] = inst_os.name
2775 result["DEBUG_LEVEL"] = "%d" % debug
2778 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2779 variant = objects.OS.GetVariant(os_name)
2781 variant = inst_os.supported_variants[0]
2784 result["OS_VARIANT"] = variant
2787 for pname, pvalue in os_params.items():
2788 result["OSP_%s" % pname.upper()] = pvalue
2790 # Set a default path otherwise programs called by OS scripts (or
2791 # even hooks called from OS scripts) might break, and we don't want
2792 # to have each script require setting a PATH variable
2793 result["PATH"] = constants.HOOKS_PATH
2798 def OSEnvironment(instance, inst_os, debug=0):
2799 """Calculate the environment for an os script.
2801 @type instance: L{objects.Instance}
2802 @param instance: target instance for the os script run
2803 @type inst_os: L{objects.OS}
2804 @param inst_os: operating system for which the environment is being built
2805 @type debug: integer
2806 @param debug: debug level (0 or 1, for OS Api 10)
2808 @return: dict of environment variables
2809 @raise errors.BlockDeviceError: if the block device
2813 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2815 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2816 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2818 result["HYPERVISOR"] = instance.hypervisor
2819 result["DISK_COUNT"] = "%d" % len(instance.disks)
2820 result["NIC_COUNT"] = "%d" % len(instance.nics)
2821 result["INSTANCE_SECONDARY_NODES"] = \
2822 ("%s" % " ".join(instance.secondary_nodes))
2825 for idx, disk in enumerate(instance.disks):
2826 real_disk = _OpenRealBD(disk)
2827 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2828 result["DISK_%d_ACCESS" % idx] = disk.mode
2829 result["DISK_%d_UUID" % idx] = disk.uuid
2831 result["DISK_%d_NAME" % idx] = disk.name
2832 if constants.HV_DISK_TYPE in instance.hvparams:
2833 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2834 instance.hvparams[constants.HV_DISK_TYPE]
2835 if disk.dev_type in constants.LDS_BLOCK:
2836 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2837 elif disk.dev_type == constants.LD_FILE:
2838 result["DISK_%d_BACKEND_TYPE" % idx] = \
2839 "file:%s" % disk.physical_id[0]
2842 for idx, nic in enumerate(instance.nics):
2843 result["NIC_%d_MAC" % idx] = nic.mac
2844 result["NIC_%d_UUID" % idx] = nic.uuid
2846 result["NIC_%d_NAME" % idx] = nic.name
2848 result["NIC_%d_IP" % idx] = nic.ip
2849 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2850 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2851 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2852 if nic.nicparams[constants.NIC_LINK]:
2853 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2855 nobj = objects.Network.FromDict(nic.netinfo)
2856 result.update(nobj.HooksDict("NIC_%d_" % idx))
2857 if constants.HV_NIC_TYPE in instance.hvparams:
2858 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2859 instance.hvparams[constants.HV_NIC_TYPE]
2862 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2863 for key, value in source.items():
2864 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2869 def DiagnoseExtStorage(top_dirs=None):
2870 """Compute the validity for all ExtStorage Providers.
2872 @type top_dirs: list
2873 @param top_dirs: the list of directories in which to
2874 search (if not given defaults to
2875 L{pathutils.ES_SEARCH_PATH})
2876 @rtype: list of L{objects.ExtStorage}
2877 @return: a list of tuples (name, path, status, diagnose, parameters)
2878 for all (potential) ExtStorage Providers under all
2879 search paths, where:
2880 - name is the (potential) ExtStorage Provider
2881 - path is the full path to the ExtStorage Provider
2882 - status True/False is the validity of the ExtStorage Provider
2883 - diagnose is the error message for an invalid ExtStorage Provider,
2885 - parameters is a list of (name, help) parameters, if any
2888 if top_dirs is None:
2889 top_dirs = pathutils.ES_SEARCH_PATH
2892 for dir_name in top_dirs:
2893 if os.path.isdir(dir_name):
2895 f_names = utils.ListVisibleFiles(dir_name)
2896 except EnvironmentError, err:
2897 logging.exception("Can't list the ExtStorage directory %s: %s",
2900 for name in f_names:
2901 es_path = utils.PathJoin(dir_name, name)
2902 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2905 parameters = es_inst.supported_parameters
2909 result.append((name, es_path, status, diagnose, parameters))
2914 def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
2915 """Grow a stack of block devices.
2917 This function is called recursively, with the childrens being the
2918 first ones to resize.
2920 @type disk: L{objects.Disk}
2921 @param disk: the disk to be grown
2922 @type amount: integer
2923 @param amount: the amount (in mebibytes) to grow with
2924 @type dryrun: boolean
2925 @param dryrun: whether to execute the operation in simulation mode
2926 only, without actually increasing the size
2927 @param backingstore: whether to execute the operation on backing storage
2928 only, or on "logical" storage only; e.g. DRBD is logical storage,
2929 whereas LVM, file, RBD are backing storage
2930 @rtype: (status, result)
2931 @type excl_stor: boolean
2932 @param excl_stor: Whether exclusive_storage is active
2933 @return: a tuple with the status of the operation (True/False), and
2934 the errors message if status is False
2937 r_dev = _RecursiveFindBD(disk)
2939 _Fail("Cannot find block device %s", disk)
2942 r_dev.Grow(amount, dryrun, backingstore, excl_stor)
2943 except errors.BlockDeviceError, err:
2944 _Fail("Failed to grow block device: %s", err, exc=True)
2947 def BlockdevSnapshot(disk):
2948 """Create a snapshot copy of a block device.
2950 This function is called recursively, and the snapshot is actually created
2951 just for the leaf lvm backend device.
2953 @type disk: L{objects.Disk}
2954 @param disk: the disk to be snapshotted
2956 @return: snapshot disk ID as (vg, lv)
2959 if disk.dev_type == constants.LD_DRBD8:
2960 if not disk.children:
2961 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2963 return BlockdevSnapshot(disk.children[0])
2964 elif disk.dev_type == constants.LD_LV:
2965 r_dev = _RecursiveFindBD(disk)
2966 if r_dev is not None:
2967 # FIXME: choose a saner value for the snapshot size
2968 # let's stay on the safe side and ask for the full size, for now
2969 return r_dev.Snapshot(disk.size)
2971 _Fail("Cannot find block device %s", disk)
2973 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2974 disk.unique_id, disk.dev_type)
2977 def BlockdevSetInfo(disk, info):
2978 """Sets 'metadata' information on block devices.
2980 This function sets 'info' metadata on block devices. Initial
2981 information is set at device creation; this function should be used
2982 for example after renames.
2984 @type disk: L{objects.Disk}
2985 @param disk: the disk to be grown
2987 @param info: new 'info' metadata
2988 @rtype: (status, result)
2989 @return: a tuple with the status of the operation (True/False), and
2990 the errors message if status is False
2993 r_dev = _RecursiveFindBD(disk)
2995 _Fail("Cannot find block device %s", disk)
2999 except errors.BlockDeviceError, err:
3000 _Fail("Failed to set information on block device: %s", err, exc=True)
3003 def FinalizeExport(instance, snap_disks):
3004 """Write out the export configuration information.
3006 @type instance: L{objects.Instance}
3007 @param instance: the instance which we export, used for
3008 saving configuration
3009 @type snap_disks: list of L{objects.Disk}
3010 @param snap_disks: list of snapshot block devices, which
3011 will be used to get the actual name of the dump file
3016 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3017 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3019 config = objects.SerializableConfigParser()
3021 config.add_section(constants.INISECT_EXP)
3022 config.set(constants.INISECT_EXP, "version", "0")
3023 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3024 config.set(constants.INISECT_EXP, "source", instance.primary_node)
3025 config.set(constants.INISECT_EXP, "os", instance.os)
3026 config.set(constants.INISECT_EXP, "compression", "none")
3028 config.add_section(constants.INISECT_INS)
3029 config.set(constants.INISECT_INS, "name", instance.name)
3030 config.set(constants.INISECT_INS, "maxmem", "%d" %
3031 instance.beparams[constants.BE_MAXMEM])
3032 config.set(constants.INISECT_INS, "minmem", "%d" %
3033 instance.beparams[constants.BE_MINMEM])
3034 # "memory" is deprecated, but useful for exporting to old ganeti versions
3035 config.set(constants.INISECT_INS, "memory", "%d" %
3036 instance.beparams[constants.BE_MAXMEM])
3037 config.set(constants.INISECT_INS, "vcpus", "%d" %
3038 instance.beparams[constants.BE_VCPUS])
3039 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3040 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3041 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3044 for nic_count, nic in enumerate(instance.nics):
3046 config.set(constants.INISECT_INS, "nic%d_mac" %
3047 nic_count, "%s" % nic.mac)
3048 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3049 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3051 for param in constants.NICS_PARAMETER_TYPES:
3052 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3053 "%s" % nic.nicparams.get(param, None))
3054 # TODO: redundant: on load can read nics until it doesn't exist
3055 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3058 for disk_count, disk in enumerate(snap_disks):
3061 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3062 ("%s" % disk.iv_name))
3063 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3064 ("%s" % disk.physical_id[1]))
3065 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3068 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3070 # New-style hypervisor/backend parameters
3072 config.add_section(constants.INISECT_HYP)
3073 for name, value in instance.hvparams.items():
3074 if name not in constants.HVC_GLOBALS:
3075 config.set(constants.INISECT_HYP, name, str(value))
3077 config.add_section(constants.INISECT_BEP)
3078 for name, value in instance.beparams.items():
3079 config.set(constants.INISECT_BEP, name, str(value))
3081 config.add_section(constants.INISECT_OSP)
3082 for name, value in instance.osparams.items():
3083 config.set(constants.INISECT_OSP, name, str(value))
3085 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3086 data=config.Dumps())
3087 shutil.rmtree(finaldestdir, ignore_errors=True)
3088 shutil.move(destdir, finaldestdir)
3091 def ExportInfo(dest):
3092 """Get export configuration information.
3095 @param dest: directory containing the export
3097 @rtype: L{objects.SerializableConfigParser}
3098 @return: a serializable config file containing the
3102 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3104 config = objects.SerializableConfigParser()
3107 if (not config.has_section(constants.INISECT_EXP) or
3108 not config.has_section(constants.INISECT_INS)):
3109 _Fail("Export info file doesn't have the required fields")
3111 return config.Dumps()
3115 """Return a list of exports currently available on this machine.
3118 @return: list of the exports
3121 if os.path.isdir(pathutils.EXPORT_DIR):
3122 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3124 _Fail("No exports directory")
3127 def RemoveExport(export):
3128 """Remove an existing export from the node.
3131 @param export: the name of the export to remove
3135 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3138 shutil.rmtree(target)
3139 except EnvironmentError, err:
3140 _Fail("Error while removing the export: %s", err, exc=True)
3143 def BlockdevRename(devlist):
3144 """Rename a list of block devices.
3146 @type devlist: list of tuples
3147 @param devlist: list of tuples of the form (disk,
3148 new_logical_id, new_physical_id); disk is an
3149 L{objects.Disk} object describing the current disk,
3150 and new logical_id/physical_id is the name we
3153 @return: True if all renames succeeded, False otherwise
3158 for disk, unique_id in devlist:
3159 dev = _RecursiveFindBD(disk)
3161 msgs.append("Can't find device %s in rename" % str(disk))
3165 old_rpath = dev.dev_path
3166 dev.Rename(unique_id)
3167 new_rpath = dev.dev_path
3168 if old_rpath != new_rpath:
3169 DevCacheManager.RemoveCache(old_rpath)
3170 # FIXME: we should add the new cache information here, like:
3171 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3172 # but we don't have the owner here - maybe parse from existing
3173 # cache? for now, we only lose lvm data when we rename, which
3174 # is less critical than DRBD or MD
3175 except errors.BlockDeviceError, err:
3176 msgs.append("Can't rename device '%s' to '%s': %s" %
3177 (dev, unique_id, err))
3178 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3181 _Fail("; ".join(msgs))
3184 def _TransformFileStorageDir(fs_dir):
3185 """Checks whether given file_storage_dir is valid.
3187 Checks wheter the given fs_dir is within the cluster-wide default
3188 file_storage_dir or the shared_file_storage_dir, which are stored in
3189 SimpleStore. Only paths under those directories are allowed.
3192 @param fs_dir: the path to check
3194 @return: the normalized path if valid, None otherwise
3197 if not (constants.ENABLE_FILE_STORAGE or
3198 constants.ENABLE_SHARED_FILE_STORAGE):
3199 _Fail("File storage disabled at configure time")
3201 bdev.CheckFileStoragePath(fs_dir)
3203 return os.path.normpath(fs_dir)
3206 def CreateFileStorageDir(file_storage_dir):
3207 """Create file storage directory.
3209 @type file_storage_dir: str
3210 @param file_storage_dir: directory to create
3213 @return: tuple with first element a boolean indicating wheter dir
3214 creation was successful or not
3217 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3218 if os.path.exists(file_storage_dir):
3219 if not os.path.isdir(file_storage_dir):
3220 _Fail("Specified storage dir '%s' is not a directory",
3224 os.makedirs(file_storage_dir, 0750)
3225 except OSError, err:
3226 _Fail("Cannot create file storage directory '%s': %s",
3227 file_storage_dir, err, exc=True)
3230 def RemoveFileStorageDir(file_storage_dir):
3231 """Remove file storage directory.
3233 Remove it only if it's empty. If not log an error and return.
3235 @type file_storage_dir: str
3236 @param file_storage_dir: the directory we should cleanup
3237 @rtype: tuple (success,)
3238 @return: tuple of one element, C{success}, denoting
3239 whether the operation was successful
3242 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3243 if os.path.exists(file_storage_dir):
3244 if not os.path.isdir(file_storage_dir):
3245 _Fail("Specified Storage directory '%s' is not a directory",
3247 # deletes dir only if empty, otherwise we want to fail the rpc call
3249 os.rmdir(file_storage_dir)
3250 except OSError, err:
3251 _Fail("Cannot remove file storage directory '%s': %s",
3252 file_storage_dir, err)
3255 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3256 """Rename the file storage directory.
3258 @type old_file_storage_dir: str
3259 @param old_file_storage_dir: the current path
3260 @type new_file_storage_dir: str
3261 @param new_file_storage_dir: the name we should rename to
3262 @rtype: tuple (success,)
3263 @return: tuple of one element, C{success}, denoting
3264 whether the operation was successful
3267 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3268 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3269 if not os.path.exists(new_file_storage_dir):
3270 if os.path.isdir(old_file_storage_dir):
3272 os.rename(old_file_storage_dir, new_file_storage_dir)
3273 except OSError, err:
3274 _Fail("Cannot rename '%s' to '%s': %s",
3275 old_file_storage_dir, new_file_storage_dir, err)
3277 _Fail("Specified storage dir '%s' is not a directory",
3278 old_file_storage_dir)
3280 if os.path.exists(old_file_storage_dir):
3281 _Fail("Cannot rename '%s' to '%s': both locations exist",
3282 old_file_storage_dir, new_file_storage_dir)
3285 def _EnsureJobQueueFile(file_name):
3286 """Checks whether the given filename is in the queue directory.
3288 @type file_name: str
3289 @param file_name: the file name we should check
3291 @raises RPCFail: if the file is not valid
3294 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3295 _Fail("Passed job queue file '%s' does not belong to"
3296 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3299 def JobQueueUpdate(file_name, content):
3300 """Updates a file in the queue directory.
3302 This is just a wrapper over L{utils.io.WriteFile}, with proper
3305 @type file_name: str
3306 @param file_name: the job file name
3308 @param content: the new job contents
3310 @return: the success of the operation
3313 file_name = vcluster.LocalizeVirtualPath(file_name)
3315 _EnsureJobQueueFile(file_name)
3316 getents = runtime.GetEnts()
3318 # Write and replace the file atomically
3319 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3320 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3323 def JobQueueRename(old, new):
3324 """Renames a job queue file.
3326 This is just a wrapper over os.rename with proper checking.
3329 @param old: the old (actual) file name
3331 @param new: the desired file name
3333 @return: the success of the operation and payload
3336 old = vcluster.LocalizeVirtualPath(old)
3337 new = vcluster.LocalizeVirtualPath(new)
3339 _EnsureJobQueueFile(old)
3340 _EnsureJobQueueFile(new)
3342 getents = runtime.GetEnts()
3344 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3345 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3348 def BlockdevClose(instance_name, disks):
3349 """Closes the given block devices.
3351 This means they will be switched to secondary mode (in case of
3354 @param instance_name: if the argument is not empty, the symlinks
3355 of this instance will be removed
3356 @type disks: list of L{objects.Disk}
3357 @param disks: the list of disks to be closed
3358 @rtype: tuple (success, message)
3359 @return: a tuple of success and message, where success
3360 indicates the succes of the operation, and message
3361 which will contain the error details in case we
3367 rd = _RecursiveFindBD(cf)
3369 _Fail("Can't find device %s", cf)
3376 except errors.BlockDeviceError, err:
3377 msg.append(str(err))
3379 _Fail("Can't make devices secondary: %s", ",".join(msg))
3382 _RemoveBlockDevLinks(instance_name, disks)
3385 def ValidateHVParams(hvname, hvparams):
3386 """Validates the given hypervisor parameters.
3388 @type hvname: string
3389 @param hvname: the hypervisor name
3390 @type hvparams: dict
3391 @param hvparams: the hypervisor parameters to be validated
3396 hv_type = hypervisor.GetHypervisor(hvname)
3397 hv_type.ValidateParameters(hvparams)
3398 except errors.HypervisorError, err:
3399 _Fail(str(err), log=False)
3402 def _CheckOSPList(os_obj, parameters):
3403 """Check whether a list of parameters is supported by the OS.
3405 @type os_obj: L{objects.OS}
3406 @param os_obj: OS object to check
3407 @type parameters: list
3408 @param parameters: the list of parameters to check
3411 supported = [v[0] for v in os_obj.supported_parameters]
3412 delta = frozenset(parameters).difference(supported)
3414 _Fail("The following parameters are not supported"
3415 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3418 def ValidateOS(required, osname, checks, osparams):
3419 """Validate the given OS' parameters.
3421 @type required: boolean
3422 @param required: whether absence of the OS should translate into
3424 @type osname: string
3425 @param osname: the OS to be validated
3427 @param checks: list of the checks to run (currently only 'parameters')
3428 @type osparams: dict
3429 @param osparams: dictionary with OS parameters
3431 @return: True if the validation passed, or False if the OS was not
3432 found and L{required} was false
3435 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3436 _Fail("Unknown checks required for OS %s: %s", osname,
3437 set(checks).difference(constants.OS_VALIDATE_CALLS))
3439 name_only = objects.OS.GetName(osname)
3440 status, tbv = _TryOSFromDisk(name_only, None)
3448 if max(tbv.api_versions) < constants.OS_API_V20:
3451 if constants.OS_VALIDATE_PARAMETERS in checks:
3452 _CheckOSPList(tbv, osparams.keys())
3454 validate_env = OSCoreEnv(osname, tbv, osparams)
3455 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3456 cwd=tbv.path, reset_env=True)
3458 logging.error("os validate command '%s' returned error: %s output: %s",
3459 result.cmd, result.fail_reason, result.output)
3460 _Fail("OS validation script failed (%s), output: %s",
3461 result.fail_reason, result.output, log=False)
3467 """Demotes the current node from master candidate role.
3470 # try to ensure we're not the master by mistake
3471 master, myself = ssconf.GetMasterAndMyself()
3472 if master == myself:
3473 _Fail("ssconf status shows I'm the master node, will not demote")
3475 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3476 if not result.failed:
3477 _Fail("The master daemon is running, will not demote")
3480 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3481 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3482 except EnvironmentError, err:
3483 if err.errno != errno.ENOENT:
3484 _Fail("Error while backing up cluster file: %s", err, exc=True)
3486 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3489 def _GetX509Filenames(cryptodir, name):
3490 """Returns the full paths for the private key and certificate.
3493 return (utils.PathJoin(cryptodir, name),
3494 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3495 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3498 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3499 """Creates a new X509 certificate for SSL/TLS.
3502 @param validity: Validity in seconds
3503 @rtype: tuple; (string, string)
3504 @return: Certificate name and public part
3507 (key_pem, cert_pem) = \
3508 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3509 min(validity, _MAX_SSL_CERT_VALIDITY))
3511 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3512 prefix="x509-%s-" % utils.TimestampForFilename())
3514 name = os.path.basename(cert_dir)
3515 assert len(name) > 5
3517 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3519 utils.WriteFile(key_file, mode=0400, data=key_pem)
3520 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3522 # Never return private key as it shouldn't leave the node
3523 return (name, cert_pem)
3525 shutil.rmtree(cert_dir, ignore_errors=True)
3529 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3530 """Removes a X509 certificate.
3533 @param name: Certificate name
3536 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3538 utils.RemoveFile(key_file)
3539 utils.RemoveFile(cert_file)
3543 except EnvironmentError, err:
3544 _Fail("Cannot remove certificate directory '%s': %s",
3548 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3549 """Returns the command for the requested input/output.
3551 @type instance: L{objects.Instance}
3552 @param instance: The instance object
3553 @param mode: Import/export mode
3554 @param ieio: Input/output type
3555 @param ieargs: Input/output arguments
3558 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3565 if ieio == constants.IEIO_FILE:
3566 (filename, ) = ieargs
3568 if not utils.IsNormAbsPath(filename):
3569 _Fail("Path '%s' is not normalized or absolute", filename)
3571 real_filename = os.path.realpath(filename)
3572 directory = os.path.dirname(real_filename)
3574 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3575 _Fail("File '%s' is not under exports directory '%s': %s",
3576 filename, pathutils.EXPORT_DIR, real_filename)
3579 utils.Makedirs(directory, mode=0750)
3581 quoted_filename = utils.ShellQuote(filename)
3583 if mode == constants.IEM_IMPORT:
3584 suffix = "> %s" % quoted_filename
3585 elif mode == constants.IEM_EXPORT:
3586 suffix = "< %s" % quoted_filename
3588 # Retrieve file size
3590 st = os.stat(filename)
3591 except EnvironmentError, err:
3592 logging.error("Can't stat(2) %s: %s", filename, err)
3594 exp_size = utils.BytesToMebibyte(st.st_size)
3596 elif ieio == constants.IEIO_RAW_DISK:
3599 real_disk = _OpenRealBD(disk)
3601 if mode == constants.IEM_IMPORT:
3602 # we set here a smaller block size as, due to transport buffering, more
3603 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3604 # is not already there or we pass a wrong path; we use notrunc to no
3605 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3606 # much memory; this means that at best, we flush every 64k, which will
3608 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3609 " bs=%s oflag=dsync"),
3613 elif mode == constants.IEM_EXPORT:
3614 # the block size on the read dd is 1MiB to match our units
3615 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3617 str(1024 * 1024), # 1 MB
3619 exp_size = disk.size
3621 elif ieio == constants.IEIO_SCRIPT:
3622 (disk, disk_index, ) = ieargs
3624 assert isinstance(disk_index, (int, long))
3626 real_disk = _OpenRealBD(disk)
3628 inst_os = OSFromDisk(instance.os)
3629 env = OSEnvironment(instance, inst_os)
3631 if mode == constants.IEM_IMPORT:
3632 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3633 env["IMPORT_INDEX"] = str(disk_index)
3634 script = inst_os.import_script
3636 elif mode == constants.IEM_EXPORT:
3637 env["EXPORT_DEVICE"] = real_disk.dev_path
3638 env["EXPORT_INDEX"] = str(disk_index)
3639 script = inst_os.export_script
3641 # TODO: Pass special environment only to script
3642 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3644 if mode == constants.IEM_IMPORT:
3645 suffix = "| %s" % script_cmd
3647 elif mode == constants.IEM_EXPORT:
3648 prefix = "%s |" % script_cmd
3650 # Let script predict size
3651 exp_size = constants.IE_CUSTOM_SIZE
3654 _Fail("Invalid %s I/O mode %r", mode, ieio)
3656 return (env, prefix, suffix, exp_size)
3659 def _CreateImportExportStatusDir(prefix):
3660 """Creates status directory for import/export.
3663 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3665 (prefix, utils.TimestampForFilename())))
3668 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3670 """Starts an import or export daemon.
3672 @param mode: Import/output mode
3673 @type opts: L{objects.ImportExportOptions}
3674 @param opts: Daemon options
3676 @param host: Remote host for export (None for import)
3678 @param port: Remote port for export (None for import)
3679 @type instance: L{objects.Instance}
3680 @param instance: Instance object
3681 @type component: string
3682 @param component: which part of the instance is transferred now,
3684 @param ieio: Input/output type
3685 @param ieioargs: Input/output arguments
3688 if mode == constants.IEM_IMPORT:
3691 if not (host is None and port is None):
3692 _Fail("Can not specify host or port on import")
3694 elif mode == constants.IEM_EXPORT:
3697 if host is None or port is None:
3698 _Fail("Host and port must be specified for an export")
3701 _Fail("Invalid mode %r", mode)
3703 if (opts.key_name is None) ^ (opts.ca_pem is None):
3704 _Fail("Cluster certificate can only be used for both key and CA")
3706 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3707 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3709 if opts.key_name is None:
3711 key_path = pathutils.NODED_CERT_FILE
3712 cert_path = pathutils.NODED_CERT_FILE
3713 assert opts.ca_pem is None
3715 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3717 assert opts.ca_pem is not None
3719 for i in [key_path, cert_path]:
3720 if not os.path.exists(i):
3721 _Fail("File '%s' does not exist" % i)
3723 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3725 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3726 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3727 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3729 if opts.ca_pem is None:
3731 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3736 utils.WriteFile(ca_file, data=ca, mode=0400)
3739 pathutils.IMPORT_EXPORT_DAEMON,
3741 "--key=%s" % key_path,
3742 "--cert=%s" % cert_path,
3743 "--ca=%s" % ca_file,
3747 cmd.append("--host=%s" % host)
3750 cmd.append("--port=%s" % port)
3753 cmd.append("--ipv6")
3755 cmd.append("--ipv4")
3758 cmd.append("--compress=%s" % opts.compress)
3761 cmd.append("--magic=%s" % opts.magic)
3763 if exp_size is not None:
3764 cmd.append("--expected-size=%s" % exp_size)
3767 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3770 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3772 if mode == constants.IEM_EXPORT:
3773 # Retry connection a few times when connecting to remote peer
3774 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3775 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3776 elif opts.connect_timeout is not None:
3777 assert mode == constants.IEM_IMPORT
3778 # Overall timeout for establishing connection while listening
3779 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3781 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3783 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3784 # support for receiving a file descriptor for output
3785 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3788 # The import/export name is simply the status directory name
3789 return os.path.basename(status_dir)
3792 shutil.rmtree(status_dir, ignore_errors=True)
3796 def GetImportExportStatus(names):
3797 """Returns import/export daemon status.
3799 @type names: sequence
3800 @param names: List of names
3801 @rtype: List of dicts
3802 @return: Returns a list of the state of each named import/export or None if a
3803 status couldn't be read
3809 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3813 data = utils.ReadFile(status_file)
3814 except EnvironmentError, err:
3815 if err.errno != errno.ENOENT:
3823 result.append(serializer.LoadJson(data))
3828 def AbortImportExport(name):
3829 """Sends SIGTERM to a running import/export daemon.
3832 logging.info("Abort import/export %s", name)
3834 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3835 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3838 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3840 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3843 def CleanupImportExport(name):
3844 """Cleanup after an import or export.
3846 If the import/export daemon is still running it's killed. Afterwards the
3847 whole status directory is removed.
3850 logging.info("Finalizing import/export %s", name)
3852 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3854 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3857 logging.info("Import/export %s is still running with PID %s",
3859 utils.KillProcess(pid, waitpid=False)
3861 shutil.rmtree(status_dir, ignore_errors=True)
3864 def _SetPhysicalId(target_node_uuid, nodes_ip, disks):
3865 """Sets the correct physical ID on all passed disks.
3869 cf.SetPhysicalID(target_node_uuid, nodes_ip)
3872 def _FindDisks(target_node_uuid, nodes_ip, disks):
3873 """Sets the physical ID on disks and returns the block devices.
3876 _SetPhysicalId(target_node_uuid, nodes_ip, disks)
3881 rd = _RecursiveFindBD(cf)
3883 _Fail("Can't find device %s", cf)
3888 def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
3889 """Disconnects the network on a list of drbd devices.
3892 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3898 except errors.BlockDeviceError, err:
3899 _Fail("Can't change network configuration to standalone mode: %s",
3903 def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
3905 """Attaches the network on a list of drbd devices.
3908 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3911 for idx, rd in enumerate(bdevs):
3913 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3914 except EnvironmentError, err:
3915 _Fail("Can't create symlink: %s", err)
3916 # reconnect disks, switch to new master configuration and if
3917 # needed primary mode
3920 rd.AttachNet(multimaster)
3921 except errors.BlockDeviceError, err:
3922 _Fail("Can't change network configuration: %s", err)
3924 # wait until the disks are connected; we need to retry the re-attach
3925 # if the device becomes standalone, as this might happen if the one
3926 # node disconnects and reconnects in a different mode before the
3927 # other node reconnects; in this case, one or both of the nodes will
3928 # decide it has wrong configuration and switch to standalone
3931 all_connected = True
3934 stats = rd.GetProcStatus()
3936 all_connected = (all_connected and
3937 (stats.is_connected or stats.is_in_resync))
3939 if stats.is_standalone:
3940 # peer had different config info and this node became
3941 # standalone, even though this should not happen with the
3942 # new staged way of changing disk configs
3944 rd.AttachNet(multimaster)
3945 except errors.BlockDeviceError, err:
3946 _Fail("Can't change network configuration: %s", err)
3948 if not all_connected:
3949 raise utils.RetryAgain()
3952 # Start with a delay of 100 miliseconds and go up to 5 seconds
3953 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3954 except utils.RetryTimeout:
3955 _Fail("Timeout in disk reconnecting")
3958 # change to primary mode
3962 except errors.BlockDeviceError, err:
3963 _Fail("Can't change to primary mode: %s", err)
3966 def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
3967 """Wait until DRBDs have synchronized.
3971 stats = rd.GetProcStatus()
3972 if not (stats.is_connected or stats.is_in_resync):
3973 raise utils.RetryAgain()
3976 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3982 # poll each second for 15 seconds
3983 stats = utils.Retry(_helper, 1, 15, args=[rd])
3984 except utils.RetryTimeout:
3985 stats = rd.GetProcStatus()
3987 if not (stats.is_connected or stats.is_in_resync):
3988 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3989 alldone = alldone and (not stats.is_in_resync)
3990 if stats.sync_percent is not None:
3991 min_resync = min(min_resync, stats.sync_percent)
3993 return (alldone, min_resync)
3996 def DrbdNeedsActivation(target_node_uuid, nodes_ip, disks):
3997 """Checks which of the passed disks needs activation and returns their UUIDs.
4000 _SetPhysicalId(target_node_uuid, nodes_ip, disks)
4004 rd = _RecursiveFindBD(disk)
4006 faulty_disks.append(disk)
4009 stats = rd.GetProcStatus()
4010 if stats.is_standalone or stats.is_diskless:
4011 faulty_disks.append(disk)
4013 return [disk.uuid for disk in faulty_disks]
4016 def GetDrbdUsermodeHelper():
4017 """Returns DRBD usermode helper currently configured.
4021 return drbd.DRBD8.GetUsermodeHelper()
4022 except errors.BlockDeviceError, err:
4026 def PowercycleNode(hypervisor_type, hvparams=None):
4027 """Hard-powercycle the node.
4029 Because we need to return first, and schedule the powercycle in the
4030 background, we won't be able to report failures nicely.
4033 hyper = hypervisor.GetHypervisor(hypervisor_type)
4037 # if we can't fork, we'll pretend that we're in the child process
4040 return "Reboot scheduled in 5 seconds"
4041 # ensure the child is running on ram
4044 except Exception: # pylint: disable=W0703
4047 hyper.PowercycleNode(hvparams=hvparams)
4050 def _VerifyRestrictedCmdName(cmd):
4051 """Verifies a restricted command name.
4054 @param cmd: Command name
4055 @rtype: tuple; (boolean, string or None)
4056 @return: The tuple's first element is the status; if C{False}, the second
4057 element is an error message string, otherwise it's C{None}
4061 return (False, "Missing command name")
4063 if os.path.basename(cmd) != cmd:
4064 return (False, "Invalid command name")
4066 if not constants.EXT_PLUGIN_MASK.match(cmd):
4067 return (False, "Command name contains forbidden characters")
4072 def _CommonRestrictedCmdCheck(path, owner):
4073 """Common checks for restricted command file system directories and files.
4076 @param path: Path to check
4077 @param owner: C{None} or tuple containing UID and GID
4078 @rtype: tuple; (boolean, string or C{os.stat} result)
4079 @return: The tuple's first element is the status; if C{False}, the second
4080 element is an error message string, otherwise it's the result of C{os.stat}
4084 # Default to root as owner
4089 except EnvironmentError, err:
4090 return (False, "Can't stat(2) '%s': %s" % (path, err))
4092 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4093 return (False, "Permissions on '%s' are too permissive" % path)
4095 if (st.st_uid, st.st_gid) != owner:
4096 (owner_uid, owner_gid) = owner
4097 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4102 def _VerifyRestrictedCmdDirectory(path, _owner=None):
4103 """Verifies restricted command directory.
4106 @param path: Path to check
4107 @rtype: tuple; (boolean, string or None)
4108 @return: The tuple's first element is the status; if C{False}, the second
4109 element is an error message string, otherwise it's C{None}
4112 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4115 return (False, value)
4117 if not stat.S_ISDIR(value.st_mode):
4118 return (False, "Path '%s' is not a directory" % path)
4123 def _VerifyRestrictedCmd(path, cmd, _owner=None):
4124 """Verifies a whole restricted command and returns its executable filename.
4127 @param path: Directory containing restricted commands
4129 @param cmd: Command name
4130 @rtype: tuple; (boolean, string)
4131 @return: The tuple's first element is the status; if C{False}, the second
4132 element is an error message string, otherwise the second element is the
4133 absolute path to the executable
4136 executable = utils.PathJoin(path, cmd)
4138 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4143 if not utils.IsExecutable(executable):
4144 return (False, "access(2) thinks '%s' can't be executed" % executable)
4146 return (True, executable)
4149 def _PrepareRestrictedCmd(path, cmd,
4150 _verify_dir=_VerifyRestrictedCmdDirectory,
4151 _verify_name=_VerifyRestrictedCmdName,
4152 _verify_cmd=_VerifyRestrictedCmd):
4153 """Performs a number of tests on a restricted command.
4156 @param path: Directory containing restricted commands
4158 @param cmd: Command name
4159 @return: Same as L{_VerifyRestrictedCmd}
4162 # Verify the directory first
4163 (status, msg) = _verify_dir(path)
4165 # Check command if everything was alright
4166 (status, msg) = _verify_name(cmd)
4171 # Check actual executable
4172 return _verify_cmd(path, cmd)
4175 def RunRestrictedCmd(cmd,
4176 _lock_timeout=_RCMD_LOCK_TIMEOUT,
4177 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4178 _path=pathutils.RESTRICTED_COMMANDS_DIR,
4179 _sleep_fn=time.sleep,
4180 _prepare_fn=_PrepareRestrictedCmd,
4181 _runcmd_fn=utils.RunCmd,
4182 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4183 """Executes a restricted command after performing strict tests.
4186 @param cmd: Command name
4188 @return: Command output
4189 @raise RPCFail: In case of an error
4192 logging.info("Preparing to run restricted command '%s'", cmd)
4195 _Fail("Restricted commands disabled at configure time")
4201 lock = utils.FileLock.Open(_lock_file)
4202 lock.Exclusive(blocking=True, timeout=_lock_timeout)
4204 (status, value) = _prepare_fn(_path, cmd)
4207 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4208 postfork_fn=lambda _: lock.Unlock())
4210 logging.error(value)
4211 except Exception: # pylint: disable=W0703
4212 # Keep original error in log
4213 logging.exception("Caught exception")
4215 if cmdresult is None:
4216 logging.info("Sleeping for %0.1f seconds before returning",
4217 _RCMD_INVALID_DELAY)
4218 _sleep_fn(_RCMD_INVALID_DELAY)
4220 # Do not include original error message in returned error
4221 _Fail("Executing command '%s' failed" % cmd)
4222 elif cmdresult.failed or cmdresult.fail_reason:
4223 _Fail("Restricted command '%s' failed: %s; output: %s",
4224 cmd, cmdresult.fail_reason, cmdresult.output)
4226 return cmdresult.output
4228 if lock is not None:
4229 # Release lock at last
4234 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4235 """Creates or removes the watcher pause file.
4237 @type until: None or number
4238 @param until: Unix timestamp saying until when the watcher shouldn't run
4242 logging.info("Received request to no longer pause watcher")
4243 utils.RemoveFile(_filename)
4245 logging.info("Received request to pause watcher until %s", until)
4247 if not ht.TNumber(until):
4248 _Fail("Duration must be numeric")
4250 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4253 class HooksRunner(object):
4256 This class is instantiated on the node side (ganeti-noded) and not
4260 def __init__(self, hooks_base_dir=None):
4261 """Constructor for hooks runner.
4263 @type hooks_base_dir: str or None
4264 @param hooks_base_dir: if not None, this overrides the
4265 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4268 if hooks_base_dir is None:
4269 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4270 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4272 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4274 def RunLocalHooks(self, node_list, hpath, phase, env):
4275 """Check that the hooks will be run only locally and then run them.
4278 assert len(node_list) == 1
4280 _, myself = ssconf.GetMasterAndMyself()
4281 assert node == myself
4283 results = self.RunHooks(hpath, phase, env)
4285 # Return values in the form expected by HooksMaster
4286 return {node: (None, False, results)}
4288 def RunHooks(self, hpath, phase, env):
4289 """Run the scripts in the hooks directory.
4292 @param hpath: the path to the hooks directory which
4295 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4296 L{constants.HOOKS_PHASE_POST}
4298 @param env: dictionary with the environment for the hook
4300 @return: list of 3-element tuples:
4302 - script result, either L{constants.HKR_SUCCESS} or
4303 L{constants.HKR_FAIL}
4304 - output of the script
4306 @raise errors.ProgrammerError: for invalid input
4310 if phase == constants.HOOKS_PHASE_PRE:
4312 elif phase == constants.HOOKS_PHASE_POST:
4315 _Fail("Unknown hooks phase '%s'", phase)
4317 subdir = "%s-%s.d" % (hpath, suffix)
4318 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4322 if not os.path.isdir(dir_name):
4323 # for non-existing/non-dirs, we simply exit instead of logging a
4324 # warning at every operation
4327 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4329 for (relname, relstatus, runresult) in runparts_results:
4330 if relstatus == constants.RUNPARTS_SKIP:
4331 rrval = constants.HKR_SKIP
4333 elif relstatus == constants.RUNPARTS_ERR:
4334 rrval = constants.HKR_FAIL
4335 output = "Hook script execution error: %s" % runresult
4336 elif relstatus == constants.RUNPARTS_RUN:
4337 if runresult.failed:
4338 rrval = constants.HKR_FAIL
4340 rrval = constants.HKR_SUCCESS
4341 output = utils.SafeEncode(runresult.output.strip())
4342 results.append(("%s/%s" % (subdir, relname), rrval, output))
4347 class IAllocatorRunner(object):
4348 """IAllocator runner.
4350 This class is instantiated on the node side (ganeti-noded) and not on
4355 def Run(name, idata):
4356 """Run an iallocator script.
4359 @param name: the iallocator script name
4361 @param idata: the allocator input data
4364 @return: two element tuple of:
4366 - either error message or stdout of allocator (for success)
4369 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4371 if alloc_script is None:
4372 _Fail("iallocator module '%s' not found in the search path", name)
4374 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4378 result = utils.RunCmd([alloc_script, fin_name])
4380 _Fail("iallocator module '%s' failed: %s, output '%s'",
4381 name, result.fail_reason, result.output)
4385 return result.stdout
4388 class DevCacheManager(object):
4389 """Simple class for managing a cache of block device information.
4392 _DEV_PREFIX = "/dev/"
4393 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4396 def _ConvertPath(cls, dev_path):
4397 """Converts a /dev/name path to the cache file name.
4399 This replaces slashes with underscores and strips the /dev
4400 prefix. It then returns the full path to the cache file.
4403 @param dev_path: the C{/dev/} path name
4405 @return: the converted path name
4408 if dev_path.startswith(cls._DEV_PREFIX):
4409 dev_path = dev_path[len(cls._DEV_PREFIX):]
4410 dev_path = dev_path.replace("/", "_")
4411 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4415 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4416 """Updates the cache information for a given device.
4419 @param dev_path: the pathname of the device
4421 @param owner: the owner (instance name) of the device
4422 @type on_primary: bool
4423 @param on_primary: whether this is the primary
4426 @param iv_name: the instance-visible name of the
4427 device, as in objects.Disk.iv_name
4432 if dev_path is None:
4433 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4435 fpath = cls._ConvertPath(dev_path)
4441 iv_name = "not_visible"
4442 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4444 utils.WriteFile(fpath, data=fdata)
4445 except EnvironmentError, err:
4446 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4449 def RemoveCache(cls, dev_path):
4450 """Remove data for a dev_path.
4452 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4453 path name and logging.
4456 @param dev_path: the pathname of the device
4461 if dev_path is None:
4462 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4464 fpath = cls._ConvertPath(dev_path)
4466 utils.RemoveFile(fpath)
4467 except EnvironmentError, err:
4468 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)