4 # Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Functions used by the node daemon
24 @var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
25 the L{UploadFile} function
26 @var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted
27 in the L{_CleanDirectory} function
31 # pylint: disable=E1103
33 # E1103: %s %r has no %r member (but some types could not be
34 # inferred), because the _TryOSFromDisk returns either (True, os_obj)
35 # or (False, "string") which confuses pylint
52 from ganeti import errors
53 from ganeti import utils
54 from ganeti import ssh
55 from ganeti import hypervisor
56 from ganeti import constants
57 from ganeti.storage import bdev
58 from ganeti.storage import drbd
59 from ganeti.storage import filestorage
60 from ganeti import objects
61 from ganeti import ssconf
62 from ganeti import serializer
63 from ganeti import netutils
64 from ganeti import runtime
65 from ganeti import compat
66 from ganeti import pathutils
67 from ganeti import vcluster
69 from ganeti.storage.base import BlockDev
70 from ganeti.storage.drbd import DRBD8
71 from ganeti import hooksmaster
74 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
75 _ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
77 pathutils.JOB_QUEUE_ARCHIVE_DIR,
79 pathutils.CRYPTO_KEYS_DIR,
81 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
82 _X509_KEY_FILE = "key"
83 _X509_CERT_FILE = "cert"
84 _IES_STATUS_FILE = "status"
88 #: Valid LVS output line regex
89 _LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
91 # Actions for the master setup script
92 _MASTER_START = "start"
95 #: Maximum file permissions for restricted command directory and executables
96 _RCMD_MAX_MODE = (stat.S_IRWXU |
97 stat.S_IRGRP | stat.S_IXGRP |
98 stat.S_IROTH | stat.S_IXOTH)
100 #: Delay before returning an error for restricted commands
101 _RCMD_INVALID_DELAY = 10
103 #: How long to wait to acquire lock for restricted commands (shorter than
104 #: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
105 #: command requests arrive
106 _RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
109 class RPCFail(Exception):
110 """Class denoting RPC failure.
112 Its argument is the error message.
117 def _GetInstReasonFilename(instance_name):
118 """Path of the file containing the reason of the instance status change.
120 @type instance_name: string
121 @param instance_name: The name of the instance
123 @return: The path of the file
126 return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
129 def _StoreInstReasonTrail(instance_name, trail):
130 """Serialize a reason trail related to an instance change of state to file.
132 The exact location of the file depends on the name of the instance and on
133 the configuration of the Ganeti cluster defined at deploy time.
135 @type instance_name: string
136 @param instance_name: The name of the instance
140 json = serializer.DumpJson(trail)
141 filename = _GetInstReasonFilename(instance_name)
142 utils.WriteFile(filename, data=json)
145 def _Fail(msg, *args, **kwargs):
146 """Log an error and the raise an RPCFail exception.
148 This exception is then handled specially in the ganeti daemon and
149 turned into a 'failed' return type. As such, this function is a
150 useful shortcut for logging the error and returning it to the master
154 @param msg: the text of the exception
160 if "log" not in kwargs or kwargs["log"]: # if we should log this error
161 if "exc" in kwargs and kwargs["exc"]:
162 logging.exception(msg)
169 """Simple wrapper to return a SimpleStore.
171 @rtype: L{ssconf.SimpleStore}
172 @return: a SimpleStore instance
175 return ssconf.SimpleStore()
178 def _GetSshRunner(cluster_name):
179 """Simple wrapper to return an SshRunner.
181 @type cluster_name: str
182 @param cluster_name: the cluster name, which is needed
183 by the SshRunner constructor
184 @rtype: L{ssh.SshRunner}
185 @return: an SshRunner instance
188 return ssh.SshRunner(cluster_name)
191 def _Decompress(data):
192 """Unpacks data compressed by the RPC client.
194 @type data: list or tuple
195 @param data: Data sent by RPC client
197 @return: Decompressed data
200 assert isinstance(data, (list, tuple))
201 assert len(data) == 2
202 (encoding, content) = data
203 if encoding == constants.RPC_ENCODING_NONE:
205 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
206 return zlib.decompress(base64.b64decode(content))
208 raise AssertionError("Unknown data encoding")
211 def _CleanDirectory(path, exclude=None):
212 """Removes all regular files in a directory.
215 @param path: the directory to clean
217 @param exclude: list of files to be excluded, defaults
221 if path not in _ALLOWED_CLEAN_DIRS:
222 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
225 if not os.path.isdir(path):
230 # Normalize excluded paths
231 exclude = [os.path.normpath(i) for i in exclude]
233 for rel_name in utils.ListVisibleFiles(path):
234 full_name = utils.PathJoin(path, rel_name)
235 if full_name in exclude:
237 if os.path.isfile(full_name) and not os.path.islink(full_name):
238 utils.RemoveFile(full_name)
241 def _BuildUploadFileList():
242 """Build the list of allowed upload files.
244 This is abstracted so that it's built only once at module import time.
247 allowed_files = set([
248 pathutils.CLUSTER_CONF_FILE,
250 pathutils.SSH_KNOWN_HOSTS_FILE,
251 pathutils.VNC_PASSWORD_FILE,
252 pathutils.RAPI_CERT_FILE,
253 pathutils.SPICE_CERT_FILE,
254 pathutils.SPICE_CACERT_FILE,
255 pathutils.RAPI_USERS_FILE,
256 pathutils.CONFD_HMAC_KEY,
257 pathutils.CLUSTER_DOMAIN_SECRET_FILE,
260 for hv_name in constants.HYPER_TYPES:
261 hv_class = hypervisor.GetHypervisorClass(hv_name)
262 allowed_files.update(hv_class.GetAncillaryFiles()[0])
264 assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
265 "Allowed file storage paths should never be uploaded via RPC"
267 return frozenset(allowed_files)
270 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
274 """Removes job queue files and archived jobs.
280 _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
281 _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
285 """Returns master information.
287 This is an utility function to compute master information, either
288 for consumption here or from the node daemon.
291 @return: master_netdev, master_ip, master_name, primary_ip_family,
293 @raise RPCFail: in case of errors
298 master_netdev = cfg.GetMasterNetdev()
299 master_ip = cfg.GetMasterIP()
300 master_netmask = cfg.GetMasterNetmask()
301 master_node = cfg.GetMasterNode()
302 primary_ip_family = cfg.GetPrimaryIPFamily()
303 except errors.ConfigurationError, err:
304 _Fail("Cluster configuration incomplete: %s", err, exc=True)
305 return (master_netdev, master_ip, master_node, primary_ip_family,
309 def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
310 """Decorator that runs hooks before and after the decorated function.
312 @type hook_opcode: string
313 @param hook_opcode: opcode of the hook
314 @type hooks_path: string
315 @param hooks_path: path of the hooks
316 @type env_builder_fn: function
317 @param env_builder_fn: function that returns a dictionary containing the
318 environment variables for the hooks. Will get all the parameters of the
320 @raise RPCFail: in case of pre-hook failure
324 def wrapper(*args, **kwargs):
325 _, myself = ssconf.GetMasterAndMyself()
326 nodes = ([myself], [myself]) # these hooks run locally
328 env_fn = compat.partial(env_builder_fn, *args, **kwargs)
332 hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
333 hr.RunLocalHooks, None, env_fn,
334 logging.warning, cfg.GetClusterName(),
336 hm.RunPhase(constants.HOOKS_PHASE_PRE)
337 result = fn(*args, **kwargs)
338 hm.RunPhase(constants.HOOKS_PHASE_POST)
345 def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
346 """Builds environment variables for master IP hooks.
348 @type master_params: L{objects.MasterNetworkParameters}
349 @param master_params: network parameters of the master
350 @type use_external_mip_script: boolean
351 @param use_external_mip_script: whether to use an external master IP
352 address setup script (unused, but necessary per the implementation of the
353 _RunLocalHooks decorator)
356 # pylint: disable=W0613
357 ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
359 "MASTER_NETDEV": master_params.netdev,
360 "MASTER_IP": master_params.ip,
361 "MASTER_NETMASK": str(master_params.netmask),
362 "CLUSTER_IP_VERSION": str(ver),
368 def _RunMasterSetupScript(master_params, action, use_external_mip_script):
369 """Execute the master IP address setup script.
371 @type master_params: L{objects.MasterNetworkParameters}
372 @param master_params: network parameters of the master
374 @param action: action to pass to the script. Must be one of
375 L{backend._MASTER_START} or L{backend._MASTER_STOP}
376 @type use_external_mip_script: boolean
377 @param use_external_mip_script: whether to use an external master IP
379 @raise backend.RPCFail: if there are errors during the execution of the
383 env = _BuildMasterIpEnv(master_params)
385 if use_external_mip_script:
386 setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
388 setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
390 result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
393 _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
394 (action, result.exit_code, result.output), log=True)
397 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
399 def ActivateMasterIp(master_params, use_external_mip_script):
400 """Activate the IP address of the master daemon.
402 @type master_params: L{objects.MasterNetworkParameters}
403 @param master_params: network parameters of the master
404 @type use_external_mip_script: boolean
405 @param use_external_mip_script: whether to use an external master IP
407 @raise RPCFail: in case of errors during the IP startup
410 _RunMasterSetupScript(master_params, _MASTER_START,
411 use_external_mip_script)
414 def StartMasterDaemons(no_voting):
415 """Activate local node as master node.
417 The function will start the master daemons (ganeti-masterd and ganeti-rapi).
419 @type no_voting: boolean
420 @param no_voting: whether to start ganeti-masterd without a node vote
421 but still non-interactively
427 masterd_args = "--no-voting --yes-do-it"
432 "EXTRA_MASTERD_ARGS": masterd_args,
435 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
437 msg = "Can't start Ganeti master: %s" % result.output
442 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
444 def DeactivateMasterIp(master_params, use_external_mip_script):
445 """Deactivate the master IP on this node.
447 @type master_params: L{objects.MasterNetworkParameters}
448 @param master_params: network parameters of the master
449 @type use_external_mip_script: boolean
450 @param use_external_mip_script: whether to use an external master IP
452 @raise RPCFail: in case of errors during the IP turndown
455 _RunMasterSetupScript(master_params, _MASTER_STOP,
456 use_external_mip_script)
459 def StopMasterDaemons():
460 """Stop the master daemons on this node.
462 Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
467 # TODO: log and report back to the caller the error failures; we
468 # need to decide in which case we fail the RPC for this
470 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
472 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
474 result.cmd, result.exit_code, result.output)
477 def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
478 """Change the netmask of the master IP.
480 @param old_netmask: the old value of the netmask
481 @param netmask: the new value of the netmask
482 @param master_ip: the master IP
483 @param master_netdev: the master network device
486 if old_netmask == netmask:
489 if not netutils.IPAddress.Own(master_ip):
490 _Fail("The master IP address is not up, not attempting to change its"
493 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
494 "%s/%s" % (master_ip, netmask),
495 "dev", master_netdev, "label",
496 "%s:0" % master_netdev])
498 _Fail("Could not set the new netmask on the master IP address")
500 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
501 "%s/%s" % (master_ip, old_netmask),
502 "dev", master_netdev, "label",
503 "%s:0" % master_netdev])
505 _Fail("Could not bring down the master IP address with the old netmask")
508 def EtcHostsModify(mode, host, ip):
509 """Modify a host entry in /etc/hosts.
511 @param mode: The mode to operate. Either add or remove entry
512 @param host: The host to operate on
513 @param ip: The ip associated with the entry
516 if mode == constants.ETC_HOSTS_ADD:
518 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
520 utils.AddHostToEtcHosts(host, ip)
521 elif mode == constants.ETC_HOSTS_REMOVE:
523 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
524 " parameter is present")
525 utils.RemoveHostFromEtcHosts(host)
527 RPCFail("Mode not supported")
530 def LeaveCluster(modify_ssh_setup):
531 """Cleans up and remove the current node.
533 This function cleans up and prepares the current node to be removed
536 If processing is successful, then it raises an
537 L{errors.QuitGanetiException} which is used as a special case to
538 shutdown the node daemon.
540 @param modify_ssh_setup: boolean
543 _CleanDirectory(pathutils.DATA_DIR)
544 _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
549 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
551 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
553 utils.RemoveFile(priv_key)
554 utils.RemoveFile(pub_key)
555 except errors.OpExecError:
556 logging.exception("Error while processing ssh files")
559 utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
560 utils.RemoveFile(pathutils.RAPI_CERT_FILE)
561 utils.RemoveFile(pathutils.SPICE_CERT_FILE)
562 utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
563 utils.RemoveFile(pathutils.NODED_CERT_FILE)
564 except: # pylint: disable=W0702
565 logging.exception("Error while removing cluster secrets")
567 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
569 logging.error("Command %s failed with exitcode %s and error %s",
570 result.cmd, result.exit_code, result.output)
572 # Raise a custom exception (handled in ganeti-noded)
573 raise errors.QuitGanetiException(True, "Shutdown scheduled")
576 def _GetVgInfo(name, excl_stor):
577 """Retrieves information about a LVM volume group.
580 # TODO: GetVGInfo supports returning information for multiple VGs at once
581 vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
583 vg_free = int(round(vginfo[0][0], 0))
584 vg_size = int(round(vginfo[0][1], 0))
590 "type": constants.ST_LVM_VG,
597 def _GetVgSpindlesInfo(name, excl_stor):
598 """Retrieves information about spindles in an LVM volume group.
602 @type excl_stor: bool
603 @param excl_stor: exclusive storage
605 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
606 free spindles, total spindles respectively
610 (vg_free, vg_size) = bdev.LogicalVolume.GetVgSpindlesInfo(name)
615 "type": constants.ST_LVM_PV,
622 def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
623 """Retrieves node information from a hypervisor.
625 The information returned depends on the hypervisor. Common items:
627 - vg_size is the size of the configured volume group in MiB
628 - vg_free is the free size of the volume group in MiB
629 - memory_dom0 is the memory allocated for domain0 in MiB
630 - memory_free is the currently available (free) ram in MiB
631 - memory_total is the total number of ram in MiB
632 - hv_version: the hypervisor version, if available
634 @type hvparams: dict of string
635 @param hvparams: the hypervisor's hvparams
638 return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
641 def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
642 """Retrieves node information for all hypervisors.
644 See C{_GetHvInfo} for information on the output.
646 @type hv_specs: list of pairs (string, dict of strings)
647 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
654 for hvname, hvparams in hv_specs:
655 result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
659 def _GetNamedNodeInfo(names, fn):
660 """Calls C{fn} for all names in C{names} and returns a dictionary.
668 return map(fn, names)
671 def GetNodeInfo(storage_units, hv_specs, excl_stor):
672 """Gives back a hash with different information about the node.
674 @type storage_units: list of pairs (string, string)
675 @param storage_units: List of pairs (storage unit, identifier) to ask for disk
676 space information. In case of lvm-vg, the identifier is
678 @type hv_specs: list of pairs (string, dict of strings)
679 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
680 @type excl_stor: boolean
681 @param excl_stor: Whether exclusive_storage is active
682 @rtype: tuple; (string, None/dict, None/dict)
683 @return: Tuple containing boot ID, volume group information and hypervisor
687 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
688 storage_info = _GetNamedNodeInfo(
690 (lambda storage_unit: _ApplyStorageInfoFunction(storage_unit[0],
693 hv_info = _GetHvInfoAll(hv_specs)
694 return (bootid, storage_info, hv_info)
697 # pylint: disable=W0613
698 def _GetFileStorageSpaceInfo(path, *args):
699 """Wrapper around filestorage.GetSpaceInfo.
701 The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
702 and ignore the *args parameter to not leak it into the filestorage
705 @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
709 return filestorage.GetFileStorageSpaceInfo(path)
712 # FIXME: implement storage reporting for all missing storage types.
713 _STORAGE_TYPE_INFO_FN = {
714 constants.ST_BLOCK: None,
715 constants.ST_DISKLESS: None,
716 constants.ST_EXT: None,
717 constants.ST_FILE: _GetFileStorageSpaceInfo,
718 constants.ST_LVM_PV: _GetVgSpindlesInfo,
719 constants.ST_LVM_VG: _GetVgInfo,
720 constants.ST_RADOS: None,
724 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
725 """Looks up and applies the correct function to calculate free and total
726 storage for the given storage type.
728 @type storage_type: string
729 @param storage_type: the storage type for which the storage shall be reported.
730 @type storage_key: string
731 @param storage_key: identifier of a storage unit, e.g. the volume group name
732 of an LVM storage unit
734 @param args: various parameters that can be used for storage reporting. These
735 parameters and their semantics vary from storage type to storage type and
736 are just propagated in this function.
737 @return: the results of the application of the storage space function (see
738 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
740 @raises NotImplementedError: for storage types who don't support space
743 fn = _STORAGE_TYPE_INFO_FN[storage_type]
745 return fn(storage_key, *args)
747 raise NotImplementedError
750 def _CheckExclusivePvs(pvi_list):
751 """Check that PVs are not shared among LVs
753 @type pvi_list: list of L{objects.LvmPvInfo} objects
754 @param pvi_list: information about the PVs
756 @rtype: list of tuples (string, list of strings)
757 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
762 if len(pvi.lv_list) > 1:
763 res.append((pvi.name, pvi.lv_list))
767 def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
768 get_hv_fn=hypervisor.GetHypervisor):
769 """Verifies the hypervisor. Appends the results to the 'results' list.
772 @param what: a dictionary of things to check
773 @type vm_capable: boolean
774 @param vm_capable: whether or not this node is vm capable
776 @param result: dictionary of verification results; results of the
777 verifications in this function will be added here
778 @type all_hvparams: dict of dict of string
779 @param all_hvparams: dictionary mapping hypervisor names to hvparams
780 @type get_hv_fn: function
781 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
787 if constants.NV_HYPERVISOR in what:
788 result[constants.NV_HYPERVISOR] = {}
789 for hv_name in what[constants.NV_HYPERVISOR]:
790 hvparams = all_hvparams[hv_name]
792 val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
793 except errors.HypervisorError, err:
794 val = "Error while checking hypervisor: %s" % str(err)
795 result[constants.NV_HYPERVISOR][hv_name] = val
798 def _VerifyHvparams(what, vm_capable, result,
799 get_hv_fn=hypervisor.GetHypervisor):
800 """Verifies the hvparams. Appends the results to the 'results' list.
803 @param what: a dictionary of things to check
804 @type vm_capable: boolean
805 @param vm_capable: whether or not this node is vm capable
807 @param result: dictionary of verification results; results of the
808 verifications in this function will be added here
809 @type get_hv_fn: function
810 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
816 if constants.NV_HVPARAMS in what:
817 result[constants.NV_HVPARAMS] = []
818 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
820 logging.info("Validating hv %s, %s", hv_name, hvparms)
821 get_hv_fn(hv_name).ValidateParameters(hvparms)
822 except errors.HypervisorError, err:
823 result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
826 def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
827 """Verifies the instance list.
830 @param what: a dictionary of things to check
831 @type vm_capable: boolean
832 @param vm_capable: whether or not this node is vm capable
834 @param result: dictionary of verification results; results of the
835 verifications in this function will be added here
836 @type all_hvparams: dict of dict of string
837 @param all_hvparams: dictionary mapping hypervisor names to hvparams
840 if constants.NV_INSTANCELIST in what and vm_capable:
841 # GetInstanceList can fail
843 val = GetInstanceList(what[constants.NV_INSTANCELIST],
844 all_hvparams=all_hvparams)
847 result[constants.NV_INSTANCELIST] = val
850 def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
851 """Verifies the node info.
854 @param what: a dictionary of things to check
855 @type vm_capable: boolean
856 @param vm_capable: whether or not this node is vm capable
858 @param result: dictionary of verification results; results of the
859 verifications in this function will be added here
860 @type all_hvparams: dict of dict of string
861 @param all_hvparams: dictionary mapping hypervisor names to hvparams
864 if constants.NV_HVINFO in what and vm_capable:
865 hvname = what[constants.NV_HVINFO]
866 hyper = hypervisor.GetHypervisor(hvname)
867 hvparams = all_hvparams[hvname]
868 result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
871 def VerifyNode(what, cluster_name, all_hvparams):
872 """Verify the status of the local node.
874 Based on the input L{what} parameter, various checks are done on the
877 If the I{filelist} key is present, this list of
878 files is checksummed and the file/checksum pairs are returned.
880 If the I{nodelist} key is present, we check that we have
881 connectivity via ssh with the target nodes (and check the hostname
884 If the I{node-net-test} key is present, we check that we have
885 connectivity to the given nodes via both primary IP and, if
886 applicable, secondary IPs.
889 @param what: a dictionary of things to check:
890 - filelist: list of files for which to compute checksums
891 - nodelist: list of nodes we should check ssh communication with
892 - node-net-test: list of nodes we should check node daemon port
894 - hypervisor: list with hypervisors to run the verify for
895 @type cluster_name: string
896 @param cluster_name: the cluster's name
897 @type all_hvparams: dict of dict of strings
898 @param all_hvparams: a dictionary mapping hypervisor names to hvparams
900 @return: a dictionary with the same keys as the input dict, and
901 values representing the result of the checks
905 my_name = netutils.Hostname.GetSysName()
906 port = netutils.GetDaemonPort(constants.NODED)
907 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
909 _VerifyHypervisors(what, vm_capable, result, all_hvparams)
910 _VerifyHvparams(what, vm_capable, result)
912 if constants.NV_FILELIST in what:
913 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
914 what[constants.NV_FILELIST]))
915 result[constants.NV_FILELIST] = \
916 dict((vcluster.MakeVirtualPath(key), value)
917 for (key, value) in fingerprints.items())
919 if constants.NV_NODELIST in what:
920 (nodes, bynode) = what[constants.NV_NODELIST]
922 # Add nodes from other groups (different for each node)
924 nodes.extend(bynode[my_name])
929 random.shuffle(nodes)
931 # Try to contact all nodes
934 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
938 result[constants.NV_NODELIST] = val
940 if constants.NV_NODENETTEST in what:
941 result[constants.NV_NODENETTEST] = tmp = {}
942 my_pip = my_sip = None
943 for name, pip, sip in what[constants.NV_NODENETTEST]:
949 tmp[my_name] = ("Can't find my own primary/secondary IP"
952 for name, pip, sip in what[constants.NV_NODENETTEST]:
954 if not netutils.TcpPing(pip, port, source=my_pip):
955 fail.append("primary")
957 if not netutils.TcpPing(sip, port, source=my_sip):
958 fail.append("secondary")
960 tmp[name] = ("failure using the %s interface(s)" %
963 if constants.NV_MASTERIP in what:
964 # FIXME: add checks on incoming data structures (here and in the
965 # rest of the function)
966 master_name, master_ip = what[constants.NV_MASTERIP]
967 if master_name == my_name:
968 source = constants.IP4_ADDRESS_LOCALHOST
971 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
974 if constants.NV_USERSCRIPTS in what:
975 result[constants.NV_USERSCRIPTS] = \
976 [script for script in what[constants.NV_USERSCRIPTS]
977 if not utils.IsExecutable(script)]
979 if constants.NV_OOB_PATHS in what:
980 result[constants.NV_OOB_PATHS] = tmp = []
981 for path in what[constants.NV_OOB_PATHS]:
985 tmp.append("error stating out of band helper: %s" % err)
987 if stat.S_ISREG(st.st_mode):
988 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
991 tmp.append("out of band helper %s is not executable" % path)
993 tmp.append("out of band helper %s is not a file" % path)
995 if constants.NV_LVLIST in what and vm_capable:
997 val = GetVolumeList(utils.ListVolumeGroups().keys())
1000 result[constants.NV_LVLIST] = val
1002 _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1004 if constants.NV_VGLIST in what and vm_capable:
1005 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1007 if constants.NV_PVLIST in what and vm_capable:
1008 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1009 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1010 filter_allocatable=False,
1011 include_lvs=check_exclusive_pvs)
1012 if check_exclusive_pvs:
1013 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1015 # Avoid sending useless data on the wire
1017 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1019 if constants.NV_VERSION in what:
1020 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1021 constants.RELEASE_VERSION)
1023 _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1025 if constants.NV_DRBDVERSION in what and vm_capable:
1027 drbd_version = DRBD8.GetProcInfo().GetVersionString()
1028 except errors.BlockDeviceError, err:
1029 logging.warning("Can't get DRBD version", exc_info=True)
1030 drbd_version = str(err)
1031 result[constants.NV_DRBDVERSION] = drbd_version
1033 if constants.NV_DRBDLIST in what and vm_capable:
1035 used_minors = drbd.DRBD8.GetUsedDevs()
1036 except errors.BlockDeviceError, err:
1037 logging.warning("Can't get used minors list", exc_info=True)
1038 used_minors = str(err)
1039 result[constants.NV_DRBDLIST] = used_minors
1041 if constants.NV_DRBDHELPER in what and vm_capable:
1044 payload = drbd.DRBD8.GetUsermodeHelper()
1045 except errors.BlockDeviceError, err:
1046 logging.error("Can't get DRBD usermode helper: %s", str(err))
1049 result[constants.NV_DRBDHELPER] = (status, payload)
1051 if constants.NV_NODESETUP in what:
1052 result[constants.NV_NODESETUP] = tmpr = []
1053 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1054 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1055 " under /sys, missing required directories /sys/block"
1056 " and /sys/class/net")
1057 if (not os.path.isdir("/proc/sys") or
1058 not os.path.isfile("/proc/sysrq-trigger")):
1059 tmpr.append("The procfs filesystem doesn't seem to be mounted"
1060 " under /proc, missing required directory /proc/sys and"
1061 " the file /proc/sysrq-trigger")
1063 if constants.NV_TIME in what:
1064 result[constants.NV_TIME] = utils.SplitTime(time.time())
1066 if constants.NV_OSLIST in what and vm_capable:
1067 result[constants.NV_OSLIST] = DiagnoseOS()
1069 if constants.NV_BRIDGES in what and vm_capable:
1070 result[constants.NV_BRIDGES] = [bridge
1071 for bridge in what[constants.NV_BRIDGES]
1072 if not utils.BridgeExists(bridge)]
1074 if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
1075 result[constants.NV_FILE_STORAGE_PATHS] = \
1076 bdev.ComputeWrongFileStoragePaths()
1081 def GetBlockDevSizes(devices):
1082 """Return the size of the given block devices
1085 @param devices: list of block device nodes to query
1088 dictionary of all block devices under /dev (key). The value is their
1091 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1094 DEV_PREFIX = "/dev/"
1097 for devpath in devices:
1098 if not utils.IsBelowDir(DEV_PREFIX, devpath):
1102 st = os.stat(devpath)
1103 except EnvironmentError, err:
1104 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1107 if stat.S_ISBLK(st.st_mode):
1108 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1110 # We don't want to fail, just do not list this device as available
1111 logging.warning("Cannot get size for block device %s", devpath)
1114 size = int(result.stdout) / (1024 * 1024)
1115 blockdevs[devpath] = size
1119 def GetVolumeList(vg_names):
1120 """Compute list of logical volumes and their size.
1122 @type vg_names: list
1123 @param vg_names: the volume groups whose LVs we should list, or
1124 empty for all volume groups
1127 dictionary of all partions (key) with value being a tuple of
1128 their size (in MiB), inactive and online status::
1130 {'xenvg/test1': ('20.06', True, True)}
1132 in case of errors, a string is returned with the error
1140 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1141 "--separator=%s" % sep,
1142 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1144 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1146 for line in result.stdout.splitlines():
1148 match = _LVSLINE_REGEX.match(line)
1150 logging.error("Invalid line returned from lvs output: '%s'", line)
1152 vg_name, name, size, attr = match.groups()
1153 inactive = attr[4] == "-"
1154 online = attr[5] == "o"
1155 virtual = attr[0] == "v"
1157 # we don't want to report such volumes as existing, since they
1158 # don't really hold data
1160 lvs[vg_name + "/" + name] = (size, inactive, online)
1165 def ListVolumeGroups():
1166 """List the volume groups and their size.
1169 @return: dictionary with keys volume name and values the
1173 return utils.ListVolumeGroups()
1177 """List all volumes on this node.
1181 A list of dictionaries, each having four keys:
1182 - name: the logical volume name,
1183 - size: the size of the logical volume
1184 - dev: the physical device on which the LV lives
1185 - vg: the volume group to which it belongs
1187 In case of errors, we return an empty list and log the
1190 Note that since a logical volume can live on multiple physical
1191 volumes, the resulting list might include a logical volume
1195 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1197 "--options=lv_name,lv_size,devices,vg_name"])
1199 _Fail("Failed to list logical volumes, lvs output: %s",
1203 return dev.split("(")[0]
1205 def handle_dev(dev):
1206 return [parse_dev(x) for x in dev.split(",")]
1209 line = [v.strip() for v in line]
1210 return [{"name": line[0], "size": line[1],
1211 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1214 for line in result.stdout.splitlines():
1215 if line.count("|") >= 3:
1216 all_devs.extend(map_line(line.split("|")))
1218 logging.warning("Strange line in the output from lvs: '%s'", line)
1222 def BridgesExist(bridges_list):
1223 """Check if a list of bridges exist on the current node.
1226 @return: C{True} if all of them exist, C{False} otherwise
1230 for bridge in bridges_list:
1231 if not utils.BridgeExists(bridge):
1232 missing.append(bridge)
1235 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1238 def GetInstanceListForHypervisor(hname, hvparams=None,
1239 get_hv_fn=hypervisor.GetHypervisor):
1240 """Provides a list of instances of the given hypervisor.
1243 @param hname: name of the hypervisor
1244 @type hvparams: dict of strings
1245 @param hvparams: hypervisor parameters for the given hypervisor
1246 @type get_hv_fn: function
1247 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1248 name; optional parameter to increase testability
1251 @return: a list of all running instances on the current node
1252 - instance1.example.com
1253 - instance2.example.com
1258 hv = get_hv_fn(hname)
1259 names = hv.ListInstances(hvparams=hvparams)
1260 results.extend(names)
1261 except errors.HypervisorError, err:
1262 _Fail("Error enumerating instances (hypervisor %s): %s",
1263 hname, err, exc=True)
1267 def GetInstanceList(hypervisor_list, all_hvparams=None,
1268 get_hv_fn=hypervisor.GetHypervisor):
1269 """Provides a list of instances.
1271 @type hypervisor_list: list
1272 @param hypervisor_list: the list of hypervisors to query information
1273 @type all_hvparams: dict of dict of strings
1274 @param all_hvparams: a dictionary mapping hypervisor types to respective
1275 cluster-wide hypervisor parameters
1276 @type get_hv_fn: function
1277 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1278 name; optional parameter to increase testability
1281 @return: a list of all running instances on the current node
1282 - instance1.example.com
1283 - instance2.example.com
1287 for hname in hypervisor_list:
1288 hvparams = all_hvparams[hname]
1289 results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1290 get_hv_fn=get_hv_fn))
1294 def GetInstanceInfo(instance, hname, hvparams=None):
1295 """Gives back the information about an instance as a dictionary.
1297 @type instance: string
1298 @param instance: the instance name
1300 @param hname: the hypervisor type of the instance
1301 @type hvparams: dict of strings
1302 @param hvparams: the instance's hvparams
1305 @return: dictionary with the following keys:
1306 - memory: memory size of instance (int)
1307 - state: xen state of instance (string)
1308 - time: cpu time of instance (float)
1309 - vcpus: the number of vcpus (int)
1314 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1316 if iinfo is not None:
1317 output["memory"] = iinfo[2]
1318 output["vcpus"] = iinfo[3]
1319 output["state"] = iinfo[4]
1320 output["time"] = iinfo[5]
1325 def GetInstanceMigratable(instance):
1326 """Computes whether an instance can be migrated.
1328 @type instance: L{objects.Instance}
1329 @param instance: object representing the instance to be checked.
1332 @return: tuple of (result, description) where:
1333 - result: whether the instance can be migrated or not
1334 - description: a description of the issue, if relevant
1337 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1338 iname = instance.name
1339 if iname not in hyper.ListInstances(instance.hvparams):
1340 _Fail("Instance %s is not running", iname)
1342 for idx in range(len(instance.disks)):
1343 link_name = _GetBlockDevSymlinkPath(iname, idx)
1344 if not os.path.islink(link_name):
1345 logging.warning("Instance %s is missing symlink %s for disk %d",
1346 iname, link_name, idx)
1349 def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1350 """Gather data about all instances.
1352 This is the equivalent of L{GetInstanceInfo}, except that it
1353 computes data for all instances at once, thus being faster if one
1354 needs data about more than one instance.
1356 @type hypervisor_list: list
1357 @param hypervisor_list: list of hypervisors to query for instance data
1358 @type all_hvparams: dict of dict of strings
1359 @param all_hvparams: mapping of hypervisor names to hvparams
1362 @return: dictionary of instance: data, with data having the following keys:
1363 - memory: memory size of instance (int)
1364 - state: xen state of instance (string)
1365 - time: cpu time of instance (float)
1366 - vcpus: the number of vcpus
1371 for hname in hypervisor_list:
1372 hvparams = all_hvparams[hname]
1373 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1375 for name, _, memory, vcpus, state, times in iinfo:
1383 # we only check static parameters, like memory and vcpus,
1384 # and not state and time which can change between the
1385 # invocations of the different hypervisors
1386 for key in "memory", "vcpus":
1387 if value[key] != output[name][key]:
1388 _Fail("Instance %s is running twice"
1389 " with different parameters", name)
1390 output[name] = value
1395 def _InstanceLogName(kind, os_name, instance, component):
1396 """Compute the OS log filename for a given instance and operation.
1398 The instance name and os name are passed in as strings since not all
1399 operations have these as part of an instance object.
1402 @param kind: the operation type (e.g. add, import, etc.)
1403 @type os_name: string
1404 @param os_name: the os name
1405 @type instance: string
1406 @param instance: the name of the instance being imported/added/etc.
1407 @type component: string or None
1408 @param component: the name of the component of the instance being
1412 # TODO: Use tempfile.mkstemp to create unique filename
1414 assert "/" not in component
1415 c_msg = "-%s" % component
1418 base = ("%s-%s-%s%s-%s.log" %
1419 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1420 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1423 def InstanceOsAdd(instance, reinstall, debug):
1424 """Add an OS to an instance.
1426 @type instance: L{objects.Instance}
1427 @param instance: Instance whose OS is to be installed
1428 @type reinstall: boolean
1429 @param reinstall: whether this is an instance reinstall
1430 @type debug: integer
1431 @param debug: debug level, passed to the OS scripts
1435 inst_os = OSFromDisk(instance.os)
1437 create_env = OSEnvironment(instance, inst_os, debug)
1439 create_env["INSTANCE_REINSTALL"] = "1"
1441 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1443 result = utils.RunCmd([inst_os.create_script], env=create_env,
1444 cwd=inst_os.path, output=logfile, reset_env=True)
1446 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1447 " output: %s", result.cmd, result.fail_reason, logfile,
1449 lines = [utils.SafeEncode(val)
1450 for val in utils.TailFile(logfile, lines=20)]
1451 _Fail("OS create script failed (%s), last lines in the"
1452 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1455 def RunRenameInstance(instance, old_name, debug):
1456 """Run the OS rename script for an instance.
1458 @type instance: L{objects.Instance}
1459 @param instance: Instance whose OS is to be installed
1460 @type old_name: string
1461 @param old_name: previous instance name
1462 @type debug: integer
1463 @param debug: debug level, passed to the OS scripts
1465 @return: the success of the operation
1468 inst_os = OSFromDisk(instance.os)
1470 rename_env = OSEnvironment(instance, inst_os, debug)
1471 rename_env["OLD_INSTANCE_NAME"] = old_name
1473 logfile = _InstanceLogName("rename", instance.os,
1474 "%s-%s" % (old_name, instance.name), None)
1476 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1477 cwd=inst_os.path, output=logfile, reset_env=True)
1480 logging.error("os create command '%s' returned error: %s output: %s",
1481 result.cmd, result.fail_reason, result.output)
1482 lines = [utils.SafeEncode(val)
1483 for val in utils.TailFile(logfile, lines=20)]
1484 _Fail("OS rename script failed (%s), last lines in the"
1485 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1488 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1489 """Returns symlink path for block device.
1493 _dir = pathutils.DISK_LINKS_DIR
1495 return utils.PathJoin(_dir,
1497 (instance_name, constants.DISK_SEPARATOR, idx)))
1500 def _SymlinkBlockDev(instance_name, device_path, idx):
1501 """Set up symlinks to a instance's block device.
1503 This is an auxiliary function run when an instance is start (on the primary
1504 node) or when an instance is migrated (on the target node).
1507 @param instance_name: the name of the target instance
1508 @param device_path: path of the physical block device, on the node
1509 @param idx: the disk index
1510 @return: absolute path to the disk's symlink
1513 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1515 os.symlink(device_path, link_name)
1516 except OSError, err:
1517 if err.errno == errno.EEXIST:
1518 if (not os.path.islink(link_name) or
1519 os.readlink(link_name) != device_path):
1520 os.remove(link_name)
1521 os.symlink(device_path, link_name)
1528 def _RemoveBlockDevLinks(instance_name, disks):
1529 """Remove the block device symlinks belonging to the given instance.
1532 for idx, _ in enumerate(disks):
1533 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1534 if os.path.islink(link_name):
1536 os.remove(link_name)
1538 logging.exception("Can't remove symlink '%s'", link_name)
1541 def _GatherAndLinkBlockDevs(instance):
1542 """Set up an instance's block device(s).
1544 This is run on the primary node at instance startup. The block
1545 devices must be already assembled.
1547 @type instance: L{objects.Instance}
1548 @param instance: the instance whose disks we shoul assemble
1550 @return: list of (disk_object, device_path)
1554 for idx, disk in enumerate(instance.disks):
1555 device = _RecursiveFindBD(disk)
1557 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1561 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1563 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1566 block_devices.append((disk, link_name))
1568 return block_devices
1571 def StartInstance(instance, startup_paused, reason, store_reason=True):
1572 """Start an instance.
1574 @type instance: L{objects.Instance}
1575 @param instance: the instance object
1576 @type startup_paused: bool
1577 @param instance: pause instance at startup?
1578 @type reason: list of reasons
1579 @param reason: the reason trail for this startup
1580 @type store_reason: boolean
1581 @param store_reason: whether to store the shutdown reason trail on file
1585 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1588 if instance.name in running_instances:
1589 logging.info("Instance %s already running, not starting", instance.name)
1593 block_devices = _GatherAndLinkBlockDevs(instance)
1594 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1595 hyper.StartInstance(instance, block_devices, startup_paused)
1597 _StoreInstReasonTrail(instance.name, reason)
1598 except errors.BlockDeviceError, err:
1599 _Fail("Block device error: %s", err, exc=True)
1600 except errors.HypervisorError, err:
1601 _RemoveBlockDevLinks(instance.name, instance.disks)
1602 _Fail("Hypervisor error: %s", err, exc=True)
1605 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1606 """Shut an instance down.
1608 @note: this functions uses polling with a hardcoded timeout.
1610 @type instance: L{objects.Instance}
1611 @param instance: the instance object
1612 @type timeout: integer
1613 @param timeout: maximum timeout for soft shutdown
1614 @type reason: list of reasons
1615 @param reason: the reason trail for this shutdown
1616 @type store_reason: boolean
1617 @param store_reason: whether to store the shutdown reason trail on file
1621 hv_name = instance.hypervisor
1622 hyper = hypervisor.GetHypervisor(hv_name)
1623 iname = instance.name
1625 if instance.name not in hyper.ListInstances(instance.hvparams):
1626 logging.info("Instance %s not running, doing nothing", iname)
1631 self.tried_once = False
1634 if iname not in hyper.ListInstances(instance.hvparams):
1638 hyper.StopInstance(instance, retry=self.tried_once)
1640 _StoreInstReasonTrail(instance.name, reason)
1641 except errors.HypervisorError, err:
1642 if iname not in hyper.ListInstances(instance.hvparams):
1643 # if the instance is no longer existing, consider this a
1644 # success and go to cleanup
1647 _Fail("Failed to stop instance %s: %s", iname, err)
1649 self.tried_once = True
1651 raise utils.RetryAgain()
1654 utils.Retry(_TryShutdown(), 5, timeout)
1655 except utils.RetryTimeout:
1656 # the shutdown did not succeed
1657 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1660 hyper.StopInstance(instance, force=True)
1661 except errors.HypervisorError, err:
1662 if iname in hyper.ListInstances(instance.hvparams):
1663 # only raise an error if the instance still exists, otherwise
1664 # the error could simply be "instance ... unknown"!
1665 _Fail("Failed to force stop instance %s: %s", iname, err)
1669 if iname in hyper.ListInstances(instance.hvparams):
1670 _Fail("Could not shutdown instance %s even by destroy", iname)
1673 hyper.CleanupInstance(instance.name)
1674 except errors.HypervisorError, err:
1675 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1677 _RemoveBlockDevLinks(iname, instance.disks)
1680 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1681 """Reboot an instance.
1683 @type instance: L{objects.Instance}
1684 @param instance: the instance object to reboot
1685 @type reboot_type: str
1686 @param reboot_type: the type of reboot, one the following
1688 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1689 instance OS, do not recreate the VM
1690 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1691 restart the VM (at the hypervisor level)
1692 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1693 not accepted here, since that mode is handled differently, in
1694 cmdlib, and translates into full stop and start of the
1695 instance (instead of a call_instance_reboot RPC)
1696 @type shutdown_timeout: integer
1697 @param shutdown_timeout: maximum timeout for soft shutdown
1698 @type reason: list of reasons
1699 @param reason: the reason trail for this reboot
1703 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1706 if instance.name not in running_instances:
1707 _Fail("Cannot reboot instance %s that is not running", instance.name)
1709 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1710 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1712 hyper.RebootInstance(instance)
1713 except errors.HypervisorError, err:
1714 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1715 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1717 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1718 result = StartInstance(instance, False, reason, store_reason=False)
1719 _StoreInstReasonTrail(instance.name, reason)
1721 except errors.HypervisorError, err:
1722 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1724 _Fail("Invalid reboot_type received: %s", reboot_type)
1727 def InstanceBalloonMemory(instance, memory):
1728 """Resize an instance's memory.
1730 @type instance: L{objects.Instance}
1731 @param instance: the instance object
1733 @param memory: new memory amount in MB
1737 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1738 running = hyper.ListInstances(instance.hvparams)
1739 if instance.name not in running:
1740 logging.info("Instance %s is not running, cannot balloon", instance.name)
1743 hyper.BalloonInstanceMemory(instance, memory)
1744 except errors.HypervisorError, err:
1745 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1748 def MigrationInfo(instance):
1749 """Gather information about an instance to be migrated.
1751 @type instance: L{objects.Instance}
1752 @param instance: the instance definition
1755 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1757 info = hyper.MigrationInfo(instance)
1758 except errors.HypervisorError, err:
1759 _Fail("Failed to fetch migration information: %s", err, exc=True)
1763 def AcceptInstance(instance, info, target):
1764 """Prepare the node to accept an instance.
1766 @type instance: L{objects.Instance}
1767 @param instance: the instance definition
1768 @type info: string/data (opaque)
1769 @param info: migration information, from the source node
1770 @type target: string
1771 @param target: target host (usually ip), on this node
1774 # TODO: why is this required only for DTS_EXT_MIRROR?
1775 if instance.disk_template in constants.DTS_EXT_MIRROR:
1776 # Create the symlinks, as the disks are not active
1779 _GatherAndLinkBlockDevs(instance)
1780 except errors.BlockDeviceError, err:
1781 _Fail("Block device error: %s", err, exc=True)
1783 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1785 hyper.AcceptInstance(instance, info, target)
1786 except errors.HypervisorError, err:
1787 if instance.disk_template in constants.DTS_EXT_MIRROR:
1788 _RemoveBlockDevLinks(instance.name, instance.disks)
1789 _Fail("Failed to accept instance: %s", err, exc=True)
1792 def FinalizeMigrationDst(instance, info, success):
1793 """Finalize any preparation to accept an instance.
1795 @type instance: L{objects.Instance}
1796 @param instance: the instance definition
1797 @type info: string/data (opaque)
1798 @param info: migration information, from the source node
1799 @type success: boolean
1800 @param success: whether the migration was a success or a failure
1803 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1805 hyper.FinalizeMigrationDst(instance, info, success)
1806 except errors.HypervisorError, err:
1807 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1810 def MigrateInstance(cluster_name, instance, target, live):
1811 """Migrates an instance to another node.
1813 @type cluster_name: string
1814 @param cluster_name: name of the cluster
1815 @type instance: L{objects.Instance}
1816 @param instance: the instance definition
1817 @type target: string
1818 @param target: the target node name
1820 @param live: whether the migration should be done live or not (the
1821 interpretation of this parameter is left to the hypervisor)
1822 @raise RPCFail: if migration fails for some reason
1825 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1828 hyper.MigrateInstance(cluster_name, instance, target, live)
1829 except errors.HypervisorError, err:
1830 _Fail("Failed to migrate instance: %s", err, exc=True)
1833 def FinalizeMigrationSource(instance, success, live):
1834 """Finalize the instance migration on the source node.
1836 @type instance: L{objects.Instance}
1837 @param instance: the instance definition of the migrated instance
1839 @param success: whether the migration succeeded or not
1841 @param live: whether the user requested a live migration or not
1842 @raise RPCFail: If the execution fails for some reason
1845 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1848 hyper.FinalizeMigrationSource(instance, success, live)
1849 except Exception, err: # pylint: disable=W0703
1850 _Fail("Failed to finalize the migration on the source node: %s", err,
1854 def GetMigrationStatus(instance):
1855 """Get the migration status
1857 @type instance: L{objects.Instance}
1858 @param instance: the instance that is being migrated
1859 @rtype: L{objects.MigrationStatus}
1860 @return: the status of the current migration (one of
1861 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1862 progress info that can be retrieved from the hypervisor
1863 @raise RPCFail: If the migration status cannot be retrieved
1866 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1868 return hyper.GetMigrationStatus(instance)
1869 except Exception, err: # pylint: disable=W0703
1870 _Fail("Failed to get migration status: %s", err, exc=True)
1873 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1874 """Creates a block device for an instance.
1876 @type disk: L{objects.Disk}
1877 @param disk: the object describing the disk we should create
1879 @param size: the size of the physical underlying device, in MiB
1881 @param owner: the name of the instance for which disk is created,
1882 used for device cache data
1883 @type on_primary: boolean
1884 @param on_primary: indicates if it is the primary node or not
1886 @param info: string that will be sent to the physical device
1887 creation, used for example to set (LVM) tags on LVs
1888 @type excl_stor: boolean
1889 @param excl_stor: Whether exclusive_storage is active
1891 @return: the new unique_id of the device (this can sometime be
1892 computed only after creation), or None. On secondary nodes,
1893 it's not required to return anything.
1896 # TODO: remove the obsolete "size" argument
1897 # pylint: disable=W0613
1900 for child in disk.children:
1902 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1903 except errors.BlockDeviceError, err:
1904 _Fail("Can't assemble device %s: %s", child, err)
1905 if on_primary or disk.AssembleOnSecondary():
1906 # we need the children open in case the device itself has to
1909 # pylint: disable=E1103
1911 except errors.BlockDeviceError, err:
1912 _Fail("Can't make child '%s' read-write: %s", child, err)
1916 device = bdev.Create(disk, clist, excl_stor)
1917 except errors.BlockDeviceError, err:
1918 _Fail("Can't create block device: %s", err)
1920 if on_primary or disk.AssembleOnSecondary():
1923 except errors.BlockDeviceError, err:
1924 _Fail("Can't assemble device after creation, unusual event: %s", err)
1925 if on_primary or disk.OpenOnSecondary():
1927 device.Open(force=True)
1928 except errors.BlockDeviceError, err:
1929 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1930 DevCacheManager.UpdateCache(device.dev_path, owner,
1931 on_primary, disk.iv_name)
1933 device.SetInfo(info)
1935 return device.unique_id
1938 def _WipeDevice(path, offset, size):
1939 """This function actually wipes the device.
1941 @param path: The path to the device to wipe
1942 @param offset: The offset in MiB in the file
1943 @param size: The size in MiB to write
1946 # Internal sizes are always in Mebibytes; if the following "dd" command
1947 # should use a different block size the offset and size given to this
1948 # function must be adjusted accordingly before being passed to "dd".
1949 block_size = 1024 * 1024
1951 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1952 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1954 result = utils.RunCmd(cmd)
1957 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1958 result.fail_reason, result.output)
1961 def BlockdevWipe(disk, offset, size):
1962 """Wipes a block device.
1964 @type disk: L{objects.Disk}
1965 @param disk: the disk object we want to wipe
1967 @param offset: The offset in MiB in the file
1969 @param size: The size in MiB to write
1973 rdev = _RecursiveFindBD(disk)
1974 except errors.BlockDeviceError:
1978 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1980 # Do cross verify some of the parameters
1982 _Fail("Negative offset")
1984 _Fail("Negative size")
1985 if offset > rdev.size:
1986 _Fail("Offset is bigger than device size")
1987 if (offset + size) > rdev.size:
1988 _Fail("The provided offset and size to wipe is bigger than device size")
1990 _WipeDevice(rdev.dev_path, offset, size)
1993 def BlockdevPauseResumeSync(disks, pause):
1994 """Pause or resume the sync of the block device.
1996 @type disks: list of L{objects.Disk}
1997 @param disks: the disks object we want to pause/resume
1999 @param pause: Wheater to pause or resume
2005 rdev = _RecursiveFindBD(disk)
2006 except errors.BlockDeviceError:
2010 success.append((False, ("Cannot change sync for device %s:"
2011 " device not found" % disk.iv_name)))
2014 result = rdev.PauseResumeSync(pause)
2017 success.append((result, None))
2023 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2028 def BlockdevRemove(disk):
2029 """Remove a block device.
2031 @note: This is intended to be called recursively.
2033 @type disk: L{objects.Disk}
2034 @param disk: the disk object we should remove
2036 @return: the success of the operation
2041 rdev = _RecursiveFindBD(disk)
2042 except errors.BlockDeviceError, err:
2043 # probably can't attach
2044 logging.info("Can't attach to device %s in remove", disk)
2046 if rdev is not None:
2047 r_path = rdev.dev_path
2050 except errors.BlockDeviceError, err:
2051 msgs.append(str(err))
2053 DevCacheManager.RemoveCache(r_path)
2056 for child in disk.children:
2058 BlockdevRemove(child)
2059 except RPCFail, err:
2060 msgs.append(str(err))
2063 _Fail("; ".join(msgs))
2066 def _RecursiveAssembleBD(disk, owner, as_primary):
2067 """Activate a block device for an instance.
2069 This is run on the primary and secondary nodes for an instance.
2071 @note: this function is called recursively.
2073 @type disk: L{objects.Disk}
2074 @param disk: the disk we try to assemble
2076 @param owner: the name of the instance which owns the disk
2077 @type as_primary: boolean
2078 @param as_primary: if we should make the block device
2081 @return: the assembled device or None (in case no device
2083 @raise errors.BlockDeviceError: in case there is an error
2084 during the activation of the children or the device
2090 mcn = disk.ChildrenNeeded()
2092 mcn = 0 # max number of Nones allowed
2094 mcn = len(disk.children) - mcn # max number of Nones
2095 for chld_disk in disk.children:
2097 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2098 except errors.BlockDeviceError, err:
2099 if children.count(None) >= mcn:
2102 logging.error("Error in child activation (but continuing): %s",
2104 children.append(cdev)
2106 if as_primary or disk.AssembleOnSecondary():
2107 r_dev = bdev.Assemble(disk, children)
2109 if as_primary or disk.OpenOnSecondary():
2111 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2112 as_primary, disk.iv_name)
2119 def BlockdevAssemble(disk, owner, as_primary, idx):
2120 """Activate a block device for an instance.
2122 This is a wrapper over _RecursiveAssembleBD.
2124 @rtype: str or boolean
2125 @return: a C{/dev/...} path for primary nodes, and
2126 C{True} for secondary nodes
2130 result = _RecursiveAssembleBD(disk, owner, as_primary)
2131 if isinstance(result, BlockDev):
2132 # pylint: disable=E1103
2133 result = result.dev_path
2135 _SymlinkBlockDev(owner, result, idx)
2136 except errors.BlockDeviceError, err:
2137 _Fail("Error while assembling disk: %s", err, exc=True)
2138 except OSError, err:
2139 _Fail("Error while symlinking disk: %s", err, exc=True)
2144 def BlockdevShutdown(disk):
2145 """Shut down a block device.
2147 First, if the device is assembled (Attach() is successful), then
2148 the device is shutdown. Then the children of the device are
2151 This function is called recursively. Note that we don't cache the
2152 children or such, as oppossed to assemble, shutdown of different
2153 devices doesn't require that the upper device was active.
2155 @type disk: L{objects.Disk}
2156 @param disk: the description of the disk we should
2162 r_dev = _RecursiveFindBD(disk)
2163 if r_dev is not None:
2164 r_path = r_dev.dev_path
2167 DevCacheManager.RemoveCache(r_path)
2168 except errors.BlockDeviceError, err:
2169 msgs.append(str(err))
2172 for child in disk.children:
2174 BlockdevShutdown(child)
2175 except RPCFail, err:
2176 msgs.append(str(err))
2179 _Fail("; ".join(msgs))
2182 def BlockdevAddchildren(parent_cdev, new_cdevs):
2183 """Extend a mirrored block device.
2185 @type parent_cdev: L{objects.Disk}
2186 @param parent_cdev: the disk to which we should add children
2187 @type new_cdevs: list of L{objects.Disk}
2188 @param new_cdevs: the list of children which we should add
2192 parent_bdev = _RecursiveFindBD(parent_cdev)
2193 if parent_bdev is None:
2194 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2195 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2196 if new_bdevs.count(None) > 0:
2197 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2198 parent_bdev.AddChildren(new_bdevs)
2201 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2202 """Shrink a mirrored block device.
2204 @type parent_cdev: L{objects.Disk}
2205 @param parent_cdev: the disk from which we should remove children
2206 @type new_cdevs: list of L{objects.Disk}
2207 @param new_cdevs: the list of children which we should remove
2211 parent_bdev = _RecursiveFindBD(parent_cdev)
2212 if parent_bdev is None:
2213 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2215 for disk in new_cdevs:
2216 rpath = disk.StaticDevPath()
2218 bd = _RecursiveFindBD(disk)
2220 _Fail("Can't find device %s while removing children", disk)
2222 devs.append(bd.dev_path)
2224 if not utils.IsNormAbsPath(rpath):
2225 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2227 parent_bdev.RemoveChildren(devs)
2230 def BlockdevGetmirrorstatus(disks):
2231 """Get the mirroring status of a list of devices.
2233 @type disks: list of L{objects.Disk}
2234 @param disks: the list of disks which we should query
2236 @return: List of L{objects.BlockDevStatus}, one for each disk
2237 @raise errors.BlockDeviceError: if any of the disks cannot be
2243 rbd = _RecursiveFindBD(dsk)
2245 _Fail("Can't find device %s", dsk)
2247 stats.append(rbd.CombinedSyncStatus())
2252 def BlockdevGetmirrorstatusMulti(disks):
2253 """Get the mirroring status of a list of devices.
2255 @type disks: list of L{objects.Disk}
2256 @param disks: the list of disks which we should query
2258 @return: List of tuples, (bool, status), one for each disk; bool denotes
2259 success/failure, status is L{objects.BlockDevStatus} on success, string
2266 rbd = _RecursiveFindBD(disk)
2268 result.append((False, "Can't find device %s" % disk))
2271 status = rbd.CombinedSyncStatus()
2272 except errors.BlockDeviceError, err:
2273 logging.exception("Error while getting disk status")
2274 result.append((False, str(err)))
2276 result.append((True, status))
2278 assert len(disks) == len(result)
2283 def _RecursiveFindBD(disk):
2284 """Check if a device is activated.
2286 If so, return information about the real device.
2288 @type disk: L{objects.Disk}
2289 @param disk: the disk object we need to find
2291 @return: None if the device can't be found,
2292 otherwise the device instance
2297 for chdisk in disk.children:
2298 children.append(_RecursiveFindBD(chdisk))
2300 return bdev.FindDevice(disk, children)
2303 def _OpenRealBD(disk):
2304 """Opens the underlying block device of a disk.
2306 @type disk: L{objects.Disk}
2307 @param disk: the disk object we want to open
2310 real_disk = _RecursiveFindBD(disk)
2311 if real_disk is None:
2312 _Fail("Block device '%s' is not set up", disk)
2319 def BlockdevFind(disk):
2320 """Check if a device is activated.
2322 If it is, return information about the real device.
2324 @type disk: L{objects.Disk}
2325 @param disk: the disk to find
2326 @rtype: None or objects.BlockDevStatus
2327 @return: None if the disk cannot be found, otherwise a the current
2332 rbd = _RecursiveFindBD(disk)
2333 except errors.BlockDeviceError, err:
2334 _Fail("Failed to find device: %s", err, exc=True)
2339 return rbd.GetSyncStatus()
2342 def BlockdevGetdimensions(disks):
2343 """Computes the size of the given disks.
2345 If a disk is not found, returns None instead.
2347 @type disks: list of L{objects.Disk}
2348 @param disks: the list of disk to compute the size for
2350 @return: list with elements None if the disk cannot be found,
2351 otherwise the pair (size, spindles), where spindles is None if the
2352 device doesn't support that
2358 rbd = _RecursiveFindBD(cf)
2359 except errors.BlockDeviceError:
2365 result.append(rbd.GetActualDimensions())
2369 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2370 """Export a block device to a remote node.
2372 @type disk: L{objects.Disk}
2373 @param disk: the description of the disk to export
2374 @type dest_node: str
2375 @param dest_node: the destination node to export to
2376 @type dest_path: str
2377 @param dest_path: the destination path on the target node
2378 @type cluster_name: str
2379 @param cluster_name: the cluster name, needed for SSH hostalias
2383 real_disk = _OpenRealBD(disk)
2385 # the block size on the read dd is 1MiB to match our units
2386 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2387 "dd if=%s bs=1048576 count=%s",
2388 real_disk.dev_path, str(disk.size))
2390 # we set here a smaller block size as, due to ssh buffering, more
2391 # than 64-128k will mostly ignored; we use nocreat to fail if the
2392 # device is not already there or we pass a wrong path; we use
2393 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2394 # to not buffer too much memory; this means that at best, we flush
2395 # every 64k, which will not be very fast
2396 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2397 " oflag=dsync", dest_path)
2399 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2400 constants.SSH_LOGIN_USER,
2403 # all commands have been checked, so we're safe to combine them
2404 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2406 result = utils.RunCmd(["bash", "-c", command])
2409 _Fail("Disk copy command '%s' returned error: %s"
2410 " output: %s", command, result.fail_reason, result.output)
2413 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2414 """Write a file to the filesystem.
2416 This allows the master to overwrite(!) a file. It will only perform
2417 the operation if the file belongs to a list of configuration files.
2419 @type file_name: str
2420 @param file_name: the target file name
2422 @param data: the new contents of the file
2424 @param mode: the mode to give the file (can be None)
2426 @param uid: the owner of the file
2428 @param gid: the group of the file
2430 @param atime: the atime to set on the file (can be None)
2432 @param mtime: the mtime to set on the file (can be None)
2436 file_name = vcluster.LocalizeVirtualPath(file_name)
2438 if not os.path.isabs(file_name):
2439 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2441 if file_name not in _ALLOWED_UPLOAD_FILES:
2442 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2445 raw_data = _Decompress(data)
2447 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2448 _Fail("Invalid username/groupname type")
2450 getents = runtime.GetEnts()
2451 uid = getents.LookupUser(uid)
2452 gid = getents.LookupGroup(gid)
2454 utils.SafeWriteFile(file_name, None,
2455 data=raw_data, mode=mode, uid=uid, gid=gid,
2456 atime=atime, mtime=mtime)
2459 def RunOob(oob_program, command, node, timeout):
2460 """Executes oob_program with given command on given node.
2462 @param oob_program: The path to the executable oob_program
2463 @param command: The command to invoke on oob_program
2464 @param node: The node given as an argument to the program
2465 @param timeout: Timeout after which we kill the oob program
2468 @raise RPCFail: If execution fails for some reason
2471 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2474 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2475 result.fail_reason, result.output)
2477 return result.stdout
2480 def _OSOndiskAPIVersion(os_dir):
2481 """Compute and return the API version of a given OS.
2483 This function will try to read the API version of the OS residing in
2484 the 'os_dir' directory.
2487 @param os_dir: the directory in which we should look for the OS
2489 @return: tuple (status, data) with status denoting the validity and
2490 data holding either the vaid versions or an error message
2493 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2496 st = os.stat(api_file)
2497 except EnvironmentError, err:
2498 return False, ("Required file '%s' not found under path %s: %s" %
2499 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2501 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2502 return False, ("File '%s' in %s is not a regular file" %
2503 (constants.OS_API_FILE, os_dir))
2506 api_versions = utils.ReadFile(api_file).splitlines()
2507 except EnvironmentError, err:
2508 return False, ("Error while reading the API version file at %s: %s" %
2509 (api_file, utils.ErrnoOrStr(err)))
2512 api_versions = [int(version.strip()) for version in api_versions]
2513 except (TypeError, ValueError), err:
2514 return False, ("API version(s) can't be converted to integer: %s" %
2517 return True, api_versions
2520 def DiagnoseOS(top_dirs=None):
2521 """Compute the validity for all OSes.
2523 @type top_dirs: list
2524 @param top_dirs: the list of directories in which to
2525 search (if not given defaults to
2526 L{pathutils.OS_SEARCH_PATH})
2527 @rtype: list of L{objects.OS}
2528 @return: a list of tuples (name, path, status, diagnose, variants,
2529 parameters, api_version) for all (potential) OSes under all
2530 search paths, where:
2531 - name is the (potential) OS name
2532 - path is the full path to the OS
2533 - status True/False is the validity of the OS
2534 - diagnose is the error message for an invalid OS, otherwise empty
2535 - variants is a list of supported OS variants, if any
2536 - parameters is a list of (name, help) parameters, if any
2537 - api_version is a list of support OS API versions
2540 if top_dirs is None:
2541 top_dirs = pathutils.OS_SEARCH_PATH
2544 for dir_name in top_dirs:
2545 if os.path.isdir(dir_name):
2547 f_names = utils.ListVisibleFiles(dir_name)
2548 except EnvironmentError, err:
2549 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2551 for name in f_names:
2552 os_path = utils.PathJoin(dir_name, name)
2553 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2556 variants = os_inst.supported_variants
2557 parameters = os_inst.supported_parameters
2558 api_versions = os_inst.api_versions
2561 variants = parameters = api_versions = []
2562 result.append((name, os_path, status, diagnose, variants,
2563 parameters, api_versions))
2568 def _TryOSFromDisk(name, base_dir=None):
2569 """Create an OS instance from disk.
2571 This function will return an OS instance if the given name is a
2574 @type base_dir: string
2575 @keyword base_dir: Base directory containing OS installations.
2576 Defaults to a search in all the OS_SEARCH_PATH dirs.
2578 @return: success and either the OS instance if we find a valid one,
2582 if base_dir is None:
2583 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2585 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2588 return False, "Directory for OS %s not found in search path" % name
2590 status, api_versions = _OSOndiskAPIVersion(os_dir)
2593 return status, api_versions
2595 if not constants.OS_API_VERSIONS.intersection(api_versions):
2596 return False, ("API version mismatch for path '%s': found %s, want %s." %
2597 (os_dir, api_versions, constants.OS_API_VERSIONS))
2599 # OS Files dictionary, we will populate it with the absolute path
2600 # names; if the value is True, then it is a required file, otherwise
2602 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2604 if max(api_versions) >= constants.OS_API_V15:
2605 os_files[constants.OS_VARIANTS_FILE] = False
2607 if max(api_versions) >= constants.OS_API_V20:
2608 os_files[constants.OS_PARAMETERS_FILE] = True
2610 del os_files[constants.OS_SCRIPT_VERIFY]
2612 for (filename, required) in os_files.items():
2613 os_files[filename] = utils.PathJoin(os_dir, filename)
2616 st = os.stat(os_files[filename])
2617 except EnvironmentError, err:
2618 if err.errno == errno.ENOENT and not required:
2619 del os_files[filename]
2621 return False, ("File '%s' under path '%s' is missing (%s)" %
2622 (filename, os_dir, utils.ErrnoOrStr(err)))
2624 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2625 return False, ("File '%s' under path '%s' is not a regular file" %
2628 if filename in constants.OS_SCRIPTS:
2629 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2630 return False, ("File '%s' under path '%s' is not executable" %
2634 if constants.OS_VARIANTS_FILE in os_files:
2635 variants_file = os_files[constants.OS_VARIANTS_FILE]
2638 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2639 except EnvironmentError, err:
2640 # we accept missing files, but not other errors
2641 if err.errno != errno.ENOENT:
2642 return False, ("Error while reading the OS variants file at %s: %s" %
2643 (variants_file, utils.ErrnoOrStr(err)))
2646 if constants.OS_PARAMETERS_FILE in os_files:
2647 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2649 parameters = utils.ReadFile(parameters_file).splitlines()
2650 except EnvironmentError, err:
2651 return False, ("Error while reading the OS parameters file at %s: %s" %
2652 (parameters_file, utils.ErrnoOrStr(err)))
2653 parameters = [v.split(None, 1) for v in parameters]
2655 os_obj = objects.OS(name=name, path=os_dir,
2656 create_script=os_files[constants.OS_SCRIPT_CREATE],
2657 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2658 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2659 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2660 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2662 supported_variants=variants,
2663 supported_parameters=parameters,
2664 api_versions=api_versions)
2668 def OSFromDisk(name, base_dir=None):
2669 """Create an OS instance from disk.
2671 This function will return an OS instance if the given name is a
2672 valid OS name. Otherwise, it will raise an appropriate
2673 L{RPCFail} exception, detailing why this is not a valid OS.
2675 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2676 an exception but returns true/false status data.
2678 @type base_dir: string
2679 @keyword base_dir: Base directory containing OS installations.
2680 Defaults to a search in all the OS_SEARCH_PATH dirs.
2681 @rtype: L{objects.OS}
2682 @return: the OS instance if we find a valid one
2683 @raise RPCFail: if we don't find a valid OS
2686 name_only = objects.OS.GetName(name)
2687 status, payload = _TryOSFromDisk(name_only, base_dir)
2695 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2696 """Calculate the basic environment for an os script.
2699 @param os_name: full operating system name (including variant)
2700 @type inst_os: L{objects.OS}
2701 @param inst_os: operating system for which the environment is being built
2702 @type os_params: dict
2703 @param os_params: the OS parameters
2704 @type debug: integer
2705 @param debug: debug level (0 or 1, for OS Api 10)
2707 @return: dict of environment variables
2708 @raise errors.BlockDeviceError: if the block device
2714 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2715 result["OS_API_VERSION"] = "%d" % api_version
2716 result["OS_NAME"] = inst_os.name
2717 result["DEBUG_LEVEL"] = "%d" % debug
2720 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2721 variant = objects.OS.GetVariant(os_name)
2723 variant = inst_os.supported_variants[0]
2726 result["OS_VARIANT"] = variant
2729 for pname, pvalue in os_params.items():
2730 result["OSP_%s" % pname.upper()] = pvalue
2732 # Set a default path otherwise programs called by OS scripts (or
2733 # even hooks called from OS scripts) might break, and we don't want
2734 # to have each script require setting a PATH variable
2735 result["PATH"] = constants.HOOKS_PATH
2740 def OSEnvironment(instance, inst_os, debug=0):
2741 """Calculate the environment for an os script.
2743 @type instance: L{objects.Instance}
2744 @param instance: target instance for the os script run
2745 @type inst_os: L{objects.OS}
2746 @param inst_os: operating system for which the environment is being built
2747 @type debug: integer
2748 @param debug: debug level (0 or 1, for OS Api 10)
2750 @return: dict of environment variables
2751 @raise errors.BlockDeviceError: if the block device
2755 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2757 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2758 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2760 result["HYPERVISOR"] = instance.hypervisor
2761 result["DISK_COUNT"] = "%d" % len(instance.disks)
2762 result["NIC_COUNT"] = "%d" % len(instance.nics)
2763 result["INSTANCE_SECONDARY_NODES"] = \
2764 ("%s" % " ".join(instance.secondary_nodes))
2767 for idx, disk in enumerate(instance.disks):
2768 real_disk = _OpenRealBD(disk)
2769 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2770 result["DISK_%d_ACCESS" % idx] = disk.mode
2771 if constants.HV_DISK_TYPE in instance.hvparams:
2772 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2773 instance.hvparams[constants.HV_DISK_TYPE]
2774 if disk.dev_type in constants.LDS_BLOCK:
2775 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2776 elif disk.dev_type == constants.LD_FILE:
2777 result["DISK_%d_BACKEND_TYPE" % idx] = \
2778 "file:%s" % disk.physical_id[0]
2781 for idx, nic in enumerate(instance.nics):
2782 result["NIC_%d_MAC" % idx] = nic.mac
2784 result["NIC_%d_IP" % idx] = nic.ip
2785 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2786 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2787 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2788 if nic.nicparams[constants.NIC_LINK]:
2789 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2791 nobj = objects.Network.FromDict(nic.netinfo)
2792 result.update(nobj.HooksDict("NIC_%d_" % idx))
2793 if constants.HV_NIC_TYPE in instance.hvparams:
2794 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2795 instance.hvparams[constants.HV_NIC_TYPE]
2798 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2799 for key, value in source.items():
2800 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2805 def DiagnoseExtStorage(top_dirs=None):
2806 """Compute the validity for all ExtStorage Providers.
2808 @type top_dirs: list
2809 @param top_dirs: the list of directories in which to
2810 search (if not given defaults to
2811 L{pathutils.ES_SEARCH_PATH})
2812 @rtype: list of L{objects.ExtStorage}
2813 @return: a list of tuples (name, path, status, diagnose, parameters)
2814 for all (potential) ExtStorage Providers under all
2815 search paths, where:
2816 - name is the (potential) ExtStorage Provider
2817 - path is the full path to the ExtStorage Provider
2818 - status True/False is the validity of the ExtStorage Provider
2819 - diagnose is the error message for an invalid ExtStorage Provider,
2821 - parameters is a list of (name, help) parameters, if any
2824 if top_dirs is None:
2825 top_dirs = pathutils.ES_SEARCH_PATH
2828 for dir_name in top_dirs:
2829 if os.path.isdir(dir_name):
2831 f_names = utils.ListVisibleFiles(dir_name)
2832 except EnvironmentError, err:
2833 logging.exception("Can't list the ExtStorage directory %s: %s",
2836 for name in f_names:
2837 es_path = utils.PathJoin(dir_name, name)
2838 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2841 parameters = es_inst.supported_parameters
2845 result.append((name, es_path, status, diagnose, parameters))
2850 def BlockdevGrow(disk, amount, dryrun, backingstore):
2851 """Grow a stack of block devices.
2853 This function is called recursively, with the childrens being the
2854 first ones to resize.
2856 @type disk: L{objects.Disk}
2857 @param disk: the disk to be grown
2858 @type amount: integer
2859 @param amount: the amount (in mebibytes) to grow with
2860 @type dryrun: boolean
2861 @param dryrun: whether to execute the operation in simulation mode
2862 only, without actually increasing the size
2863 @param backingstore: whether to execute the operation on backing storage
2864 only, or on "logical" storage only; e.g. DRBD is logical storage,
2865 whereas LVM, file, RBD are backing storage
2866 @rtype: (status, result)
2867 @return: a tuple with the status of the operation (True/False), and
2868 the errors message if status is False
2871 r_dev = _RecursiveFindBD(disk)
2873 _Fail("Cannot find block device %s", disk)
2876 r_dev.Grow(amount, dryrun, backingstore)
2877 except errors.BlockDeviceError, err:
2878 _Fail("Failed to grow block device: %s", err, exc=True)
2881 def BlockdevSnapshot(disk):
2882 """Create a snapshot copy of a block device.
2884 This function is called recursively, and the snapshot is actually created
2885 just for the leaf lvm backend device.
2887 @type disk: L{objects.Disk}
2888 @param disk: the disk to be snapshotted
2890 @return: snapshot disk ID as (vg, lv)
2893 if disk.dev_type == constants.LD_DRBD8:
2894 if not disk.children:
2895 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2897 return BlockdevSnapshot(disk.children[0])
2898 elif disk.dev_type == constants.LD_LV:
2899 r_dev = _RecursiveFindBD(disk)
2900 if r_dev is not None:
2901 # FIXME: choose a saner value for the snapshot size
2902 # let's stay on the safe side and ask for the full size, for now
2903 return r_dev.Snapshot(disk.size)
2905 _Fail("Cannot find block device %s", disk)
2907 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2908 disk.unique_id, disk.dev_type)
2911 def BlockdevSetInfo(disk, info):
2912 """Sets 'metadata' information on block devices.
2914 This function sets 'info' metadata on block devices. Initial
2915 information is set at device creation; this function should be used
2916 for example after renames.
2918 @type disk: L{objects.Disk}
2919 @param disk: the disk to be grown
2921 @param info: new 'info' metadata
2922 @rtype: (status, result)
2923 @return: a tuple with the status of the operation (True/False), and
2924 the errors message if status is False
2927 r_dev = _RecursiveFindBD(disk)
2929 _Fail("Cannot find block device %s", disk)
2933 except errors.BlockDeviceError, err:
2934 _Fail("Failed to set information on block device: %s", err, exc=True)
2937 def FinalizeExport(instance, snap_disks):
2938 """Write out the export configuration information.
2940 @type instance: L{objects.Instance}
2941 @param instance: the instance which we export, used for
2942 saving configuration
2943 @type snap_disks: list of L{objects.Disk}
2944 @param snap_disks: list of snapshot block devices, which
2945 will be used to get the actual name of the dump file
2950 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2951 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2953 config = objects.SerializableConfigParser()
2955 config.add_section(constants.INISECT_EXP)
2956 config.set(constants.INISECT_EXP, "version", "0")
2957 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2958 config.set(constants.INISECT_EXP, "source", instance.primary_node)
2959 config.set(constants.INISECT_EXP, "os", instance.os)
2960 config.set(constants.INISECT_EXP, "compression", "none")
2962 config.add_section(constants.INISECT_INS)
2963 config.set(constants.INISECT_INS, "name", instance.name)
2964 config.set(constants.INISECT_INS, "maxmem", "%d" %
2965 instance.beparams[constants.BE_MAXMEM])
2966 config.set(constants.INISECT_INS, "minmem", "%d" %
2967 instance.beparams[constants.BE_MINMEM])
2968 # "memory" is deprecated, but useful for exporting to old ganeti versions
2969 config.set(constants.INISECT_INS, "memory", "%d" %
2970 instance.beparams[constants.BE_MAXMEM])
2971 config.set(constants.INISECT_INS, "vcpus", "%d" %
2972 instance.beparams[constants.BE_VCPUS])
2973 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2974 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2975 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2978 for nic_count, nic in enumerate(instance.nics):
2980 config.set(constants.INISECT_INS, "nic%d_mac" %
2981 nic_count, "%s" % nic.mac)
2982 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2983 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2985 for param in constants.NICS_PARAMETER_TYPES:
2986 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2987 "%s" % nic.nicparams.get(param, None))
2988 # TODO: redundant: on load can read nics until it doesn't exist
2989 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2992 for disk_count, disk in enumerate(snap_disks):
2995 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2996 ("%s" % disk.iv_name))
2997 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2998 ("%s" % disk.physical_id[1]))
2999 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3002 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3004 # New-style hypervisor/backend parameters
3006 config.add_section(constants.INISECT_HYP)
3007 for name, value in instance.hvparams.items():
3008 if name not in constants.HVC_GLOBALS:
3009 config.set(constants.INISECT_HYP, name, str(value))
3011 config.add_section(constants.INISECT_BEP)
3012 for name, value in instance.beparams.items():
3013 config.set(constants.INISECT_BEP, name, str(value))
3015 config.add_section(constants.INISECT_OSP)
3016 for name, value in instance.osparams.items():
3017 config.set(constants.INISECT_OSP, name, str(value))
3019 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3020 data=config.Dumps())
3021 shutil.rmtree(finaldestdir, ignore_errors=True)
3022 shutil.move(destdir, finaldestdir)
3025 def ExportInfo(dest):
3026 """Get export configuration information.
3029 @param dest: directory containing the export
3031 @rtype: L{objects.SerializableConfigParser}
3032 @return: a serializable config file containing the
3036 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3038 config = objects.SerializableConfigParser()
3041 if (not config.has_section(constants.INISECT_EXP) or
3042 not config.has_section(constants.INISECT_INS)):
3043 _Fail("Export info file doesn't have the required fields")
3045 return config.Dumps()
3049 """Return a list of exports currently available on this machine.
3052 @return: list of the exports
3055 if os.path.isdir(pathutils.EXPORT_DIR):
3056 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3058 _Fail("No exports directory")
3061 def RemoveExport(export):
3062 """Remove an existing export from the node.
3065 @param export: the name of the export to remove
3069 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3072 shutil.rmtree(target)
3073 except EnvironmentError, err:
3074 _Fail("Error while removing the export: %s", err, exc=True)
3077 def BlockdevRename(devlist):
3078 """Rename a list of block devices.
3080 @type devlist: list of tuples
3081 @param devlist: list of tuples of the form (disk,
3082 new_logical_id, new_physical_id); disk is an
3083 L{objects.Disk} object describing the current disk,
3084 and new logical_id/physical_id is the name we
3087 @return: True if all renames succeeded, False otherwise
3092 for disk, unique_id in devlist:
3093 dev = _RecursiveFindBD(disk)
3095 msgs.append("Can't find device %s in rename" % str(disk))
3099 old_rpath = dev.dev_path
3100 dev.Rename(unique_id)
3101 new_rpath = dev.dev_path
3102 if old_rpath != new_rpath:
3103 DevCacheManager.RemoveCache(old_rpath)
3104 # FIXME: we should add the new cache information here, like:
3105 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3106 # but we don't have the owner here - maybe parse from existing
3107 # cache? for now, we only lose lvm data when we rename, which
3108 # is less critical than DRBD or MD
3109 except errors.BlockDeviceError, err:
3110 msgs.append("Can't rename device '%s' to '%s': %s" %
3111 (dev, unique_id, err))
3112 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3115 _Fail("; ".join(msgs))
3118 def _TransformFileStorageDir(fs_dir):
3119 """Checks whether given file_storage_dir is valid.
3121 Checks wheter the given fs_dir is within the cluster-wide default
3122 file_storage_dir or the shared_file_storage_dir, which are stored in
3123 SimpleStore. Only paths under those directories are allowed.
3126 @param fs_dir: the path to check
3128 @return: the normalized path if valid, None otherwise
3131 if not (constants.ENABLE_FILE_STORAGE or
3132 constants.ENABLE_SHARED_FILE_STORAGE):
3133 _Fail("File storage disabled at configure time")
3135 bdev.CheckFileStoragePath(fs_dir)
3137 return os.path.normpath(fs_dir)
3140 def CreateFileStorageDir(file_storage_dir):
3141 """Create file storage directory.
3143 @type file_storage_dir: str
3144 @param file_storage_dir: directory to create
3147 @return: tuple with first element a boolean indicating wheter dir
3148 creation was successful or not
3151 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3152 if os.path.exists(file_storage_dir):
3153 if not os.path.isdir(file_storage_dir):
3154 _Fail("Specified storage dir '%s' is not a directory",
3158 os.makedirs(file_storage_dir, 0750)
3159 except OSError, err:
3160 _Fail("Cannot create file storage directory '%s': %s",
3161 file_storage_dir, err, exc=True)
3164 def RemoveFileStorageDir(file_storage_dir):
3165 """Remove file storage directory.
3167 Remove it only if it's empty. If not log an error and return.
3169 @type file_storage_dir: str
3170 @param file_storage_dir: the directory we should cleanup
3171 @rtype: tuple (success,)
3172 @return: tuple of one element, C{success}, denoting
3173 whether the operation was successful
3176 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3177 if os.path.exists(file_storage_dir):
3178 if not os.path.isdir(file_storage_dir):
3179 _Fail("Specified Storage directory '%s' is not a directory",
3181 # deletes dir only if empty, otherwise we want to fail the rpc call
3183 os.rmdir(file_storage_dir)
3184 except OSError, err:
3185 _Fail("Cannot remove file storage directory '%s': %s",
3186 file_storage_dir, err)
3189 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3190 """Rename the file storage directory.
3192 @type old_file_storage_dir: str
3193 @param old_file_storage_dir: the current path
3194 @type new_file_storage_dir: str
3195 @param new_file_storage_dir: the name we should rename to
3196 @rtype: tuple (success,)
3197 @return: tuple of one element, C{success}, denoting
3198 whether the operation was successful
3201 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3202 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3203 if not os.path.exists(new_file_storage_dir):
3204 if os.path.isdir(old_file_storage_dir):
3206 os.rename(old_file_storage_dir, new_file_storage_dir)
3207 except OSError, err:
3208 _Fail("Cannot rename '%s' to '%s': %s",
3209 old_file_storage_dir, new_file_storage_dir, err)
3211 _Fail("Specified storage dir '%s' is not a directory",
3212 old_file_storage_dir)
3214 if os.path.exists(old_file_storage_dir):
3215 _Fail("Cannot rename '%s' to '%s': both locations exist",
3216 old_file_storage_dir, new_file_storage_dir)
3219 def _EnsureJobQueueFile(file_name):
3220 """Checks whether the given filename is in the queue directory.
3222 @type file_name: str
3223 @param file_name: the file name we should check
3225 @raises RPCFail: if the file is not valid
3228 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3229 _Fail("Passed job queue file '%s' does not belong to"
3230 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3233 def JobQueueUpdate(file_name, content):
3234 """Updates a file in the queue directory.
3236 This is just a wrapper over L{utils.io.WriteFile}, with proper
3239 @type file_name: str
3240 @param file_name: the job file name
3242 @param content: the new job contents
3244 @return: the success of the operation
3247 file_name = vcluster.LocalizeVirtualPath(file_name)
3249 _EnsureJobQueueFile(file_name)
3250 getents = runtime.GetEnts()
3252 # Write and replace the file atomically
3253 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3254 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3257 def JobQueueRename(old, new):
3258 """Renames a job queue file.
3260 This is just a wrapper over os.rename with proper checking.
3263 @param old: the old (actual) file name
3265 @param new: the desired file name
3267 @return: the success of the operation and payload
3270 old = vcluster.LocalizeVirtualPath(old)
3271 new = vcluster.LocalizeVirtualPath(new)
3273 _EnsureJobQueueFile(old)
3274 _EnsureJobQueueFile(new)
3276 getents = runtime.GetEnts()
3278 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3279 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3282 def BlockdevClose(instance_name, disks):
3283 """Closes the given block devices.
3285 This means they will be switched to secondary mode (in case of
3288 @param instance_name: if the argument is not empty, the symlinks
3289 of this instance will be removed
3290 @type disks: list of L{objects.Disk}
3291 @param disks: the list of disks to be closed
3292 @rtype: tuple (success, message)
3293 @return: a tuple of success and message, where success
3294 indicates the succes of the operation, and message
3295 which will contain the error details in case we
3301 rd = _RecursiveFindBD(cf)
3303 _Fail("Can't find device %s", cf)
3310 except errors.BlockDeviceError, err:
3311 msg.append(str(err))
3313 _Fail("Can't make devices secondary: %s", ",".join(msg))
3316 _RemoveBlockDevLinks(instance_name, disks)
3319 def ValidateHVParams(hvname, hvparams):
3320 """Validates the given hypervisor parameters.
3322 @type hvname: string
3323 @param hvname: the hypervisor name
3324 @type hvparams: dict
3325 @param hvparams: the hypervisor parameters to be validated
3330 hv_type = hypervisor.GetHypervisor(hvname)
3331 hv_type.ValidateParameters(hvparams)
3332 except errors.HypervisorError, err:
3333 _Fail(str(err), log=False)
3336 def _CheckOSPList(os_obj, parameters):
3337 """Check whether a list of parameters is supported by the OS.
3339 @type os_obj: L{objects.OS}
3340 @param os_obj: OS object to check
3341 @type parameters: list
3342 @param parameters: the list of parameters to check
3345 supported = [v[0] for v in os_obj.supported_parameters]
3346 delta = frozenset(parameters).difference(supported)
3348 _Fail("The following parameters are not supported"
3349 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3352 def ValidateOS(required, osname, checks, osparams):
3353 """Validate the given OS' parameters.
3355 @type required: boolean
3356 @param required: whether absence of the OS should translate into
3358 @type osname: string
3359 @param osname: the OS to be validated
3361 @param checks: list of the checks to run (currently only 'parameters')
3362 @type osparams: dict
3363 @param osparams: dictionary with OS parameters
3365 @return: True if the validation passed, or False if the OS was not
3366 found and L{required} was false
3369 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3370 _Fail("Unknown checks required for OS %s: %s", osname,
3371 set(checks).difference(constants.OS_VALIDATE_CALLS))
3373 name_only = objects.OS.GetName(osname)
3374 status, tbv = _TryOSFromDisk(name_only, None)
3382 if max(tbv.api_versions) < constants.OS_API_V20:
3385 if constants.OS_VALIDATE_PARAMETERS in checks:
3386 _CheckOSPList(tbv, osparams.keys())
3388 validate_env = OSCoreEnv(osname, tbv, osparams)
3389 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3390 cwd=tbv.path, reset_env=True)
3392 logging.error("os validate command '%s' returned error: %s output: %s",
3393 result.cmd, result.fail_reason, result.output)
3394 _Fail("OS validation script failed (%s), output: %s",
3395 result.fail_reason, result.output, log=False)
3401 """Demotes the current node from master candidate role.
3404 # try to ensure we're not the master by mistake
3405 master, myself = ssconf.GetMasterAndMyself()
3406 if master == myself:
3407 _Fail("ssconf status shows I'm the master node, will not demote")
3409 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3410 if not result.failed:
3411 _Fail("The master daemon is running, will not demote")
3414 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3415 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3416 except EnvironmentError, err:
3417 if err.errno != errno.ENOENT:
3418 _Fail("Error while backing up cluster file: %s", err, exc=True)
3420 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3423 def _GetX509Filenames(cryptodir, name):
3424 """Returns the full paths for the private key and certificate.
3427 return (utils.PathJoin(cryptodir, name),
3428 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3429 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3432 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3433 """Creates a new X509 certificate for SSL/TLS.
3436 @param validity: Validity in seconds
3437 @rtype: tuple; (string, string)
3438 @return: Certificate name and public part
3441 (key_pem, cert_pem) = \
3442 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3443 min(validity, _MAX_SSL_CERT_VALIDITY))
3445 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3446 prefix="x509-%s-" % utils.TimestampForFilename())
3448 name = os.path.basename(cert_dir)
3449 assert len(name) > 5
3451 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3453 utils.WriteFile(key_file, mode=0400, data=key_pem)
3454 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3456 # Never return private key as it shouldn't leave the node
3457 return (name, cert_pem)
3459 shutil.rmtree(cert_dir, ignore_errors=True)
3463 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3464 """Removes a X509 certificate.
3467 @param name: Certificate name
3470 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3472 utils.RemoveFile(key_file)
3473 utils.RemoveFile(cert_file)
3477 except EnvironmentError, err:
3478 _Fail("Cannot remove certificate directory '%s': %s",
3482 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3483 """Returns the command for the requested input/output.
3485 @type instance: L{objects.Instance}
3486 @param instance: The instance object
3487 @param mode: Import/export mode
3488 @param ieio: Input/output type
3489 @param ieargs: Input/output arguments
3492 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3499 if ieio == constants.IEIO_FILE:
3500 (filename, ) = ieargs
3502 if not utils.IsNormAbsPath(filename):
3503 _Fail("Path '%s' is not normalized or absolute", filename)
3505 real_filename = os.path.realpath(filename)
3506 directory = os.path.dirname(real_filename)
3508 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3509 _Fail("File '%s' is not under exports directory '%s': %s",
3510 filename, pathutils.EXPORT_DIR, real_filename)
3513 utils.Makedirs(directory, mode=0750)
3515 quoted_filename = utils.ShellQuote(filename)
3517 if mode == constants.IEM_IMPORT:
3518 suffix = "> %s" % quoted_filename
3519 elif mode == constants.IEM_EXPORT:
3520 suffix = "< %s" % quoted_filename
3522 # Retrieve file size
3524 st = os.stat(filename)
3525 except EnvironmentError, err:
3526 logging.error("Can't stat(2) %s: %s", filename, err)
3528 exp_size = utils.BytesToMebibyte(st.st_size)
3530 elif ieio == constants.IEIO_RAW_DISK:
3533 real_disk = _OpenRealBD(disk)
3535 if mode == constants.IEM_IMPORT:
3536 # we set here a smaller block size as, due to transport buffering, more
3537 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3538 # is not already there or we pass a wrong path; we use notrunc to no
3539 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3540 # much memory; this means that at best, we flush every 64k, which will
3542 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3543 " bs=%s oflag=dsync"),
3547 elif mode == constants.IEM_EXPORT:
3548 # the block size on the read dd is 1MiB to match our units
3549 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3551 str(1024 * 1024), # 1 MB
3553 exp_size = disk.size
3555 elif ieio == constants.IEIO_SCRIPT:
3556 (disk, disk_index, ) = ieargs
3558 assert isinstance(disk_index, (int, long))
3560 real_disk = _OpenRealBD(disk)
3562 inst_os = OSFromDisk(instance.os)
3563 env = OSEnvironment(instance, inst_os)
3565 if mode == constants.IEM_IMPORT:
3566 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3567 env["IMPORT_INDEX"] = str(disk_index)
3568 script = inst_os.import_script
3570 elif mode == constants.IEM_EXPORT:
3571 env["EXPORT_DEVICE"] = real_disk.dev_path
3572 env["EXPORT_INDEX"] = str(disk_index)
3573 script = inst_os.export_script
3575 # TODO: Pass special environment only to script
3576 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3578 if mode == constants.IEM_IMPORT:
3579 suffix = "| %s" % script_cmd
3581 elif mode == constants.IEM_EXPORT:
3582 prefix = "%s |" % script_cmd
3584 # Let script predict size
3585 exp_size = constants.IE_CUSTOM_SIZE
3588 _Fail("Invalid %s I/O mode %r", mode, ieio)
3590 return (env, prefix, suffix, exp_size)
3593 def _CreateImportExportStatusDir(prefix):
3594 """Creates status directory for import/export.
3597 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3599 (prefix, utils.TimestampForFilename())))
3602 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3604 """Starts an import or export daemon.
3606 @param mode: Import/output mode
3607 @type opts: L{objects.ImportExportOptions}
3608 @param opts: Daemon options
3610 @param host: Remote host for export (None for import)
3612 @param port: Remote port for export (None for import)
3613 @type instance: L{objects.Instance}
3614 @param instance: Instance object
3615 @type component: string
3616 @param component: which part of the instance is transferred now,
3618 @param ieio: Input/output type
3619 @param ieioargs: Input/output arguments
3622 if mode == constants.IEM_IMPORT:
3625 if not (host is None and port is None):
3626 _Fail("Can not specify host or port on import")
3628 elif mode == constants.IEM_EXPORT:
3631 if host is None or port is None:
3632 _Fail("Host and port must be specified for an export")
3635 _Fail("Invalid mode %r", mode)
3637 if (opts.key_name is None) ^ (opts.ca_pem is None):
3638 _Fail("Cluster certificate can only be used for both key and CA")
3640 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3641 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3643 if opts.key_name is None:
3645 key_path = pathutils.NODED_CERT_FILE
3646 cert_path = pathutils.NODED_CERT_FILE
3647 assert opts.ca_pem is None
3649 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3651 assert opts.ca_pem is not None
3653 for i in [key_path, cert_path]:
3654 if not os.path.exists(i):
3655 _Fail("File '%s' does not exist" % i)
3657 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3659 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3660 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3661 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3663 if opts.ca_pem is None:
3665 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3670 utils.WriteFile(ca_file, data=ca, mode=0400)
3673 pathutils.IMPORT_EXPORT_DAEMON,
3675 "--key=%s" % key_path,
3676 "--cert=%s" % cert_path,
3677 "--ca=%s" % ca_file,
3681 cmd.append("--host=%s" % host)
3684 cmd.append("--port=%s" % port)
3687 cmd.append("--ipv6")
3689 cmd.append("--ipv4")
3692 cmd.append("--compress=%s" % opts.compress)
3695 cmd.append("--magic=%s" % opts.magic)
3697 if exp_size is not None:
3698 cmd.append("--expected-size=%s" % exp_size)
3701 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3704 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3706 if mode == constants.IEM_EXPORT:
3707 # Retry connection a few times when connecting to remote peer
3708 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3709 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3710 elif opts.connect_timeout is not None:
3711 assert mode == constants.IEM_IMPORT
3712 # Overall timeout for establishing connection while listening
3713 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3715 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3717 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3718 # support for receiving a file descriptor for output
3719 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3722 # The import/export name is simply the status directory name
3723 return os.path.basename(status_dir)
3726 shutil.rmtree(status_dir, ignore_errors=True)
3730 def GetImportExportStatus(names):
3731 """Returns import/export daemon status.
3733 @type names: sequence
3734 @param names: List of names
3735 @rtype: List of dicts
3736 @return: Returns a list of the state of each named import/export or None if a
3737 status couldn't be read
3743 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3747 data = utils.ReadFile(status_file)
3748 except EnvironmentError, err:
3749 if err.errno != errno.ENOENT:
3757 result.append(serializer.LoadJson(data))
3762 def AbortImportExport(name):
3763 """Sends SIGTERM to a running import/export daemon.
3766 logging.info("Abort import/export %s", name)
3768 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3769 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3772 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3774 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3777 def CleanupImportExport(name):
3778 """Cleanup after an import or export.
3780 If the import/export daemon is still running it's killed. Afterwards the
3781 whole status directory is removed.
3784 logging.info("Finalizing import/export %s", name)
3786 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3788 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3791 logging.info("Import/export %s is still running with PID %s",
3793 utils.KillProcess(pid, waitpid=False)
3795 shutil.rmtree(status_dir, ignore_errors=True)
3798 def _FindDisks(target_node_uuid, nodes_ip, disks):
3799 """Sets the physical ID on disks and returns the block devices.
3802 # set the correct physical ID
3804 cf.SetPhysicalID(target_node_uuid, nodes_ip)
3809 rd = _RecursiveFindBD(cf)
3811 _Fail("Can't find device %s", cf)
3816 def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
3817 """Disconnects the network on a list of drbd devices.
3820 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3826 except errors.BlockDeviceError, err:
3827 _Fail("Can't change network configuration to standalone mode: %s",
3831 def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
3833 """Attaches the network on a list of drbd devices.
3836 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3839 for idx, rd in enumerate(bdevs):
3841 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3842 except EnvironmentError, err:
3843 _Fail("Can't create symlink: %s", err)
3844 # reconnect disks, switch to new master configuration and if
3845 # needed primary mode
3848 rd.AttachNet(multimaster)
3849 except errors.BlockDeviceError, err:
3850 _Fail("Can't change network configuration: %s", err)
3852 # wait until the disks are connected; we need to retry the re-attach
3853 # if the device becomes standalone, as this might happen if the one
3854 # node disconnects and reconnects in a different mode before the
3855 # other node reconnects; in this case, one or both of the nodes will
3856 # decide it has wrong configuration and switch to standalone
3859 all_connected = True
3862 stats = rd.GetProcStatus()
3864 all_connected = (all_connected and
3865 (stats.is_connected or stats.is_in_resync))
3867 if stats.is_standalone:
3868 # peer had different config info and this node became
3869 # standalone, even though this should not happen with the
3870 # new staged way of changing disk configs
3872 rd.AttachNet(multimaster)
3873 except errors.BlockDeviceError, err:
3874 _Fail("Can't change network configuration: %s", err)
3876 if not all_connected:
3877 raise utils.RetryAgain()
3880 # Start with a delay of 100 miliseconds and go up to 5 seconds
3881 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3882 except utils.RetryTimeout:
3883 _Fail("Timeout in disk reconnecting")
3886 # change to primary mode
3890 except errors.BlockDeviceError, err:
3891 _Fail("Can't change to primary mode: %s", err)
3894 def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
3895 """Wait until DRBDs have synchronized.
3899 stats = rd.GetProcStatus()
3900 if not (stats.is_connected or stats.is_in_resync):
3901 raise utils.RetryAgain()
3904 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3910 # poll each second for 15 seconds
3911 stats = utils.Retry(_helper, 1, 15, args=[rd])
3912 except utils.RetryTimeout:
3913 stats = rd.GetProcStatus()
3915 if not (stats.is_connected or stats.is_in_resync):
3916 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3917 alldone = alldone and (not stats.is_in_resync)
3918 if stats.sync_percent is not None:
3919 min_resync = min(min_resync, stats.sync_percent)
3921 return (alldone, min_resync)
3924 def GetDrbdUsermodeHelper():
3925 """Returns DRBD usermode helper currently configured.
3929 return drbd.DRBD8.GetUsermodeHelper()
3930 except errors.BlockDeviceError, err:
3934 def PowercycleNode(hypervisor_type, hvparams=None):
3935 """Hard-powercycle the node.
3937 Because we need to return first, and schedule the powercycle in the
3938 background, we won't be able to report failures nicely.
3941 hyper = hypervisor.GetHypervisor(hypervisor_type)
3945 # if we can't fork, we'll pretend that we're in the child process
3948 return "Reboot scheduled in 5 seconds"
3949 # ensure the child is running on ram
3952 except Exception: # pylint: disable=W0703
3955 hyper.PowercycleNode(hvparams=hvparams)
3958 def _VerifyRestrictedCmdName(cmd):
3959 """Verifies a restricted command name.
3962 @param cmd: Command name
3963 @rtype: tuple; (boolean, string or None)
3964 @return: The tuple's first element is the status; if C{False}, the second
3965 element is an error message string, otherwise it's C{None}
3969 return (False, "Missing command name")
3971 if os.path.basename(cmd) != cmd:
3972 return (False, "Invalid command name")
3974 if not constants.EXT_PLUGIN_MASK.match(cmd):
3975 return (False, "Command name contains forbidden characters")
3980 def _CommonRestrictedCmdCheck(path, owner):
3981 """Common checks for restricted command file system directories and files.
3984 @param path: Path to check
3985 @param owner: C{None} or tuple containing UID and GID
3986 @rtype: tuple; (boolean, string or C{os.stat} result)
3987 @return: The tuple's first element is the status; if C{False}, the second
3988 element is an error message string, otherwise it's the result of C{os.stat}
3992 # Default to root as owner
3997 except EnvironmentError, err:
3998 return (False, "Can't stat(2) '%s': %s" % (path, err))
4000 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4001 return (False, "Permissions on '%s' are too permissive" % path)
4003 if (st.st_uid, st.st_gid) != owner:
4004 (owner_uid, owner_gid) = owner
4005 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4010 def _VerifyRestrictedCmdDirectory(path, _owner=None):
4011 """Verifies restricted command directory.
4014 @param path: Path to check
4015 @rtype: tuple; (boolean, string or None)
4016 @return: The tuple's first element is the status; if C{False}, the second
4017 element is an error message string, otherwise it's C{None}
4020 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4023 return (False, value)
4025 if not stat.S_ISDIR(value.st_mode):
4026 return (False, "Path '%s' is not a directory" % path)
4031 def _VerifyRestrictedCmd(path, cmd, _owner=None):
4032 """Verifies a whole restricted command and returns its executable filename.
4035 @param path: Directory containing restricted commands
4037 @param cmd: Command name
4038 @rtype: tuple; (boolean, string)
4039 @return: The tuple's first element is the status; if C{False}, the second
4040 element is an error message string, otherwise the second element is the
4041 absolute path to the executable
4044 executable = utils.PathJoin(path, cmd)
4046 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4051 if not utils.IsExecutable(executable):
4052 return (False, "access(2) thinks '%s' can't be executed" % executable)
4054 return (True, executable)
4057 def _PrepareRestrictedCmd(path, cmd,
4058 _verify_dir=_VerifyRestrictedCmdDirectory,
4059 _verify_name=_VerifyRestrictedCmdName,
4060 _verify_cmd=_VerifyRestrictedCmd):
4061 """Performs a number of tests on a restricted command.
4064 @param path: Directory containing restricted commands
4066 @param cmd: Command name
4067 @return: Same as L{_VerifyRestrictedCmd}
4070 # Verify the directory first
4071 (status, msg) = _verify_dir(path)
4073 # Check command if everything was alright
4074 (status, msg) = _verify_name(cmd)
4079 # Check actual executable
4080 return _verify_cmd(path, cmd)
4083 def RunRestrictedCmd(cmd,
4084 _lock_timeout=_RCMD_LOCK_TIMEOUT,
4085 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4086 _path=pathutils.RESTRICTED_COMMANDS_DIR,
4087 _sleep_fn=time.sleep,
4088 _prepare_fn=_PrepareRestrictedCmd,
4089 _runcmd_fn=utils.RunCmd,
4090 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4091 """Executes a restricted command after performing strict tests.
4094 @param cmd: Command name
4096 @return: Command output
4097 @raise RPCFail: In case of an error
4100 logging.info("Preparing to run restricted command '%s'", cmd)
4103 _Fail("Restricted commands disabled at configure time")
4109 lock = utils.FileLock.Open(_lock_file)
4110 lock.Exclusive(blocking=True, timeout=_lock_timeout)
4112 (status, value) = _prepare_fn(_path, cmd)
4115 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4116 postfork_fn=lambda _: lock.Unlock())
4118 logging.error(value)
4119 except Exception: # pylint: disable=W0703
4120 # Keep original error in log
4121 logging.exception("Caught exception")
4123 if cmdresult is None:
4124 logging.info("Sleeping for %0.1f seconds before returning",
4125 _RCMD_INVALID_DELAY)
4126 _sleep_fn(_RCMD_INVALID_DELAY)
4128 # Do not include original error message in returned error
4129 _Fail("Executing command '%s' failed" % cmd)
4130 elif cmdresult.failed or cmdresult.fail_reason:
4131 _Fail("Restricted command '%s' failed: %s; output: %s",
4132 cmd, cmdresult.fail_reason, cmdresult.output)
4134 return cmdresult.output
4136 if lock is not None:
4137 # Release lock at last
4142 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4143 """Creates or removes the watcher pause file.
4145 @type until: None or number
4146 @param until: Unix timestamp saying until when the watcher shouldn't run
4150 logging.info("Received request to no longer pause watcher")
4151 utils.RemoveFile(_filename)
4153 logging.info("Received request to pause watcher until %s", until)
4155 if not ht.TNumber(until):
4156 _Fail("Duration must be numeric")
4158 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4161 class HooksRunner(object):
4164 This class is instantiated on the node side (ganeti-noded) and not
4168 def __init__(self, hooks_base_dir=None):
4169 """Constructor for hooks runner.
4171 @type hooks_base_dir: str or None
4172 @param hooks_base_dir: if not None, this overrides the
4173 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4176 if hooks_base_dir is None:
4177 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4178 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4180 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4182 def RunLocalHooks(self, node_list, hpath, phase, env):
4183 """Check that the hooks will be run only locally and then run them.
4186 assert len(node_list) == 1
4188 _, myself = ssconf.GetMasterAndMyself()
4189 assert node == myself
4191 results = self.RunHooks(hpath, phase, env)
4193 # Return values in the form expected by HooksMaster
4194 return {node: (None, False, results)}
4196 def RunHooks(self, hpath, phase, env):
4197 """Run the scripts in the hooks directory.
4200 @param hpath: the path to the hooks directory which
4203 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4204 L{constants.HOOKS_PHASE_POST}
4206 @param env: dictionary with the environment for the hook
4208 @return: list of 3-element tuples:
4210 - script result, either L{constants.HKR_SUCCESS} or
4211 L{constants.HKR_FAIL}
4212 - output of the script
4214 @raise errors.ProgrammerError: for invalid input
4218 if phase == constants.HOOKS_PHASE_PRE:
4220 elif phase == constants.HOOKS_PHASE_POST:
4223 _Fail("Unknown hooks phase '%s'", phase)
4225 subdir = "%s-%s.d" % (hpath, suffix)
4226 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4230 if not os.path.isdir(dir_name):
4231 # for non-existing/non-dirs, we simply exit instead of logging a
4232 # warning at every operation
4235 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4237 for (relname, relstatus, runresult) in runparts_results:
4238 if relstatus == constants.RUNPARTS_SKIP:
4239 rrval = constants.HKR_SKIP
4241 elif relstatus == constants.RUNPARTS_ERR:
4242 rrval = constants.HKR_FAIL
4243 output = "Hook script execution error: %s" % runresult
4244 elif relstatus == constants.RUNPARTS_RUN:
4245 if runresult.failed:
4246 rrval = constants.HKR_FAIL
4248 rrval = constants.HKR_SUCCESS
4249 output = utils.SafeEncode(runresult.output.strip())
4250 results.append(("%s/%s" % (subdir, relname), rrval, output))
4255 class IAllocatorRunner(object):
4256 """IAllocator runner.
4258 This class is instantiated on the node side (ganeti-noded) and not on
4263 def Run(name, idata):
4264 """Run an iallocator script.
4267 @param name: the iallocator script name
4269 @param idata: the allocator input data
4272 @return: two element tuple of:
4274 - either error message or stdout of allocator (for success)
4277 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4279 if alloc_script is None:
4280 _Fail("iallocator module '%s' not found in the search path", name)
4282 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4286 result = utils.RunCmd([alloc_script, fin_name])
4288 _Fail("iallocator module '%s' failed: %s, output '%s'",
4289 name, result.fail_reason, result.output)
4293 return result.stdout
4296 class DevCacheManager(object):
4297 """Simple class for managing a cache of block device information.
4300 _DEV_PREFIX = "/dev/"
4301 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4304 def _ConvertPath(cls, dev_path):
4305 """Converts a /dev/name path to the cache file name.
4307 This replaces slashes with underscores and strips the /dev
4308 prefix. It then returns the full path to the cache file.
4311 @param dev_path: the C{/dev/} path name
4313 @return: the converted path name
4316 if dev_path.startswith(cls._DEV_PREFIX):
4317 dev_path = dev_path[len(cls._DEV_PREFIX):]
4318 dev_path = dev_path.replace("/", "_")
4319 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4323 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4324 """Updates the cache information for a given device.
4327 @param dev_path: the pathname of the device
4329 @param owner: the owner (instance name) of the device
4330 @type on_primary: bool
4331 @param on_primary: whether this is the primary
4334 @param iv_name: the instance-visible name of the
4335 device, as in objects.Disk.iv_name
4340 if dev_path is None:
4341 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4343 fpath = cls._ConvertPath(dev_path)
4349 iv_name = "not_visible"
4350 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4352 utils.WriteFile(fpath, data=fdata)
4353 except EnvironmentError, err:
4354 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4357 def RemoveCache(cls, dev_path):
4358 """Remove data for a dev_path.
4360 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4361 path name and logging.
4364 @param dev_path: the pathname of the device
4369 if dev_path is None:
4370 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4372 fpath = cls._ConvertPath(dev_path)
4374 utils.RemoveFile(fpath)
4375 except EnvironmentError, err:
4376 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)