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,C0302
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
37 # C0302: This module has become too big and should be split up
54 from ganeti import errors
55 from ganeti import utils
56 from ganeti import ssh
57 from ganeti import hypervisor
58 from ganeti import constants
59 from ganeti.storage import bdev
60 from ganeti.storage import drbd
61 from ganeti.storage import filestorage
62 from ganeti import objects
63 from ganeti import ssconf
64 from ganeti import serializer
65 from ganeti import netutils
66 from ganeti import runtime
67 from ganeti import compat
68 from ganeti import pathutils
69 from ganeti import vcluster
71 from ganeti.storage.base import BlockDev
72 from ganeti.storage.drbd import DRBD8
73 from ganeti import hooksmaster
76 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
77 _ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
79 pathutils.JOB_QUEUE_ARCHIVE_DIR,
81 pathutils.CRYPTO_KEYS_DIR,
83 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
84 _X509_KEY_FILE = "key"
85 _X509_CERT_FILE = "cert"
86 _IES_STATUS_FILE = "status"
90 #: Valid LVS output line regex
91 _LVSLINE_REGEX = re.compile(r"^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
93 # Actions for the master setup script
94 _MASTER_START = "start"
97 #: Maximum file permissions for restricted command directory and executables
98 _RCMD_MAX_MODE = (stat.S_IRWXU |
99 stat.S_IRGRP | stat.S_IXGRP |
100 stat.S_IROTH | stat.S_IXOTH)
102 #: Delay before returning an error for restricted commands
103 _RCMD_INVALID_DELAY = 10
105 #: How long to wait to acquire lock for restricted commands (shorter than
106 #: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
107 #: command requests arrive
108 _RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
111 class RPCFail(Exception):
112 """Class denoting RPC failure.
114 Its argument is the error message.
119 def _GetInstReasonFilename(instance_name):
120 """Path of the file containing the reason of the instance status change.
122 @type instance_name: string
123 @param instance_name: The name of the instance
125 @return: The path of the file
128 return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
131 def _StoreInstReasonTrail(instance_name, trail):
132 """Serialize a reason trail related to an instance change of state to file.
134 The exact location of the file depends on the name of the instance and on
135 the configuration of the Ganeti cluster defined at deploy time.
137 @type instance_name: string
138 @param instance_name: The name of the instance
142 json = serializer.DumpJson(trail)
143 filename = _GetInstReasonFilename(instance_name)
144 utils.WriteFile(filename, data=json)
147 def _Fail(msg, *args, **kwargs):
148 """Log an error and the raise an RPCFail exception.
150 This exception is then handled specially in the ganeti daemon and
151 turned into a 'failed' return type. As such, this function is a
152 useful shortcut for logging the error and returning it to the master
156 @param msg: the text of the exception
162 if "log" not in kwargs or kwargs["log"]: # if we should log this error
163 if "exc" in kwargs and kwargs["exc"]:
164 logging.exception(msg)
171 """Simple wrapper to return a SimpleStore.
173 @rtype: L{ssconf.SimpleStore}
174 @return: a SimpleStore instance
177 return ssconf.SimpleStore()
180 def _GetSshRunner(cluster_name):
181 """Simple wrapper to return an SshRunner.
183 @type cluster_name: str
184 @param cluster_name: the cluster name, which is needed
185 by the SshRunner constructor
186 @rtype: L{ssh.SshRunner}
187 @return: an SshRunner instance
190 return ssh.SshRunner(cluster_name)
193 def _Decompress(data):
194 """Unpacks data compressed by the RPC client.
196 @type data: list or tuple
197 @param data: Data sent by RPC client
199 @return: Decompressed data
202 assert isinstance(data, (list, tuple))
203 assert len(data) == 2
204 (encoding, content) = data
205 if encoding == constants.RPC_ENCODING_NONE:
207 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
208 return zlib.decompress(base64.b64decode(content))
210 raise AssertionError("Unknown data encoding")
213 def _CleanDirectory(path, exclude=None):
214 """Removes all regular files in a directory.
217 @param path: the directory to clean
219 @param exclude: list of files to be excluded, defaults
223 if path not in _ALLOWED_CLEAN_DIRS:
224 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
227 if not os.path.isdir(path):
232 # Normalize excluded paths
233 exclude = [os.path.normpath(i) for i in exclude]
235 for rel_name in utils.ListVisibleFiles(path):
236 full_name = utils.PathJoin(path, rel_name)
237 if full_name in exclude:
239 if os.path.isfile(full_name) and not os.path.islink(full_name):
240 utils.RemoveFile(full_name)
243 def _BuildUploadFileList():
244 """Build the list of allowed upload files.
246 This is abstracted so that it's built only once at module import time.
249 allowed_files = set([
250 pathutils.CLUSTER_CONF_FILE,
252 pathutils.SSH_KNOWN_HOSTS_FILE,
253 pathutils.VNC_PASSWORD_FILE,
254 pathutils.RAPI_CERT_FILE,
255 pathutils.SPICE_CERT_FILE,
256 pathutils.SPICE_CACERT_FILE,
257 pathutils.RAPI_USERS_FILE,
258 pathutils.CONFD_HMAC_KEY,
259 pathutils.CLUSTER_DOMAIN_SECRET_FILE,
262 for hv_name in constants.HYPER_TYPES:
263 hv_class = hypervisor.GetHypervisorClass(hv_name)
264 allowed_files.update(hv_class.GetAncillaryFiles()[0])
266 assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
267 "Allowed file storage paths should never be uploaded via RPC"
269 return frozenset(allowed_files)
272 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
276 """Removes job queue files and archived jobs.
282 _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
283 _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
287 """Returns master information.
289 This is an utility function to compute master information, either
290 for consumption here or from the node daemon.
293 @return: master_netdev, master_ip, master_name, primary_ip_family,
295 @raise RPCFail: in case of errors
300 master_netdev = cfg.GetMasterNetdev()
301 master_ip = cfg.GetMasterIP()
302 master_netmask = cfg.GetMasterNetmask()
303 master_node = cfg.GetMasterNode()
304 primary_ip_family = cfg.GetPrimaryIPFamily()
305 except errors.ConfigurationError, err:
306 _Fail("Cluster configuration incomplete: %s", err, exc=True)
307 return (master_netdev, master_ip, master_node, primary_ip_family,
311 def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
312 """Decorator that runs hooks before and after the decorated function.
314 @type hook_opcode: string
315 @param hook_opcode: opcode of the hook
316 @type hooks_path: string
317 @param hooks_path: path of the hooks
318 @type env_builder_fn: function
319 @param env_builder_fn: function that returns a dictionary containing the
320 environment variables for the hooks. Will get all the parameters of the
322 @raise RPCFail: in case of pre-hook failure
326 def wrapper(*args, **kwargs):
327 _, myself = ssconf.GetMasterAndMyself()
328 nodes = ([myself], [myself]) # these hooks run locally
330 env_fn = compat.partial(env_builder_fn, *args, **kwargs)
334 hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
335 hr.RunLocalHooks, None, env_fn, None,
336 logging.warning, cfg.GetClusterName(),
338 hm.RunPhase(constants.HOOKS_PHASE_PRE)
339 result = fn(*args, **kwargs)
340 hm.RunPhase(constants.HOOKS_PHASE_POST)
347 def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
348 """Builds environment variables for master IP hooks.
350 @type master_params: L{objects.MasterNetworkParameters}
351 @param master_params: network parameters of the master
352 @type use_external_mip_script: boolean
353 @param use_external_mip_script: whether to use an external master IP
354 address setup script (unused, but necessary per the implementation of the
355 _RunLocalHooks decorator)
358 # pylint: disable=W0613
359 ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
361 "MASTER_NETDEV": master_params.netdev,
362 "MASTER_IP": master_params.ip,
363 "MASTER_NETMASK": str(master_params.netmask),
364 "CLUSTER_IP_VERSION": str(ver),
370 def _RunMasterSetupScript(master_params, action, use_external_mip_script):
371 """Execute the master IP address setup script.
373 @type master_params: L{objects.MasterNetworkParameters}
374 @param master_params: network parameters of the master
376 @param action: action to pass to the script. Must be one of
377 L{backend._MASTER_START} or L{backend._MASTER_STOP}
378 @type use_external_mip_script: boolean
379 @param use_external_mip_script: whether to use an external master IP
381 @raise backend.RPCFail: if there are errors during the execution of the
385 env = _BuildMasterIpEnv(master_params)
387 if use_external_mip_script:
388 setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
390 setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
392 result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
395 _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
396 (action, result.exit_code, result.output), log=True)
399 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
401 def ActivateMasterIp(master_params, use_external_mip_script):
402 """Activate the IP address of the master daemon.
404 @type master_params: L{objects.MasterNetworkParameters}
405 @param master_params: network parameters of the master
406 @type use_external_mip_script: boolean
407 @param use_external_mip_script: whether to use an external master IP
409 @raise RPCFail: in case of errors during the IP startup
412 _RunMasterSetupScript(master_params, _MASTER_START,
413 use_external_mip_script)
416 def StartMasterDaemons(no_voting):
417 """Activate local node as master node.
419 The function will start the master daemons (ganeti-masterd and ganeti-rapi).
421 @type no_voting: boolean
422 @param no_voting: whether to start ganeti-masterd without a node vote
423 but still non-interactively
429 masterd_args = "--no-voting --yes-do-it"
434 "EXTRA_MASTERD_ARGS": masterd_args,
437 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
439 msg = "Can't start Ganeti master: %s" % result.output
444 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
446 def DeactivateMasterIp(master_params, use_external_mip_script):
447 """Deactivate the master IP on this node.
449 @type master_params: L{objects.MasterNetworkParameters}
450 @param master_params: network parameters of the master
451 @type use_external_mip_script: boolean
452 @param use_external_mip_script: whether to use an external master IP
454 @raise RPCFail: in case of errors during the IP turndown
457 _RunMasterSetupScript(master_params, _MASTER_STOP,
458 use_external_mip_script)
461 def StopMasterDaemons():
462 """Stop the master daemons on this node.
464 Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
469 # TODO: log and report back to the caller the error failures; we
470 # need to decide in which case we fail the RPC for this
472 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
474 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
476 result.cmd, result.exit_code, result.output)
479 def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
480 """Change the netmask of the master IP.
482 @param old_netmask: the old value of the netmask
483 @param netmask: the new value of the netmask
484 @param master_ip: the master IP
485 @param master_netdev: the master network device
488 if old_netmask == netmask:
491 if not netutils.IPAddress.Own(master_ip):
492 _Fail("The master IP address is not up, not attempting to change its"
495 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
496 "%s/%s" % (master_ip, netmask),
497 "dev", master_netdev, "label",
498 "%s:0" % master_netdev])
500 _Fail("Could not set the new netmask on the master IP address")
502 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
503 "%s/%s" % (master_ip, old_netmask),
504 "dev", master_netdev, "label",
505 "%s:0" % master_netdev])
507 _Fail("Could not bring down the master IP address with the old netmask")
510 def EtcHostsModify(mode, host, ip):
511 """Modify a host entry in /etc/hosts.
513 @param mode: The mode to operate. Either add or remove entry
514 @param host: The host to operate on
515 @param ip: The ip associated with the entry
518 if mode == constants.ETC_HOSTS_ADD:
520 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
522 utils.AddHostToEtcHosts(host, ip)
523 elif mode == constants.ETC_HOSTS_REMOVE:
525 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
526 " parameter is present")
527 utils.RemoveHostFromEtcHosts(host)
529 RPCFail("Mode not supported")
532 def LeaveCluster(modify_ssh_setup):
533 """Cleans up and remove the current node.
535 This function cleans up and prepares the current node to be removed
538 If processing is successful, then it raises an
539 L{errors.QuitGanetiException} which is used as a special case to
540 shutdown the node daemon.
542 @param modify_ssh_setup: boolean
545 _CleanDirectory(pathutils.DATA_DIR)
546 _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
551 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
553 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
555 utils.RemoveFile(priv_key)
556 utils.RemoveFile(pub_key)
557 except errors.OpExecError:
558 logging.exception("Error while processing ssh files")
561 utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
562 utils.RemoveFile(pathutils.RAPI_CERT_FILE)
563 utils.RemoveFile(pathutils.SPICE_CERT_FILE)
564 utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
565 utils.RemoveFile(pathutils.NODED_CERT_FILE)
566 except: # pylint: disable=W0702
567 logging.exception("Error while removing cluster secrets")
569 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
571 logging.error("Command %s failed with exitcode %s and error %s",
572 result.cmd, result.exit_code, result.output)
574 # Raise a custom exception (handled in ganeti-noded)
575 raise errors.QuitGanetiException(True, "Shutdown scheduled")
578 def _CheckStorageParams(params, num_params):
579 """Performs sanity checks for storage parameters.
582 @param params: list of storage parameters
583 @type num_params: int
584 @param num_params: expected number of parameters
588 raise errors.ProgrammerError("No storage parameters for storage"
589 " reporting is provided.")
590 if not isinstance(params, list):
591 raise errors.ProgrammerError("The storage parameters are not of type"
592 " list: '%s'" % params)
593 if not len(params) == num_params:
594 raise errors.ProgrammerError("Did not receive the expected number of"
595 "storage parameters: expected %s,"
596 " received '%s'" % (num_params, len(params)))
599 def _CheckLvmStorageParams(params):
600 """Performs sanity check for the 'exclusive storage' flag.
602 @see: C{_CheckStorageParams}
605 _CheckStorageParams(params, 1)
606 excl_stor = params[0]
607 if not isinstance(params[0], bool):
608 raise errors.ProgrammerError("Exclusive storage parameter is not"
609 " boolean: '%s'." % excl_stor)
613 def _GetLvmVgSpaceInfo(name, params):
614 """Wrapper around C{_GetVgInfo} which checks the storage parameters.
617 @param name: name of the volume group
619 @param params: list of storage parameters, which in this case should be
620 containing only one for exclusive storage
623 excl_stor = _CheckLvmStorageParams(params)
624 return _GetVgInfo(name, excl_stor)
628 name, excl_stor, info_fn=bdev.LogicalVolume.GetVGInfo):
629 """Retrieves information about a LVM volume group.
632 # TODO: GetVGInfo supports returning information for multiple VGs at once
633 vginfo = info_fn([name], excl_stor)
635 vg_free = int(round(vginfo[0][0], 0))
636 vg_size = int(round(vginfo[0][1], 0))
642 "type": constants.ST_LVM_VG,
644 "storage_free": vg_free,
645 "storage_size": vg_size,
649 def _GetLvmPvSpaceInfo(name, params):
650 """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
652 @see: C{_GetLvmVgSpaceInfo}
655 excl_stor = _CheckLvmStorageParams(params)
656 return _GetVgSpindlesInfo(name, excl_stor)
659 def _GetVgSpindlesInfo(
660 name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
661 """Retrieves information about spindles in an LVM volume group.
665 @type excl_stor: bool
666 @param excl_stor: exclusive storage
668 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
669 free spindles, total spindles respectively
673 (vg_free, vg_size) = info_fn(name)
678 "type": constants.ST_LVM_PV,
680 "storage_free": vg_free,
681 "storage_size": vg_size,
685 def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
686 """Retrieves node information from a hypervisor.
688 The information returned depends on the hypervisor. Common items:
690 - vg_size is the size of the configured volume group in MiB
691 - vg_free is the free size of the volume group in MiB
692 - memory_dom0 is the memory allocated for domain0 in MiB
693 - memory_free is the currently available (free) ram in MiB
694 - memory_total is the total number of ram in MiB
695 - hv_version: the hypervisor version, if available
697 @type hvparams: dict of string
698 @param hvparams: the hypervisor's hvparams
701 return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
704 def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
705 """Retrieves node information for all hypervisors.
707 See C{_GetHvInfo} for information on the output.
709 @type hv_specs: list of pairs (string, dict of strings)
710 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
717 for hvname, hvparams in hv_specs:
718 result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
722 def _GetNamedNodeInfo(names, fn):
723 """Calls C{fn} for all names in C{names} and returns a dictionary.
731 return map(fn, names)
734 def GetNodeInfo(storage_units, hv_specs):
735 """Gives back a hash with different information about the node.
737 @type storage_units: list of tuples (string, string, list)
738 @param storage_units: List of tuples (storage unit, identifier, parameters) to
739 ask for disk space information. In case of lvm-vg, the identifier is
740 the VG name. The parameters can contain additional, storage-type-specific
741 parameters, for example exclusive storage for lvm storage.
742 @type hv_specs: list of pairs (string, dict of strings)
743 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
744 @rtype: tuple; (string, None/dict, None/dict)
745 @return: Tuple containing boot ID, volume group information and hypervisor
749 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
750 storage_info = _GetNamedNodeInfo(
752 (lambda (storage_type, storage_key, storage_params):
753 _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
754 hv_info = _GetHvInfoAll(hv_specs)
755 return (bootid, storage_info, hv_info)
758 def _GetFileStorageSpaceInfo(path, params):
759 """Wrapper around filestorage.GetSpaceInfo.
761 The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
762 and ignore the *args parameter to not leak it into the filestorage
765 @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
769 _CheckStorageParams(params, 0)
770 return filestorage.GetFileStorageSpaceInfo(path)
773 # FIXME: implement storage reporting for all missing storage types.
774 _STORAGE_TYPE_INFO_FN = {
775 constants.ST_BLOCK: None,
776 constants.ST_DISKLESS: None,
777 constants.ST_EXT: None,
778 constants.ST_FILE: _GetFileStorageSpaceInfo,
779 constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
780 constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
781 constants.ST_RADOS: None,
785 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
786 """Looks up and applies the correct function to calculate free and total
787 storage for the given storage type.
789 @type storage_type: string
790 @param storage_type: the storage type for which the storage shall be reported.
791 @type storage_key: string
792 @param storage_key: identifier of a storage unit, e.g. the volume group name
793 of an LVM storage unit
795 @param args: various parameters that can be used for storage reporting. These
796 parameters and their semantics vary from storage type to storage type and
797 are just propagated in this function.
798 @return: the results of the application of the storage space function (see
799 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
801 @raises NotImplementedError: for storage types who don't support space
804 fn = _STORAGE_TYPE_INFO_FN[storage_type]
806 return fn(storage_key, *args)
808 raise NotImplementedError
811 def _CheckExclusivePvs(pvi_list):
812 """Check that PVs are not shared among LVs
814 @type pvi_list: list of L{objects.LvmPvInfo} objects
815 @param pvi_list: information about the PVs
817 @rtype: list of tuples (string, list of strings)
818 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
823 if len(pvi.lv_list) > 1:
824 res.append((pvi.name, pvi.lv_list))
828 def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
829 get_hv_fn=hypervisor.GetHypervisor):
830 """Verifies the hypervisor. Appends the results to the 'results' list.
833 @param what: a dictionary of things to check
834 @type vm_capable: boolean
835 @param vm_capable: whether or not this node is vm capable
837 @param result: dictionary of verification results; results of the
838 verifications in this function will be added here
839 @type all_hvparams: dict of dict of string
840 @param all_hvparams: dictionary mapping hypervisor names to hvparams
841 @type get_hv_fn: function
842 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
848 if constants.NV_HYPERVISOR in what:
849 result[constants.NV_HYPERVISOR] = {}
850 for hv_name in what[constants.NV_HYPERVISOR]:
851 hvparams = all_hvparams[hv_name]
853 val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
854 except errors.HypervisorError, err:
855 val = "Error while checking hypervisor: %s" % str(err)
856 result[constants.NV_HYPERVISOR][hv_name] = val
859 def _VerifyHvparams(what, vm_capable, result,
860 get_hv_fn=hypervisor.GetHypervisor):
861 """Verifies the hvparams. Appends the results to the 'results' list.
864 @param what: a dictionary of things to check
865 @type vm_capable: boolean
866 @param vm_capable: whether or not this node is vm capable
868 @param result: dictionary of verification results; results of the
869 verifications in this function will be added here
870 @type get_hv_fn: function
871 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
877 if constants.NV_HVPARAMS in what:
878 result[constants.NV_HVPARAMS] = []
879 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
881 logging.info("Validating hv %s, %s", hv_name, hvparms)
882 get_hv_fn(hv_name).ValidateParameters(hvparms)
883 except errors.HypervisorError, err:
884 result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
887 def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
888 """Verifies the instance list.
891 @param what: a dictionary of things to check
892 @type vm_capable: boolean
893 @param vm_capable: whether or not this node is vm capable
895 @param result: dictionary of verification results; results of the
896 verifications in this function will be added here
897 @type all_hvparams: dict of dict of string
898 @param all_hvparams: dictionary mapping hypervisor names to hvparams
901 if constants.NV_INSTANCELIST in what and vm_capable:
902 # GetInstanceList can fail
904 val = GetInstanceList(what[constants.NV_INSTANCELIST],
905 all_hvparams=all_hvparams)
908 result[constants.NV_INSTANCELIST] = val
911 def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
912 """Verifies the node info.
915 @param what: a dictionary of things to check
916 @type vm_capable: boolean
917 @param vm_capable: whether or not this node is vm capable
919 @param result: dictionary of verification results; results of the
920 verifications in this function will be added here
921 @type all_hvparams: dict of dict of string
922 @param all_hvparams: dictionary mapping hypervisor names to hvparams
925 if constants.NV_HVINFO in what and vm_capable:
926 hvname = what[constants.NV_HVINFO]
927 hyper = hypervisor.GetHypervisor(hvname)
928 hvparams = all_hvparams[hvname]
929 result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
932 def VerifyNode(what, cluster_name, all_hvparams):
933 """Verify the status of the local node.
935 Based on the input L{what} parameter, various checks are done on the
938 If the I{filelist} key is present, this list of
939 files is checksummed and the file/checksum pairs are returned.
941 If the I{nodelist} key is present, we check that we have
942 connectivity via ssh with the target nodes (and check the hostname
945 If the I{node-net-test} key is present, we check that we have
946 connectivity to the given nodes via both primary IP and, if
947 applicable, secondary IPs.
950 @param what: a dictionary of things to check:
951 - filelist: list of files for which to compute checksums
952 - nodelist: list of nodes we should check ssh communication with
953 - node-net-test: list of nodes we should check node daemon port
955 - hypervisor: list with hypervisors to run the verify for
956 @type cluster_name: string
957 @param cluster_name: the cluster's name
958 @type all_hvparams: dict of dict of strings
959 @param all_hvparams: a dictionary mapping hypervisor names to hvparams
961 @return: a dictionary with the same keys as the input dict, and
962 values representing the result of the checks
966 my_name = netutils.Hostname.GetSysName()
967 port = netutils.GetDaemonPort(constants.NODED)
968 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
970 _VerifyHypervisors(what, vm_capable, result, all_hvparams)
971 _VerifyHvparams(what, vm_capable, result)
973 if constants.NV_FILELIST in what:
974 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
975 what[constants.NV_FILELIST]))
976 result[constants.NV_FILELIST] = \
977 dict((vcluster.MakeVirtualPath(key), value)
978 for (key, value) in fingerprints.items())
980 if constants.NV_NODELIST in what:
981 (nodes, bynode) = what[constants.NV_NODELIST]
983 # Add nodes from other groups (different for each node)
985 nodes.extend(bynode[my_name])
990 random.shuffle(nodes)
992 # Try to contact all nodes
995 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
999 result[constants.NV_NODELIST] = val
1001 if constants.NV_NODENETTEST in what:
1002 result[constants.NV_NODENETTEST] = tmp = {}
1003 my_pip = my_sip = None
1004 for name, pip, sip in what[constants.NV_NODENETTEST]:
1010 tmp[my_name] = ("Can't find my own primary/secondary IP"
1011 " in the node list")
1013 for name, pip, sip in what[constants.NV_NODENETTEST]:
1015 if not netutils.TcpPing(pip, port, source=my_pip):
1016 fail.append("primary")
1018 if not netutils.TcpPing(sip, port, source=my_sip):
1019 fail.append("secondary")
1021 tmp[name] = ("failure using the %s interface(s)" %
1024 if constants.NV_MASTERIP in what:
1025 # FIXME: add checks on incoming data structures (here and in the
1026 # rest of the function)
1027 master_name, master_ip = what[constants.NV_MASTERIP]
1028 if master_name == my_name:
1029 source = constants.IP4_ADDRESS_LOCALHOST
1032 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
1035 if constants.NV_USERSCRIPTS in what:
1036 result[constants.NV_USERSCRIPTS] = \
1037 [script for script in what[constants.NV_USERSCRIPTS]
1038 if not utils.IsExecutable(script)]
1040 if constants.NV_OOB_PATHS in what:
1041 result[constants.NV_OOB_PATHS] = tmp = []
1042 for path in what[constants.NV_OOB_PATHS]:
1045 except OSError, err:
1046 tmp.append("error stating out of band helper: %s" % err)
1048 if stat.S_ISREG(st.st_mode):
1049 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
1052 tmp.append("out of band helper %s is not executable" % path)
1054 tmp.append("out of band helper %s is not a file" % path)
1056 if constants.NV_LVLIST in what and vm_capable:
1058 val = GetVolumeList(utils.ListVolumeGroups().keys())
1059 except RPCFail, err:
1061 result[constants.NV_LVLIST] = val
1063 _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1065 if constants.NV_VGLIST in what and vm_capable:
1066 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1068 if constants.NV_PVLIST in what and vm_capable:
1069 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1070 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1071 filter_allocatable=False,
1072 include_lvs=check_exclusive_pvs)
1073 if check_exclusive_pvs:
1074 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1076 # Avoid sending useless data on the wire
1078 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1080 if constants.NV_VERSION in what:
1081 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1082 constants.RELEASE_VERSION)
1084 _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1086 if constants.NV_DRBDVERSION in what and vm_capable:
1088 drbd_version = DRBD8.GetProcInfo().GetVersionString()
1089 except errors.BlockDeviceError, err:
1090 logging.warning("Can't get DRBD version", exc_info=True)
1091 drbd_version = str(err)
1092 result[constants.NV_DRBDVERSION] = drbd_version
1094 if constants.NV_DRBDLIST in what and vm_capable:
1096 used_minors = drbd.DRBD8.GetUsedDevs()
1097 except errors.BlockDeviceError, err:
1098 logging.warning("Can't get used minors list", exc_info=True)
1099 used_minors = str(err)
1100 result[constants.NV_DRBDLIST] = used_minors
1102 if constants.NV_DRBDHELPER in what and vm_capable:
1105 payload = drbd.DRBD8.GetUsermodeHelper()
1106 except errors.BlockDeviceError, err:
1107 logging.error("Can't get DRBD usermode helper: %s", str(err))
1110 result[constants.NV_DRBDHELPER] = (status, payload)
1112 if constants.NV_NODESETUP in what:
1113 result[constants.NV_NODESETUP] = tmpr = []
1114 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1115 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1116 " under /sys, missing required directories /sys/block"
1117 " and /sys/class/net")
1118 if (not os.path.isdir("/proc/sys") or
1119 not os.path.isfile("/proc/sysrq-trigger")):
1120 tmpr.append("The procfs filesystem doesn't seem to be mounted"
1121 " under /proc, missing required directory /proc/sys and"
1122 " the file /proc/sysrq-trigger")
1124 if constants.NV_TIME in what:
1125 result[constants.NV_TIME] = utils.SplitTime(time.time())
1127 if constants.NV_OSLIST in what and vm_capable:
1128 result[constants.NV_OSLIST] = DiagnoseOS()
1130 if constants.NV_BRIDGES in what and vm_capable:
1131 result[constants.NV_BRIDGES] = [bridge
1132 for bridge in what[constants.NV_BRIDGES]
1133 if not utils.BridgeExists(bridge)]
1135 if what.get(constants.NV_ACCEPTED_STORAGE_PATHS) == my_name:
1136 result[constants.NV_ACCEPTED_STORAGE_PATHS] = \
1137 filestorage.ComputeWrongFileStoragePaths()
1139 if what.get(constants.NV_FILE_STORAGE_PATH):
1140 pathresult = filestorage.CheckFileStoragePath(
1141 what[constants.NV_FILE_STORAGE_PATH])
1143 result[constants.NV_FILE_STORAGE_PATH] = pathresult
1145 if what.get(constants.NV_SHARED_FILE_STORAGE_PATH):
1146 pathresult = filestorage.CheckFileStoragePath(
1147 what[constants.NV_SHARED_FILE_STORAGE_PATH])
1149 result[constants.NV_SHARED_FILE_STORAGE_PATH] = pathresult
1154 def GetBlockDevSizes(devices):
1155 """Return the size of the given block devices
1158 @param devices: list of block device nodes to query
1161 dictionary of all block devices under /dev (key). The value is their
1164 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1167 DEV_PREFIX = "/dev/"
1170 for devpath in devices:
1171 if not utils.IsBelowDir(DEV_PREFIX, devpath):
1175 st = os.stat(devpath)
1176 except EnvironmentError, err:
1177 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1180 if stat.S_ISBLK(st.st_mode):
1181 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1183 # We don't want to fail, just do not list this device as available
1184 logging.warning("Cannot get size for block device %s", devpath)
1187 size = int(result.stdout) / (1024 * 1024)
1188 blockdevs[devpath] = size
1192 def GetVolumeList(vg_names):
1193 """Compute list of logical volumes and their size.
1195 @type vg_names: list
1196 @param vg_names: the volume groups whose LVs we should list, or
1197 empty for all volume groups
1200 dictionary of all partions (key) with value being a tuple of
1201 their size (in MiB), inactive and online status::
1203 {'xenvg/test1': ('20.06', True, True)}
1205 in case of errors, a string is returned with the error
1213 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1214 "--separator=%s" % sep,
1215 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1217 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1219 for line in result.stdout.splitlines():
1221 match = _LVSLINE_REGEX.match(line)
1223 logging.error("Invalid line returned from lvs output: '%s'", line)
1225 vg_name, name, size, attr = match.groups()
1226 inactive = attr[4] == "-"
1227 online = attr[5] == "o"
1228 virtual = attr[0] == "v"
1230 # we don't want to report such volumes as existing, since they
1231 # don't really hold data
1233 lvs[vg_name + "/" + name] = (size, inactive, online)
1238 def ListVolumeGroups():
1239 """List the volume groups and their size.
1242 @return: dictionary with keys volume name and values the
1246 return utils.ListVolumeGroups()
1250 """List all volumes on this node.
1254 A list of dictionaries, each having four keys:
1255 - name: the logical volume name,
1256 - size: the size of the logical volume
1257 - dev: the physical device on which the LV lives
1258 - vg: the volume group to which it belongs
1260 In case of errors, we return an empty list and log the
1263 Note that since a logical volume can live on multiple physical
1264 volumes, the resulting list might include a logical volume
1268 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1270 "--options=lv_name,lv_size,devices,vg_name"])
1272 _Fail("Failed to list logical volumes, lvs output: %s",
1276 return dev.split("(")[0]
1278 def handle_dev(dev):
1279 return [parse_dev(x) for x in dev.split(",")]
1282 line = [v.strip() for v in line]
1283 return [{"name": line[0], "size": line[1],
1284 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1287 for line in result.stdout.splitlines():
1288 if line.count("|") >= 3:
1289 all_devs.extend(map_line(line.split("|")))
1291 logging.warning("Strange line in the output from lvs: '%s'", line)
1295 def BridgesExist(bridges_list):
1296 """Check if a list of bridges exist on the current node.
1299 @return: C{True} if all of them exist, C{False} otherwise
1303 for bridge in bridges_list:
1304 if not utils.BridgeExists(bridge):
1305 missing.append(bridge)
1308 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1311 def GetInstanceListForHypervisor(hname, hvparams=None,
1312 get_hv_fn=hypervisor.GetHypervisor):
1313 """Provides a list of instances of the given hypervisor.
1316 @param hname: name of the hypervisor
1317 @type hvparams: dict of strings
1318 @param hvparams: hypervisor parameters for the given hypervisor
1319 @type get_hv_fn: function
1320 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1321 name; optional parameter to increase testability
1324 @return: a list of all running instances on the current node
1325 - instance1.example.com
1326 - instance2.example.com
1331 hv = get_hv_fn(hname)
1332 names = hv.ListInstances(hvparams=hvparams)
1333 results.extend(names)
1334 except errors.HypervisorError, err:
1335 _Fail("Error enumerating instances (hypervisor %s): %s",
1336 hname, err, exc=True)
1340 def GetInstanceList(hypervisor_list, all_hvparams=None,
1341 get_hv_fn=hypervisor.GetHypervisor):
1342 """Provides a list of instances.
1344 @type hypervisor_list: list
1345 @param hypervisor_list: the list of hypervisors to query information
1346 @type all_hvparams: dict of dict of strings
1347 @param all_hvparams: a dictionary mapping hypervisor types to respective
1348 cluster-wide hypervisor parameters
1349 @type get_hv_fn: function
1350 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1351 name; optional parameter to increase testability
1354 @return: a list of all running instances on the current node
1355 - instance1.example.com
1356 - instance2.example.com
1360 for hname in hypervisor_list:
1361 hvparams = all_hvparams[hname]
1362 results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1363 get_hv_fn=get_hv_fn))
1367 def GetInstanceInfo(instance, hname, hvparams=None):
1368 """Gives back the information about an instance as a dictionary.
1370 @type instance: string
1371 @param instance: the instance name
1373 @param hname: the hypervisor type of the instance
1374 @type hvparams: dict of strings
1375 @param hvparams: the instance's hvparams
1378 @return: dictionary with the following keys:
1379 - memory: memory size of instance (int)
1380 - state: xen state of instance (string)
1381 - time: cpu time of instance (float)
1382 - vcpus: the number of vcpus (int)
1387 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1389 if iinfo is not None:
1390 output["memory"] = iinfo[2]
1391 output["vcpus"] = iinfo[3]
1392 output["state"] = iinfo[4]
1393 output["time"] = iinfo[5]
1398 def GetInstanceMigratable(instance):
1399 """Computes whether an instance can be migrated.
1401 @type instance: L{objects.Instance}
1402 @param instance: object representing the instance to be checked.
1405 @return: tuple of (result, description) where:
1406 - result: whether the instance can be migrated or not
1407 - description: a description of the issue, if relevant
1410 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1411 iname = instance.name
1412 if iname not in hyper.ListInstances(instance.hvparams):
1413 _Fail("Instance %s is not running", iname)
1415 for idx in range(len(instance.disks)):
1416 link_name = _GetBlockDevSymlinkPath(iname, idx)
1417 if not os.path.islink(link_name):
1418 logging.warning("Instance %s is missing symlink %s for disk %d",
1419 iname, link_name, idx)
1422 def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1423 """Gather data about all instances.
1425 This is the equivalent of L{GetInstanceInfo}, except that it
1426 computes data for all instances at once, thus being faster if one
1427 needs data about more than one instance.
1429 @type hypervisor_list: list
1430 @param hypervisor_list: list of hypervisors to query for instance data
1431 @type all_hvparams: dict of dict of strings
1432 @param all_hvparams: mapping of hypervisor names to hvparams
1435 @return: dictionary of instance: data, with data having the following keys:
1436 - memory: memory size of instance (int)
1437 - state: xen state of instance (string)
1438 - time: cpu time of instance (float)
1439 - vcpus: the number of vcpus
1444 for hname in hypervisor_list:
1445 hvparams = all_hvparams[hname]
1446 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1448 for name, _, memory, vcpus, state, times in iinfo:
1456 # we only check static parameters, like memory and vcpus,
1457 # and not state and time which can change between the
1458 # invocations of the different hypervisors
1459 for key in "memory", "vcpus":
1460 if value[key] != output[name][key]:
1461 _Fail("Instance %s is running twice"
1462 " with different parameters", name)
1463 output[name] = value
1468 def _InstanceLogName(kind, os_name, instance, component):
1469 """Compute the OS log filename for a given instance and operation.
1471 The instance name and os name are passed in as strings since not all
1472 operations have these as part of an instance object.
1475 @param kind: the operation type (e.g. add, import, etc.)
1476 @type os_name: string
1477 @param os_name: the os name
1478 @type instance: string
1479 @param instance: the name of the instance being imported/added/etc.
1480 @type component: string or None
1481 @param component: the name of the component of the instance being
1485 # TODO: Use tempfile.mkstemp to create unique filename
1487 assert "/" not in component
1488 c_msg = "-%s" % component
1491 base = ("%s-%s-%s%s-%s.log" %
1492 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1493 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1496 def InstanceOsAdd(instance, reinstall, debug):
1497 """Add an OS to an instance.
1499 @type instance: L{objects.Instance}
1500 @param instance: Instance whose OS is to be installed
1501 @type reinstall: boolean
1502 @param reinstall: whether this is an instance reinstall
1503 @type debug: integer
1504 @param debug: debug level, passed to the OS scripts
1508 inst_os = OSFromDisk(instance.os)
1510 create_env = OSEnvironment(instance, inst_os, debug)
1512 create_env["INSTANCE_REINSTALL"] = "1"
1514 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1516 result = utils.RunCmd([inst_os.create_script], env=create_env,
1517 cwd=inst_os.path, output=logfile, reset_env=True)
1519 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1520 " output: %s", result.cmd, result.fail_reason, logfile,
1522 lines = [utils.SafeEncode(val)
1523 for val in utils.TailFile(logfile, lines=20)]
1524 _Fail("OS create script failed (%s), last lines in the"
1525 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1528 def RunRenameInstance(instance, old_name, debug):
1529 """Run the OS rename script for an instance.
1531 @type instance: L{objects.Instance}
1532 @param instance: Instance whose OS is to be installed
1533 @type old_name: string
1534 @param old_name: previous instance name
1535 @type debug: integer
1536 @param debug: debug level, passed to the OS scripts
1538 @return: the success of the operation
1541 inst_os = OSFromDisk(instance.os)
1543 rename_env = OSEnvironment(instance, inst_os, debug)
1544 rename_env["OLD_INSTANCE_NAME"] = old_name
1546 logfile = _InstanceLogName("rename", instance.os,
1547 "%s-%s" % (old_name, instance.name), None)
1549 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1550 cwd=inst_os.path, output=logfile, reset_env=True)
1553 logging.error("os create command '%s' returned error: %s output: %s",
1554 result.cmd, result.fail_reason, result.output)
1555 lines = [utils.SafeEncode(val)
1556 for val in utils.TailFile(logfile, lines=20)]
1557 _Fail("OS rename script failed (%s), last lines in the"
1558 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1561 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1562 """Returns symlink path for block device.
1566 _dir = pathutils.DISK_LINKS_DIR
1568 return utils.PathJoin(_dir,
1570 (instance_name, constants.DISK_SEPARATOR, idx)))
1573 def _SymlinkBlockDev(instance_name, device_path, idx):
1574 """Set up symlinks to a instance's block device.
1576 This is an auxiliary function run when an instance is start (on the primary
1577 node) or when an instance is migrated (on the target node).
1580 @param instance_name: the name of the target instance
1581 @param device_path: path of the physical block device, on the node
1582 @param idx: the disk index
1583 @return: absolute path to the disk's symlink
1586 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1588 os.symlink(device_path, link_name)
1589 except OSError, err:
1590 if err.errno == errno.EEXIST:
1591 if (not os.path.islink(link_name) or
1592 os.readlink(link_name) != device_path):
1593 os.remove(link_name)
1594 os.symlink(device_path, link_name)
1601 def _RemoveBlockDevLinks(instance_name, disks):
1602 """Remove the block device symlinks belonging to the given instance.
1605 for idx, _ in enumerate(disks):
1606 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1607 if os.path.islink(link_name):
1609 os.remove(link_name)
1611 logging.exception("Can't remove symlink '%s'", link_name)
1614 def _CalculateDeviceURI(instance, disk, device):
1615 """Get the URI for the device.
1617 @type instance: L{objects.Instance}
1618 @param instance: the instance which disk belongs to
1619 @type disk: L{objects.Disk}
1620 @param disk: the target disk object
1621 @type device: L{bdev.BlockDev}
1622 @param device: the corresponding BlockDevice
1624 @return: the device uri if any else None
1627 access_mode = disk.params.get(constants.LDP_ACCESS,
1628 constants.DISK_KERNELSPACE)
1629 if access_mode == constants.DISK_USERSPACE:
1630 # This can raise errors.BlockDeviceError
1631 return device.GetUserspaceAccessUri(instance.hypervisor)
1636 def _GatherAndLinkBlockDevs(instance):
1637 """Set up an instance's block device(s).
1639 This is run on the primary node at instance startup. The block
1640 devices must be already assembled.
1642 @type instance: L{objects.Instance}
1643 @param instance: the instance whose disks we should assemble
1645 @return: list of (disk_object, link_name, drive_uri)
1649 for idx, disk in enumerate(instance.disks):
1650 device = _RecursiveFindBD(disk)
1652 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1656 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1658 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1660 uri = _CalculateDeviceURI(instance, disk, device)
1662 block_devices.append((disk, link_name, uri))
1664 return block_devices
1667 def StartInstance(instance, startup_paused, reason, store_reason=True):
1668 """Start an instance.
1670 @type instance: L{objects.Instance}
1671 @param instance: the instance object
1672 @type startup_paused: bool
1673 @param instance: pause instance at startup?
1674 @type reason: list of reasons
1675 @param reason: the reason trail for this startup
1676 @type store_reason: boolean
1677 @param store_reason: whether to store the shutdown reason trail on file
1681 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1684 if instance.name in running_instances:
1685 logging.info("Instance %s already running, not starting", instance.name)
1689 block_devices = _GatherAndLinkBlockDevs(instance)
1690 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1691 hyper.StartInstance(instance, block_devices, startup_paused)
1693 _StoreInstReasonTrail(instance.name, reason)
1694 except errors.BlockDeviceError, err:
1695 _Fail("Block device error: %s", err, exc=True)
1696 except errors.HypervisorError, err:
1697 _RemoveBlockDevLinks(instance.name, instance.disks)
1698 _Fail("Hypervisor error: %s", err, exc=True)
1701 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1702 """Shut an instance down.
1704 @note: this functions uses polling with a hardcoded timeout.
1706 @type instance: L{objects.Instance}
1707 @param instance: the instance object
1708 @type timeout: integer
1709 @param timeout: maximum timeout for soft shutdown
1710 @type reason: list of reasons
1711 @param reason: the reason trail for this shutdown
1712 @type store_reason: boolean
1713 @param store_reason: whether to store the shutdown reason trail on file
1717 hv_name = instance.hypervisor
1718 hyper = hypervisor.GetHypervisor(hv_name)
1719 iname = instance.name
1721 if instance.name not in hyper.ListInstances(instance.hvparams):
1722 logging.info("Instance %s not running, doing nothing", iname)
1727 self.tried_once = False
1730 if iname not in hyper.ListInstances(instance.hvparams):
1734 hyper.StopInstance(instance, retry=self.tried_once, timeout=timeout)
1736 _StoreInstReasonTrail(instance.name, reason)
1737 except errors.HypervisorError, err:
1738 if iname not in hyper.ListInstances(instance.hvparams):
1739 # if the instance is no longer existing, consider this a
1740 # success and go to cleanup
1743 _Fail("Failed to stop instance %s: %s", iname, err)
1745 self.tried_once = True
1747 raise utils.RetryAgain()
1750 utils.Retry(_TryShutdown(), 5, timeout)
1751 except utils.RetryTimeout:
1752 # the shutdown did not succeed
1753 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1756 hyper.StopInstance(instance, force=True)
1757 except errors.HypervisorError, err:
1758 if iname in hyper.ListInstances(instance.hvparams):
1759 # only raise an error if the instance still exists, otherwise
1760 # the error could simply be "instance ... unknown"!
1761 _Fail("Failed to force stop instance %s: %s", iname, err)
1765 if iname in hyper.ListInstances(instance.hvparams):
1766 _Fail("Could not shutdown instance %s even by destroy", iname)
1769 hyper.CleanupInstance(instance.name)
1770 except errors.HypervisorError, err:
1771 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1773 _RemoveBlockDevLinks(iname, instance.disks)
1776 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1777 """Reboot an instance.
1779 @type instance: L{objects.Instance}
1780 @param instance: the instance object to reboot
1781 @type reboot_type: str
1782 @param reboot_type: the type of reboot, one the following
1784 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1785 instance OS, do not recreate the VM
1786 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1787 restart the VM (at the hypervisor level)
1788 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1789 not accepted here, since that mode is handled differently, in
1790 cmdlib, and translates into full stop and start of the
1791 instance (instead of a call_instance_reboot RPC)
1792 @type shutdown_timeout: integer
1793 @param shutdown_timeout: maximum timeout for soft shutdown
1794 @type reason: list of reasons
1795 @param reason: the reason trail for this reboot
1799 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1802 if instance.name not in running_instances:
1803 _Fail("Cannot reboot instance %s that is not running", instance.name)
1805 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1806 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1808 hyper.RebootInstance(instance)
1809 except errors.HypervisorError, err:
1810 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1811 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1813 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1814 result = StartInstance(instance, False, reason, store_reason=False)
1815 _StoreInstReasonTrail(instance.name, reason)
1817 except errors.HypervisorError, err:
1818 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1820 _Fail("Invalid reboot_type received: %s", reboot_type)
1823 def InstanceBalloonMemory(instance, memory):
1824 """Resize an instance's memory.
1826 @type instance: L{objects.Instance}
1827 @param instance: the instance object
1829 @param memory: new memory amount in MB
1833 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1834 running = hyper.ListInstances(instance.hvparams)
1835 if instance.name not in running:
1836 logging.info("Instance %s is not running, cannot balloon", instance.name)
1839 hyper.BalloonInstanceMemory(instance, memory)
1840 except errors.HypervisorError, err:
1841 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1844 def MigrationInfo(instance):
1845 """Gather information about an instance to be migrated.
1847 @type instance: L{objects.Instance}
1848 @param instance: the instance definition
1851 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1853 info = hyper.MigrationInfo(instance)
1854 except errors.HypervisorError, err:
1855 _Fail("Failed to fetch migration information: %s", err, exc=True)
1859 def AcceptInstance(instance, info, target):
1860 """Prepare the node to accept an instance.
1862 @type instance: L{objects.Instance}
1863 @param instance: the instance definition
1864 @type info: string/data (opaque)
1865 @param info: migration information, from the source node
1866 @type target: string
1867 @param target: target host (usually ip), on this node
1870 # TODO: why is this required only for DTS_EXT_MIRROR?
1871 if instance.disk_template in constants.DTS_EXT_MIRROR:
1872 # Create the symlinks, as the disks are not active
1875 _GatherAndLinkBlockDevs(instance)
1876 except errors.BlockDeviceError, err:
1877 _Fail("Block device error: %s", err, exc=True)
1879 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1881 hyper.AcceptInstance(instance, info, target)
1882 except errors.HypervisorError, err:
1883 if instance.disk_template in constants.DTS_EXT_MIRROR:
1884 _RemoveBlockDevLinks(instance.name, instance.disks)
1885 _Fail("Failed to accept instance: %s", err, exc=True)
1888 def FinalizeMigrationDst(instance, info, success):
1889 """Finalize any preparation to accept an instance.
1891 @type instance: L{objects.Instance}
1892 @param instance: the instance definition
1893 @type info: string/data (opaque)
1894 @param info: migration information, from the source node
1895 @type success: boolean
1896 @param success: whether the migration was a success or a failure
1899 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1901 hyper.FinalizeMigrationDst(instance, info, success)
1902 except errors.HypervisorError, err:
1903 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1906 def MigrateInstance(cluster_name, instance, target, live):
1907 """Migrates an instance to another node.
1909 @type cluster_name: string
1910 @param cluster_name: name of the cluster
1911 @type instance: L{objects.Instance}
1912 @param instance: the instance definition
1913 @type target: string
1914 @param target: the target node name
1916 @param live: whether the migration should be done live or not (the
1917 interpretation of this parameter is left to the hypervisor)
1918 @raise RPCFail: if migration fails for some reason
1921 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1924 hyper.MigrateInstance(cluster_name, instance, target, live)
1925 except errors.HypervisorError, err:
1926 _Fail("Failed to migrate instance: %s", err, exc=True)
1929 def FinalizeMigrationSource(instance, success, live):
1930 """Finalize the instance migration on the source node.
1932 @type instance: L{objects.Instance}
1933 @param instance: the instance definition of the migrated instance
1935 @param success: whether the migration succeeded or not
1937 @param live: whether the user requested a live migration or not
1938 @raise RPCFail: If the execution fails for some reason
1941 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1944 hyper.FinalizeMigrationSource(instance, success, live)
1945 except Exception, err: # pylint: disable=W0703
1946 _Fail("Failed to finalize the migration on the source node: %s", err,
1950 def GetMigrationStatus(instance):
1951 """Get the migration status
1953 @type instance: L{objects.Instance}
1954 @param instance: the instance that is being migrated
1955 @rtype: L{objects.MigrationStatus}
1956 @return: the status of the current migration (one of
1957 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1958 progress info that can be retrieved from the hypervisor
1959 @raise RPCFail: If the migration status cannot be retrieved
1962 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1964 return hyper.GetMigrationStatus(instance)
1965 except Exception, err: # pylint: disable=W0703
1966 _Fail("Failed to get migration status: %s", err, exc=True)
1969 def HotplugDevice(instance, action, dev_type, device, extra, seq):
1972 Hotplug is currently supported only for KVM Hypervisor.
1973 @type instance: L{objects.Instance}
1974 @param instance: the instance to which we hotplug a device
1975 @type action: string
1976 @param action: the hotplug action to perform
1977 @type dev_type: string
1978 @param dev_type: the device type to hotplug
1979 @type device: either L{objects.NIC} or L{objects.Disk}
1980 @param device: the device object to hotplug
1982 @param extra: extra info used by hotplug code (e.g. disk link)
1984 @param seq: the index of the device from master perspective
1985 @raise RPCFail: in case instance does not have KVM hypervisor
1988 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1990 hyper.VerifyHotplugSupport(instance, action, dev_type)
1991 except errors.HotplugError, err:
1992 _Fail("Hotplug is not supported: %s", err)
1994 if action == constants.HOTPLUG_ACTION_ADD:
1995 fn = hyper.HotAddDevice
1996 elif action == constants.HOTPLUG_ACTION_REMOVE:
1997 fn = hyper.HotDelDevice
1998 elif action == constants.HOTPLUG_ACTION_MODIFY:
1999 fn = hyper.HotModDevice
2001 assert action in constants.HOTPLUG_ALL_ACTIONS
2003 return fn(instance, dev_type, device, extra, seq)
2006 def HotplugSupported(instance):
2007 """Checks if hotplug is generally supported.
2010 hyper = hypervisor.GetHypervisor(instance.hypervisor)
2012 hyper.HotplugSupported(instance)
2013 except errors.HotplugError, err:
2014 _Fail("Hotplug is not supported: %s", err)
2017 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
2018 """Creates a block device for an instance.
2020 @type disk: L{objects.Disk}
2021 @param disk: the object describing the disk we should create
2023 @param size: the size of the physical underlying device, in MiB
2025 @param owner: the name of the instance for which disk is created,
2026 used for device cache data
2027 @type on_primary: boolean
2028 @param on_primary: indicates if it is the primary node or not
2030 @param info: string that will be sent to the physical device
2031 creation, used for example to set (LVM) tags on LVs
2032 @type excl_stor: boolean
2033 @param excl_stor: Whether exclusive_storage is active
2035 @return: the new unique_id of the device (this can sometime be
2036 computed only after creation), or None. On secondary nodes,
2037 it's not required to return anything.
2040 # TODO: remove the obsolete "size" argument
2041 # pylint: disable=W0613
2044 for child in disk.children:
2046 crdev = _RecursiveAssembleBD(child, owner, on_primary)
2047 except errors.BlockDeviceError, err:
2048 _Fail("Can't assemble device %s: %s", child, err)
2049 if on_primary or disk.AssembleOnSecondary():
2050 # we need the children open in case the device itself has to
2053 # pylint: disable=E1103
2055 except errors.BlockDeviceError, err:
2056 _Fail("Can't make child '%s' read-write: %s", child, err)
2060 device = bdev.Create(disk, clist, excl_stor)
2061 except errors.BlockDeviceError, err:
2062 _Fail("Can't create block device: %s", err)
2064 if on_primary or disk.AssembleOnSecondary():
2067 except errors.BlockDeviceError, err:
2068 _Fail("Can't assemble device after creation, unusual event: %s", err)
2069 if on_primary or disk.OpenOnSecondary():
2071 device.Open(force=True)
2072 except errors.BlockDeviceError, err:
2073 _Fail("Can't make device r/w after creation, unusual event: %s", err)
2074 DevCacheManager.UpdateCache(device.dev_path, owner,
2075 on_primary, disk.iv_name)
2077 device.SetInfo(info)
2079 return device.unique_id
2082 def _WipeDevice(path, offset, size):
2083 """This function actually wipes the device.
2085 @param path: The path to the device to wipe
2086 @param offset: The offset in MiB in the file
2087 @param size: The size in MiB to write
2090 # Internal sizes are always in Mebibytes; if the following "dd" command
2091 # should use a different block size the offset and size given to this
2092 # function must be adjusted accordingly before being passed to "dd".
2093 block_size = 1024 * 1024
2095 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2096 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2098 result = utils.RunCmd(cmd)
2101 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2102 result.fail_reason, result.output)
2105 def BlockdevWipe(disk, offset, size):
2106 """Wipes a block device.
2108 @type disk: L{objects.Disk}
2109 @param disk: the disk object we want to wipe
2111 @param offset: The offset in MiB in the file
2113 @param size: The size in MiB to write
2117 rdev = _RecursiveFindBD(disk)
2118 except errors.BlockDeviceError:
2122 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2124 # Do cross verify some of the parameters
2126 _Fail("Negative offset")
2128 _Fail("Negative size")
2129 if offset > rdev.size:
2130 _Fail("Offset is bigger than device size")
2131 if (offset + size) > rdev.size:
2132 _Fail("The provided offset and size to wipe is bigger than device size")
2134 _WipeDevice(rdev.dev_path, offset, size)
2137 def BlockdevPauseResumeSync(disks, pause):
2138 """Pause or resume the sync of the block device.
2140 @type disks: list of L{objects.Disk}
2141 @param disks: the disks object we want to pause/resume
2143 @param pause: Wheater to pause or resume
2149 rdev = _RecursiveFindBD(disk)
2150 except errors.BlockDeviceError:
2154 success.append((False, ("Cannot change sync for device %s:"
2155 " device not found" % disk.iv_name)))
2158 result = rdev.PauseResumeSync(pause)
2161 success.append((result, None))
2167 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2172 def BlockdevRemove(disk):
2173 """Remove a block device.
2175 @note: This is intended to be called recursively.
2177 @type disk: L{objects.Disk}
2178 @param disk: the disk object we should remove
2180 @return: the success of the operation
2185 rdev = _RecursiveFindBD(disk)
2186 except errors.BlockDeviceError, err:
2187 # probably can't attach
2188 logging.info("Can't attach to device %s in remove", disk)
2190 if rdev is not None:
2191 r_path = rdev.dev_path
2197 except errors.BlockDeviceError, err:
2200 msgs.extend(utils.SimpleRetry([], _TryRemove,
2201 constants.DISK_REMOVE_RETRY_INTERVAL,
2202 constants.DISK_REMOVE_RETRY_TIMEOUT))
2205 DevCacheManager.RemoveCache(r_path)
2208 for child in disk.children:
2210 BlockdevRemove(child)
2211 except RPCFail, err:
2212 msgs.append(str(err))
2215 _Fail("; ".join(msgs))
2218 def _RecursiveAssembleBD(disk, owner, as_primary):
2219 """Activate a block device for an instance.
2221 This is run on the primary and secondary nodes for an instance.
2223 @note: this function is called recursively.
2225 @type disk: L{objects.Disk}
2226 @param disk: the disk we try to assemble
2228 @param owner: the name of the instance which owns the disk
2229 @type as_primary: boolean
2230 @param as_primary: if we should make the block device
2233 @return: the assembled device or None (in case no device
2235 @raise errors.BlockDeviceError: in case there is an error
2236 during the activation of the children or the device
2242 mcn = disk.ChildrenNeeded()
2244 mcn = 0 # max number of Nones allowed
2246 mcn = len(disk.children) - mcn # max number of Nones
2247 for chld_disk in disk.children:
2249 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2250 except errors.BlockDeviceError, err:
2251 if children.count(None) >= mcn:
2254 logging.error("Error in child activation (but continuing): %s",
2256 children.append(cdev)
2258 if as_primary or disk.AssembleOnSecondary():
2259 r_dev = bdev.Assemble(disk, children)
2261 if as_primary or disk.OpenOnSecondary():
2263 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2264 as_primary, disk.iv_name)
2271 def BlockdevAssemble(disk, owner, as_primary, idx):
2272 """Activate a block device for an instance.
2274 This is a wrapper over _RecursiveAssembleBD.
2276 @rtype: str or boolean
2277 @return: a tuple with the C{/dev/...} path and the created symlink
2278 for primary nodes, and (C{True}, C{True}) for secondary nodes
2282 result = _RecursiveAssembleBD(disk, owner, as_primary)
2283 if isinstance(result, BlockDev):
2284 # pylint: disable=E1103
2285 dev_path = result.dev_path
2288 link_name = _SymlinkBlockDev(owner, dev_path, idx)
2290 return result, result
2292 _Fail("Unexpected result from _RecursiveAssembleBD")
2293 except errors.BlockDeviceError, err:
2294 _Fail("Error while assembling disk: %s", err, exc=True)
2295 except OSError, err:
2296 _Fail("Error while symlinking disk: %s", err, exc=True)
2298 return dev_path, link_name
2301 def BlockdevShutdown(disk):
2302 """Shut down a block device.
2304 First, if the device is assembled (Attach() is successful), then
2305 the device is shutdown. Then the children of the device are
2308 This function is called recursively. Note that we don't cache the
2309 children or such, as oppossed to assemble, shutdown of different
2310 devices doesn't require that the upper device was active.
2312 @type disk: L{objects.Disk}
2313 @param disk: the description of the disk we should
2319 r_dev = _RecursiveFindBD(disk)
2320 if r_dev is not None:
2321 r_path = r_dev.dev_path
2324 DevCacheManager.RemoveCache(r_path)
2325 except errors.BlockDeviceError, err:
2326 msgs.append(str(err))
2329 for child in disk.children:
2331 BlockdevShutdown(child)
2332 except RPCFail, err:
2333 msgs.append(str(err))
2336 _Fail("; ".join(msgs))
2339 def BlockdevAddchildren(parent_cdev, new_cdevs):
2340 """Extend a mirrored block device.
2342 @type parent_cdev: L{objects.Disk}
2343 @param parent_cdev: the disk to which we should add children
2344 @type new_cdevs: list of L{objects.Disk}
2345 @param new_cdevs: the list of children which we should add
2349 parent_bdev = _RecursiveFindBD(parent_cdev)
2350 if parent_bdev is None:
2351 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2352 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2353 if new_bdevs.count(None) > 0:
2354 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2355 parent_bdev.AddChildren(new_bdevs)
2358 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2359 """Shrink a mirrored block device.
2361 @type parent_cdev: L{objects.Disk}
2362 @param parent_cdev: the disk from which we should remove children
2363 @type new_cdevs: list of L{objects.Disk}
2364 @param new_cdevs: the list of children which we should remove
2368 parent_bdev = _RecursiveFindBD(parent_cdev)
2369 if parent_bdev is None:
2370 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2372 for disk in new_cdevs:
2373 rpath = disk.StaticDevPath()
2375 bd = _RecursiveFindBD(disk)
2377 _Fail("Can't find device %s while removing children", disk)
2379 devs.append(bd.dev_path)
2381 if not utils.IsNormAbsPath(rpath):
2382 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2384 parent_bdev.RemoveChildren(devs)
2387 def BlockdevGetmirrorstatus(disks):
2388 """Get the mirroring status of a list of devices.
2390 @type disks: list of L{objects.Disk}
2391 @param disks: the list of disks which we should query
2393 @return: List of L{objects.BlockDevStatus}, one for each disk
2394 @raise errors.BlockDeviceError: if any of the disks cannot be
2400 rbd = _RecursiveFindBD(dsk)
2402 _Fail("Can't find device %s", dsk)
2404 stats.append(rbd.CombinedSyncStatus())
2409 def BlockdevGetmirrorstatusMulti(disks):
2410 """Get the mirroring status of a list of devices.
2412 @type disks: list of L{objects.Disk}
2413 @param disks: the list of disks which we should query
2415 @return: List of tuples, (bool, status), one for each disk; bool denotes
2416 success/failure, status is L{objects.BlockDevStatus} on success, string
2423 rbd = _RecursiveFindBD(disk)
2425 result.append((False, "Can't find device %s" % disk))
2428 status = rbd.CombinedSyncStatus()
2429 except errors.BlockDeviceError, err:
2430 logging.exception("Error while getting disk status")
2431 result.append((False, str(err)))
2433 result.append((True, status))
2435 assert len(disks) == len(result)
2440 def _RecursiveFindBD(disk):
2441 """Check if a device is activated.
2443 If so, return information about the real device.
2445 @type disk: L{objects.Disk}
2446 @param disk: the disk object we need to find
2448 @return: None if the device can't be found,
2449 otherwise the device instance
2454 for chdisk in disk.children:
2455 children.append(_RecursiveFindBD(chdisk))
2457 return bdev.FindDevice(disk, children)
2460 def _OpenRealBD(disk):
2461 """Opens the underlying block device of a disk.
2463 @type disk: L{objects.Disk}
2464 @param disk: the disk object we want to open
2467 real_disk = _RecursiveFindBD(disk)
2468 if real_disk is None:
2469 _Fail("Block device '%s' is not set up", disk)
2476 def BlockdevFind(disk):
2477 """Check if a device is activated.
2479 If it is, return information about the real device.
2481 @type disk: L{objects.Disk}
2482 @param disk: the disk to find
2483 @rtype: None or objects.BlockDevStatus
2484 @return: None if the disk cannot be found, otherwise a the current
2489 rbd = _RecursiveFindBD(disk)
2490 except errors.BlockDeviceError, err:
2491 _Fail("Failed to find device: %s", err, exc=True)
2496 return rbd.GetSyncStatus()
2499 def BlockdevGetdimensions(disks):
2500 """Computes the size of the given disks.
2502 If a disk is not found, returns None instead.
2504 @type disks: list of L{objects.Disk}
2505 @param disks: the list of disk to compute the size for
2507 @return: list with elements None if the disk cannot be found,
2508 otherwise the pair (size, spindles), where spindles is None if the
2509 device doesn't support that
2515 rbd = _RecursiveFindBD(cf)
2516 except errors.BlockDeviceError:
2522 result.append(rbd.GetActualDimensions())
2526 def BlockdevExport(disk, dest_node_ip, dest_path, cluster_name):
2527 """Export a block device to a remote node.
2529 @type disk: L{objects.Disk}
2530 @param disk: the description of the disk to export
2531 @type dest_node_ip: str
2532 @param dest_node_ip: the destination node IP to export to
2533 @type dest_path: str
2534 @param dest_path: the destination path on the target node
2535 @type cluster_name: str
2536 @param cluster_name: the cluster name, needed for SSH hostalias
2540 real_disk = _OpenRealBD(disk)
2542 # the block size on the read dd is 1MiB to match our units
2543 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2544 "dd if=%s bs=1048576 count=%s",
2545 real_disk.dev_path, str(disk.size))
2547 # we set here a smaller block size as, due to ssh buffering, more
2548 # than 64-128k will mostly ignored; we use nocreat to fail if the
2549 # device is not already there or we pass a wrong path; we use
2550 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2551 # to not buffer too much memory; this means that at best, we flush
2552 # every 64k, which will not be very fast
2553 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2554 " oflag=dsync", dest_path)
2556 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node_ip,
2557 constants.SSH_LOGIN_USER,
2560 # all commands have been checked, so we're safe to combine them
2561 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2563 result = utils.RunCmd(["bash", "-c", command])
2566 _Fail("Disk copy command '%s' returned error: %s"
2567 " output: %s", command, result.fail_reason, result.output)
2570 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2571 """Write a file to the filesystem.
2573 This allows the master to overwrite(!) a file. It will only perform
2574 the operation if the file belongs to a list of configuration files.
2576 @type file_name: str
2577 @param file_name: the target file name
2579 @param data: the new contents of the file
2581 @param mode: the mode to give the file (can be None)
2583 @param uid: the owner of the file
2585 @param gid: the group of the file
2587 @param atime: the atime to set on the file (can be None)
2589 @param mtime: the mtime to set on the file (can be None)
2593 file_name = vcluster.LocalizeVirtualPath(file_name)
2595 if not os.path.isabs(file_name):
2596 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2598 if file_name not in _ALLOWED_UPLOAD_FILES:
2599 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2602 raw_data = _Decompress(data)
2604 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2605 _Fail("Invalid username/groupname type")
2607 getents = runtime.GetEnts()
2608 uid = getents.LookupUser(uid)
2609 gid = getents.LookupGroup(gid)
2611 utils.SafeWriteFile(file_name, None,
2612 data=raw_data, mode=mode, uid=uid, gid=gid,
2613 atime=atime, mtime=mtime)
2616 def RunOob(oob_program, command, node, timeout):
2617 """Executes oob_program with given command on given node.
2619 @param oob_program: The path to the executable oob_program
2620 @param command: The command to invoke on oob_program
2621 @param node: The node given as an argument to the program
2622 @param timeout: Timeout after which we kill the oob program
2625 @raise RPCFail: If execution fails for some reason
2628 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2631 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2632 result.fail_reason, result.output)
2634 return result.stdout
2637 def _OSOndiskAPIVersion(os_dir):
2638 """Compute and return the API version of a given OS.
2640 This function will try to read the API version of the OS residing in
2641 the 'os_dir' directory.
2644 @param os_dir: the directory in which we should look for the OS
2646 @return: tuple (status, data) with status denoting the validity and
2647 data holding either the vaid versions or an error message
2650 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2653 st = os.stat(api_file)
2654 except EnvironmentError, err:
2655 return False, ("Required file '%s' not found under path %s: %s" %
2656 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2658 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2659 return False, ("File '%s' in %s is not a regular file" %
2660 (constants.OS_API_FILE, os_dir))
2663 api_versions = utils.ReadFile(api_file).splitlines()
2664 except EnvironmentError, err:
2665 return False, ("Error while reading the API version file at %s: %s" %
2666 (api_file, utils.ErrnoOrStr(err)))
2669 api_versions = [int(version.strip()) for version in api_versions]
2670 except (TypeError, ValueError), err:
2671 return False, ("API version(s) can't be converted to integer: %s" %
2674 return True, api_versions
2677 def DiagnoseOS(top_dirs=None):
2678 """Compute the validity for all OSes.
2680 @type top_dirs: list
2681 @param top_dirs: the list of directories in which to
2682 search (if not given defaults to
2683 L{pathutils.OS_SEARCH_PATH})
2684 @rtype: list of L{objects.OS}
2685 @return: a list of tuples (name, path, status, diagnose, variants,
2686 parameters, api_version) for all (potential) OSes under all
2687 search paths, where:
2688 - name is the (potential) OS name
2689 - path is the full path to the OS
2690 - status True/False is the validity of the OS
2691 - diagnose is the error message for an invalid OS, otherwise empty
2692 - variants is a list of supported OS variants, if any
2693 - parameters is a list of (name, help) parameters, if any
2694 - api_version is a list of support OS API versions
2697 if top_dirs is None:
2698 top_dirs = pathutils.OS_SEARCH_PATH
2701 for dir_name in top_dirs:
2702 if os.path.isdir(dir_name):
2704 f_names = utils.ListVisibleFiles(dir_name)
2705 except EnvironmentError, err:
2706 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2708 for name in f_names:
2709 os_path = utils.PathJoin(dir_name, name)
2710 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2713 variants = os_inst.supported_variants
2714 parameters = os_inst.supported_parameters
2715 api_versions = os_inst.api_versions
2718 variants = parameters = api_versions = []
2719 result.append((name, os_path, status, diagnose, variants,
2720 parameters, api_versions))
2725 def _TryOSFromDisk(name, base_dir=None):
2726 """Create an OS instance from disk.
2728 This function will return an OS instance if the given name is a
2731 @type base_dir: string
2732 @keyword base_dir: Base directory containing OS installations.
2733 Defaults to a search in all the OS_SEARCH_PATH dirs.
2735 @return: success and either the OS instance if we find a valid one,
2739 if base_dir is None:
2740 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2742 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2745 return False, "Directory for OS %s not found in search path" % name
2747 status, api_versions = _OSOndiskAPIVersion(os_dir)
2750 return status, api_versions
2752 if not constants.OS_API_VERSIONS.intersection(api_versions):
2753 return False, ("API version mismatch for path '%s': found %s, want %s." %
2754 (os_dir, api_versions, constants.OS_API_VERSIONS))
2756 # OS Files dictionary, we will populate it with the absolute path
2757 # names; if the value is True, then it is a required file, otherwise
2759 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2761 if max(api_versions) >= constants.OS_API_V15:
2762 os_files[constants.OS_VARIANTS_FILE] = False
2764 if max(api_versions) >= constants.OS_API_V20:
2765 os_files[constants.OS_PARAMETERS_FILE] = True
2767 del os_files[constants.OS_SCRIPT_VERIFY]
2769 for (filename, required) in os_files.items():
2770 os_files[filename] = utils.PathJoin(os_dir, filename)
2773 st = os.stat(os_files[filename])
2774 except EnvironmentError, err:
2775 if err.errno == errno.ENOENT and not required:
2776 del os_files[filename]
2778 return False, ("File '%s' under path '%s' is missing (%s)" %
2779 (filename, os_dir, utils.ErrnoOrStr(err)))
2781 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2782 return False, ("File '%s' under path '%s' is not a regular file" %
2785 if filename in constants.OS_SCRIPTS:
2786 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2787 return False, ("File '%s' under path '%s' is not executable" %
2791 if constants.OS_VARIANTS_FILE in os_files:
2792 variants_file = os_files[constants.OS_VARIANTS_FILE]
2795 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2796 except EnvironmentError, err:
2797 # we accept missing files, but not other errors
2798 if err.errno != errno.ENOENT:
2799 return False, ("Error while reading the OS variants file at %s: %s" %
2800 (variants_file, utils.ErrnoOrStr(err)))
2803 if constants.OS_PARAMETERS_FILE in os_files:
2804 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2806 parameters = utils.ReadFile(parameters_file).splitlines()
2807 except EnvironmentError, err:
2808 return False, ("Error while reading the OS parameters file at %s: %s" %
2809 (parameters_file, utils.ErrnoOrStr(err)))
2810 parameters = [v.split(None, 1) for v in parameters]
2812 os_obj = objects.OS(name=name, path=os_dir,
2813 create_script=os_files[constants.OS_SCRIPT_CREATE],
2814 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2815 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2816 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2817 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2819 supported_variants=variants,
2820 supported_parameters=parameters,
2821 api_versions=api_versions)
2825 def OSFromDisk(name, base_dir=None):
2826 """Create an OS instance from disk.
2828 This function will return an OS instance if the given name is a
2829 valid OS name. Otherwise, it will raise an appropriate
2830 L{RPCFail} exception, detailing why this is not a valid OS.
2832 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2833 an exception but returns true/false status data.
2835 @type base_dir: string
2836 @keyword base_dir: Base directory containing OS installations.
2837 Defaults to a search in all the OS_SEARCH_PATH dirs.
2838 @rtype: L{objects.OS}
2839 @return: the OS instance if we find a valid one
2840 @raise RPCFail: if we don't find a valid OS
2843 name_only = objects.OS.GetName(name)
2844 status, payload = _TryOSFromDisk(name_only, base_dir)
2852 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2853 """Calculate the basic environment for an os script.
2856 @param os_name: full operating system name (including variant)
2857 @type inst_os: L{objects.OS}
2858 @param inst_os: operating system for which the environment is being built
2859 @type os_params: dict
2860 @param os_params: the OS parameters
2861 @type debug: integer
2862 @param debug: debug level (0 or 1, for OS Api 10)
2864 @return: dict of environment variables
2865 @raise errors.BlockDeviceError: if the block device
2871 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2872 result["OS_API_VERSION"] = "%d" % api_version
2873 result["OS_NAME"] = inst_os.name
2874 result["DEBUG_LEVEL"] = "%d" % debug
2877 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2878 variant = objects.OS.GetVariant(os_name)
2880 variant = inst_os.supported_variants[0]
2883 result["OS_VARIANT"] = variant
2886 for pname, pvalue in os_params.items():
2887 result["OSP_%s" % pname.upper()] = pvalue
2889 # Set a default path otherwise programs called by OS scripts (or
2890 # even hooks called from OS scripts) might break, and we don't want
2891 # to have each script require setting a PATH variable
2892 result["PATH"] = constants.HOOKS_PATH
2897 def OSEnvironment(instance, inst_os, debug=0):
2898 """Calculate the environment for an os script.
2900 @type instance: L{objects.Instance}
2901 @param instance: target instance for the os script run
2902 @type inst_os: L{objects.OS}
2903 @param inst_os: operating system for which the environment is being built
2904 @type debug: integer
2905 @param debug: debug level (0 or 1, for OS Api 10)
2907 @return: dict of environment variables
2908 @raise errors.BlockDeviceError: if the block device
2912 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2914 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2915 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2917 result["HYPERVISOR"] = instance.hypervisor
2918 result["DISK_COUNT"] = "%d" % len(instance.disks)
2919 result["NIC_COUNT"] = "%d" % len(instance.nics)
2920 result["INSTANCE_SECONDARY_NODES"] = \
2921 ("%s" % " ".join(instance.secondary_nodes))
2924 for idx, disk in enumerate(instance.disks):
2925 real_disk = _OpenRealBD(disk)
2926 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2927 result["DISK_%d_ACCESS" % idx] = disk.mode
2928 result["DISK_%d_UUID" % idx] = disk.uuid
2930 result["DISK_%d_NAME" % idx] = disk.name
2931 if constants.HV_DISK_TYPE in instance.hvparams:
2932 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2933 instance.hvparams[constants.HV_DISK_TYPE]
2934 if disk.dev_type in constants.DTS_BLOCK:
2935 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2936 elif disk.dev_type in [constants.DT_FILE, constants.DT_SHARED_FILE]:
2937 result["DISK_%d_BACKEND_TYPE" % idx] = \
2938 "file:%s" % disk.logical_id[0]
2941 for idx, nic in enumerate(instance.nics):
2942 result["NIC_%d_MAC" % idx] = nic.mac
2943 result["NIC_%d_UUID" % idx] = nic.uuid
2945 result["NIC_%d_NAME" % idx] = nic.name
2947 result["NIC_%d_IP" % idx] = nic.ip
2948 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2949 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2950 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2951 if nic.nicparams[constants.NIC_LINK]:
2952 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2954 nobj = objects.Network.FromDict(nic.netinfo)
2955 result.update(nobj.HooksDict("NIC_%d_" % idx))
2956 if constants.HV_NIC_TYPE in instance.hvparams:
2957 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2958 instance.hvparams[constants.HV_NIC_TYPE]
2961 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2962 for key, value in source.items():
2963 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2968 def DiagnoseExtStorage(top_dirs=None):
2969 """Compute the validity for all ExtStorage Providers.
2971 @type top_dirs: list
2972 @param top_dirs: the list of directories in which to
2973 search (if not given defaults to
2974 L{pathutils.ES_SEARCH_PATH})
2975 @rtype: list of L{objects.ExtStorage}
2976 @return: a list of tuples (name, path, status, diagnose, parameters)
2977 for all (potential) ExtStorage Providers under all
2978 search paths, where:
2979 - name is the (potential) ExtStorage Provider
2980 - path is the full path to the ExtStorage Provider
2981 - status True/False is the validity of the ExtStorage Provider
2982 - diagnose is the error message for an invalid ExtStorage Provider,
2984 - parameters is a list of (name, help) parameters, if any
2987 if top_dirs is None:
2988 top_dirs = pathutils.ES_SEARCH_PATH
2991 for dir_name in top_dirs:
2992 if os.path.isdir(dir_name):
2994 f_names = utils.ListVisibleFiles(dir_name)
2995 except EnvironmentError, err:
2996 logging.exception("Can't list the ExtStorage directory %s: %s",
2999 for name in f_names:
3000 es_path = utils.PathJoin(dir_name, name)
3001 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
3004 parameters = es_inst.supported_parameters
3008 result.append((name, es_path, status, diagnose, parameters))
3013 def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
3014 """Grow a stack of block devices.
3016 This function is called recursively, with the childrens being the
3017 first ones to resize.
3019 @type disk: L{objects.Disk}
3020 @param disk: the disk to be grown
3021 @type amount: integer
3022 @param amount: the amount (in mebibytes) to grow with
3023 @type dryrun: boolean
3024 @param dryrun: whether to execute the operation in simulation mode
3025 only, without actually increasing the size
3026 @param backingstore: whether to execute the operation on backing storage
3027 only, or on "logical" storage only; e.g. DRBD is logical storage,
3028 whereas LVM, file, RBD are backing storage
3029 @rtype: (status, result)
3030 @type excl_stor: boolean
3031 @param excl_stor: Whether exclusive_storage is active
3032 @return: a tuple with the status of the operation (True/False), and
3033 the errors message if status is False
3036 r_dev = _RecursiveFindBD(disk)
3038 _Fail("Cannot find block device %s", disk)
3041 r_dev.Grow(amount, dryrun, backingstore, excl_stor)
3042 except errors.BlockDeviceError, err:
3043 _Fail("Failed to grow block device: %s", err, exc=True)
3046 def BlockdevSnapshot(disk):
3047 """Create a snapshot copy of a block device.
3049 This function is called recursively, and the snapshot is actually created
3050 just for the leaf lvm backend device.
3052 @type disk: L{objects.Disk}
3053 @param disk: the disk to be snapshotted
3055 @return: snapshot disk ID as (vg, lv)
3058 if disk.dev_type == constants.DT_DRBD8:
3059 if not disk.children:
3060 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
3062 return BlockdevSnapshot(disk.children[0])
3063 elif disk.dev_type == constants.DT_PLAIN:
3064 r_dev = _RecursiveFindBD(disk)
3065 if r_dev is not None:
3066 # FIXME: choose a saner value for the snapshot size
3067 # let's stay on the safe side and ask for the full size, for now
3068 return r_dev.Snapshot(disk.size)
3070 _Fail("Cannot find block device %s", disk)
3072 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
3073 disk.unique_id, disk.dev_type)
3076 def BlockdevSetInfo(disk, info):
3077 """Sets 'metadata' information on block devices.
3079 This function sets 'info' metadata on block devices. Initial
3080 information is set at device creation; this function should be used
3081 for example after renames.
3083 @type disk: L{objects.Disk}
3084 @param disk: the disk to be grown
3086 @param info: new 'info' metadata
3087 @rtype: (status, result)
3088 @return: a tuple with the status of the operation (True/False), and
3089 the errors message if status is False
3092 r_dev = _RecursiveFindBD(disk)
3094 _Fail("Cannot find block device %s", disk)
3098 except errors.BlockDeviceError, err:
3099 _Fail("Failed to set information on block device: %s", err, exc=True)
3102 def FinalizeExport(instance, snap_disks):
3103 """Write out the export configuration information.
3105 @type instance: L{objects.Instance}
3106 @param instance: the instance which we export, used for
3107 saving configuration
3108 @type snap_disks: list of L{objects.Disk}
3109 @param snap_disks: list of snapshot block devices, which
3110 will be used to get the actual name of the dump file
3115 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3116 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3118 config = objects.SerializableConfigParser()
3120 config.add_section(constants.INISECT_EXP)
3121 config.set(constants.INISECT_EXP, "version", "0")
3122 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3123 config.set(constants.INISECT_EXP, "source", instance.primary_node)
3124 config.set(constants.INISECT_EXP, "os", instance.os)
3125 config.set(constants.INISECT_EXP, "compression", "none")
3127 config.add_section(constants.INISECT_INS)
3128 config.set(constants.INISECT_INS, "name", instance.name)
3129 config.set(constants.INISECT_INS, "maxmem", "%d" %
3130 instance.beparams[constants.BE_MAXMEM])
3131 config.set(constants.INISECT_INS, "minmem", "%d" %
3132 instance.beparams[constants.BE_MINMEM])
3133 # "memory" is deprecated, but useful for exporting to old ganeti versions
3134 config.set(constants.INISECT_INS, "memory", "%d" %
3135 instance.beparams[constants.BE_MAXMEM])
3136 config.set(constants.INISECT_INS, "vcpus", "%d" %
3137 instance.beparams[constants.BE_VCPUS])
3138 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3139 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3140 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3143 for nic_count, nic in enumerate(instance.nics):
3145 config.set(constants.INISECT_INS, "nic%d_mac" %
3146 nic_count, "%s" % nic.mac)
3147 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3148 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3150 config.set(constants.INISECT_INS, "nic%d_name" % nic_count,
3152 for param in constants.NICS_PARAMETER_TYPES:
3153 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3154 "%s" % nic.nicparams.get(param, None))
3155 # TODO: redundant: on load can read nics until it doesn't exist
3156 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3159 for disk_count, disk in enumerate(snap_disks):
3162 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3163 ("%s" % disk.iv_name))
3164 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3165 ("%s" % disk.logical_id[1]))
3166 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3168 config.set(constants.INISECT_INS, "disk%d_name" % disk_count,
3171 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3173 # New-style hypervisor/backend parameters
3175 config.add_section(constants.INISECT_HYP)
3176 for name, value in instance.hvparams.items():
3177 if name not in constants.HVC_GLOBALS:
3178 config.set(constants.INISECT_HYP, name, str(value))
3180 config.add_section(constants.INISECT_BEP)
3181 for name, value in instance.beparams.items():
3182 config.set(constants.INISECT_BEP, name, str(value))
3184 config.add_section(constants.INISECT_OSP)
3185 for name, value in instance.osparams.items():
3186 config.set(constants.INISECT_OSP, name, str(value))
3188 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3189 data=config.Dumps())
3190 shutil.rmtree(finaldestdir, ignore_errors=True)
3191 shutil.move(destdir, finaldestdir)
3194 def ExportInfo(dest):
3195 """Get export configuration information.
3198 @param dest: directory containing the export
3200 @rtype: L{objects.SerializableConfigParser}
3201 @return: a serializable config file containing the
3205 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3207 config = objects.SerializableConfigParser()
3210 if (not config.has_section(constants.INISECT_EXP) or
3211 not config.has_section(constants.INISECT_INS)):
3212 _Fail("Export info file doesn't have the required fields")
3214 return config.Dumps()
3218 """Return a list of exports currently available on this machine.
3221 @return: list of the exports
3224 if os.path.isdir(pathutils.EXPORT_DIR):
3225 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3227 _Fail("No exports directory")
3230 def RemoveExport(export):
3231 """Remove an existing export from the node.
3234 @param export: the name of the export to remove
3238 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3241 shutil.rmtree(target)
3242 except EnvironmentError, err:
3243 _Fail("Error while removing the export: %s", err, exc=True)
3246 def BlockdevRename(devlist):
3247 """Rename a list of block devices.
3249 @type devlist: list of tuples
3250 @param devlist: list of tuples of the form (disk, new_unique_id); disk is
3251 an L{objects.Disk} object describing the current disk, and new
3252 unique_id is the name we rename it to
3254 @return: True if all renames succeeded, False otherwise
3259 for disk, unique_id in devlist:
3260 dev = _RecursiveFindBD(disk)
3262 msgs.append("Can't find device %s in rename" % str(disk))
3266 old_rpath = dev.dev_path
3267 dev.Rename(unique_id)
3268 new_rpath = dev.dev_path
3269 if old_rpath != new_rpath:
3270 DevCacheManager.RemoveCache(old_rpath)
3271 # FIXME: we should add the new cache information here, like:
3272 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3273 # but we don't have the owner here - maybe parse from existing
3274 # cache? for now, we only lose lvm data when we rename, which
3275 # is less critical than DRBD or MD
3276 except errors.BlockDeviceError, err:
3277 msgs.append("Can't rename device '%s' to '%s': %s" %
3278 (dev, unique_id, err))
3279 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3282 _Fail("; ".join(msgs))
3285 def _TransformFileStorageDir(fs_dir):
3286 """Checks whether given file_storage_dir is valid.
3288 Checks wheter the given fs_dir is within the cluster-wide default
3289 file_storage_dir or the shared_file_storage_dir, which are stored in
3290 SimpleStore. Only paths under those directories are allowed.
3293 @param fs_dir: the path to check
3295 @return: the normalized path if valid, None otherwise
3298 filestorage.CheckFileStoragePath(fs_dir)
3300 return os.path.normpath(fs_dir)
3303 def CreateFileStorageDir(file_storage_dir):
3304 """Create file storage directory.
3306 @type file_storage_dir: str
3307 @param file_storage_dir: directory to create
3310 @return: tuple with first element a boolean indicating wheter dir
3311 creation was successful or not
3314 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3315 if os.path.exists(file_storage_dir):
3316 if not os.path.isdir(file_storage_dir):
3317 _Fail("Specified storage dir '%s' is not a directory",
3321 os.makedirs(file_storage_dir, 0750)
3322 except OSError, err:
3323 _Fail("Cannot create file storage directory '%s': %s",
3324 file_storage_dir, err, exc=True)
3327 def RemoveFileStorageDir(file_storage_dir):
3328 """Remove file storage directory.
3330 Remove it only if it's empty. If not log an error and return.
3332 @type file_storage_dir: str
3333 @param file_storage_dir: the directory we should cleanup
3334 @rtype: tuple (success,)
3335 @return: tuple of one element, C{success}, denoting
3336 whether the operation was successful
3339 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3340 if os.path.exists(file_storage_dir):
3341 if not os.path.isdir(file_storage_dir):
3342 _Fail("Specified Storage directory '%s' is not a directory",
3344 # deletes dir only if empty, otherwise we want to fail the rpc call
3346 os.rmdir(file_storage_dir)
3347 except OSError, err:
3348 _Fail("Cannot remove file storage directory '%s': %s",
3349 file_storage_dir, err)
3352 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3353 """Rename the file storage directory.
3355 @type old_file_storage_dir: str
3356 @param old_file_storage_dir: the current path
3357 @type new_file_storage_dir: str
3358 @param new_file_storage_dir: the name we should rename to
3359 @rtype: tuple (success,)
3360 @return: tuple of one element, C{success}, denoting
3361 whether the operation was successful
3364 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3365 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3366 if not os.path.exists(new_file_storage_dir):
3367 if os.path.isdir(old_file_storage_dir):
3369 os.rename(old_file_storage_dir, new_file_storage_dir)
3370 except OSError, err:
3371 _Fail("Cannot rename '%s' to '%s': %s",
3372 old_file_storage_dir, new_file_storage_dir, err)
3374 _Fail("Specified storage dir '%s' is not a directory",
3375 old_file_storage_dir)
3377 if os.path.exists(old_file_storage_dir):
3378 _Fail("Cannot rename '%s' to '%s': both locations exist",
3379 old_file_storage_dir, new_file_storage_dir)
3382 def _EnsureJobQueueFile(file_name):
3383 """Checks whether the given filename is in the queue directory.
3385 @type file_name: str
3386 @param file_name: the file name we should check
3388 @raises RPCFail: if the file is not valid
3391 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3392 _Fail("Passed job queue file '%s' does not belong to"
3393 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3396 def JobQueueUpdate(file_name, content):
3397 """Updates a file in the queue directory.
3399 This is just a wrapper over L{utils.io.WriteFile}, with proper
3402 @type file_name: str
3403 @param file_name: the job file name
3405 @param content: the new job contents
3407 @return: the success of the operation
3410 file_name = vcluster.LocalizeVirtualPath(file_name)
3412 _EnsureJobQueueFile(file_name)
3413 getents = runtime.GetEnts()
3415 # Write and replace the file atomically
3416 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3417 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3420 def JobQueueRename(old, new):
3421 """Renames a job queue file.
3423 This is just a wrapper over os.rename with proper checking.
3426 @param old: the old (actual) file name
3428 @param new: the desired file name
3430 @return: the success of the operation and payload
3433 old = vcluster.LocalizeVirtualPath(old)
3434 new = vcluster.LocalizeVirtualPath(new)
3436 _EnsureJobQueueFile(old)
3437 _EnsureJobQueueFile(new)
3439 getents = runtime.GetEnts()
3441 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3442 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3445 def BlockdevClose(instance_name, disks):
3446 """Closes the given block devices.
3448 This means they will be switched to secondary mode (in case of
3451 @param instance_name: if the argument is not empty, the symlinks
3452 of this instance will be removed
3453 @type disks: list of L{objects.Disk}
3454 @param disks: the list of disks to be closed
3455 @rtype: tuple (success, message)
3456 @return: a tuple of success and message, where success
3457 indicates the succes of the operation, and message
3458 which will contain the error details in case we
3464 rd = _RecursiveFindBD(cf)
3466 _Fail("Can't find device %s", cf)
3473 except errors.BlockDeviceError, err:
3474 msg.append(str(err))
3476 _Fail("Can't make devices secondary: %s", ",".join(msg))
3479 _RemoveBlockDevLinks(instance_name, disks)
3482 def ValidateHVParams(hvname, hvparams):
3483 """Validates the given hypervisor parameters.
3485 @type hvname: string
3486 @param hvname: the hypervisor name
3487 @type hvparams: dict
3488 @param hvparams: the hypervisor parameters to be validated
3493 hv_type = hypervisor.GetHypervisor(hvname)
3494 hv_type.ValidateParameters(hvparams)
3495 except errors.HypervisorError, err:
3496 _Fail(str(err), log=False)
3499 def _CheckOSPList(os_obj, parameters):
3500 """Check whether a list of parameters is supported by the OS.
3502 @type os_obj: L{objects.OS}
3503 @param os_obj: OS object to check
3504 @type parameters: list
3505 @param parameters: the list of parameters to check
3508 supported = [v[0] for v in os_obj.supported_parameters]
3509 delta = frozenset(parameters).difference(supported)
3511 _Fail("The following parameters are not supported"
3512 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3515 def ValidateOS(required, osname, checks, osparams):
3516 """Validate the given OS' parameters.
3518 @type required: boolean
3519 @param required: whether absence of the OS should translate into
3521 @type osname: string
3522 @param osname: the OS to be validated
3524 @param checks: list of the checks to run (currently only 'parameters')
3525 @type osparams: dict
3526 @param osparams: dictionary with OS parameters
3528 @return: True if the validation passed, or False if the OS was not
3529 found and L{required} was false
3532 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3533 _Fail("Unknown checks required for OS %s: %s", osname,
3534 set(checks).difference(constants.OS_VALIDATE_CALLS))
3536 name_only = objects.OS.GetName(osname)
3537 status, tbv = _TryOSFromDisk(name_only, None)
3545 if max(tbv.api_versions) < constants.OS_API_V20:
3548 if constants.OS_VALIDATE_PARAMETERS in checks:
3549 _CheckOSPList(tbv, osparams.keys())
3551 validate_env = OSCoreEnv(osname, tbv, osparams)
3552 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3553 cwd=tbv.path, reset_env=True)
3555 logging.error("os validate command '%s' returned error: %s output: %s",
3556 result.cmd, result.fail_reason, result.output)
3557 _Fail("OS validation script failed (%s), output: %s",
3558 result.fail_reason, result.output, log=False)
3564 """Demotes the current node from master candidate role.
3567 # try to ensure we're not the master by mistake
3568 master, myself = ssconf.GetMasterAndMyself()
3569 if master == myself:
3570 _Fail("ssconf status shows I'm the master node, will not demote")
3572 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3573 if not result.failed:
3574 _Fail("The master daemon is running, will not demote")
3577 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3578 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3579 except EnvironmentError, err:
3580 if err.errno != errno.ENOENT:
3581 _Fail("Error while backing up cluster file: %s", err, exc=True)
3583 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3586 def _GetX509Filenames(cryptodir, name):
3587 """Returns the full paths for the private key and certificate.
3590 return (utils.PathJoin(cryptodir, name),
3591 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3592 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3595 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3596 """Creates a new X509 certificate for SSL/TLS.
3599 @param validity: Validity in seconds
3600 @rtype: tuple; (string, string)
3601 @return: Certificate name and public part
3604 (key_pem, cert_pem) = \
3605 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3606 min(validity, _MAX_SSL_CERT_VALIDITY))
3608 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3609 prefix="x509-%s-" % utils.TimestampForFilename())
3611 name = os.path.basename(cert_dir)
3612 assert len(name) > 5
3614 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3616 utils.WriteFile(key_file, mode=0400, data=key_pem)
3617 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3619 # Never return private key as it shouldn't leave the node
3620 return (name, cert_pem)
3622 shutil.rmtree(cert_dir, ignore_errors=True)
3626 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3627 """Removes a X509 certificate.
3630 @param name: Certificate name
3633 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3635 utils.RemoveFile(key_file)
3636 utils.RemoveFile(cert_file)
3640 except EnvironmentError, err:
3641 _Fail("Cannot remove certificate directory '%s': %s",
3645 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3646 """Returns the command for the requested input/output.
3648 @type instance: L{objects.Instance}
3649 @param instance: The instance object
3650 @param mode: Import/export mode
3651 @param ieio: Input/output type
3652 @param ieargs: Input/output arguments
3655 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3662 if ieio == constants.IEIO_FILE:
3663 (filename, ) = ieargs
3665 if not utils.IsNormAbsPath(filename):
3666 _Fail("Path '%s' is not normalized or absolute", filename)
3668 real_filename = os.path.realpath(filename)
3669 directory = os.path.dirname(real_filename)
3671 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3672 _Fail("File '%s' is not under exports directory '%s': %s",
3673 filename, pathutils.EXPORT_DIR, real_filename)
3676 utils.Makedirs(directory, mode=0750)
3678 quoted_filename = utils.ShellQuote(filename)
3680 if mode == constants.IEM_IMPORT:
3681 suffix = "> %s" % quoted_filename
3682 elif mode == constants.IEM_EXPORT:
3683 suffix = "< %s" % quoted_filename
3685 # Retrieve file size
3687 st = os.stat(filename)
3688 except EnvironmentError, err:
3689 logging.error("Can't stat(2) %s: %s", filename, err)
3691 exp_size = utils.BytesToMebibyte(st.st_size)
3693 elif ieio == constants.IEIO_RAW_DISK:
3696 real_disk = _OpenRealBD(disk)
3698 if mode == constants.IEM_IMPORT:
3699 # we set here a smaller block size as, due to transport buffering, more
3700 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3701 # is not already there or we pass a wrong path; we use notrunc to no
3702 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3703 # much memory; this means that at best, we flush every 64k, which will
3705 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3706 " bs=%s oflag=dsync"),
3710 elif mode == constants.IEM_EXPORT:
3711 # the block size on the read dd is 1MiB to match our units
3712 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3714 str(1024 * 1024), # 1 MB
3716 exp_size = disk.size
3718 elif ieio == constants.IEIO_SCRIPT:
3719 (disk, disk_index, ) = ieargs
3721 assert isinstance(disk_index, (int, long))
3723 inst_os = OSFromDisk(instance.os)
3724 env = OSEnvironment(instance, inst_os)
3726 if mode == constants.IEM_IMPORT:
3727 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3728 env["IMPORT_INDEX"] = str(disk_index)
3729 script = inst_os.import_script
3731 elif mode == constants.IEM_EXPORT:
3732 real_disk = _OpenRealBD(disk)
3733 env["EXPORT_DEVICE"] = real_disk.dev_path
3734 env["EXPORT_INDEX"] = str(disk_index)
3735 script = inst_os.export_script
3737 # TODO: Pass special environment only to script
3738 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3740 if mode == constants.IEM_IMPORT:
3741 suffix = "| %s" % script_cmd
3743 elif mode == constants.IEM_EXPORT:
3744 prefix = "%s |" % script_cmd
3746 # Let script predict size
3747 exp_size = constants.IE_CUSTOM_SIZE
3750 _Fail("Invalid %s I/O mode %r", mode, ieio)
3752 return (env, prefix, suffix, exp_size)
3755 def _CreateImportExportStatusDir(prefix):
3756 """Creates status directory for import/export.
3759 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3761 (prefix, utils.TimestampForFilename())))
3764 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3766 """Starts an import or export daemon.
3768 @param mode: Import/output mode
3769 @type opts: L{objects.ImportExportOptions}
3770 @param opts: Daemon options
3772 @param host: Remote host for export (None for import)
3774 @param port: Remote port for export (None for import)
3775 @type instance: L{objects.Instance}
3776 @param instance: Instance object
3777 @type component: string
3778 @param component: which part of the instance is transferred now,
3780 @param ieio: Input/output type
3781 @param ieioargs: Input/output arguments
3784 if mode == constants.IEM_IMPORT:
3787 if not (host is None and port is None):
3788 _Fail("Can not specify host or port on import")
3790 elif mode == constants.IEM_EXPORT:
3793 if host is None or port is None:
3794 _Fail("Host and port must be specified for an export")
3797 _Fail("Invalid mode %r", mode)
3799 if (opts.key_name is None) ^ (opts.ca_pem is None):
3800 _Fail("Cluster certificate can only be used for both key and CA")
3802 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3803 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3805 if opts.key_name is None:
3807 key_path = pathutils.NODED_CERT_FILE
3808 cert_path = pathutils.NODED_CERT_FILE
3809 assert opts.ca_pem is None
3811 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3813 assert opts.ca_pem is not None
3815 for i in [key_path, cert_path]:
3816 if not os.path.exists(i):
3817 _Fail("File '%s' does not exist" % i)
3819 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3821 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3822 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3823 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3825 if opts.ca_pem is None:
3827 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3832 utils.WriteFile(ca_file, data=ca, mode=0400)
3835 pathutils.IMPORT_EXPORT_DAEMON,
3837 "--key=%s" % key_path,
3838 "--cert=%s" % cert_path,
3839 "--ca=%s" % ca_file,
3843 cmd.append("--host=%s" % host)
3846 cmd.append("--port=%s" % port)
3849 cmd.append("--ipv6")
3851 cmd.append("--ipv4")
3854 cmd.append("--compress=%s" % opts.compress)
3857 cmd.append("--magic=%s" % opts.magic)
3859 if exp_size is not None:
3860 cmd.append("--expected-size=%s" % exp_size)
3863 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3866 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3868 if mode == constants.IEM_EXPORT:
3869 # Retry connection a few times when connecting to remote peer
3870 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3871 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3872 elif opts.connect_timeout is not None:
3873 assert mode == constants.IEM_IMPORT
3874 # Overall timeout for establishing connection while listening
3875 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3877 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3879 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3880 # support for receiving a file descriptor for output
3881 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3884 # The import/export name is simply the status directory name
3885 return os.path.basename(status_dir)
3888 shutil.rmtree(status_dir, ignore_errors=True)
3892 def GetImportExportStatus(names):
3893 """Returns import/export daemon status.
3895 @type names: sequence
3896 @param names: List of names
3897 @rtype: List of dicts
3898 @return: Returns a list of the state of each named import/export or None if a
3899 status couldn't be read
3905 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3909 data = utils.ReadFile(status_file)
3910 except EnvironmentError, err:
3911 if err.errno != errno.ENOENT:
3919 result.append(serializer.LoadJson(data))
3924 def AbortImportExport(name):
3925 """Sends SIGTERM to a running import/export daemon.
3928 logging.info("Abort import/export %s", name)
3930 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3931 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3934 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3936 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3939 def CleanupImportExport(name):
3940 """Cleanup after an import or export.
3942 If the import/export daemon is still running it's killed. Afterwards the
3943 whole status directory is removed.
3946 logging.info("Finalizing import/export %s", name)
3948 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3950 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3953 logging.info("Import/export %s is still running with PID %s",
3955 utils.KillProcess(pid, waitpid=False)
3957 shutil.rmtree(status_dir, ignore_errors=True)
3960 def _FindDisks(disks):
3961 """Finds attached L{BlockDev}s for the given disks.
3963 @type disks: list of L{objects.Disk}
3964 @param disks: the disk objects we need to find
3966 @return: list of L{BlockDev} objects or C{None} if a given disk
3967 was not found or was no attached.
3973 rd = _RecursiveFindBD(disk)
3975 _Fail("Can't find device %s", disk)
3980 def DrbdDisconnectNet(disks):
3981 """Disconnects the network on a list of drbd devices.
3984 bdevs = _FindDisks(disks)
3990 except errors.BlockDeviceError, err:
3991 _Fail("Can't change network configuration to standalone mode: %s",
3995 def DrbdAttachNet(disks, instance_name, multimaster):
3996 """Attaches the network on a list of drbd devices.
3999 bdevs = _FindDisks(disks)
4002 for idx, rd in enumerate(bdevs):
4004 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
4005 except EnvironmentError, err:
4006 _Fail("Can't create symlink: %s", err)
4007 # reconnect disks, switch to new master configuration and if
4008 # needed primary mode
4011 rd.AttachNet(multimaster)
4012 except errors.BlockDeviceError, err:
4013 _Fail("Can't change network configuration: %s", err)
4015 # wait until the disks are connected; we need to retry the re-attach
4016 # if the device becomes standalone, as this might happen if the one
4017 # node disconnects and reconnects in a different mode before the
4018 # other node reconnects; in this case, one or both of the nodes will
4019 # decide it has wrong configuration and switch to standalone
4022 all_connected = True
4025 stats = rd.GetProcStatus()
4028 # In the multimaster case we have to wait explicitly until
4029 # the resource is Connected and UpToDate/UpToDate, because
4030 # we promote *both nodes* to primary directly afterwards.
4031 # Being in resync is not enough, since there is a race during which we
4032 # may promote a node with an Outdated disk to primary, effectively
4033 # tearing down the connection.
4034 all_connected = (all_connected and
4035 stats.is_connected and
4036 stats.is_disk_uptodate and
4037 stats.peer_disk_uptodate)
4039 all_connected = (all_connected and
4040 (stats.is_connected or stats.is_in_resync))
4042 if stats.is_standalone:
4043 # peer had different config info and this node became
4044 # standalone, even though this should not happen with the
4045 # new staged way of changing disk configs
4047 rd.AttachNet(multimaster)
4048 except errors.BlockDeviceError, err:
4049 _Fail("Can't change network configuration: %s", err)
4051 if not all_connected:
4052 raise utils.RetryAgain()
4055 # Start with a delay of 100 miliseconds and go up to 5 seconds
4056 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
4057 except utils.RetryTimeout:
4058 _Fail("Timeout in disk reconnecting")
4061 # change to primary mode
4065 except errors.BlockDeviceError, err:
4066 _Fail("Can't change to primary mode: %s", err)
4069 def DrbdWaitSync(disks):
4070 """Wait until DRBDs have synchronized.
4074 stats = rd.GetProcStatus()
4075 if not (stats.is_connected or stats.is_in_resync):
4076 raise utils.RetryAgain()
4079 bdevs = _FindDisks(disks)
4085 # poll each second for 15 seconds
4086 stats = utils.Retry(_helper, 1, 15, args=[rd])
4087 except utils.RetryTimeout:
4088 stats = rd.GetProcStatus()
4090 if not (stats.is_connected or stats.is_in_resync):
4091 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
4092 alldone = alldone and (not stats.is_in_resync)
4093 if stats.sync_percent is not None:
4094 min_resync = min(min_resync, stats.sync_percent)
4096 return (alldone, min_resync)
4099 def DrbdNeedsActivation(disks):
4100 """Checks which of the passed disks needs activation and returns their UUIDs.
4106 rd = _RecursiveFindBD(disk)
4108 faulty_disks.append(disk)
4111 stats = rd.GetProcStatus()
4112 if stats.is_standalone or stats.is_diskless:
4113 faulty_disks.append(disk)
4115 return [disk.uuid for disk in faulty_disks]
4118 def GetDrbdUsermodeHelper():
4119 """Returns DRBD usermode helper currently configured.
4123 return drbd.DRBD8.GetUsermodeHelper()
4124 except errors.BlockDeviceError, err:
4128 def PowercycleNode(hypervisor_type, hvparams=None):
4129 """Hard-powercycle the node.
4131 Because we need to return first, and schedule the powercycle in the
4132 background, we won't be able to report failures nicely.
4135 hyper = hypervisor.GetHypervisor(hypervisor_type)
4139 # if we can't fork, we'll pretend that we're in the child process
4142 return "Reboot scheduled in 5 seconds"
4143 # ensure the child is running on ram
4146 except Exception: # pylint: disable=W0703
4149 hyper.PowercycleNode(hvparams=hvparams)
4152 def _VerifyRestrictedCmdName(cmd):
4153 """Verifies a restricted command name.
4156 @param cmd: Command name
4157 @rtype: tuple; (boolean, string or None)
4158 @return: The tuple's first element is the status; if C{False}, the second
4159 element is an error message string, otherwise it's C{None}
4163 return (False, "Missing command name")
4165 if os.path.basename(cmd) != cmd:
4166 return (False, "Invalid command name")
4168 if not constants.EXT_PLUGIN_MASK.match(cmd):
4169 return (False, "Command name contains forbidden characters")
4174 def _CommonRestrictedCmdCheck(path, owner):
4175 """Common checks for restricted command file system directories and files.
4178 @param path: Path to check
4179 @param owner: C{None} or tuple containing UID and GID
4180 @rtype: tuple; (boolean, string or C{os.stat} result)
4181 @return: The tuple's first element is the status; if C{False}, the second
4182 element is an error message string, otherwise it's the result of C{os.stat}
4186 # Default to root as owner
4191 except EnvironmentError, err:
4192 return (False, "Can't stat(2) '%s': %s" % (path, err))
4194 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4195 return (False, "Permissions on '%s' are too permissive" % path)
4197 if (st.st_uid, st.st_gid) != owner:
4198 (owner_uid, owner_gid) = owner
4199 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4204 def _VerifyRestrictedCmdDirectory(path, _owner=None):
4205 """Verifies restricted command directory.
4208 @param path: Path to check
4209 @rtype: tuple; (boolean, string or None)
4210 @return: The tuple's first element is the status; if C{False}, the second
4211 element is an error message string, otherwise it's C{None}
4214 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4217 return (False, value)
4219 if not stat.S_ISDIR(value.st_mode):
4220 return (False, "Path '%s' is not a directory" % path)
4225 def _VerifyRestrictedCmd(path, cmd, _owner=None):
4226 """Verifies a whole restricted command and returns its executable filename.
4229 @param path: Directory containing restricted commands
4231 @param cmd: Command name
4232 @rtype: tuple; (boolean, string)
4233 @return: The tuple's first element is the status; if C{False}, the second
4234 element is an error message string, otherwise the second element is the
4235 absolute path to the executable
4238 executable = utils.PathJoin(path, cmd)
4240 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4245 if not utils.IsExecutable(executable):
4246 return (False, "access(2) thinks '%s' can't be executed" % executable)
4248 return (True, executable)
4251 def _PrepareRestrictedCmd(path, cmd,
4252 _verify_dir=_VerifyRestrictedCmdDirectory,
4253 _verify_name=_VerifyRestrictedCmdName,
4254 _verify_cmd=_VerifyRestrictedCmd):
4255 """Performs a number of tests on a restricted command.
4258 @param path: Directory containing restricted commands
4260 @param cmd: Command name
4261 @return: Same as L{_VerifyRestrictedCmd}
4264 # Verify the directory first
4265 (status, msg) = _verify_dir(path)
4267 # Check command if everything was alright
4268 (status, msg) = _verify_name(cmd)
4273 # Check actual executable
4274 return _verify_cmd(path, cmd)
4277 def RunRestrictedCmd(cmd,
4278 _lock_timeout=_RCMD_LOCK_TIMEOUT,
4279 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4280 _path=pathutils.RESTRICTED_COMMANDS_DIR,
4281 _sleep_fn=time.sleep,
4282 _prepare_fn=_PrepareRestrictedCmd,
4283 _runcmd_fn=utils.RunCmd,
4284 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4285 """Executes a restricted command after performing strict tests.
4288 @param cmd: Command name
4290 @return: Command output
4291 @raise RPCFail: In case of an error
4294 logging.info("Preparing to run restricted command '%s'", cmd)
4297 _Fail("Restricted commands disabled at configure time")
4303 lock = utils.FileLock.Open(_lock_file)
4304 lock.Exclusive(blocking=True, timeout=_lock_timeout)
4306 (status, value) = _prepare_fn(_path, cmd)
4309 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4310 postfork_fn=lambda _: lock.Unlock())
4312 logging.error(value)
4313 except Exception: # pylint: disable=W0703
4314 # Keep original error in log
4315 logging.exception("Caught exception")
4317 if cmdresult is None:
4318 logging.info("Sleeping for %0.1f seconds before returning",
4319 _RCMD_INVALID_DELAY)
4320 _sleep_fn(_RCMD_INVALID_DELAY)
4322 # Do not include original error message in returned error
4323 _Fail("Executing command '%s' failed" % cmd)
4324 elif cmdresult.failed or cmdresult.fail_reason:
4325 _Fail("Restricted command '%s' failed: %s; output: %s",
4326 cmd, cmdresult.fail_reason, cmdresult.output)
4328 return cmdresult.output
4330 if lock is not None:
4331 # Release lock at last
4336 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4337 """Creates or removes the watcher pause file.
4339 @type until: None or number
4340 @param until: Unix timestamp saying until when the watcher shouldn't run
4344 logging.info("Received request to no longer pause watcher")
4345 utils.RemoveFile(_filename)
4347 logging.info("Received request to pause watcher until %s", until)
4349 if not ht.TNumber(until):
4350 _Fail("Duration must be numeric")
4352 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4355 def ConfigureOVS(ovs_name, ovs_link):
4356 """Creates a OpenvSwitch on the node.
4358 This function sets up a OpenvSwitch on the node with given name nad
4359 connects it via a given eth device.
4361 @type ovs_name: string
4362 @param ovs_name: Name of the OpenvSwitch to create.
4363 @type ovs_link: None or string
4364 @param ovs_link: Ethernet device for outside connection (can be missing)
4367 # Initialize the OpenvSwitch
4368 result = utils.RunCmd(["ovs-vsctl", "add-br", ovs_name])
4370 _Fail("Failed to create openvswitch. Script return value: %s, output: '%s'"
4371 % (result.exit_code, result.output), log=True)
4373 # And connect it to a physical interface, if given
4375 result = utils.RunCmd(["ovs-vsctl", "add-port", ovs_name, ovs_link])
4377 _Fail("Failed to connect openvswitch to interface %s. Script return"
4378 " value: %s, output: '%s'" % (ovs_link, result.exit_code,
4379 result.output), log=True)
4382 class HooksRunner(object):
4385 This class is instantiated on the node side (ganeti-noded) and not
4389 def __init__(self, hooks_base_dir=None):
4390 """Constructor for hooks runner.
4392 @type hooks_base_dir: str or None
4393 @param hooks_base_dir: if not None, this overrides the
4394 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4397 if hooks_base_dir is None:
4398 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4399 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4401 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4403 def RunLocalHooks(self, node_list, hpath, phase, env):
4404 """Check that the hooks will be run only locally and then run them.
4407 assert len(node_list) == 1
4409 _, myself = ssconf.GetMasterAndMyself()
4410 assert node == myself
4412 results = self.RunHooks(hpath, phase, env)
4414 # Return values in the form expected by HooksMaster
4415 return {node: (None, False, results)}
4417 def RunHooks(self, hpath, phase, env):
4418 """Run the scripts in the hooks directory.
4421 @param hpath: the path to the hooks directory which
4424 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4425 L{constants.HOOKS_PHASE_POST}
4427 @param env: dictionary with the environment for the hook
4429 @return: list of 3-element tuples:
4431 - script result, either L{constants.HKR_SUCCESS} or
4432 L{constants.HKR_FAIL}
4433 - output of the script
4435 @raise errors.ProgrammerError: for invalid input
4439 if phase == constants.HOOKS_PHASE_PRE:
4441 elif phase == constants.HOOKS_PHASE_POST:
4444 _Fail("Unknown hooks phase '%s'", phase)
4446 subdir = "%s-%s.d" % (hpath, suffix)
4447 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4451 if not os.path.isdir(dir_name):
4452 # for non-existing/non-dirs, we simply exit instead of logging a
4453 # warning at every operation
4456 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4458 for (relname, relstatus, runresult) in runparts_results:
4459 if relstatus == constants.RUNPARTS_SKIP:
4460 rrval = constants.HKR_SKIP
4462 elif relstatus == constants.RUNPARTS_ERR:
4463 rrval = constants.HKR_FAIL
4464 output = "Hook script execution error: %s" % runresult
4465 elif relstatus == constants.RUNPARTS_RUN:
4466 if runresult.failed:
4467 rrval = constants.HKR_FAIL
4469 rrval = constants.HKR_SUCCESS
4470 output = utils.SafeEncode(runresult.output.strip())
4471 results.append(("%s/%s" % (subdir, relname), rrval, output))
4476 class IAllocatorRunner(object):
4477 """IAllocator runner.
4479 This class is instantiated on the node side (ganeti-noded) and not on
4484 def Run(name, idata):
4485 """Run an iallocator script.
4488 @param name: the iallocator script name
4490 @param idata: the allocator input data
4493 @return: two element tuple of:
4495 - either error message or stdout of allocator (for success)
4498 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4500 if alloc_script is None:
4501 _Fail("iallocator module '%s' not found in the search path", name)
4503 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4507 result = utils.RunCmd([alloc_script, fin_name])
4509 _Fail("iallocator module '%s' failed: %s, output '%s'",
4510 name, result.fail_reason, result.output)
4514 return result.stdout
4517 class DevCacheManager(object):
4518 """Simple class for managing a cache of block device information.
4521 _DEV_PREFIX = "/dev/"
4522 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4525 def _ConvertPath(cls, dev_path):
4526 """Converts a /dev/name path to the cache file name.
4528 This replaces slashes with underscores and strips the /dev
4529 prefix. It then returns the full path to the cache file.
4532 @param dev_path: the C{/dev/} path name
4534 @return: the converted path name
4537 if dev_path.startswith(cls._DEV_PREFIX):
4538 dev_path = dev_path[len(cls._DEV_PREFIX):]
4539 dev_path = dev_path.replace("/", "_")
4540 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4544 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4545 """Updates the cache information for a given device.
4548 @param dev_path: the pathname of the device
4550 @param owner: the owner (instance name) of the device
4551 @type on_primary: bool
4552 @param on_primary: whether this is the primary
4555 @param iv_name: the instance-visible name of the
4556 device, as in objects.Disk.iv_name
4561 if dev_path is None:
4562 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4564 fpath = cls._ConvertPath(dev_path)
4570 iv_name = "not_visible"
4571 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4573 utils.WriteFile(fpath, data=fdata)
4574 except EnvironmentError, err:
4575 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4578 def RemoveCache(cls, dev_path):
4579 """Remove data for a dev_path.
4581 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4582 path name and logging.
4585 @param dev_path: the pathname of the device
4590 if dev_path is None:
4591 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4593 fpath = cls._ConvertPath(dev_path)
4595 utils.RemoveFile(fpath)
4596 except EnvironmentError, err:
4597 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)