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(r"^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
91 # Actions for the master setup script
92 _MASTER_START = "start"
95 #: Maximum file permissions for restricted command directory and executables
96 _RCMD_MAX_MODE = (stat.S_IRWXU |
97 stat.S_IRGRP | stat.S_IXGRP |
98 stat.S_IROTH | stat.S_IXOTH)
100 #: Delay before returning an error for restricted commands
101 _RCMD_INVALID_DELAY = 10
103 #: How long to wait to acquire lock for restricted commands (shorter than
104 #: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
105 #: command requests arrive
106 _RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
109 class RPCFail(Exception):
110 """Class denoting RPC failure.
112 Its argument is the error message.
117 def _GetInstReasonFilename(instance_name):
118 """Path of the file containing the reason of the instance status change.
120 @type instance_name: string
121 @param instance_name: The name of the instance
123 @return: The path of the file
126 return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
129 def _StoreInstReasonTrail(instance_name, trail):
130 """Serialize a reason trail related to an instance change of state to file.
132 The exact location of the file depends on the name of the instance and on
133 the configuration of the Ganeti cluster defined at deploy time.
135 @type instance_name: string
136 @param instance_name: The name of the instance
140 json = serializer.DumpJson(trail)
141 filename = _GetInstReasonFilename(instance_name)
142 utils.WriteFile(filename, data=json)
145 def _Fail(msg, *args, **kwargs):
146 """Log an error and the raise an RPCFail exception.
148 This exception is then handled specially in the ganeti daemon and
149 turned into a 'failed' return type. As such, this function is a
150 useful shortcut for logging the error and returning it to the master
154 @param msg: the text of the exception
160 if "log" not in kwargs or kwargs["log"]: # if we should log this error
161 if "exc" in kwargs and kwargs["exc"]:
162 logging.exception(msg)
169 """Simple wrapper to return a SimpleStore.
171 @rtype: L{ssconf.SimpleStore}
172 @return: a SimpleStore instance
175 return ssconf.SimpleStore()
178 def _GetSshRunner(cluster_name):
179 """Simple wrapper to return an SshRunner.
181 @type cluster_name: str
182 @param cluster_name: the cluster name, which is needed
183 by the SshRunner constructor
184 @rtype: L{ssh.SshRunner}
185 @return: an SshRunner instance
188 return ssh.SshRunner(cluster_name)
191 def _Decompress(data):
192 """Unpacks data compressed by the RPC client.
194 @type data: list or tuple
195 @param data: Data sent by RPC client
197 @return: Decompressed data
200 assert isinstance(data, (list, tuple))
201 assert len(data) == 2
202 (encoding, content) = data
203 if encoding == constants.RPC_ENCODING_NONE:
205 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
206 return zlib.decompress(base64.b64decode(content))
208 raise AssertionError("Unknown data encoding")
211 def _CleanDirectory(path, exclude=None):
212 """Removes all regular files in a directory.
215 @param path: the directory to clean
217 @param exclude: list of files to be excluded, defaults
221 if path not in _ALLOWED_CLEAN_DIRS:
222 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
225 if not os.path.isdir(path):
230 # Normalize excluded paths
231 exclude = [os.path.normpath(i) for i in exclude]
233 for rel_name in utils.ListVisibleFiles(path):
234 full_name = utils.PathJoin(path, rel_name)
235 if full_name in exclude:
237 if os.path.isfile(full_name) and not os.path.islink(full_name):
238 utils.RemoveFile(full_name)
241 def _BuildUploadFileList():
242 """Build the list of allowed upload files.
244 This is abstracted so that it's built only once at module import time.
247 allowed_files = set([
248 pathutils.CLUSTER_CONF_FILE,
250 pathutils.SSH_KNOWN_HOSTS_FILE,
251 pathutils.VNC_PASSWORD_FILE,
252 pathutils.RAPI_CERT_FILE,
253 pathutils.SPICE_CERT_FILE,
254 pathutils.SPICE_CACERT_FILE,
255 pathutils.RAPI_USERS_FILE,
256 pathutils.CONFD_HMAC_KEY,
257 pathutils.CLUSTER_DOMAIN_SECRET_FILE,
260 for hv_name in constants.HYPER_TYPES:
261 hv_class = hypervisor.GetHypervisorClass(hv_name)
262 allowed_files.update(hv_class.GetAncillaryFiles()[0])
264 assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
265 "Allowed file storage paths should never be uploaded via RPC"
267 return frozenset(allowed_files)
270 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
274 """Removes job queue files and archived jobs.
280 _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
281 _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
285 """Returns master information.
287 This is an utility function to compute master information, either
288 for consumption here or from the node daemon.
291 @return: master_netdev, master_ip, master_name, primary_ip_family,
293 @raise RPCFail: in case of errors
298 master_netdev = cfg.GetMasterNetdev()
299 master_ip = cfg.GetMasterIP()
300 master_netmask = cfg.GetMasterNetmask()
301 master_node = cfg.GetMasterNode()
302 primary_ip_family = cfg.GetPrimaryIPFamily()
303 except errors.ConfigurationError, err:
304 _Fail("Cluster configuration incomplete: %s", err, exc=True)
305 return (master_netdev, master_ip, master_node, primary_ip_family,
309 def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
310 """Decorator that runs hooks before and after the decorated function.
312 @type hook_opcode: string
313 @param hook_opcode: opcode of the hook
314 @type hooks_path: string
315 @param hooks_path: path of the hooks
316 @type env_builder_fn: function
317 @param env_builder_fn: function that returns a dictionary containing the
318 environment variables for the hooks. Will get all the parameters of the
320 @raise RPCFail: in case of pre-hook failure
324 def wrapper(*args, **kwargs):
325 _, myself = ssconf.GetMasterAndMyself()
326 nodes = ([myself], [myself]) # these hooks run locally
328 env_fn = compat.partial(env_builder_fn, *args, **kwargs)
332 hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
333 hr.RunLocalHooks, None, env_fn,
334 logging.warning, cfg.GetClusterName(),
336 hm.RunPhase(constants.HOOKS_PHASE_PRE)
337 result = fn(*args, **kwargs)
338 hm.RunPhase(constants.HOOKS_PHASE_POST)
345 def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
346 """Builds environment variables for master IP hooks.
348 @type master_params: L{objects.MasterNetworkParameters}
349 @param master_params: network parameters of the master
350 @type use_external_mip_script: boolean
351 @param use_external_mip_script: whether to use an external master IP
352 address setup script (unused, but necessary per the implementation of the
353 _RunLocalHooks decorator)
356 # pylint: disable=W0613
357 ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
359 "MASTER_NETDEV": master_params.netdev,
360 "MASTER_IP": master_params.ip,
361 "MASTER_NETMASK": str(master_params.netmask),
362 "CLUSTER_IP_VERSION": str(ver),
368 def _RunMasterSetupScript(master_params, action, use_external_mip_script):
369 """Execute the master IP address setup script.
371 @type master_params: L{objects.MasterNetworkParameters}
372 @param master_params: network parameters of the master
374 @param action: action to pass to the script. Must be one of
375 L{backend._MASTER_START} or L{backend._MASTER_STOP}
376 @type use_external_mip_script: boolean
377 @param use_external_mip_script: whether to use an external master IP
379 @raise backend.RPCFail: if there are errors during the execution of the
383 env = _BuildMasterIpEnv(master_params)
385 if use_external_mip_script:
386 setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
388 setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
390 result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
393 _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
394 (action, result.exit_code, result.output), log=True)
397 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
399 def ActivateMasterIp(master_params, use_external_mip_script):
400 """Activate the IP address of the master daemon.
402 @type master_params: L{objects.MasterNetworkParameters}
403 @param master_params: network parameters of the master
404 @type use_external_mip_script: boolean
405 @param use_external_mip_script: whether to use an external master IP
407 @raise RPCFail: in case of errors during the IP startup
410 _RunMasterSetupScript(master_params, _MASTER_START,
411 use_external_mip_script)
414 def StartMasterDaemons(no_voting):
415 """Activate local node as master node.
417 The function will start the master daemons (ganeti-masterd and ganeti-rapi).
419 @type no_voting: boolean
420 @param no_voting: whether to start ganeti-masterd without a node vote
421 but still non-interactively
427 masterd_args = "--no-voting --yes-do-it"
432 "EXTRA_MASTERD_ARGS": masterd_args,
435 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
437 msg = "Can't start Ganeti master: %s" % result.output
442 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
444 def DeactivateMasterIp(master_params, use_external_mip_script):
445 """Deactivate the master IP on this node.
447 @type master_params: L{objects.MasterNetworkParameters}
448 @param master_params: network parameters of the master
449 @type use_external_mip_script: boolean
450 @param use_external_mip_script: whether to use an external master IP
452 @raise RPCFail: in case of errors during the IP turndown
455 _RunMasterSetupScript(master_params, _MASTER_STOP,
456 use_external_mip_script)
459 def StopMasterDaemons():
460 """Stop the master daemons on this node.
462 Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
467 # TODO: log and report back to the caller the error failures; we
468 # need to decide in which case we fail the RPC for this
470 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
472 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
474 result.cmd, result.exit_code, result.output)
477 def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
478 """Change the netmask of the master IP.
480 @param old_netmask: the old value of the netmask
481 @param netmask: the new value of the netmask
482 @param master_ip: the master IP
483 @param master_netdev: the master network device
486 if old_netmask == netmask:
489 if not netutils.IPAddress.Own(master_ip):
490 _Fail("The master IP address is not up, not attempting to change its"
493 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
494 "%s/%s" % (master_ip, netmask),
495 "dev", master_netdev, "label",
496 "%s:0" % master_netdev])
498 _Fail("Could not set the new netmask on the master IP address")
500 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
501 "%s/%s" % (master_ip, old_netmask),
502 "dev", master_netdev, "label",
503 "%s:0" % master_netdev])
505 _Fail("Could not bring down the master IP address with the old netmask")
508 def EtcHostsModify(mode, host, ip):
509 """Modify a host entry in /etc/hosts.
511 @param mode: The mode to operate. Either add or remove entry
512 @param host: The host to operate on
513 @param ip: The ip associated with the entry
516 if mode == constants.ETC_HOSTS_ADD:
518 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
520 utils.AddHostToEtcHosts(host, ip)
521 elif mode == constants.ETC_HOSTS_REMOVE:
523 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
524 " parameter is present")
525 utils.RemoveHostFromEtcHosts(host)
527 RPCFail("Mode not supported")
530 def LeaveCluster(modify_ssh_setup):
531 """Cleans up and remove the current node.
533 This function cleans up and prepares the current node to be removed
536 If processing is successful, then it raises an
537 L{errors.QuitGanetiException} which is used as a special case to
538 shutdown the node daemon.
540 @param modify_ssh_setup: boolean
543 _CleanDirectory(pathutils.DATA_DIR)
544 _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
549 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
551 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
553 utils.RemoveFile(priv_key)
554 utils.RemoveFile(pub_key)
555 except errors.OpExecError:
556 logging.exception("Error while processing ssh files")
559 utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
560 utils.RemoveFile(pathutils.RAPI_CERT_FILE)
561 utils.RemoveFile(pathutils.SPICE_CERT_FILE)
562 utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
563 utils.RemoveFile(pathutils.NODED_CERT_FILE)
564 except: # pylint: disable=W0702
565 logging.exception("Error while removing cluster secrets")
567 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
569 logging.error("Command %s failed with exitcode %s and error %s",
570 result.cmd, result.exit_code, result.output)
572 # Raise a custom exception (handled in ganeti-noded)
573 raise errors.QuitGanetiException(True, "Shutdown scheduled")
576 def _CheckStorageParams(params, num_params):
577 """Performs sanity checks for storage parameters.
580 @param params: list of storage parameters
581 @type num_params: int
582 @param num_params: expected number of parameters
586 raise errors.ProgrammerError("No storage parameters for storage"
587 " reporting is provided.")
588 if not isinstance(params, list):
589 raise errors.ProgrammerError("The storage parameters are not of type"
590 " list: '%s'" % params)
591 if not len(params) == num_params:
592 raise errors.ProgrammerError("Did not receive the expected number of"
593 "storage parameters: expected %s,"
594 " received '%s'" % (num_params, len(params)))
597 def _CheckLvmStorageParams(params):
598 """Performs sanity check for the 'exclusive storage' flag.
600 @see: C{_CheckStorageParams}
603 _CheckStorageParams(params, 1)
604 excl_stor = params[0]
605 if not isinstance(params[0], bool):
606 raise errors.ProgrammerError("Exclusive storage parameter is not"
607 " boolean: '%s'." % excl_stor)
611 def _GetLvmVgSpaceInfo(name, params):
612 """Wrapper around C{_GetVgInfo} which checks the storage parameters.
615 @param name: name of the volume group
617 @param params: list of storage parameters, which in this case should be
618 containing only one for exclusive storage
621 excl_stor = _CheckLvmStorageParams(params)
622 return _GetVgInfo(name, excl_stor)
626 name, excl_stor, info_fn=bdev.LogicalVolume.GetVGInfo):
627 """Retrieves information about a LVM volume group.
630 # TODO: GetVGInfo supports returning information for multiple VGs at once
631 vginfo = info_fn([name], excl_stor)
633 vg_free = int(round(vginfo[0][0], 0))
634 vg_size = int(round(vginfo[0][1], 0))
640 "type": constants.ST_LVM_VG,
642 "storage_free": vg_free,
643 "storage_size": vg_size,
647 def _GetLvmPvSpaceInfo(name, params):
648 """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
650 @see: C{_GetLvmVgSpaceInfo}
653 excl_stor = _CheckLvmStorageParams(params)
654 return _GetVgSpindlesInfo(name, excl_stor)
657 def _GetVgSpindlesInfo(
658 name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
659 """Retrieves information about spindles in an LVM volume group.
663 @type excl_stor: bool
664 @param excl_stor: exclusive storage
666 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
667 free spindles, total spindles respectively
671 (vg_free, vg_size) = info_fn(name)
676 "type": constants.ST_LVM_PV,
678 "storage_free": vg_free,
679 "storage_size": vg_size,
683 def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
684 """Retrieves node information from a hypervisor.
686 The information returned depends on the hypervisor. Common items:
688 - vg_size is the size of the configured volume group in MiB
689 - vg_free is the free size of the volume group in MiB
690 - memory_dom0 is the memory allocated for domain0 in MiB
691 - memory_free is the currently available (free) ram in MiB
692 - memory_total is the total number of ram in MiB
693 - hv_version: the hypervisor version, if available
695 @type hvparams: dict of string
696 @param hvparams: the hypervisor's hvparams
699 return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
702 def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
703 """Retrieves node information for all hypervisors.
705 See C{_GetHvInfo} for information on the output.
707 @type hv_specs: list of pairs (string, dict of strings)
708 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
715 for hvname, hvparams in hv_specs:
716 result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
720 def _GetNamedNodeInfo(names, fn):
721 """Calls C{fn} for all names in C{names} and returns a dictionary.
729 return map(fn, names)
732 def GetNodeInfo(storage_units, hv_specs):
733 """Gives back a hash with different information about the node.
735 @type storage_units: list of tuples (string, string, list)
736 @param storage_units: List of tuples (storage unit, identifier, parameters) to
737 ask for disk space information. In case of lvm-vg, the identifier is
738 the VG name. The parameters can contain additional, storage-type-specific
739 parameters, for example exclusive storage for lvm storage.
740 @type hv_specs: list of pairs (string, dict of strings)
741 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
742 @rtype: tuple; (string, None/dict, None/dict)
743 @return: Tuple containing boot ID, volume group information and hypervisor
747 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
748 storage_info = _GetNamedNodeInfo(
750 (lambda (storage_type, storage_key, storage_params):
751 _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
752 hv_info = _GetHvInfoAll(hv_specs)
753 return (bootid, storage_info, hv_info)
756 def _GetFileStorageSpaceInfo(path, params):
757 """Wrapper around filestorage.GetSpaceInfo.
759 The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
760 and ignore the *args parameter to not leak it into the filestorage
763 @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
767 _CheckStorageParams(params, 0)
768 return filestorage.GetFileStorageSpaceInfo(path)
771 # FIXME: implement storage reporting for all missing storage types.
772 _STORAGE_TYPE_INFO_FN = {
773 constants.ST_BLOCK: None,
774 constants.ST_DISKLESS: None,
775 constants.ST_EXT: None,
776 constants.ST_FILE: _GetFileStorageSpaceInfo,
777 constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
778 constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
779 constants.ST_RADOS: None,
783 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
784 """Looks up and applies the correct function to calculate free and total
785 storage for the given storage type.
787 @type storage_type: string
788 @param storage_type: the storage type for which the storage shall be reported.
789 @type storage_key: string
790 @param storage_key: identifier of a storage unit, e.g. the volume group name
791 of an LVM storage unit
793 @param args: various parameters that can be used for storage reporting. These
794 parameters and their semantics vary from storage type to storage type and
795 are just propagated in this function.
796 @return: the results of the application of the storage space function (see
797 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
799 @raises NotImplementedError: for storage types who don't support space
802 fn = _STORAGE_TYPE_INFO_FN[storage_type]
804 return fn(storage_key, *args)
806 raise NotImplementedError
809 def _CheckExclusivePvs(pvi_list):
810 """Check that PVs are not shared among LVs
812 @type pvi_list: list of L{objects.LvmPvInfo} objects
813 @param pvi_list: information about the PVs
815 @rtype: list of tuples (string, list of strings)
816 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
821 if len(pvi.lv_list) > 1:
822 res.append((pvi.name, pvi.lv_list))
826 def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
827 get_hv_fn=hypervisor.GetHypervisor):
828 """Verifies the hypervisor. Appends the results to the 'results' list.
831 @param what: a dictionary of things to check
832 @type vm_capable: boolean
833 @param vm_capable: whether or not this node is vm capable
835 @param result: dictionary of verification results; results of the
836 verifications in this function will be added here
837 @type all_hvparams: dict of dict of string
838 @param all_hvparams: dictionary mapping hypervisor names to hvparams
839 @type get_hv_fn: function
840 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
846 if constants.NV_HYPERVISOR in what:
847 result[constants.NV_HYPERVISOR] = {}
848 for hv_name in what[constants.NV_HYPERVISOR]:
849 hvparams = all_hvparams[hv_name]
851 val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
852 except errors.HypervisorError, err:
853 val = "Error while checking hypervisor: %s" % str(err)
854 result[constants.NV_HYPERVISOR][hv_name] = val
857 def _VerifyHvparams(what, vm_capable, result,
858 get_hv_fn=hypervisor.GetHypervisor):
859 """Verifies the hvparams. Appends the results to the 'results' list.
862 @param what: a dictionary of things to check
863 @type vm_capable: boolean
864 @param vm_capable: whether or not this node is vm capable
866 @param result: dictionary of verification results; results of the
867 verifications in this function will be added here
868 @type get_hv_fn: function
869 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
875 if constants.NV_HVPARAMS in what:
876 result[constants.NV_HVPARAMS] = []
877 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
879 logging.info("Validating hv %s, %s", hv_name, hvparms)
880 get_hv_fn(hv_name).ValidateParameters(hvparms)
881 except errors.HypervisorError, err:
882 result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
885 def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
886 """Verifies the instance list.
889 @param what: a dictionary of things to check
890 @type vm_capable: boolean
891 @param vm_capable: whether or not this node is vm capable
893 @param result: dictionary of verification results; results of the
894 verifications in this function will be added here
895 @type all_hvparams: dict of dict of string
896 @param all_hvparams: dictionary mapping hypervisor names to hvparams
899 if constants.NV_INSTANCELIST in what and vm_capable:
900 # GetInstanceList can fail
902 val = GetInstanceList(what[constants.NV_INSTANCELIST],
903 all_hvparams=all_hvparams)
906 result[constants.NV_INSTANCELIST] = val
909 def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
910 """Verifies the node info.
913 @param what: a dictionary of things to check
914 @type vm_capable: boolean
915 @param vm_capable: whether or not this node is vm capable
917 @param result: dictionary of verification results; results of the
918 verifications in this function will be added here
919 @type all_hvparams: dict of dict of string
920 @param all_hvparams: dictionary mapping hypervisor names to hvparams
923 if constants.NV_HVINFO in what and vm_capable:
924 hvname = what[constants.NV_HVINFO]
925 hyper = hypervisor.GetHypervisor(hvname)
926 hvparams = all_hvparams[hvname]
927 result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
930 def VerifyNode(what, cluster_name, all_hvparams):
931 """Verify the status of the local node.
933 Based on the input L{what} parameter, various checks are done on the
936 If the I{filelist} key is present, this list of
937 files is checksummed and the file/checksum pairs are returned.
939 If the I{nodelist} key is present, we check that we have
940 connectivity via ssh with the target nodes (and check the hostname
943 If the I{node-net-test} key is present, we check that we have
944 connectivity to the given nodes via both primary IP and, if
945 applicable, secondary IPs.
948 @param what: a dictionary of things to check:
949 - filelist: list of files for which to compute checksums
950 - nodelist: list of nodes we should check ssh communication with
951 - node-net-test: list of nodes we should check node daemon port
953 - hypervisor: list with hypervisors to run the verify for
954 @type cluster_name: string
955 @param cluster_name: the cluster's name
956 @type all_hvparams: dict of dict of strings
957 @param all_hvparams: a dictionary mapping hypervisor names to hvparams
959 @return: a dictionary with the same keys as the input dict, and
960 values representing the result of the checks
964 my_name = netutils.Hostname.GetSysName()
965 port = netutils.GetDaemonPort(constants.NODED)
966 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
968 _VerifyHypervisors(what, vm_capable, result, all_hvparams)
969 _VerifyHvparams(what, vm_capable, result)
971 if constants.NV_FILELIST in what:
972 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
973 what[constants.NV_FILELIST]))
974 result[constants.NV_FILELIST] = \
975 dict((vcluster.MakeVirtualPath(key), value)
976 for (key, value) in fingerprints.items())
978 if constants.NV_NODELIST in what:
979 (nodes, bynode) = what[constants.NV_NODELIST]
981 # Add nodes from other groups (different for each node)
983 nodes.extend(bynode[my_name])
988 random.shuffle(nodes)
990 # Try to contact all nodes
993 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
997 result[constants.NV_NODELIST] = val
999 if constants.NV_NODENETTEST in what:
1000 result[constants.NV_NODENETTEST] = tmp = {}
1001 my_pip = my_sip = None
1002 for name, pip, sip in what[constants.NV_NODENETTEST]:
1008 tmp[my_name] = ("Can't find my own primary/secondary IP"
1009 " in the node list")
1011 for name, pip, sip in what[constants.NV_NODENETTEST]:
1013 if not netutils.TcpPing(pip, port, source=my_pip):
1014 fail.append("primary")
1016 if not netutils.TcpPing(sip, port, source=my_sip):
1017 fail.append("secondary")
1019 tmp[name] = ("failure using the %s interface(s)" %
1022 if constants.NV_MASTERIP in what:
1023 # FIXME: add checks on incoming data structures (here and in the
1024 # rest of the function)
1025 master_name, master_ip = what[constants.NV_MASTERIP]
1026 if master_name == my_name:
1027 source = constants.IP4_ADDRESS_LOCALHOST
1030 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
1033 if constants.NV_USERSCRIPTS in what:
1034 result[constants.NV_USERSCRIPTS] = \
1035 [script for script in what[constants.NV_USERSCRIPTS]
1036 if not utils.IsExecutable(script)]
1038 if constants.NV_OOB_PATHS in what:
1039 result[constants.NV_OOB_PATHS] = tmp = []
1040 for path in what[constants.NV_OOB_PATHS]:
1043 except OSError, err:
1044 tmp.append("error stating out of band helper: %s" % err)
1046 if stat.S_ISREG(st.st_mode):
1047 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
1050 tmp.append("out of band helper %s is not executable" % path)
1052 tmp.append("out of band helper %s is not a file" % path)
1054 if constants.NV_LVLIST in what and vm_capable:
1056 val = GetVolumeList(utils.ListVolumeGroups().keys())
1057 except RPCFail, err:
1059 result[constants.NV_LVLIST] = val
1061 _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1063 if constants.NV_VGLIST in what and vm_capable:
1064 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1066 if constants.NV_PVLIST in what and vm_capable:
1067 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1068 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1069 filter_allocatable=False,
1070 include_lvs=check_exclusive_pvs)
1071 if check_exclusive_pvs:
1072 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1074 # Avoid sending useless data on the wire
1076 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1078 if constants.NV_VERSION in what:
1079 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1080 constants.RELEASE_VERSION)
1082 _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1084 if constants.NV_DRBDVERSION in what and vm_capable:
1086 drbd_version = DRBD8.GetProcInfo().GetVersionString()
1087 except errors.BlockDeviceError, err:
1088 logging.warning("Can't get DRBD version", exc_info=True)
1089 drbd_version = str(err)
1090 result[constants.NV_DRBDVERSION] = drbd_version
1092 if constants.NV_DRBDLIST in what and vm_capable:
1094 used_minors = drbd.DRBD8.GetUsedDevs()
1095 except errors.BlockDeviceError, err:
1096 logging.warning("Can't get used minors list", exc_info=True)
1097 used_minors = str(err)
1098 result[constants.NV_DRBDLIST] = used_minors
1100 if constants.NV_DRBDHELPER in what and vm_capable:
1103 payload = drbd.DRBD8.GetUsermodeHelper()
1104 except errors.BlockDeviceError, err:
1105 logging.error("Can't get DRBD usermode helper: %s", str(err))
1108 result[constants.NV_DRBDHELPER] = (status, payload)
1110 if constants.NV_NODESETUP in what:
1111 result[constants.NV_NODESETUP] = tmpr = []
1112 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1113 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1114 " under /sys, missing required directories /sys/block"
1115 " and /sys/class/net")
1116 if (not os.path.isdir("/proc/sys") or
1117 not os.path.isfile("/proc/sysrq-trigger")):
1118 tmpr.append("The procfs filesystem doesn't seem to be mounted"
1119 " under /proc, missing required directory /proc/sys and"
1120 " the file /proc/sysrq-trigger")
1122 if constants.NV_TIME in what:
1123 result[constants.NV_TIME] = utils.SplitTime(time.time())
1125 if constants.NV_OSLIST in what and vm_capable:
1126 result[constants.NV_OSLIST] = DiagnoseOS()
1128 if constants.NV_BRIDGES in what and vm_capable:
1129 result[constants.NV_BRIDGES] = [bridge
1130 for bridge in what[constants.NV_BRIDGES]
1131 if not utils.BridgeExists(bridge)]
1133 if what.get(constants.NV_ACCEPTED_STORAGE_PATHS) == my_name:
1134 result[constants.NV_ACCEPTED_STORAGE_PATHS] = \
1135 filestorage.ComputeWrongFileStoragePaths()
1137 if what.get(constants.NV_FILE_STORAGE_PATH):
1138 pathresult = filestorage.CheckFileStoragePath(
1139 what[constants.NV_FILE_STORAGE_PATH])
1141 result[constants.NV_FILE_STORAGE_PATH] = pathresult
1143 if what.get(constants.NV_SHARED_FILE_STORAGE_PATH):
1144 pathresult = filestorage.CheckFileStoragePath(
1145 what[constants.NV_SHARED_FILE_STORAGE_PATH])
1147 result[constants.NV_SHARED_FILE_STORAGE_PATH] = pathresult
1152 def GetBlockDevSizes(devices):
1153 """Return the size of the given block devices
1156 @param devices: list of block device nodes to query
1159 dictionary of all block devices under /dev (key). The value is their
1162 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1165 DEV_PREFIX = "/dev/"
1168 for devpath in devices:
1169 if not utils.IsBelowDir(DEV_PREFIX, devpath):
1173 st = os.stat(devpath)
1174 except EnvironmentError, err:
1175 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1178 if stat.S_ISBLK(st.st_mode):
1179 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1181 # We don't want to fail, just do not list this device as available
1182 logging.warning("Cannot get size for block device %s", devpath)
1185 size = int(result.stdout) / (1024 * 1024)
1186 blockdevs[devpath] = size
1190 def GetVolumeList(vg_names):
1191 """Compute list of logical volumes and their size.
1193 @type vg_names: list
1194 @param vg_names: the volume groups whose LVs we should list, or
1195 empty for all volume groups
1198 dictionary of all partions (key) with value being a tuple of
1199 their size (in MiB), inactive and online status::
1201 {'xenvg/test1': ('20.06', True, True)}
1203 in case of errors, a string is returned with the error
1211 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1212 "--separator=%s" % sep,
1213 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1215 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1217 for line in result.stdout.splitlines():
1219 match = _LVSLINE_REGEX.match(line)
1221 logging.error("Invalid line returned from lvs output: '%s'", line)
1223 vg_name, name, size, attr = match.groups()
1224 inactive = attr[4] == "-"
1225 online = attr[5] == "o"
1226 virtual = attr[0] == "v"
1228 # we don't want to report such volumes as existing, since they
1229 # don't really hold data
1231 lvs[vg_name + "/" + name] = (size, inactive, online)
1236 def ListVolumeGroups():
1237 """List the volume groups and their size.
1240 @return: dictionary with keys volume name and values the
1244 return utils.ListVolumeGroups()
1248 """List all volumes on this node.
1252 A list of dictionaries, each having four keys:
1253 - name: the logical volume name,
1254 - size: the size of the logical volume
1255 - dev: the physical device on which the LV lives
1256 - vg: the volume group to which it belongs
1258 In case of errors, we return an empty list and log the
1261 Note that since a logical volume can live on multiple physical
1262 volumes, the resulting list might include a logical volume
1266 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1268 "--options=lv_name,lv_size,devices,vg_name"])
1270 _Fail("Failed to list logical volumes, lvs output: %s",
1274 return dev.split("(")[0]
1276 def handle_dev(dev):
1277 return [parse_dev(x) for x in dev.split(",")]
1280 line = [v.strip() for v in line]
1281 return [{"name": line[0], "size": line[1],
1282 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1285 for line in result.stdout.splitlines():
1286 if line.count("|") >= 3:
1287 all_devs.extend(map_line(line.split("|")))
1289 logging.warning("Strange line in the output from lvs: '%s'", line)
1293 def BridgesExist(bridges_list):
1294 """Check if a list of bridges exist on the current node.
1297 @return: C{True} if all of them exist, C{False} otherwise
1301 for bridge in bridges_list:
1302 if not utils.BridgeExists(bridge):
1303 missing.append(bridge)
1306 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1309 def GetInstanceListForHypervisor(hname, hvparams=None,
1310 get_hv_fn=hypervisor.GetHypervisor):
1311 """Provides a list of instances of the given hypervisor.
1314 @param hname: name of the hypervisor
1315 @type hvparams: dict of strings
1316 @param hvparams: hypervisor parameters for the given hypervisor
1317 @type get_hv_fn: function
1318 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1319 name; optional parameter to increase testability
1322 @return: a list of all running instances on the current node
1323 - instance1.example.com
1324 - instance2.example.com
1329 hv = get_hv_fn(hname)
1330 names = hv.ListInstances(hvparams=hvparams)
1331 results.extend(names)
1332 except errors.HypervisorError, err:
1333 _Fail("Error enumerating instances (hypervisor %s): %s",
1334 hname, err, exc=True)
1338 def GetInstanceList(hypervisor_list, all_hvparams=None,
1339 get_hv_fn=hypervisor.GetHypervisor):
1340 """Provides a list of instances.
1342 @type hypervisor_list: list
1343 @param hypervisor_list: the list of hypervisors to query information
1344 @type all_hvparams: dict of dict of strings
1345 @param all_hvparams: a dictionary mapping hypervisor types to respective
1346 cluster-wide hypervisor parameters
1347 @type get_hv_fn: function
1348 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1349 name; optional parameter to increase testability
1352 @return: a list of all running instances on the current node
1353 - instance1.example.com
1354 - instance2.example.com
1358 for hname in hypervisor_list:
1359 hvparams = all_hvparams[hname]
1360 results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1361 get_hv_fn=get_hv_fn))
1365 def GetInstanceInfo(instance, hname, hvparams=None):
1366 """Gives back the information about an instance as a dictionary.
1368 @type instance: string
1369 @param instance: the instance name
1371 @param hname: the hypervisor type of the instance
1372 @type hvparams: dict of strings
1373 @param hvparams: the instance's hvparams
1376 @return: dictionary with the following keys:
1377 - memory: memory size of instance (int)
1378 - state: xen state of instance (string)
1379 - time: cpu time of instance (float)
1380 - vcpus: the number of vcpus (int)
1385 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1387 if iinfo is not None:
1388 output["memory"] = iinfo[2]
1389 output["vcpus"] = iinfo[3]
1390 output["state"] = iinfo[4]
1391 output["time"] = iinfo[5]
1396 def GetInstanceMigratable(instance):
1397 """Computes whether an instance can be migrated.
1399 @type instance: L{objects.Instance}
1400 @param instance: object representing the instance to be checked.
1403 @return: tuple of (result, description) where:
1404 - result: whether the instance can be migrated or not
1405 - description: a description of the issue, if relevant
1408 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1409 iname = instance.name
1410 if iname not in hyper.ListInstances(instance.hvparams):
1411 _Fail("Instance %s is not running", iname)
1413 for idx in range(len(instance.disks)):
1414 link_name = _GetBlockDevSymlinkPath(iname, idx)
1415 if not os.path.islink(link_name):
1416 logging.warning("Instance %s is missing symlink %s for disk %d",
1417 iname, link_name, idx)
1420 def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1421 """Gather data about all instances.
1423 This is the equivalent of L{GetInstanceInfo}, except that it
1424 computes data for all instances at once, thus being faster if one
1425 needs data about more than one instance.
1427 @type hypervisor_list: list
1428 @param hypervisor_list: list of hypervisors to query for instance data
1429 @type all_hvparams: dict of dict of strings
1430 @param all_hvparams: mapping of hypervisor names to hvparams
1433 @return: dictionary of instance: data, with data having the following keys:
1434 - memory: memory size of instance (int)
1435 - state: xen state of instance (string)
1436 - time: cpu time of instance (float)
1437 - vcpus: the number of vcpus
1442 for hname in hypervisor_list:
1443 hvparams = all_hvparams[hname]
1444 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1446 for name, _, memory, vcpus, state, times in iinfo:
1454 # we only check static parameters, like memory and vcpus,
1455 # and not state and time which can change between the
1456 # invocations of the different hypervisors
1457 for key in "memory", "vcpus":
1458 if value[key] != output[name][key]:
1459 _Fail("Instance %s is running twice"
1460 " with different parameters", name)
1461 output[name] = value
1466 def _InstanceLogName(kind, os_name, instance, component):
1467 """Compute the OS log filename for a given instance and operation.
1469 The instance name and os name are passed in as strings since not all
1470 operations have these as part of an instance object.
1473 @param kind: the operation type (e.g. add, import, etc.)
1474 @type os_name: string
1475 @param os_name: the os name
1476 @type instance: string
1477 @param instance: the name of the instance being imported/added/etc.
1478 @type component: string or None
1479 @param component: the name of the component of the instance being
1483 # TODO: Use tempfile.mkstemp to create unique filename
1485 assert "/" not in component
1486 c_msg = "-%s" % component
1489 base = ("%s-%s-%s%s-%s.log" %
1490 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1491 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1494 def InstanceOsAdd(instance, reinstall, debug):
1495 """Add an OS to an instance.
1497 @type instance: L{objects.Instance}
1498 @param instance: Instance whose OS is to be installed
1499 @type reinstall: boolean
1500 @param reinstall: whether this is an instance reinstall
1501 @type debug: integer
1502 @param debug: debug level, passed to the OS scripts
1506 inst_os = OSFromDisk(instance.os)
1508 create_env = OSEnvironment(instance, inst_os, debug)
1510 create_env["INSTANCE_REINSTALL"] = "1"
1512 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1514 result = utils.RunCmd([inst_os.create_script], env=create_env,
1515 cwd=inst_os.path, output=logfile, reset_env=True)
1517 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1518 " output: %s", result.cmd, result.fail_reason, logfile,
1520 lines = [utils.SafeEncode(val)
1521 for val in utils.TailFile(logfile, lines=20)]
1522 _Fail("OS create script failed (%s), last lines in the"
1523 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1526 def RunRenameInstance(instance, old_name, debug):
1527 """Run the OS rename script for an instance.
1529 @type instance: L{objects.Instance}
1530 @param instance: Instance whose OS is to be installed
1531 @type old_name: string
1532 @param old_name: previous instance name
1533 @type debug: integer
1534 @param debug: debug level, passed to the OS scripts
1536 @return: the success of the operation
1539 inst_os = OSFromDisk(instance.os)
1541 rename_env = OSEnvironment(instance, inst_os, debug)
1542 rename_env["OLD_INSTANCE_NAME"] = old_name
1544 logfile = _InstanceLogName("rename", instance.os,
1545 "%s-%s" % (old_name, instance.name), None)
1547 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1548 cwd=inst_os.path, output=logfile, reset_env=True)
1551 logging.error("os create command '%s' returned error: %s output: %s",
1552 result.cmd, result.fail_reason, result.output)
1553 lines = [utils.SafeEncode(val)
1554 for val in utils.TailFile(logfile, lines=20)]
1555 _Fail("OS rename script failed (%s), last lines in the"
1556 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1559 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1560 """Returns symlink path for block device.
1564 _dir = pathutils.DISK_LINKS_DIR
1566 return utils.PathJoin(_dir,
1568 (instance_name, constants.DISK_SEPARATOR, idx)))
1571 def _SymlinkBlockDev(instance_name, device_path, idx):
1572 """Set up symlinks to a instance's block device.
1574 This is an auxiliary function run when an instance is start (on the primary
1575 node) or when an instance is migrated (on the target node).
1578 @param instance_name: the name of the target instance
1579 @param device_path: path of the physical block device, on the node
1580 @param idx: the disk index
1581 @return: absolute path to the disk's symlink
1584 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1586 os.symlink(device_path, link_name)
1587 except OSError, err:
1588 if err.errno == errno.EEXIST:
1589 if (not os.path.islink(link_name) or
1590 os.readlink(link_name) != device_path):
1591 os.remove(link_name)
1592 os.symlink(device_path, link_name)
1599 def _RemoveBlockDevLinks(instance_name, disks):
1600 """Remove the block device symlinks belonging to the given instance.
1603 for idx, _ in enumerate(disks):
1604 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1605 if os.path.islink(link_name):
1607 os.remove(link_name)
1609 logging.exception("Can't remove symlink '%s'", link_name)
1612 def _GatherAndLinkBlockDevs(instance):
1613 """Set up an instance's block device(s).
1615 This is run on the primary node at instance startup. The block
1616 devices must be already assembled.
1618 @type instance: L{objects.Instance}
1619 @param instance: the instance whose disks we shoul assemble
1621 @return: list of (disk_object, device_path)
1625 for idx, disk in enumerate(instance.disks):
1626 device = _RecursiveFindBD(disk)
1628 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1632 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1634 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1637 block_devices.append((disk, link_name))
1639 return block_devices
1642 def StartInstance(instance, startup_paused, reason, store_reason=True):
1643 """Start an instance.
1645 @type instance: L{objects.Instance}
1646 @param instance: the instance object
1647 @type startup_paused: bool
1648 @param instance: pause instance at startup?
1649 @type reason: list of reasons
1650 @param reason: the reason trail for this startup
1651 @type store_reason: boolean
1652 @param store_reason: whether to store the shutdown reason trail on file
1656 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1659 if instance.name in running_instances:
1660 logging.info("Instance %s already running, not starting", instance.name)
1664 block_devices = _GatherAndLinkBlockDevs(instance)
1665 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1666 hyper.StartInstance(instance, block_devices, startup_paused)
1668 _StoreInstReasonTrail(instance.name, reason)
1669 except errors.BlockDeviceError, err:
1670 _Fail("Block device error: %s", err, exc=True)
1671 except errors.HypervisorError, err:
1672 _RemoveBlockDevLinks(instance.name, instance.disks)
1673 _Fail("Hypervisor error: %s", err, exc=True)
1676 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1677 """Shut an instance down.
1679 @note: this functions uses polling with a hardcoded timeout.
1681 @type instance: L{objects.Instance}
1682 @param instance: the instance object
1683 @type timeout: integer
1684 @param timeout: maximum timeout for soft shutdown
1685 @type reason: list of reasons
1686 @param reason: the reason trail for this shutdown
1687 @type store_reason: boolean
1688 @param store_reason: whether to store the shutdown reason trail on file
1692 hv_name = instance.hypervisor
1693 hyper = hypervisor.GetHypervisor(hv_name)
1694 iname = instance.name
1696 if instance.name not in hyper.ListInstances(instance.hvparams):
1697 logging.info("Instance %s not running, doing nothing", iname)
1702 self.tried_once = False
1705 if iname not in hyper.ListInstances(instance.hvparams):
1709 hyper.StopInstance(instance, retry=self.tried_once)
1711 _StoreInstReasonTrail(instance.name, reason)
1712 except errors.HypervisorError, err:
1713 if iname not in hyper.ListInstances(instance.hvparams):
1714 # if the instance is no longer existing, consider this a
1715 # success and go to cleanup
1718 _Fail("Failed to stop instance %s: %s", iname, err)
1720 self.tried_once = True
1722 raise utils.RetryAgain()
1725 utils.Retry(_TryShutdown(), 5, timeout)
1726 except utils.RetryTimeout:
1727 # the shutdown did not succeed
1728 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1731 hyper.StopInstance(instance, force=True)
1732 except errors.HypervisorError, err:
1733 if iname in hyper.ListInstances(instance.hvparams):
1734 # only raise an error if the instance still exists, otherwise
1735 # the error could simply be "instance ... unknown"!
1736 _Fail("Failed to force stop instance %s: %s", iname, err)
1740 if iname in hyper.ListInstances(instance.hvparams):
1741 _Fail("Could not shutdown instance %s even by destroy", iname)
1744 hyper.CleanupInstance(instance.name)
1745 except errors.HypervisorError, err:
1746 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1748 _RemoveBlockDevLinks(iname, instance.disks)
1751 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1752 """Reboot an instance.
1754 @type instance: L{objects.Instance}
1755 @param instance: the instance object to reboot
1756 @type reboot_type: str
1757 @param reboot_type: the type of reboot, one the following
1759 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1760 instance OS, do not recreate the VM
1761 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1762 restart the VM (at the hypervisor level)
1763 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1764 not accepted here, since that mode is handled differently, in
1765 cmdlib, and translates into full stop and start of the
1766 instance (instead of a call_instance_reboot RPC)
1767 @type shutdown_timeout: integer
1768 @param shutdown_timeout: maximum timeout for soft shutdown
1769 @type reason: list of reasons
1770 @param reason: the reason trail for this reboot
1774 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1777 if instance.name not in running_instances:
1778 _Fail("Cannot reboot instance %s that is not running", instance.name)
1780 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1781 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1783 hyper.RebootInstance(instance)
1784 except errors.HypervisorError, err:
1785 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1786 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1788 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1789 result = StartInstance(instance, False, reason, store_reason=False)
1790 _StoreInstReasonTrail(instance.name, reason)
1792 except errors.HypervisorError, err:
1793 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1795 _Fail("Invalid reboot_type received: %s", reboot_type)
1798 def InstanceBalloonMemory(instance, memory):
1799 """Resize an instance's memory.
1801 @type instance: L{objects.Instance}
1802 @param instance: the instance object
1804 @param memory: new memory amount in MB
1808 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1809 running = hyper.ListInstances(instance.hvparams)
1810 if instance.name not in running:
1811 logging.info("Instance %s is not running, cannot balloon", instance.name)
1814 hyper.BalloonInstanceMemory(instance, memory)
1815 except errors.HypervisorError, err:
1816 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1819 def MigrationInfo(instance):
1820 """Gather information about an instance to be migrated.
1822 @type instance: L{objects.Instance}
1823 @param instance: the instance definition
1826 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1828 info = hyper.MigrationInfo(instance)
1829 except errors.HypervisorError, err:
1830 _Fail("Failed to fetch migration information: %s", err, exc=True)
1834 def AcceptInstance(instance, info, target):
1835 """Prepare the node to accept an instance.
1837 @type instance: L{objects.Instance}
1838 @param instance: the instance definition
1839 @type info: string/data (opaque)
1840 @param info: migration information, from the source node
1841 @type target: string
1842 @param target: target host (usually ip), on this node
1845 # TODO: why is this required only for DTS_EXT_MIRROR?
1846 if instance.disk_template in constants.DTS_EXT_MIRROR:
1847 # Create the symlinks, as the disks are not active
1850 _GatherAndLinkBlockDevs(instance)
1851 except errors.BlockDeviceError, err:
1852 _Fail("Block device error: %s", err, exc=True)
1854 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1856 hyper.AcceptInstance(instance, info, target)
1857 except errors.HypervisorError, err:
1858 if instance.disk_template in constants.DTS_EXT_MIRROR:
1859 _RemoveBlockDevLinks(instance.name, instance.disks)
1860 _Fail("Failed to accept instance: %s", err, exc=True)
1863 def FinalizeMigrationDst(instance, info, success):
1864 """Finalize any preparation to accept an instance.
1866 @type instance: L{objects.Instance}
1867 @param instance: the instance definition
1868 @type info: string/data (opaque)
1869 @param info: migration information, from the source node
1870 @type success: boolean
1871 @param success: whether the migration was a success or a failure
1874 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1876 hyper.FinalizeMigrationDst(instance, info, success)
1877 except errors.HypervisorError, err:
1878 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1881 def MigrateInstance(cluster_name, instance, target, live):
1882 """Migrates an instance to another node.
1884 @type cluster_name: string
1885 @param cluster_name: name of the cluster
1886 @type instance: L{objects.Instance}
1887 @param instance: the instance definition
1888 @type target: string
1889 @param target: the target node name
1891 @param live: whether the migration should be done live or not (the
1892 interpretation of this parameter is left to the hypervisor)
1893 @raise RPCFail: if migration fails for some reason
1896 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1899 hyper.MigrateInstance(cluster_name, instance, target, live)
1900 except errors.HypervisorError, err:
1901 _Fail("Failed to migrate instance: %s", err, exc=True)
1904 def FinalizeMigrationSource(instance, success, live):
1905 """Finalize the instance migration on the source node.
1907 @type instance: L{objects.Instance}
1908 @param instance: the instance definition of the migrated instance
1910 @param success: whether the migration succeeded or not
1912 @param live: whether the user requested a live migration or not
1913 @raise RPCFail: If the execution fails for some reason
1916 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1919 hyper.FinalizeMigrationSource(instance, success, live)
1920 except Exception, err: # pylint: disable=W0703
1921 _Fail("Failed to finalize the migration on the source node: %s", err,
1925 def GetMigrationStatus(instance):
1926 """Get the migration status
1928 @type instance: L{objects.Instance}
1929 @param instance: the instance that is being migrated
1930 @rtype: L{objects.MigrationStatus}
1931 @return: the status of the current migration (one of
1932 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1933 progress info that can be retrieved from the hypervisor
1934 @raise RPCFail: If the migration status cannot be retrieved
1937 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1939 return hyper.GetMigrationStatus(instance)
1940 except Exception, err: # pylint: disable=W0703
1941 _Fail("Failed to get migration status: %s", err, exc=True)
1944 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1945 """Creates a block device for an instance.
1947 @type disk: L{objects.Disk}
1948 @param disk: the object describing the disk we should create
1950 @param size: the size of the physical underlying device, in MiB
1952 @param owner: the name of the instance for which disk is created,
1953 used for device cache data
1954 @type on_primary: boolean
1955 @param on_primary: indicates if it is the primary node or not
1957 @param info: string that will be sent to the physical device
1958 creation, used for example to set (LVM) tags on LVs
1959 @type excl_stor: boolean
1960 @param excl_stor: Whether exclusive_storage is active
1962 @return: the new unique_id of the device (this can sometime be
1963 computed only after creation), or None. On secondary nodes,
1964 it's not required to return anything.
1967 # TODO: remove the obsolete "size" argument
1968 # pylint: disable=W0613
1971 for child in disk.children:
1973 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1974 except errors.BlockDeviceError, err:
1975 _Fail("Can't assemble device %s: %s", child, err)
1976 if on_primary or disk.AssembleOnSecondary():
1977 # we need the children open in case the device itself has to
1980 # pylint: disable=E1103
1982 except errors.BlockDeviceError, err:
1983 _Fail("Can't make child '%s' read-write: %s", child, err)
1987 device = bdev.Create(disk, clist, excl_stor)
1988 except errors.BlockDeviceError, err:
1989 _Fail("Can't create block device: %s", err)
1991 if on_primary or disk.AssembleOnSecondary():
1994 except errors.BlockDeviceError, err:
1995 _Fail("Can't assemble device after creation, unusual event: %s", err)
1996 if on_primary or disk.OpenOnSecondary():
1998 device.Open(force=True)
1999 except errors.BlockDeviceError, err:
2000 _Fail("Can't make device r/w after creation, unusual event: %s", err)
2001 DevCacheManager.UpdateCache(device.dev_path, owner,
2002 on_primary, disk.iv_name)
2004 device.SetInfo(info)
2006 return device.unique_id
2009 def _WipeDevice(path, offset, size):
2010 """This function actually wipes the device.
2012 @param path: The path to the device to wipe
2013 @param offset: The offset in MiB in the file
2014 @param size: The size in MiB to write
2017 # Internal sizes are always in Mebibytes; if the following "dd" command
2018 # should use a different block size the offset and size given to this
2019 # function must be adjusted accordingly before being passed to "dd".
2020 block_size = 1024 * 1024
2022 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2023 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2025 result = utils.RunCmd(cmd)
2028 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2029 result.fail_reason, result.output)
2032 def BlockdevWipe(disk, offset, size):
2033 """Wipes a block device.
2035 @type disk: L{objects.Disk}
2036 @param disk: the disk object we want to wipe
2038 @param offset: The offset in MiB in the file
2040 @param size: The size in MiB to write
2044 rdev = _RecursiveFindBD(disk)
2045 except errors.BlockDeviceError:
2049 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2051 # Do cross verify some of the parameters
2053 _Fail("Negative offset")
2055 _Fail("Negative size")
2056 if offset > rdev.size:
2057 _Fail("Offset is bigger than device size")
2058 if (offset + size) > rdev.size:
2059 _Fail("The provided offset and size to wipe is bigger than device size")
2061 _WipeDevice(rdev.dev_path, offset, size)
2064 def BlockdevPauseResumeSync(disks, pause):
2065 """Pause or resume the sync of the block device.
2067 @type disks: list of L{objects.Disk}
2068 @param disks: the disks object we want to pause/resume
2070 @param pause: Wheater to pause or resume
2076 rdev = _RecursiveFindBD(disk)
2077 except errors.BlockDeviceError:
2081 success.append((False, ("Cannot change sync for device %s:"
2082 " device not found" % disk.iv_name)))
2085 result = rdev.PauseResumeSync(pause)
2088 success.append((result, None))
2094 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2099 def BlockdevRemove(disk):
2100 """Remove a block device.
2102 @note: This is intended to be called recursively.
2104 @type disk: L{objects.Disk}
2105 @param disk: the disk object we should remove
2107 @return: the success of the operation
2112 rdev = _RecursiveFindBD(disk)
2113 except errors.BlockDeviceError, err:
2114 # probably can't attach
2115 logging.info("Can't attach to device %s in remove", disk)
2117 if rdev is not None:
2118 r_path = rdev.dev_path
2121 except errors.BlockDeviceError, err:
2122 msgs.append(str(err))
2124 DevCacheManager.RemoveCache(r_path)
2127 for child in disk.children:
2129 BlockdevRemove(child)
2130 except RPCFail, err:
2131 msgs.append(str(err))
2134 _Fail("; ".join(msgs))
2137 def _RecursiveAssembleBD(disk, owner, as_primary):
2138 """Activate a block device for an instance.
2140 This is run on the primary and secondary nodes for an instance.
2142 @note: this function is called recursively.
2144 @type disk: L{objects.Disk}
2145 @param disk: the disk we try to assemble
2147 @param owner: the name of the instance which owns the disk
2148 @type as_primary: boolean
2149 @param as_primary: if we should make the block device
2152 @return: the assembled device or None (in case no device
2154 @raise errors.BlockDeviceError: in case there is an error
2155 during the activation of the children or the device
2161 mcn = disk.ChildrenNeeded()
2163 mcn = 0 # max number of Nones allowed
2165 mcn = len(disk.children) - mcn # max number of Nones
2166 for chld_disk in disk.children:
2168 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2169 except errors.BlockDeviceError, err:
2170 if children.count(None) >= mcn:
2173 logging.error("Error in child activation (but continuing): %s",
2175 children.append(cdev)
2177 if as_primary or disk.AssembleOnSecondary():
2178 r_dev = bdev.Assemble(disk, children)
2180 if as_primary or disk.OpenOnSecondary():
2182 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2183 as_primary, disk.iv_name)
2190 def BlockdevAssemble(disk, owner, as_primary, idx):
2191 """Activate a block device for an instance.
2193 This is a wrapper over _RecursiveAssembleBD.
2195 @rtype: str or boolean
2196 @return: a C{/dev/...} path for primary nodes, and
2197 C{True} for secondary nodes
2201 result = _RecursiveAssembleBD(disk, owner, as_primary)
2202 if isinstance(result, BlockDev):
2203 # pylint: disable=E1103
2204 result = result.dev_path
2206 _SymlinkBlockDev(owner, result, idx)
2207 except errors.BlockDeviceError, err:
2208 _Fail("Error while assembling disk: %s", err, exc=True)
2209 except OSError, err:
2210 _Fail("Error while symlinking disk: %s", err, exc=True)
2215 def BlockdevShutdown(disk):
2216 """Shut down a block device.
2218 First, if the device is assembled (Attach() is successful), then
2219 the device is shutdown. Then the children of the device are
2222 This function is called recursively. Note that we don't cache the
2223 children or such, as oppossed to assemble, shutdown of different
2224 devices doesn't require that the upper device was active.
2226 @type disk: L{objects.Disk}
2227 @param disk: the description of the disk we should
2233 r_dev = _RecursiveFindBD(disk)
2234 if r_dev is not None:
2235 r_path = r_dev.dev_path
2238 DevCacheManager.RemoveCache(r_path)
2239 except errors.BlockDeviceError, err:
2240 msgs.append(str(err))
2243 for child in disk.children:
2245 BlockdevShutdown(child)
2246 except RPCFail, err:
2247 msgs.append(str(err))
2250 _Fail("; ".join(msgs))
2253 def BlockdevAddchildren(parent_cdev, new_cdevs):
2254 """Extend a mirrored block device.
2256 @type parent_cdev: L{objects.Disk}
2257 @param parent_cdev: the disk to which we should add children
2258 @type new_cdevs: list of L{objects.Disk}
2259 @param new_cdevs: the list of children which we should add
2263 parent_bdev = _RecursiveFindBD(parent_cdev)
2264 if parent_bdev is None:
2265 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2266 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2267 if new_bdevs.count(None) > 0:
2268 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2269 parent_bdev.AddChildren(new_bdevs)
2272 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2273 """Shrink a mirrored block device.
2275 @type parent_cdev: L{objects.Disk}
2276 @param parent_cdev: the disk from which we should remove children
2277 @type new_cdevs: list of L{objects.Disk}
2278 @param new_cdevs: the list of children which we should remove
2282 parent_bdev = _RecursiveFindBD(parent_cdev)
2283 if parent_bdev is None:
2284 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2286 for disk in new_cdevs:
2287 rpath = disk.StaticDevPath()
2289 bd = _RecursiveFindBD(disk)
2291 _Fail("Can't find device %s while removing children", disk)
2293 devs.append(bd.dev_path)
2295 if not utils.IsNormAbsPath(rpath):
2296 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2298 parent_bdev.RemoveChildren(devs)
2301 def BlockdevGetmirrorstatus(disks):
2302 """Get the mirroring status of a list of devices.
2304 @type disks: list of L{objects.Disk}
2305 @param disks: the list of disks which we should query
2307 @return: List of L{objects.BlockDevStatus}, one for each disk
2308 @raise errors.BlockDeviceError: if any of the disks cannot be
2314 rbd = _RecursiveFindBD(dsk)
2316 _Fail("Can't find device %s", dsk)
2318 stats.append(rbd.CombinedSyncStatus())
2323 def BlockdevGetmirrorstatusMulti(disks):
2324 """Get the mirroring status of a list of devices.
2326 @type disks: list of L{objects.Disk}
2327 @param disks: the list of disks which we should query
2329 @return: List of tuples, (bool, status), one for each disk; bool denotes
2330 success/failure, status is L{objects.BlockDevStatus} on success, string
2337 rbd = _RecursiveFindBD(disk)
2339 result.append((False, "Can't find device %s" % disk))
2342 status = rbd.CombinedSyncStatus()
2343 except errors.BlockDeviceError, err:
2344 logging.exception("Error while getting disk status")
2345 result.append((False, str(err)))
2347 result.append((True, status))
2349 assert len(disks) == len(result)
2354 def _RecursiveFindBD(disk):
2355 """Check if a device is activated.
2357 If so, return information about the real device.
2359 @type disk: L{objects.Disk}
2360 @param disk: the disk object we need to find
2362 @return: None if the device can't be found,
2363 otherwise the device instance
2368 for chdisk in disk.children:
2369 children.append(_RecursiveFindBD(chdisk))
2371 return bdev.FindDevice(disk, children)
2374 def _OpenRealBD(disk):
2375 """Opens the underlying block device of a disk.
2377 @type disk: L{objects.Disk}
2378 @param disk: the disk object we want to open
2381 real_disk = _RecursiveFindBD(disk)
2382 if real_disk is None:
2383 _Fail("Block device '%s' is not set up", disk)
2390 def BlockdevFind(disk):
2391 """Check if a device is activated.
2393 If it is, return information about the real device.
2395 @type disk: L{objects.Disk}
2396 @param disk: the disk to find
2397 @rtype: None or objects.BlockDevStatus
2398 @return: None if the disk cannot be found, otherwise a the current
2403 rbd = _RecursiveFindBD(disk)
2404 except errors.BlockDeviceError, err:
2405 _Fail("Failed to find device: %s", err, exc=True)
2410 return rbd.GetSyncStatus()
2413 def BlockdevGetdimensions(disks):
2414 """Computes the size of the given disks.
2416 If a disk is not found, returns None instead.
2418 @type disks: list of L{objects.Disk}
2419 @param disks: the list of disk to compute the size for
2421 @return: list with elements None if the disk cannot be found,
2422 otherwise the pair (size, spindles), where spindles is None if the
2423 device doesn't support that
2429 rbd = _RecursiveFindBD(cf)
2430 except errors.BlockDeviceError:
2436 result.append(rbd.GetActualDimensions())
2440 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2441 """Export a block device to a remote node.
2443 @type disk: L{objects.Disk}
2444 @param disk: the description of the disk to export
2445 @type dest_node: str
2446 @param dest_node: the destination node to export to
2447 @type dest_path: str
2448 @param dest_path: the destination path on the target node
2449 @type cluster_name: str
2450 @param cluster_name: the cluster name, needed for SSH hostalias
2454 real_disk = _OpenRealBD(disk)
2456 # the block size on the read dd is 1MiB to match our units
2457 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2458 "dd if=%s bs=1048576 count=%s",
2459 real_disk.dev_path, str(disk.size))
2461 # we set here a smaller block size as, due to ssh buffering, more
2462 # than 64-128k will mostly ignored; we use nocreat to fail if the
2463 # device is not already there or we pass a wrong path; we use
2464 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2465 # to not buffer too much memory; this means that at best, we flush
2466 # every 64k, which will not be very fast
2467 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2468 " oflag=dsync", dest_path)
2470 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2471 constants.SSH_LOGIN_USER,
2474 # all commands have been checked, so we're safe to combine them
2475 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2477 result = utils.RunCmd(["bash", "-c", command])
2480 _Fail("Disk copy command '%s' returned error: %s"
2481 " output: %s", command, result.fail_reason, result.output)
2484 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2485 """Write a file to the filesystem.
2487 This allows the master to overwrite(!) a file. It will only perform
2488 the operation if the file belongs to a list of configuration files.
2490 @type file_name: str
2491 @param file_name: the target file name
2493 @param data: the new contents of the file
2495 @param mode: the mode to give the file (can be None)
2497 @param uid: the owner of the file
2499 @param gid: the group of the file
2501 @param atime: the atime to set on the file (can be None)
2503 @param mtime: the mtime to set on the file (can be None)
2507 file_name = vcluster.LocalizeVirtualPath(file_name)
2509 if not os.path.isabs(file_name):
2510 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2512 if file_name not in _ALLOWED_UPLOAD_FILES:
2513 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2516 raw_data = _Decompress(data)
2518 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2519 _Fail("Invalid username/groupname type")
2521 getents = runtime.GetEnts()
2522 uid = getents.LookupUser(uid)
2523 gid = getents.LookupGroup(gid)
2525 utils.SafeWriteFile(file_name, None,
2526 data=raw_data, mode=mode, uid=uid, gid=gid,
2527 atime=atime, mtime=mtime)
2530 def RunOob(oob_program, command, node, timeout):
2531 """Executes oob_program with given command on given node.
2533 @param oob_program: The path to the executable oob_program
2534 @param command: The command to invoke on oob_program
2535 @param node: The node given as an argument to the program
2536 @param timeout: Timeout after which we kill the oob program
2539 @raise RPCFail: If execution fails for some reason
2542 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2545 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2546 result.fail_reason, result.output)
2548 return result.stdout
2551 def _OSOndiskAPIVersion(os_dir):
2552 """Compute and return the API version of a given OS.
2554 This function will try to read the API version of the OS residing in
2555 the 'os_dir' directory.
2558 @param os_dir: the directory in which we should look for the OS
2560 @return: tuple (status, data) with status denoting the validity and
2561 data holding either the vaid versions or an error message
2564 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2567 st = os.stat(api_file)
2568 except EnvironmentError, err:
2569 return False, ("Required file '%s' not found under path %s: %s" %
2570 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2572 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2573 return False, ("File '%s' in %s is not a regular file" %
2574 (constants.OS_API_FILE, os_dir))
2577 api_versions = utils.ReadFile(api_file).splitlines()
2578 except EnvironmentError, err:
2579 return False, ("Error while reading the API version file at %s: %s" %
2580 (api_file, utils.ErrnoOrStr(err)))
2583 api_versions = [int(version.strip()) for version in api_versions]
2584 except (TypeError, ValueError), err:
2585 return False, ("API version(s) can't be converted to integer: %s" %
2588 return True, api_versions
2591 def DiagnoseOS(top_dirs=None):
2592 """Compute the validity for all OSes.
2594 @type top_dirs: list
2595 @param top_dirs: the list of directories in which to
2596 search (if not given defaults to
2597 L{pathutils.OS_SEARCH_PATH})
2598 @rtype: list of L{objects.OS}
2599 @return: a list of tuples (name, path, status, diagnose, variants,
2600 parameters, api_version) for all (potential) OSes under all
2601 search paths, where:
2602 - name is the (potential) OS name
2603 - path is the full path to the OS
2604 - status True/False is the validity of the OS
2605 - diagnose is the error message for an invalid OS, otherwise empty
2606 - variants is a list of supported OS variants, if any
2607 - parameters is a list of (name, help) parameters, if any
2608 - api_version is a list of support OS API versions
2611 if top_dirs is None:
2612 top_dirs = pathutils.OS_SEARCH_PATH
2615 for dir_name in top_dirs:
2616 if os.path.isdir(dir_name):
2618 f_names = utils.ListVisibleFiles(dir_name)
2619 except EnvironmentError, err:
2620 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2622 for name in f_names:
2623 os_path = utils.PathJoin(dir_name, name)
2624 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2627 variants = os_inst.supported_variants
2628 parameters = os_inst.supported_parameters
2629 api_versions = os_inst.api_versions
2632 variants = parameters = api_versions = []
2633 result.append((name, os_path, status, diagnose, variants,
2634 parameters, api_versions))
2639 def _TryOSFromDisk(name, base_dir=None):
2640 """Create an OS instance from disk.
2642 This function will return an OS instance if the given name is a
2645 @type base_dir: string
2646 @keyword base_dir: Base directory containing OS installations.
2647 Defaults to a search in all the OS_SEARCH_PATH dirs.
2649 @return: success and either the OS instance if we find a valid one,
2653 if base_dir is None:
2654 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2656 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2659 return False, "Directory for OS %s not found in search path" % name
2661 status, api_versions = _OSOndiskAPIVersion(os_dir)
2664 return status, api_versions
2666 if not constants.OS_API_VERSIONS.intersection(api_versions):
2667 return False, ("API version mismatch for path '%s': found %s, want %s." %
2668 (os_dir, api_versions, constants.OS_API_VERSIONS))
2670 # OS Files dictionary, we will populate it with the absolute path
2671 # names; if the value is True, then it is a required file, otherwise
2673 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2675 if max(api_versions) >= constants.OS_API_V15:
2676 os_files[constants.OS_VARIANTS_FILE] = False
2678 if max(api_versions) >= constants.OS_API_V20:
2679 os_files[constants.OS_PARAMETERS_FILE] = True
2681 del os_files[constants.OS_SCRIPT_VERIFY]
2683 for (filename, required) in os_files.items():
2684 os_files[filename] = utils.PathJoin(os_dir, filename)
2687 st = os.stat(os_files[filename])
2688 except EnvironmentError, err:
2689 if err.errno == errno.ENOENT and not required:
2690 del os_files[filename]
2692 return False, ("File '%s' under path '%s' is missing (%s)" %
2693 (filename, os_dir, utils.ErrnoOrStr(err)))
2695 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2696 return False, ("File '%s' under path '%s' is not a regular file" %
2699 if filename in constants.OS_SCRIPTS:
2700 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2701 return False, ("File '%s' under path '%s' is not executable" %
2705 if constants.OS_VARIANTS_FILE in os_files:
2706 variants_file = os_files[constants.OS_VARIANTS_FILE]
2709 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2710 except EnvironmentError, err:
2711 # we accept missing files, but not other errors
2712 if err.errno != errno.ENOENT:
2713 return False, ("Error while reading the OS variants file at %s: %s" %
2714 (variants_file, utils.ErrnoOrStr(err)))
2717 if constants.OS_PARAMETERS_FILE in os_files:
2718 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2720 parameters = utils.ReadFile(parameters_file).splitlines()
2721 except EnvironmentError, err:
2722 return False, ("Error while reading the OS parameters file at %s: %s" %
2723 (parameters_file, utils.ErrnoOrStr(err)))
2724 parameters = [v.split(None, 1) for v in parameters]
2726 os_obj = objects.OS(name=name, path=os_dir,
2727 create_script=os_files[constants.OS_SCRIPT_CREATE],
2728 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2729 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2730 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2731 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2733 supported_variants=variants,
2734 supported_parameters=parameters,
2735 api_versions=api_versions)
2739 def OSFromDisk(name, base_dir=None):
2740 """Create an OS instance from disk.
2742 This function will return an OS instance if the given name is a
2743 valid OS name. Otherwise, it will raise an appropriate
2744 L{RPCFail} exception, detailing why this is not a valid OS.
2746 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2747 an exception but returns true/false status data.
2749 @type base_dir: string
2750 @keyword base_dir: Base directory containing OS installations.
2751 Defaults to a search in all the OS_SEARCH_PATH dirs.
2752 @rtype: L{objects.OS}
2753 @return: the OS instance if we find a valid one
2754 @raise RPCFail: if we don't find a valid OS
2757 name_only = objects.OS.GetName(name)
2758 status, payload = _TryOSFromDisk(name_only, base_dir)
2766 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2767 """Calculate the basic environment for an os script.
2770 @param os_name: full operating system name (including variant)
2771 @type inst_os: L{objects.OS}
2772 @param inst_os: operating system for which the environment is being built
2773 @type os_params: dict
2774 @param os_params: the OS parameters
2775 @type debug: integer
2776 @param debug: debug level (0 or 1, for OS Api 10)
2778 @return: dict of environment variables
2779 @raise errors.BlockDeviceError: if the block device
2785 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2786 result["OS_API_VERSION"] = "%d" % api_version
2787 result["OS_NAME"] = inst_os.name
2788 result["DEBUG_LEVEL"] = "%d" % debug
2791 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2792 variant = objects.OS.GetVariant(os_name)
2794 variant = inst_os.supported_variants[0]
2797 result["OS_VARIANT"] = variant
2800 for pname, pvalue in os_params.items():
2801 result["OSP_%s" % pname.upper()] = pvalue
2803 # Set a default path otherwise programs called by OS scripts (or
2804 # even hooks called from OS scripts) might break, and we don't want
2805 # to have each script require setting a PATH variable
2806 result["PATH"] = constants.HOOKS_PATH
2811 def OSEnvironment(instance, inst_os, debug=0):
2812 """Calculate the environment for an os script.
2814 @type instance: L{objects.Instance}
2815 @param instance: target instance for the os script run
2816 @type inst_os: L{objects.OS}
2817 @param inst_os: operating system for which the environment is being built
2818 @type debug: integer
2819 @param debug: debug level (0 or 1, for OS Api 10)
2821 @return: dict of environment variables
2822 @raise errors.BlockDeviceError: if the block device
2826 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2828 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2829 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2831 result["HYPERVISOR"] = instance.hypervisor
2832 result["DISK_COUNT"] = "%d" % len(instance.disks)
2833 result["NIC_COUNT"] = "%d" % len(instance.nics)
2834 result["INSTANCE_SECONDARY_NODES"] = \
2835 ("%s" % " ".join(instance.secondary_nodes))
2838 for idx, disk in enumerate(instance.disks):
2839 real_disk = _OpenRealBD(disk)
2840 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2841 result["DISK_%d_ACCESS" % idx] = disk.mode
2842 result["DISK_%d_UUID" % idx] = disk.uuid
2844 result["DISK_%d_NAME" % idx] = disk.name
2845 if constants.HV_DISK_TYPE in instance.hvparams:
2846 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2847 instance.hvparams[constants.HV_DISK_TYPE]
2848 if disk.dev_type in constants.LDS_BLOCK:
2849 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2850 elif disk.dev_type == constants.LD_FILE:
2851 result["DISK_%d_BACKEND_TYPE" % idx] = \
2852 "file:%s" % disk.physical_id[0]
2855 for idx, nic in enumerate(instance.nics):
2856 result["NIC_%d_MAC" % idx] = nic.mac
2857 result["NIC_%d_UUID" % idx] = nic.uuid
2859 result["NIC_%d_NAME" % idx] = nic.name
2861 result["NIC_%d_IP" % idx] = nic.ip
2862 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2863 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2864 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2865 if nic.nicparams[constants.NIC_LINK]:
2866 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2868 nobj = objects.Network.FromDict(nic.netinfo)
2869 result.update(nobj.HooksDict("NIC_%d_" % idx))
2870 if constants.HV_NIC_TYPE in instance.hvparams:
2871 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2872 instance.hvparams[constants.HV_NIC_TYPE]
2875 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2876 for key, value in source.items():
2877 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2882 def DiagnoseExtStorage(top_dirs=None):
2883 """Compute the validity for all ExtStorage Providers.
2885 @type top_dirs: list
2886 @param top_dirs: the list of directories in which to
2887 search (if not given defaults to
2888 L{pathutils.ES_SEARCH_PATH})
2889 @rtype: list of L{objects.ExtStorage}
2890 @return: a list of tuples (name, path, status, diagnose, parameters)
2891 for all (potential) ExtStorage Providers under all
2892 search paths, where:
2893 - name is the (potential) ExtStorage Provider
2894 - path is the full path to the ExtStorage Provider
2895 - status True/False is the validity of the ExtStorage Provider
2896 - diagnose is the error message for an invalid ExtStorage Provider,
2898 - parameters is a list of (name, help) parameters, if any
2901 if top_dirs is None:
2902 top_dirs = pathutils.ES_SEARCH_PATH
2905 for dir_name in top_dirs:
2906 if os.path.isdir(dir_name):
2908 f_names = utils.ListVisibleFiles(dir_name)
2909 except EnvironmentError, err:
2910 logging.exception("Can't list the ExtStorage directory %s: %s",
2913 for name in f_names:
2914 es_path = utils.PathJoin(dir_name, name)
2915 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2918 parameters = es_inst.supported_parameters
2922 result.append((name, es_path, status, diagnose, parameters))
2927 def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
2928 """Grow a stack of block devices.
2930 This function is called recursively, with the childrens being the
2931 first ones to resize.
2933 @type disk: L{objects.Disk}
2934 @param disk: the disk to be grown
2935 @type amount: integer
2936 @param amount: the amount (in mebibytes) to grow with
2937 @type dryrun: boolean
2938 @param dryrun: whether to execute the operation in simulation mode
2939 only, without actually increasing the size
2940 @param backingstore: whether to execute the operation on backing storage
2941 only, or on "logical" storage only; e.g. DRBD is logical storage,
2942 whereas LVM, file, RBD are backing storage
2943 @rtype: (status, result)
2944 @type excl_stor: boolean
2945 @param excl_stor: Whether exclusive_storage is active
2946 @return: a tuple with the status of the operation (True/False), and
2947 the errors message if status is False
2950 r_dev = _RecursiveFindBD(disk)
2952 _Fail("Cannot find block device %s", disk)
2955 r_dev.Grow(amount, dryrun, backingstore, excl_stor)
2956 except errors.BlockDeviceError, err:
2957 _Fail("Failed to grow block device: %s", err, exc=True)
2960 def BlockdevSnapshot(disk):
2961 """Create a snapshot copy of a block device.
2963 This function is called recursively, and the snapshot is actually created
2964 just for the leaf lvm backend device.
2966 @type disk: L{objects.Disk}
2967 @param disk: the disk to be snapshotted
2969 @return: snapshot disk ID as (vg, lv)
2972 if disk.dev_type == constants.LD_DRBD8:
2973 if not disk.children:
2974 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2976 return BlockdevSnapshot(disk.children[0])
2977 elif disk.dev_type == constants.LD_LV:
2978 r_dev = _RecursiveFindBD(disk)
2979 if r_dev is not None:
2980 # FIXME: choose a saner value for the snapshot size
2981 # let's stay on the safe side and ask for the full size, for now
2982 return r_dev.Snapshot(disk.size)
2984 _Fail("Cannot find block device %s", disk)
2986 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2987 disk.unique_id, disk.dev_type)
2990 def BlockdevSetInfo(disk, info):
2991 """Sets 'metadata' information on block devices.
2993 This function sets 'info' metadata on block devices. Initial
2994 information is set at device creation; this function should be used
2995 for example after renames.
2997 @type disk: L{objects.Disk}
2998 @param disk: the disk to be grown
3000 @param info: new 'info' metadata
3001 @rtype: (status, result)
3002 @return: a tuple with the status of the operation (True/False), and
3003 the errors message if status is False
3006 r_dev = _RecursiveFindBD(disk)
3008 _Fail("Cannot find block device %s", disk)
3012 except errors.BlockDeviceError, err:
3013 _Fail("Failed to set information on block device: %s", err, exc=True)
3016 def FinalizeExport(instance, snap_disks):
3017 """Write out the export configuration information.
3019 @type instance: L{objects.Instance}
3020 @param instance: the instance which we export, used for
3021 saving configuration
3022 @type snap_disks: list of L{objects.Disk}
3023 @param snap_disks: list of snapshot block devices, which
3024 will be used to get the actual name of the dump file
3029 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3030 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3032 config = objects.SerializableConfigParser()
3034 config.add_section(constants.INISECT_EXP)
3035 config.set(constants.INISECT_EXP, "version", "0")
3036 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3037 config.set(constants.INISECT_EXP, "source", instance.primary_node)
3038 config.set(constants.INISECT_EXP, "os", instance.os)
3039 config.set(constants.INISECT_EXP, "compression", "none")
3041 config.add_section(constants.INISECT_INS)
3042 config.set(constants.INISECT_INS, "name", instance.name)
3043 config.set(constants.INISECT_INS, "maxmem", "%d" %
3044 instance.beparams[constants.BE_MAXMEM])
3045 config.set(constants.INISECT_INS, "minmem", "%d" %
3046 instance.beparams[constants.BE_MINMEM])
3047 # "memory" is deprecated, but useful for exporting to old ganeti versions
3048 config.set(constants.INISECT_INS, "memory", "%d" %
3049 instance.beparams[constants.BE_MAXMEM])
3050 config.set(constants.INISECT_INS, "vcpus", "%d" %
3051 instance.beparams[constants.BE_VCPUS])
3052 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3053 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3054 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3057 for nic_count, nic in enumerate(instance.nics):
3059 config.set(constants.INISECT_INS, "nic%d_mac" %
3060 nic_count, "%s" % nic.mac)
3061 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3062 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3064 for param in constants.NICS_PARAMETER_TYPES:
3065 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3066 "%s" % nic.nicparams.get(param, None))
3067 # TODO: redundant: on load can read nics until it doesn't exist
3068 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3071 for disk_count, disk in enumerate(snap_disks):
3074 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3075 ("%s" % disk.iv_name))
3076 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3077 ("%s" % disk.physical_id[1]))
3078 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3081 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3083 # New-style hypervisor/backend parameters
3085 config.add_section(constants.INISECT_HYP)
3086 for name, value in instance.hvparams.items():
3087 if name not in constants.HVC_GLOBALS:
3088 config.set(constants.INISECT_HYP, name, str(value))
3090 config.add_section(constants.INISECT_BEP)
3091 for name, value in instance.beparams.items():
3092 config.set(constants.INISECT_BEP, name, str(value))
3094 config.add_section(constants.INISECT_OSP)
3095 for name, value in instance.osparams.items():
3096 config.set(constants.INISECT_OSP, name, str(value))
3098 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3099 data=config.Dumps())
3100 shutil.rmtree(finaldestdir, ignore_errors=True)
3101 shutil.move(destdir, finaldestdir)
3104 def ExportInfo(dest):
3105 """Get export configuration information.
3108 @param dest: directory containing the export
3110 @rtype: L{objects.SerializableConfigParser}
3111 @return: a serializable config file containing the
3115 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3117 config = objects.SerializableConfigParser()
3120 if (not config.has_section(constants.INISECT_EXP) or
3121 not config.has_section(constants.INISECT_INS)):
3122 _Fail("Export info file doesn't have the required fields")
3124 return config.Dumps()
3128 """Return a list of exports currently available on this machine.
3131 @return: list of the exports
3134 if os.path.isdir(pathutils.EXPORT_DIR):
3135 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3137 _Fail("No exports directory")
3140 def RemoveExport(export):
3141 """Remove an existing export from the node.
3144 @param export: the name of the export to remove
3148 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3151 shutil.rmtree(target)
3152 except EnvironmentError, err:
3153 _Fail("Error while removing the export: %s", err, exc=True)
3156 def BlockdevRename(devlist):
3157 """Rename a list of block devices.
3159 @type devlist: list of tuples
3160 @param devlist: list of tuples of the form (disk,
3161 new_logical_id, new_physical_id); disk is an
3162 L{objects.Disk} object describing the current disk,
3163 and new logical_id/physical_id is the name we
3166 @return: True if all renames succeeded, False otherwise
3171 for disk, unique_id in devlist:
3172 dev = _RecursiveFindBD(disk)
3174 msgs.append("Can't find device %s in rename" % str(disk))
3178 old_rpath = dev.dev_path
3179 dev.Rename(unique_id)
3180 new_rpath = dev.dev_path
3181 if old_rpath != new_rpath:
3182 DevCacheManager.RemoveCache(old_rpath)
3183 # FIXME: we should add the new cache information here, like:
3184 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3185 # but we don't have the owner here - maybe parse from existing
3186 # cache? for now, we only lose lvm data when we rename, which
3187 # is less critical than DRBD or MD
3188 except errors.BlockDeviceError, err:
3189 msgs.append("Can't rename device '%s' to '%s': %s" %
3190 (dev, unique_id, err))
3191 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3194 _Fail("; ".join(msgs))
3197 def _TransformFileStorageDir(fs_dir):
3198 """Checks whether given file_storage_dir is valid.
3200 Checks wheter the given fs_dir is within the cluster-wide default
3201 file_storage_dir or the shared_file_storage_dir, which are stored in
3202 SimpleStore. Only paths under those directories are allowed.
3205 @param fs_dir: the path to check
3207 @return: the normalized path if valid, None otherwise
3210 filestorage.CheckFileStoragePath(fs_dir)
3212 return os.path.normpath(fs_dir)
3215 def CreateFileStorageDir(file_storage_dir):
3216 """Create file storage directory.
3218 @type file_storage_dir: str
3219 @param file_storage_dir: directory to create
3222 @return: tuple with first element a boolean indicating wheter dir
3223 creation was successful or not
3226 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3227 if os.path.exists(file_storage_dir):
3228 if not os.path.isdir(file_storage_dir):
3229 _Fail("Specified storage dir '%s' is not a directory",
3233 os.makedirs(file_storage_dir, 0750)
3234 except OSError, err:
3235 _Fail("Cannot create file storage directory '%s': %s",
3236 file_storage_dir, err, exc=True)
3239 def RemoveFileStorageDir(file_storage_dir):
3240 """Remove file storage directory.
3242 Remove it only if it's empty. If not log an error and return.
3244 @type file_storage_dir: str
3245 @param file_storage_dir: the directory we should cleanup
3246 @rtype: tuple (success,)
3247 @return: tuple of one element, C{success}, denoting
3248 whether the operation was successful
3251 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3252 if os.path.exists(file_storage_dir):
3253 if not os.path.isdir(file_storage_dir):
3254 _Fail("Specified Storage directory '%s' is not a directory",
3256 # deletes dir only if empty, otherwise we want to fail the rpc call
3258 os.rmdir(file_storage_dir)
3259 except OSError, err:
3260 _Fail("Cannot remove file storage directory '%s': %s",
3261 file_storage_dir, err)
3264 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3265 """Rename the file storage directory.
3267 @type old_file_storage_dir: str
3268 @param old_file_storage_dir: the current path
3269 @type new_file_storage_dir: str
3270 @param new_file_storage_dir: the name we should rename to
3271 @rtype: tuple (success,)
3272 @return: tuple of one element, C{success}, denoting
3273 whether the operation was successful
3276 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3277 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3278 if not os.path.exists(new_file_storage_dir):
3279 if os.path.isdir(old_file_storage_dir):
3281 os.rename(old_file_storage_dir, new_file_storage_dir)
3282 except OSError, err:
3283 _Fail("Cannot rename '%s' to '%s': %s",
3284 old_file_storage_dir, new_file_storage_dir, err)
3286 _Fail("Specified storage dir '%s' is not a directory",
3287 old_file_storage_dir)
3289 if os.path.exists(old_file_storage_dir):
3290 _Fail("Cannot rename '%s' to '%s': both locations exist",
3291 old_file_storage_dir, new_file_storage_dir)
3294 def _EnsureJobQueueFile(file_name):
3295 """Checks whether the given filename is in the queue directory.
3297 @type file_name: str
3298 @param file_name: the file name we should check
3300 @raises RPCFail: if the file is not valid
3303 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3304 _Fail("Passed job queue file '%s' does not belong to"
3305 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3308 def JobQueueUpdate(file_name, content):
3309 """Updates a file in the queue directory.
3311 This is just a wrapper over L{utils.io.WriteFile}, with proper
3314 @type file_name: str
3315 @param file_name: the job file name
3317 @param content: the new job contents
3319 @return: the success of the operation
3322 file_name = vcluster.LocalizeVirtualPath(file_name)
3324 _EnsureJobQueueFile(file_name)
3325 getents = runtime.GetEnts()
3327 # Write and replace the file atomically
3328 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3329 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3332 def JobQueueRename(old, new):
3333 """Renames a job queue file.
3335 This is just a wrapper over os.rename with proper checking.
3338 @param old: the old (actual) file name
3340 @param new: the desired file name
3342 @return: the success of the operation and payload
3345 old = vcluster.LocalizeVirtualPath(old)
3346 new = vcluster.LocalizeVirtualPath(new)
3348 _EnsureJobQueueFile(old)
3349 _EnsureJobQueueFile(new)
3351 getents = runtime.GetEnts()
3353 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3354 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3357 def BlockdevClose(instance_name, disks):
3358 """Closes the given block devices.
3360 This means they will be switched to secondary mode (in case of
3363 @param instance_name: if the argument is not empty, the symlinks
3364 of this instance will be removed
3365 @type disks: list of L{objects.Disk}
3366 @param disks: the list of disks to be closed
3367 @rtype: tuple (success, message)
3368 @return: a tuple of success and message, where success
3369 indicates the succes of the operation, and message
3370 which will contain the error details in case we
3376 rd = _RecursiveFindBD(cf)
3378 _Fail("Can't find device %s", cf)
3385 except errors.BlockDeviceError, err:
3386 msg.append(str(err))
3388 _Fail("Can't make devices secondary: %s", ",".join(msg))
3391 _RemoveBlockDevLinks(instance_name, disks)
3394 def ValidateHVParams(hvname, hvparams):
3395 """Validates the given hypervisor parameters.
3397 @type hvname: string
3398 @param hvname: the hypervisor name
3399 @type hvparams: dict
3400 @param hvparams: the hypervisor parameters to be validated
3405 hv_type = hypervisor.GetHypervisor(hvname)
3406 hv_type.ValidateParameters(hvparams)
3407 except errors.HypervisorError, err:
3408 _Fail(str(err), log=False)
3411 def _CheckOSPList(os_obj, parameters):
3412 """Check whether a list of parameters is supported by the OS.
3414 @type os_obj: L{objects.OS}
3415 @param os_obj: OS object to check
3416 @type parameters: list
3417 @param parameters: the list of parameters to check
3420 supported = [v[0] for v in os_obj.supported_parameters]
3421 delta = frozenset(parameters).difference(supported)
3423 _Fail("The following parameters are not supported"
3424 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3427 def ValidateOS(required, osname, checks, osparams):
3428 """Validate the given OS' parameters.
3430 @type required: boolean
3431 @param required: whether absence of the OS should translate into
3433 @type osname: string
3434 @param osname: the OS to be validated
3436 @param checks: list of the checks to run (currently only 'parameters')
3437 @type osparams: dict
3438 @param osparams: dictionary with OS parameters
3440 @return: True if the validation passed, or False if the OS was not
3441 found and L{required} was false
3444 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3445 _Fail("Unknown checks required for OS %s: %s", osname,
3446 set(checks).difference(constants.OS_VALIDATE_CALLS))
3448 name_only = objects.OS.GetName(osname)
3449 status, tbv = _TryOSFromDisk(name_only, None)
3457 if max(tbv.api_versions) < constants.OS_API_V20:
3460 if constants.OS_VALIDATE_PARAMETERS in checks:
3461 _CheckOSPList(tbv, osparams.keys())
3463 validate_env = OSCoreEnv(osname, tbv, osparams)
3464 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3465 cwd=tbv.path, reset_env=True)
3467 logging.error("os validate command '%s' returned error: %s output: %s",
3468 result.cmd, result.fail_reason, result.output)
3469 _Fail("OS validation script failed (%s), output: %s",
3470 result.fail_reason, result.output, log=False)
3476 """Demotes the current node from master candidate role.
3479 # try to ensure we're not the master by mistake
3480 master, myself = ssconf.GetMasterAndMyself()
3481 if master == myself:
3482 _Fail("ssconf status shows I'm the master node, will not demote")
3484 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3485 if not result.failed:
3486 _Fail("The master daemon is running, will not demote")
3489 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3490 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3491 except EnvironmentError, err:
3492 if err.errno != errno.ENOENT:
3493 _Fail("Error while backing up cluster file: %s", err, exc=True)
3495 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3498 def _GetX509Filenames(cryptodir, name):
3499 """Returns the full paths for the private key and certificate.
3502 return (utils.PathJoin(cryptodir, name),
3503 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3504 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3507 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3508 """Creates a new X509 certificate for SSL/TLS.
3511 @param validity: Validity in seconds
3512 @rtype: tuple; (string, string)
3513 @return: Certificate name and public part
3516 (key_pem, cert_pem) = \
3517 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3518 min(validity, _MAX_SSL_CERT_VALIDITY))
3520 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3521 prefix="x509-%s-" % utils.TimestampForFilename())
3523 name = os.path.basename(cert_dir)
3524 assert len(name) > 5
3526 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3528 utils.WriteFile(key_file, mode=0400, data=key_pem)
3529 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3531 # Never return private key as it shouldn't leave the node
3532 return (name, cert_pem)
3534 shutil.rmtree(cert_dir, ignore_errors=True)
3538 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3539 """Removes a X509 certificate.
3542 @param name: Certificate name
3545 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3547 utils.RemoveFile(key_file)
3548 utils.RemoveFile(cert_file)
3552 except EnvironmentError, err:
3553 _Fail("Cannot remove certificate directory '%s': %s",
3557 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3558 """Returns the command for the requested input/output.
3560 @type instance: L{objects.Instance}
3561 @param instance: The instance object
3562 @param mode: Import/export mode
3563 @param ieio: Input/output type
3564 @param ieargs: Input/output arguments
3567 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3574 if ieio == constants.IEIO_FILE:
3575 (filename, ) = ieargs
3577 if not utils.IsNormAbsPath(filename):
3578 _Fail("Path '%s' is not normalized or absolute", filename)
3580 real_filename = os.path.realpath(filename)
3581 directory = os.path.dirname(real_filename)
3583 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3584 _Fail("File '%s' is not under exports directory '%s': %s",
3585 filename, pathutils.EXPORT_DIR, real_filename)
3588 utils.Makedirs(directory, mode=0750)
3590 quoted_filename = utils.ShellQuote(filename)
3592 if mode == constants.IEM_IMPORT:
3593 suffix = "> %s" % quoted_filename
3594 elif mode == constants.IEM_EXPORT:
3595 suffix = "< %s" % quoted_filename
3597 # Retrieve file size
3599 st = os.stat(filename)
3600 except EnvironmentError, err:
3601 logging.error("Can't stat(2) %s: %s", filename, err)
3603 exp_size = utils.BytesToMebibyte(st.st_size)
3605 elif ieio == constants.IEIO_RAW_DISK:
3608 real_disk = _OpenRealBD(disk)
3610 if mode == constants.IEM_IMPORT:
3611 # we set here a smaller block size as, due to transport buffering, more
3612 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3613 # is not already there or we pass a wrong path; we use notrunc to no
3614 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3615 # much memory; this means that at best, we flush every 64k, which will
3617 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3618 " bs=%s oflag=dsync"),
3622 elif mode == constants.IEM_EXPORT:
3623 # the block size on the read dd is 1MiB to match our units
3624 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3626 str(1024 * 1024), # 1 MB
3628 exp_size = disk.size
3630 elif ieio == constants.IEIO_SCRIPT:
3631 (disk, disk_index, ) = ieargs
3633 assert isinstance(disk_index, (int, long))
3635 real_disk = _OpenRealBD(disk)
3637 inst_os = OSFromDisk(instance.os)
3638 env = OSEnvironment(instance, inst_os)
3640 if mode == constants.IEM_IMPORT:
3641 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3642 env["IMPORT_INDEX"] = str(disk_index)
3643 script = inst_os.import_script
3645 elif mode == constants.IEM_EXPORT:
3646 env["EXPORT_DEVICE"] = real_disk.dev_path
3647 env["EXPORT_INDEX"] = str(disk_index)
3648 script = inst_os.export_script
3650 # TODO: Pass special environment only to script
3651 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3653 if mode == constants.IEM_IMPORT:
3654 suffix = "| %s" % script_cmd
3656 elif mode == constants.IEM_EXPORT:
3657 prefix = "%s |" % script_cmd
3659 # Let script predict size
3660 exp_size = constants.IE_CUSTOM_SIZE
3663 _Fail("Invalid %s I/O mode %r", mode, ieio)
3665 return (env, prefix, suffix, exp_size)
3668 def _CreateImportExportStatusDir(prefix):
3669 """Creates status directory for import/export.
3672 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3674 (prefix, utils.TimestampForFilename())))
3677 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3679 """Starts an import or export daemon.
3681 @param mode: Import/output mode
3682 @type opts: L{objects.ImportExportOptions}
3683 @param opts: Daemon options
3685 @param host: Remote host for export (None for import)
3687 @param port: Remote port for export (None for import)
3688 @type instance: L{objects.Instance}
3689 @param instance: Instance object
3690 @type component: string
3691 @param component: which part of the instance is transferred now,
3693 @param ieio: Input/output type
3694 @param ieioargs: Input/output arguments
3697 if mode == constants.IEM_IMPORT:
3700 if not (host is None and port is None):
3701 _Fail("Can not specify host or port on import")
3703 elif mode == constants.IEM_EXPORT:
3706 if host is None or port is None:
3707 _Fail("Host and port must be specified for an export")
3710 _Fail("Invalid mode %r", mode)
3712 if (opts.key_name is None) ^ (opts.ca_pem is None):
3713 _Fail("Cluster certificate can only be used for both key and CA")
3715 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3716 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3718 if opts.key_name is None:
3720 key_path = pathutils.NODED_CERT_FILE
3721 cert_path = pathutils.NODED_CERT_FILE
3722 assert opts.ca_pem is None
3724 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3726 assert opts.ca_pem is not None
3728 for i in [key_path, cert_path]:
3729 if not os.path.exists(i):
3730 _Fail("File '%s' does not exist" % i)
3732 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3734 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3735 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3736 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3738 if opts.ca_pem is None:
3740 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3745 utils.WriteFile(ca_file, data=ca, mode=0400)
3748 pathutils.IMPORT_EXPORT_DAEMON,
3750 "--key=%s" % key_path,
3751 "--cert=%s" % cert_path,
3752 "--ca=%s" % ca_file,
3756 cmd.append("--host=%s" % host)
3759 cmd.append("--port=%s" % port)
3762 cmd.append("--ipv6")
3764 cmd.append("--ipv4")
3767 cmd.append("--compress=%s" % opts.compress)
3770 cmd.append("--magic=%s" % opts.magic)
3772 if exp_size is not None:
3773 cmd.append("--expected-size=%s" % exp_size)
3776 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3779 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3781 if mode == constants.IEM_EXPORT:
3782 # Retry connection a few times when connecting to remote peer
3783 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3784 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3785 elif opts.connect_timeout is not None:
3786 assert mode == constants.IEM_IMPORT
3787 # Overall timeout for establishing connection while listening
3788 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3790 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3792 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3793 # support for receiving a file descriptor for output
3794 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3797 # The import/export name is simply the status directory name
3798 return os.path.basename(status_dir)
3801 shutil.rmtree(status_dir, ignore_errors=True)
3805 def GetImportExportStatus(names):
3806 """Returns import/export daemon status.
3808 @type names: sequence
3809 @param names: List of names
3810 @rtype: List of dicts
3811 @return: Returns a list of the state of each named import/export or None if a
3812 status couldn't be read
3818 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3822 data = utils.ReadFile(status_file)
3823 except EnvironmentError, err:
3824 if err.errno != errno.ENOENT:
3832 result.append(serializer.LoadJson(data))
3837 def AbortImportExport(name):
3838 """Sends SIGTERM to a running import/export daemon.
3841 logging.info("Abort import/export %s", name)
3843 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3844 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3847 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3849 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3852 def CleanupImportExport(name):
3853 """Cleanup after an import or export.
3855 If the import/export daemon is still running it's killed. Afterwards the
3856 whole status directory is removed.
3859 logging.info("Finalizing import/export %s", name)
3861 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3863 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3866 logging.info("Import/export %s is still running with PID %s",
3868 utils.KillProcess(pid, waitpid=False)
3870 shutil.rmtree(status_dir, ignore_errors=True)
3873 def _SetPhysicalId(target_node_uuid, nodes_ip, disks):
3874 """Sets the correct physical ID on all passed disks.
3878 cf.SetPhysicalID(target_node_uuid, nodes_ip)
3881 def _FindDisks(target_node_uuid, nodes_ip, disks):
3882 """Sets the physical ID on disks and returns the block devices.
3885 _SetPhysicalId(target_node_uuid, nodes_ip, disks)
3890 rd = _RecursiveFindBD(cf)
3892 _Fail("Can't find device %s", cf)
3897 def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
3898 """Disconnects the network on a list of drbd devices.
3901 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3907 except errors.BlockDeviceError, err:
3908 _Fail("Can't change network configuration to standalone mode: %s",
3912 def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
3914 """Attaches the network on a list of drbd devices.
3917 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3920 for idx, rd in enumerate(bdevs):
3922 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3923 except EnvironmentError, err:
3924 _Fail("Can't create symlink: %s", err)
3925 # reconnect disks, switch to new master configuration and if
3926 # needed primary mode
3929 rd.AttachNet(multimaster)
3930 except errors.BlockDeviceError, err:
3931 _Fail("Can't change network configuration: %s", err)
3933 # wait until the disks are connected; we need to retry the re-attach
3934 # if the device becomes standalone, as this might happen if the one
3935 # node disconnects and reconnects in a different mode before the
3936 # other node reconnects; in this case, one or both of the nodes will
3937 # decide it has wrong configuration and switch to standalone
3940 all_connected = True
3943 stats = rd.GetProcStatus()
3945 all_connected = (all_connected and
3946 (stats.is_connected or stats.is_in_resync))
3948 if stats.is_standalone:
3949 # peer had different config info and this node became
3950 # standalone, even though this should not happen with the
3951 # new staged way of changing disk configs
3953 rd.AttachNet(multimaster)
3954 except errors.BlockDeviceError, err:
3955 _Fail("Can't change network configuration: %s", err)
3957 if not all_connected:
3958 raise utils.RetryAgain()
3961 # Start with a delay of 100 miliseconds and go up to 5 seconds
3962 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3963 except utils.RetryTimeout:
3964 _Fail("Timeout in disk reconnecting")
3967 # change to primary mode
3971 except errors.BlockDeviceError, err:
3972 _Fail("Can't change to primary mode: %s", err)
3975 def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
3976 """Wait until DRBDs have synchronized.
3980 stats = rd.GetProcStatus()
3981 if not (stats.is_connected or stats.is_in_resync):
3982 raise utils.RetryAgain()
3985 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3991 # poll each second for 15 seconds
3992 stats = utils.Retry(_helper, 1, 15, args=[rd])
3993 except utils.RetryTimeout:
3994 stats = rd.GetProcStatus()
3996 if not (stats.is_connected or stats.is_in_resync):
3997 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3998 alldone = alldone and (not stats.is_in_resync)
3999 if stats.sync_percent is not None:
4000 min_resync = min(min_resync, stats.sync_percent)
4002 return (alldone, min_resync)
4005 def DrbdNeedsActivation(target_node_uuid, nodes_ip, disks):
4006 """Checks which of the passed disks needs activation and returns their UUIDs.
4009 _SetPhysicalId(target_node_uuid, nodes_ip, disks)
4013 rd = _RecursiveFindBD(disk)
4015 faulty_disks.append(disk)
4018 stats = rd.GetProcStatus()
4019 if stats.is_standalone or stats.is_diskless:
4020 faulty_disks.append(disk)
4022 return [disk.uuid for disk in faulty_disks]
4025 def GetDrbdUsermodeHelper():
4026 """Returns DRBD usermode helper currently configured.
4030 return drbd.DRBD8.GetUsermodeHelper()
4031 except errors.BlockDeviceError, err:
4035 def PowercycleNode(hypervisor_type, hvparams=None):
4036 """Hard-powercycle the node.
4038 Because we need to return first, and schedule the powercycle in the
4039 background, we won't be able to report failures nicely.
4042 hyper = hypervisor.GetHypervisor(hypervisor_type)
4046 # if we can't fork, we'll pretend that we're in the child process
4049 return "Reboot scheduled in 5 seconds"
4050 # ensure the child is running on ram
4053 except Exception: # pylint: disable=W0703
4056 hyper.PowercycleNode(hvparams=hvparams)
4059 def _VerifyRestrictedCmdName(cmd):
4060 """Verifies a restricted command name.
4063 @param cmd: Command name
4064 @rtype: tuple; (boolean, string or None)
4065 @return: The tuple's first element is the status; if C{False}, the second
4066 element is an error message string, otherwise it's C{None}
4070 return (False, "Missing command name")
4072 if os.path.basename(cmd) != cmd:
4073 return (False, "Invalid command name")
4075 if not constants.EXT_PLUGIN_MASK.match(cmd):
4076 return (False, "Command name contains forbidden characters")
4081 def _CommonRestrictedCmdCheck(path, owner):
4082 """Common checks for restricted command file system directories and files.
4085 @param path: Path to check
4086 @param owner: C{None} or tuple containing UID and GID
4087 @rtype: tuple; (boolean, string or C{os.stat} result)
4088 @return: The tuple's first element is the status; if C{False}, the second
4089 element is an error message string, otherwise it's the result of C{os.stat}
4093 # Default to root as owner
4098 except EnvironmentError, err:
4099 return (False, "Can't stat(2) '%s': %s" % (path, err))
4101 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4102 return (False, "Permissions on '%s' are too permissive" % path)
4104 if (st.st_uid, st.st_gid) != owner:
4105 (owner_uid, owner_gid) = owner
4106 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4111 def _VerifyRestrictedCmdDirectory(path, _owner=None):
4112 """Verifies restricted command directory.
4115 @param path: Path to check
4116 @rtype: tuple; (boolean, string or None)
4117 @return: The tuple's first element is the status; if C{False}, the second
4118 element is an error message string, otherwise it's C{None}
4121 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4124 return (False, value)
4126 if not stat.S_ISDIR(value.st_mode):
4127 return (False, "Path '%s' is not a directory" % path)
4132 def _VerifyRestrictedCmd(path, cmd, _owner=None):
4133 """Verifies a whole restricted command and returns its executable filename.
4136 @param path: Directory containing restricted commands
4138 @param cmd: Command name
4139 @rtype: tuple; (boolean, string)
4140 @return: The tuple's first element is the status; if C{False}, the second
4141 element is an error message string, otherwise the second element is the
4142 absolute path to the executable
4145 executable = utils.PathJoin(path, cmd)
4147 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4152 if not utils.IsExecutable(executable):
4153 return (False, "access(2) thinks '%s' can't be executed" % executable)
4155 return (True, executable)
4158 def _PrepareRestrictedCmd(path, cmd,
4159 _verify_dir=_VerifyRestrictedCmdDirectory,
4160 _verify_name=_VerifyRestrictedCmdName,
4161 _verify_cmd=_VerifyRestrictedCmd):
4162 """Performs a number of tests on a restricted command.
4165 @param path: Directory containing restricted commands
4167 @param cmd: Command name
4168 @return: Same as L{_VerifyRestrictedCmd}
4171 # Verify the directory first
4172 (status, msg) = _verify_dir(path)
4174 # Check command if everything was alright
4175 (status, msg) = _verify_name(cmd)
4180 # Check actual executable
4181 return _verify_cmd(path, cmd)
4184 def RunRestrictedCmd(cmd,
4185 _lock_timeout=_RCMD_LOCK_TIMEOUT,
4186 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4187 _path=pathutils.RESTRICTED_COMMANDS_DIR,
4188 _sleep_fn=time.sleep,
4189 _prepare_fn=_PrepareRestrictedCmd,
4190 _runcmd_fn=utils.RunCmd,
4191 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4192 """Executes a restricted command after performing strict tests.
4195 @param cmd: Command name
4197 @return: Command output
4198 @raise RPCFail: In case of an error
4201 logging.info("Preparing to run restricted command '%s'", cmd)
4204 _Fail("Restricted commands disabled at configure time")
4210 lock = utils.FileLock.Open(_lock_file)
4211 lock.Exclusive(blocking=True, timeout=_lock_timeout)
4213 (status, value) = _prepare_fn(_path, cmd)
4216 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4217 postfork_fn=lambda _: lock.Unlock())
4219 logging.error(value)
4220 except Exception: # pylint: disable=W0703
4221 # Keep original error in log
4222 logging.exception("Caught exception")
4224 if cmdresult is None:
4225 logging.info("Sleeping for %0.1f seconds before returning",
4226 _RCMD_INVALID_DELAY)
4227 _sleep_fn(_RCMD_INVALID_DELAY)
4229 # Do not include original error message in returned error
4230 _Fail("Executing command '%s' failed" % cmd)
4231 elif cmdresult.failed or cmdresult.fail_reason:
4232 _Fail("Restricted command '%s' failed: %s; output: %s",
4233 cmd, cmdresult.fail_reason, cmdresult.output)
4235 return cmdresult.output
4237 if lock is not None:
4238 # Release lock at last
4243 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4244 """Creates or removes the watcher pause file.
4246 @type until: None or number
4247 @param until: Unix timestamp saying until when the watcher shouldn't run
4251 logging.info("Received request to no longer pause watcher")
4252 utils.RemoveFile(_filename)
4254 logging.info("Received request to pause watcher until %s", until)
4256 if not ht.TNumber(until):
4257 _Fail("Duration must be numeric")
4259 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4262 class HooksRunner(object):
4265 This class is instantiated on the node side (ganeti-noded) and not
4269 def __init__(self, hooks_base_dir=None):
4270 """Constructor for hooks runner.
4272 @type hooks_base_dir: str or None
4273 @param hooks_base_dir: if not None, this overrides the
4274 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4277 if hooks_base_dir is None:
4278 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4279 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4281 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4283 def RunLocalHooks(self, node_list, hpath, phase, env):
4284 """Check that the hooks will be run only locally and then run them.
4287 assert len(node_list) == 1
4289 _, myself = ssconf.GetMasterAndMyself()
4290 assert node == myself
4292 results = self.RunHooks(hpath, phase, env)
4294 # Return values in the form expected by HooksMaster
4295 return {node: (None, False, results)}
4297 def RunHooks(self, hpath, phase, env):
4298 """Run the scripts in the hooks directory.
4301 @param hpath: the path to the hooks directory which
4304 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4305 L{constants.HOOKS_PHASE_POST}
4307 @param env: dictionary with the environment for the hook
4309 @return: list of 3-element tuples:
4311 - script result, either L{constants.HKR_SUCCESS} or
4312 L{constants.HKR_FAIL}
4313 - output of the script
4315 @raise errors.ProgrammerError: for invalid input
4319 if phase == constants.HOOKS_PHASE_PRE:
4321 elif phase == constants.HOOKS_PHASE_POST:
4324 _Fail("Unknown hooks phase '%s'", phase)
4326 subdir = "%s-%s.d" % (hpath, suffix)
4327 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4331 if not os.path.isdir(dir_name):
4332 # for non-existing/non-dirs, we simply exit instead of logging a
4333 # warning at every operation
4336 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4338 for (relname, relstatus, runresult) in runparts_results:
4339 if relstatus == constants.RUNPARTS_SKIP:
4340 rrval = constants.HKR_SKIP
4342 elif relstatus == constants.RUNPARTS_ERR:
4343 rrval = constants.HKR_FAIL
4344 output = "Hook script execution error: %s" % runresult
4345 elif relstatus == constants.RUNPARTS_RUN:
4346 if runresult.failed:
4347 rrval = constants.HKR_FAIL
4349 rrval = constants.HKR_SUCCESS
4350 output = utils.SafeEncode(runresult.output.strip())
4351 results.append(("%s/%s" % (subdir, relname), rrval, output))
4356 class IAllocatorRunner(object):
4357 """IAllocator runner.
4359 This class is instantiated on the node side (ganeti-noded) and not on
4364 def Run(name, idata):
4365 """Run an iallocator script.
4368 @param name: the iallocator script name
4370 @param idata: the allocator input data
4373 @return: two element tuple of:
4375 - either error message or stdout of allocator (for success)
4378 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4380 if alloc_script is None:
4381 _Fail("iallocator module '%s' not found in the search path", name)
4383 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4387 result = utils.RunCmd([alloc_script, fin_name])
4389 _Fail("iallocator module '%s' failed: %s, output '%s'",
4390 name, result.fail_reason, result.output)
4394 return result.stdout
4397 class DevCacheManager(object):
4398 """Simple class for managing a cache of block device information.
4401 _DEV_PREFIX = "/dev/"
4402 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4405 def _ConvertPath(cls, dev_path):
4406 """Converts a /dev/name path to the cache file name.
4408 This replaces slashes with underscores and strips the /dev
4409 prefix. It then returns the full path to the cache file.
4412 @param dev_path: the C{/dev/} path name
4414 @return: the converted path name
4417 if dev_path.startswith(cls._DEV_PREFIX):
4418 dev_path = dev_path[len(cls._DEV_PREFIX):]
4419 dev_path = dev_path.replace("/", "_")
4420 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4424 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4425 """Updates the cache information for a given device.
4428 @param dev_path: the pathname of the device
4430 @param owner: the owner (instance name) of the device
4431 @type on_primary: bool
4432 @param on_primary: whether this is the primary
4435 @param iv_name: the instance-visible name of the
4436 device, as in objects.Disk.iv_name
4441 if dev_path is None:
4442 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4444 fpath = cls._ConvertPath(dev_path)
4450 iv_name = "not_visible"
4451 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4453 utils.WriteFile(fpath, data=fdata)
4454 except EnvironmentError, err:
4455 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4458 def RemoveCache(cls, dev_path):
4459 """Remove data for a dev_path.
4461 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4462 path name and logging.
4465 @param dev_path: the pathname of the device
4470 if dev_path is None:
4471 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4473 fpath = cls._ConvertPath(dev_path)
4475 utils.RemoveFile(fpath)
4476 except EnvironmentError, err:
4477 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)