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 import objects
60 from ganeti import ssconf
61 from ganeti import serializer
62 from ganeti import netutils
63 from ganeti import runtime
64 from ganeti import compat
65 from ganeti import pathutils
66 from ganeti import vcluster
68 from ganeti.storage.base import BlockDev
69 from ganeti.storage.drbd import DRBD8
70 from ganeti import hooksmaster
73 _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
74 _ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
76 pathutils.JOB_QUEUE_ARCHIVE_DIR,
78 pathutils.CRYPTO_KEYS_DIR,
80 _MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
81 _X509_KEY_FILE = "key"
82 _X509_CERT_FILE = "cert"
83 _IES_STATUS_FILE = "status"
87 #: Valid LVS output line regex
88 _LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
90 # Actions for the master setup script
91 _MASTER_START = "start"
94 #: Maximum file permissions for restricted command directory and executables
95 _RCMD_MAX_MODE = (stat.S_IRWXU |
96 stat.S_IRGRP | stat.S_IXGRP |
97 stat.S_IROTH | stat.S_IXOTH)
99 #: Delay before returning an error for restricted commands
100 _RCMD_INVALID_DELAY = 10
102 #: How long to wait to acquire lock for restricted commands (shorter than
103 #: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
104 #: command requests arrive
105 _RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
108 class RPCFail(Exception):
109 """Class denoting RPC failure.
111 Its argument is the error message.
116 def _GetInstReasonFilename(instance_name):
117 """Path of the file containing the reason of the instance status change.
119 @type instance_name: string
120 @param instance_name: The name of the instance
122 @return: The path of the file
125 return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
128 def _StoreInstReasonTrail(instance_name, trail):
129 """Serialize a reason trail related to an instance change of state to file.
131 The exact location of the file depends on the name of the instance and on
132 the configuration of the Ganeti cluster defined at deploy time.
134 @type instance_name: string
135 @param instance_name: The name of the instance
139 json = serializer.DumpJson(trail)
140 filename = _GetInstReasonFilename(instance_name)
141 utils.WriteFile(filename, data=json)
144 def _Fail(msg, *args, **kwargs):
145 """Log an error and the raise an RPCFail exception.
147 This exception is then handled specially in the ganeti daemon and
148 turned into a 'failed' return type. As such, this function is a
149 useful shortcut for logging the error and returning it to the master
153 @param msg: the text of the exception
159 if "log" not in kwargs or kwargs["log"]: # if we should log this error
160 if "exc" in kwargs and kwargs["exc"]:
161 logging.exception(msg)
168 """Simple wrapper to return a SimpleStore.
170 @rtype: L{ssconf.SimpleStore}
171 @return: a SimpleStore instance
174 return ssconf.SimpleStore()
177 def _GetSshRunner(cluster_name):
178 """Simple wrapper to return an SshRunner.
180 @type cluster_name: str
181 @param cluster_name: the cluster name, which is needed
182 by the SshRunner constructor
183 @rtype: L{ssh.SshRunner}
184 @return: an SshRunner instance
187 return ssh.SshRunner(cluster_name)
190 def _Decompress(data):
191 """Unpacks data compressed by the RPC client.
193 @type data: list or tuple
194 @param data: Data sent by RPC client
196 @return: Decompressed data
199 assert isinstance(data, (list, tuple))
200 assert len(data) == 2
201 (encoding, content) = data
202 if encoding == constants.RPC_ENCODING_NONE:
204 elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
205 return zlib.decompress(base64.b64decode(content))
207 raise AssertionError("Unknown data encoding")
210 def _CleanDirectory(path, exclude=None):
211 """Removes all regular files in a directory.
214 @param path: the directory to clean
216 @param exclude: list of files to be excluded, defaults
220 if path not in _ALLOWED_CLEAN_DIRS:
221 _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
224 if not os.path.isdir(path):
229 # Normalize excluded paths
230 exclude = [os.path.normpath(i) for i in exclude]
232 for rel_name in utils.ListVisibleFiles(path):
233 full_name = utils.PathJoin(path, rel_name)
234 if full_name in exclude:
236 if os.path.isfile(full_name) and not os.path.islink(full_name):
237 utils.RemoveFile(full_name)
240 def _BuildUploadFileList():
241 """Build the list of allowed upload files.
243 This is abstracted so that it's built only once at module import time.
246 allowed_files = set([
247 pathutils.CLUSTER_CONF_FILE,
249 pathutils.SSH_KNOWN_HOSTS_FILE,
250 pathutils.VNC_PASSWORD_FILE,
251 pathutils.RAPI_CERT_FILE,
252 pathutils.SPICE_CERT_FILE,
253 pathutils.SPICE_CACERT_FILE,
254 pathutils.RAPI_USERS_FILE,
255 pathutils.CONFD_HMAC_KEY,
256 pathutils.CLUSTER_DOMAIN_SECRET_FILE,
259 for hv_name in constants.HYPER_TYPES:
260 hv_class = hypervisor.GetHypervisorClass(hv_name)
261 allowed_files.update(hv_class.GetAncillaryFiles()[0])
263 assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
264 "Allowed file storage paths should never be uploaded via RPC"
266 return frozenset(allowed_files)
269 _ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
273 """Removes job queue files and archived jobs.
279 _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
280 _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
284 """Returns master information.
286 This is an utility function to compute master information, either
287 for consumption here or from the node daemon.
290 @return: master_netdev, master_ip, master_name, primary_ip_family,
292 @raise RPCFail: in case of errors
297 master_netdev = cfg.GetMasterNetdev()
298 master_ip = cfg.GetMasterIP()
299 master_netmask = cfg.GetMasterNetmask()
300 master_node = cfg.GetMasterNode()
301 primary_ip_family = cfg.GetPrimaryIPFamily()
302 except errors.ConfigurationError, err:
303 _Fail("Cluster configuration incomplete: %s", err, exc=True)
304 return (master_netdev, master_ip, master_node, primary_ip_family,
308 def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
309 """Decorator that runs hooks before and after the decorated function.
311 @type hook_opcode: string
312 @param hook_opcode: opcode of the hook
313 @type hooks_path: string
314 @param hooks_path: path of the hooks
315 @type env_builder_fn: function
316 @param env_builder_fn: function that returns a dictionary containing the
317 environment variables for the hooks. Will get all the parameters of the
319 @raise RPCFail: in case of pre-hook failure
323 def wrapper(*args, **kwargs):
324 _, myself = ssconf.GetMasterAndMyself()
325 nodes = ([myself], [myself]) # these hooks run locally
327 env_fn = compat.partial(env_builder_fn, *args, **kwargs)
331 hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
332 hr.RunLocalHooks, None, env_fn,
333 logging.warning, cfg.GetClusterName(),
335 hm.RunPhase(constants.HOOKS_PHASE_PRE)
336 result = fn(*args, **kwargs)
337 hm.RunPhase(constants.HOOKS_PHASE_POST)
344 def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
345 """Builds environment variables for master IP hooks.
347 @type master_params: L{objects.MasterNetworkParameters}
348 @param master_params: network parameters of the master
349 @type use_external_mip_script: boolean
350 @param use_external_mip_script: whether to use an external master IP
351 address setup script (unused, but necessary per the implementation of the
352 _RunLocalHooks decorator)
355 # pylint: disable=W0613
356 ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
358 "MASTER_NETDEV": master_params.netdev,
359 "MASTER_IP": master_params.ip,
360 "MASTER_NETMASK": str(master_params.netmask),
361 "CLUSTER_IP_VERSION": str(ver),
367 def _RunMasterSetupScript(master_params, action, use_external_mip_script):
368 """Execute the master IP address setup script.
370 @type master_params: L{objects.MasterNetworkParameters}
371 @param master_params: network parameters of the master
373 @param action: action to pass to the script. Must be one of
374 L{backend._MASTER_START} or L{backend._MASTER_STOP}
375 @type use_external_mip_script: boolean
376 @param use_external_mip_script: whether to use an external master IP
378 @raise backend.RPCFail: if there are errors during the execution of the
382 env = _BuildMasterIpEnv(master_params)
384 if use_external_mip_script:
385 setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
387 setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
389 result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
392 _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
393 (action, result.exit_code, result.output), log=True)
396 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
398 def ActivateMasterIp(master_params, use_external_mip_script):
399 """Activate the IP address of the master daemon.
401 @type master_params: L{objects.MasterNetworkParameters}
402 @param master_params: network parameters of the master
403 @type use_external_mip_script: boolean
404 @param use_external_mip_script: whether to use an external master IP
406 @raise RPCFail: in case of errors during the IP startup
409 _RunMasterSetupScript(master_params, _MASTER_START,
410 use_external_mip_script)
413 def StartMasterDaemons(no_voting):
414 """Activate local node as master node.
416 The function will start the master daemons (ganeti-masterd and ganeti-rapi).
418 @type no_voting: boolean
419 @param no_voting: whether to start ganeti-masterd without a node vote
420 but still non-interactively
426 masterd_args = "--no-voting --yes-do-it"
431 "EXTRA_MASTERD_ARGS": masterd_args,
434 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
436 msg = "Can't start Ganeti master: %s" % result.output
441 @RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
443 def DeactivateMasterIp(master_params, use_external_mip_script):
444 """Deactivate the master IP on this node.
446 @type master_params: L{objects.MasterNetworkParameters}
447 @param master_params: network parameters of the master
448 @type use_external_mip_script: boolean
449 @param use_external_mip_script: whether to use an external master IP
451 @raise RPCFail: in case of errors during the IP turndown
454 _RunMasterSetupScript(master_params, _MASTER_STOP,
455 use_external_mip_script)
458 def StopMasterDaemons():
459 """Stop the master daemons on this node.
461 Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
466 # TODO: log and report back to the caller the error failures; we
467 # need to decide in which case we fail the RPC for this
469 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
471 logging.error("Could not stop Ganeti master, command %s had exitcode %s"
473 result.cmd, result.exit_code, result.output)
476 def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
477 """Change the netmask of the master IP.
479 @param old_netmask: the old value of the netmask
480 @param netmask: the new value of the netmask
481 @param master_ip: the master IP
482 @param master_netdev: the master network device
485 if old_netmask == netmask:
488 if not netutils.IPAddress.Own(master_ip):
489 _Fail("The master IP address is not up, not attempting to change its"
492 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
493 "%s/%s" % (master_ip, netmask),
494 "dev", master_netdev, "label",
495 "%s:0" % master_netdev])
497 _Fail("Could not set the new netmask on the master IP address")
499 result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
500 "%s/%s" % (master_ip, old_netmask),
501 "dev", master_netdev, "label",
502 "%s:0" % master_netdev])
504 _Fail("Could not bring down the master IP address with the old netmask")
507 def EtcHostsModify(mode, host, ip):
508 """Modify a host entry in /etc/hosts.
510 @param mode: The mode to operate. Either add or remove entry
511 @param host: The host to operate on
512 @param ip: The ip associated with the entry
515 if mode == constants.ETC_HOSTS_ADD:
517 RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
519 utils.AddHostToEtcHosts(host, ip)
520 elif mode == constants.ETC_HOSTS_REMOVE:
522 RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
523 " parameter is present")
524 utils.RemoveHostFromEtcHosts(host)
526 RPCFail("Mode not supported")
529 def LeaveCluster(modify_ssh_setup):
530 """Cleans up and remove the current node.
532 This function cleans up and prepares the current node to be removed
535 If processing is successful, then it raises an
536 L{errors.QuitGanetiException} which is used as a special case to
537 shutdown the node daemon.
539 @param modify_ssh_setup: boolean
542 _CleanDirectory(pathutils.DATA_DIR)
543 _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
548 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
550 utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
552 utils.RemoveFile(priv_key)
553 utils.RemoveFile(pub_key)
554 except errors.OpExecError:
555 logging.exception("Error while processing ssh files")
558 utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
559 utils.RemoveFile(pathutils.RAPI_CERT_FILE)
560 utils.RemoveFile(pathutils.SPICE_CERT_FILE)
561 utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
562 utils.RemoveFile(pathutils.NODED_CERT_FILE)
563 except: # pylint: disable=W0702
564 logging.exception("Error while removing cluster secrets")
566 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
568 logging.error("Command %s failed with exitcode %s and error %s",
569 result.cmd, result.exit_code, result.output)
571 # Raise a custom exception (handled in ganeti-noded)
572 raise errors.QuitGanetiException(True, "Shutdown scheduled")
575 def _GetVgInfo(name, excl_stor):
576 """Retrieves information about a LVM volume group.
579 # TODO: GetVGInfo supports returning information for multiple VGs at once
580 vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
582 vg_free = int(round(vginfo[0][0], 0))
583 vg_size = int(round(vginfo[0][1], 0))
589 "type": constants.ST_LVM_VG,
596 def _GetVgSpindlesInfo(name, excl_stor):
597 """Retrieves information about spindles in an LVM volume group.
601 @type excl_stor: bool
602 @param excl_stor: exclusive storage
604 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
605 free spindles, total spindles respectively
609 (vg_free, vg_size) = bdev.LogicalVolume.GetVgSpindlesInfo(name)
614 "type": constants.ST_LVM_PV,
621 def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
622 """Retrieves node information from a hypervisor.
624 The information returned depends on the hypervisor. Common items:
626 - vg_size is the size of the configured volume group in MiB
627 - vg_free is the free size of the volume group in MiB
628 - memory_dom0 is the memory allocated for domain0 in MiB
629 - memory_free is the currently available (free) ram in MiB
630 - memory_total is the total number of ram in MiB
631 - hv_version: the hypervisor version, if available
633 @type hvparams: dict of string
634 @param hvparams: the hypervisor's hvparams
637 return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
640 def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
641 """Retrieves node information for all hypervisors.
643 See C{_GetHvInfo} for information on the output.
645 @type hv_specs: list of pairs (string, dict of strings)
646 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
653 for hvname, hvparams in hv_specs:
654 result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
658 def _GetNamedNodeInfo(names, fn):
659 """Calls C{fn} for all names in C{names} and returns a dictionary.
667 return map(fn, names)
670 def GetNodeInfo(storage_units, hv_specs, excl_stor):
671 """Gives back a hash with different information about the node.
673 @type storage_units: list of pairs (string, string)
674 @param storage_units: List of pairs (storage unit, identifier) to ask for disk
675 space information. In case of lvm-vg, the identifier is
677 @type hv_specs: list of pairs (string, dict of strings)
678 @param hv_specs: list of pairs of a hypervisor's name and its hvparams
679 @type excl_stor: boolean
680 @param excl_stor: Whether exclusive_storage is active
681 @rtype: tuple; (string, None/dict, None/dict)
682 @return: Tuple containing boot ID, volume group information and hypervisor
686 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
687 storage_info = _GetNamedNodeInfo(
689 (lambda storage_unit: _ApplyStorageInfoFunction(storage_unit[0],
692 hv_info = _GetHvInfoAll(hv_specs)
693 return (bootid, storage_info, hv_info)
696 # FIXME: implement storage reporting for all missing storage types.
697 _STORAGE_TYPE_INFO_FN = {
698 constants.ST_BLOCK: None,
699 constants.ST_DISKLESS: None,
700 constants.ST_EXT: None,
701 constants.ST_FILE: None,
702 constants.ST_LVM_PV: _GetVgSpindlesInfo,
703 constants.ST_LVM_VG: _GetVgInfo,
704 constants.ST_RADOS: None,
708 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
709 """Looks up and applies the correct function to calculate free and total
710 storage for the given storage type.
712 @type storage_type: string
713 @param storage_type: the storage type for which the storage shall be reported.
714 @type storage_key: string
715 @param storage_key: identifier of a storage unit, e.g. the volume group name
716 of an LVM storage unit
718 @param args: various parameters that can be used for storage reporting. These
719 parameters and their semantics vary from storage type to storage type and
720 are just propagated in this function.
721 @return: the results of the application of the storage space function (see
722 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
724 @raises NotImplementedError: for storage types who don't support space
727 fn = _STORAGE_TYPE_INFO_FN[storage_type]
729 return fn(storage_key, *args)
731 raise NotImplementedError
734 def _CheckExclusivePvs(pvi_list):
735 """Check that PVs are not shared among LVs
737 @type pvi_list: list of L{objects.LvmPvInfo} objects
738 @param pvi_list: information about the PVs
740 @rtype: list of tuples (string, list of strings)
741 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
746 if len(pvi.lv_list) > 1:
747 res.append((pvi.name, pvi.lv_list))
751 def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
752 get_hv_fn=hypervisor.GetHypervisor):
753 """Verifies the hypervisor. Appends the results to the 'results' list.
756 @param what: a dictionary of things to check
757 @type vm_capable: boolean
758 @param vm_capable: whether or not this node is vm capable
760 @param result: dictionary of verification results; results of the
761 verifications in this function will be added here
762 @type all_hvparams: dict of dict of string
763 @param all_hvparams: dictionary mapping hypervisor names to hvparams
764 @type get_hv_fn: function
765 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
771 if constants.NV_HYPERVISOR in what:
772 result[constants.NV_HYPERVISOR] = {}
773 for hv_name in what[constants.NV_HYPERVISOR]:
774 hvparams = all_hvparams[hv_name]
776 val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
777 except errors.HypervisorError, err:
778 val = "Error while checking hypervisor: %s" % str(err)
779 result[constants.NV_HYPERVISOR][hv_name] = val
782 def _VerifyHvparams(what, vm_capable, result,
783 get_hv_fn=hypervisor.GetHypervisor):
784 """Verifies the hvparams. Appends the results to the 'results' list.
787 @param what: a dictionary of things to check
788 @type vm_capable: boolean
789 @param vm_capable: whether or not this node is vm capable
791 @param result: dictionary of verification results; results of the
792 verifications in this function will be added here
793 @type get_hv_fn: function
794 @param get_hv_fn: function to retrieve the hypervisor, to improve testability
800 if constants.NV_HVPARAMS in what:
801 result[constants.NV_HVPARAMS] = []
802 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
804 logging.info("Validating hv %s, %s", hv_name, hvparms)
805 get_hv_fn(hv_name).ValidateParameters(hvparms)
806 except errors.HypervisorError, err:
807 result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
810 def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
811 """Verifies the instance list.
814 @param what: a dictionary of things to check
815 @type vm_capable: boolean
816 @param vm_capable: whether or not this node is vm capable
818 @param result: dictionary of verification results; results of the
819 verifications in this function will be added here
820 @type all_hvparams: dict of dict of string
821 @param all_hvparams: dictionary mapping hypervisor names to hvparams
824 if constants.NV_INSTANCELIST in what and vm_capable:
825 # GetInstanceList can fail
827 val = GetInstanceList(what[constants.NV_INSTANCELIST],
828 all_hvparams=all_hvparams)
831 result[constants.NV_INSTANCELIST] = val
834 def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
835 """Verifies the node info.
838 @param what: a dictionary of things to check
839 @type vm_capable: boolean
840 @param vm_capable: whether or not this node is vm capable
842 @param result: dictionary of verification results; results of the
843 verifications in this function will be added here
844 @type all_hvparams: dict of dict of string
845 @param all_hvparams: dictionary mapping hypervisor names to hvparams
848 if constants.NV_HVINFO in what and vm_capable:
849 hvname = what[constants.NV_HVINFO]
850 hyper = hypervisor.GetHypervisor(hvname)
851 hvparams = all_hvparams[hvname]
852 result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
855 def VerifyNode(what, cluster_name, all_hvparams):
856 """Verify the status of the local node.
858 Based on the input L{what} parameter, various checks are done on the
861 If the I{filelist} key is present, this list of
862 files is checksummed and the file/checksum pairs are returned.
864 If the I{nodelist} key is present, we check that we have
865 connectivity via ssh with the target nodes (and check the hostname
868 If the I{node-net-test} key is present, we check that we have
869 connectivity to the given nodes via both primary IP and, if
870 applicable, secondary IPs.
873 @param what: a dictionary of things to check:
874 - filelist: list of files for which to compute checksums
875 - nodelist: list of nodes we should check ssh communication with
876 - node-net-test: list of nodes we should check node daemon port
878 - hypervisor: list with hypervisors to run the verify for
879 @type cluster_name: string
880 @param cluster_name: the cluster's name
881 @type all_hvparams: dict of dict of strings
882 @param all_hvparams: a dictionary mapping hypervisor names to hvparams
884 @return: a dictionary with the same keys as the input dict, and
885 values representing the result of the checks
889 my_name = netutils.Hostname.GetSysName()
890 port = netutils.GetDaemonPort(constants.NODED)
891 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
893 _VerifyHypervisors(what, vm_capable, result, all_hvparams)
894 _VerifyHvparams(what, vm_capable, result)
896 if constants.NV_FILELIST in what:
897 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
898 what[constants.NV_FILELIST]))
899 result[constants.NV_FILELIST] = \
900 dict((vcluster.MakeVirtualPath(key), value)
901 for (key, value) in fingerprints.items())
903 if constants.NV_NODELIST in what:
904 (nodes, bynode) = what[constants.NV_NODELIST]
906 # Add nodes from other groups (different for each node)
908 nodes.extend(bynode[my_name])
913 random.shuffle(nodes)
915 # Try to contact all nodes
918 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
922 result[constants.NV_NODELIST] = val
924 if constants.NV_NODENETTEST in what:
925 result[constants.NV_NODENETTEST] = tmp = {}
926 my_pip = my_sip = None
927 for name, pip, sip in what[constants.NV_NODENETTEST]:
933 tmp[my_name] = ("Can't find my own primary/secondary IP"
936 for name, pip, sip in what[constants.NV_NODENETTEST]:
938 if not netutils.TcpPing(pip, port, source=my_pip):
939 fail.append("primary")
941 if not netutils.TcpPing(sip, port, source=my_sip):
942 fail.append("secondary")
944 tmp[name] = ("failure using the %s interface(s)" %
947 if constants.NV_MASTERIP in what:
948 # FIXME: add checks on incoming data structures (here and in the
949 # rest of the function)
950 master_name, master_ip = what[constants.NV_MASTERIP]
951 if master_name == my_name:
952 source = constants.IP4_ADDRESS_LOCALHOST
955 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
958 if constants.NV_USERSCRIPTS in what:
959 result[constants.NV_USERSCRIPTS] = \
960 [script for script in what[constants.NV_USERSCRIPTS]
961 if not utils.IsExecutable(script)]
963 if constants.NV_OOB_PATHS in what:
964 result[constants.NV_OOB_PATHS] = tmp = []
965 for path in what[constants.NV_OOB_PATHS]:
969 tmp.append("error stating out of band helper: %s" % err)
971 if stat.S_ISREG(st.st_mode):
972 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
975 tmp.append("out of band helper %s is not executable" % path)
977 tmp.append("out of band helper %s is not a file" % path)
979 if constants.NV_LVLIST in what and vm_capable:
981 val = GetVolumeList(utils.ListVolumeGroups().keys())
984 result[constants.NV_LVLIST] = val
986 _VerifyInstanceList(what, vm_capable, result, all_hvparams)
988 if constants.NV_VGLIST in what and vm_capable:
989 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
991 if constants.NV_PVLIST in what and vm_capable:
992 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
993 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
994 filter_allocatable=False,
995 include_lvs=check_exclusive_pvs)
996 if check_exclusive_pvs:
997 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
999 # Avoid sending useless data on the wire
1001 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1003 if constants.NV_VERSION in what:
1004 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1005 constants.RELEASE_VERSION)
1007 _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1009 if constants.NV_DRBDVERSION in what and vm_capable:
1011 drbd_version = DRBD8.GetProcInfo().GetVersionString()
1012 except errors.BlockDeviceError, err:
1013 logging.warning("Can't get DRBD version", exc_info=True)
1014 drbd_version = str(err)
1015 result[constants.NV_DRBDVERSION] = drbd_version
1017 if constants.NV_DRBDLIST in what and vm_capable:
1019 used_minors = drbd.DRBD8.GetUsedDevs()
1020 except errors.BlockDeviceError, err:
1021 logging.warning("Can't get used minors list", exc_info=True)
1022 used_minors = str(err)
1023 result[constants.NV_DRBDLIST] = used_minors
1025 if constants.NV_DRBDHELPER in what and vm_capable:
1028 payload = drbd.DRBD8.GetUsermodeHelper()
1029 except errors.BlockDeviceError, err:
1030 logging.error("Can't get DRBD usermode helper: %s", str(err))
1033 result[constants.NV_DRBDHELPER] = (status, payload)
1035 if constants.NV_NODESETUP in what:
1036 result[constants.NV_NODESETUP] = tmpr = []
1037 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1038 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1039 " under /sys, missing required directories /sys/block"
1040 " and /sys/class/net")
1041 if (not os.path.isdir("/proc/sys") or
1042 not os.path.isfile("/proc/sysrq-trigger")):
1043 tmpr.append("The procfs filesystem doesn't seem to be mounted"
1044 " under /proc, missing required directory /proc/sys and"
1045 " the file /proc/sysrq-trigger")
1047 if constants.NV_TIME in what:
1048 result[constants.NV_TIME] = utils.SplitTime(time.time())
1050 if constants.NV_OSLIST in what and vm_capable:
1051 result[constants.NV_OSLIST] = DiagnoseOS()
1053 if constants.NV_BRIDGES in what and vm_capable:
1054 result[constants.NV_BRIDGES] = [bridge
1055 for bridge in what[constants.NV_BRIDGES]
1056 if not utils.BridgeExists(bridge)]
1058 if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
1059 result[constants.NV_FILE_STORAGE_PATHS] = \
1060 bdev.ComputeWrongFileStoragePaths()
1065 def GetBlockDevSizes(devices):
1066 """Return the size of the given block devices
1069 @param devices: list of block device nodes to query
1072 dictionary of all block devices under /dev (key). The value is their
1075 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1078 DEV_PREFIX = "/dev/"
1081 for devpath in devices:
1082 if not utils.IsBelowDir(DEV_PREFIX, devpath):
1086 st = os.stat(devpath)
1087 except EnvironmentError, err:
1088 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1091 if stat.S_ISBLK(st.st_mode):
1092 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1094 # We don't want to fail, just do not list this device as available
1095 logging.warning("Cannot get size for block device %s", devpath)
1098 size = int(result.stdout) / (1024 * 1024)
1099 blockdevs[devpath] = size
1103 def GetVolumeList(vg_names):
1104 """Compute list of logical volumes and their size.
1106 @type vg_names: list
1107 @param vg_names: the volume groups whose LVs we should list, or
1108 empty for all volume groups
1111 dictionary of all partions (key) with value being a tuple of
1112 their size (in MiB), inactive and online status::
1114 {'xenvg/test1': ('20.06', True, True)}
1116 in case of errors, a string is returned with the error
1124 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1125 "--separator=%s" % sep,
1126 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1128 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1130 for line in result.stdout.splitlines():
1132 match = _LVSLINE_REGEX.match(line)
1134 logging.error("Invalid line returned from lvs output: '%s'", line)
1136 vg_name, name, size, attr = match.groups()
1137 inactive = attr[4] == "-"
1138 online = attr[5] == "o"
1139 virtual = attr[0] == "v"
1141 # we don't want to report such volumes as existing, since they
1142 # don't really hold data
1144 lvs[vg_name + "/" + name] = (size, inactive, online)
1149 def ListVolumeGroups():
1150 """List the volume groups and their size.
1153 @return: dictionary with keys volume name and values the
1157 return utils.ListVolumeGroups()
1161 """List all volumes on this node.
1165 A list of dictionaries, each having four keys:
1166 - name: the logical volume name,
1167 - size: the size of the logical volume
1168 - dev: the physical device on which the LV lives
1169 - vg: the volume group to which it belongs
1171 In case of errors, we return an empty list and log the
1174 Note that since a logical volume can live on multiple physical
1175 volumes, the resulting list might include a logical volume
1179 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1181 "--options=lv_name,lv_size,devices,vg_name"])
1183 _Fail("Failed to list logical volumes, lvs output: %s",
1187 return dev.split("(")[0]
1189 def handle_dev(dev):
1190 return [parse_dev(x) for x in dev.split(",")]
1193 line = [v.strip() for v in line]
1194 return [{"name": line[0], "size": line[1],
1195 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1198 for line in result.stdout.splitlines():
1199 if line.count("|") >= 3:
1200 all_devs.extend(map_line(line.split("|")))
1202 logging.warning("Strange line in the output from lvs: '%s'", line)
1206 def BridgesExist(bridges_list):
1207 """Check if a list of bridges exist on the current node.
1210 @return: C{True} if all of them exist, C{False} otherwise
1214 for bridge in bridges_list:
1215 if not utils.BridgeExists(bridge):
1216 missing.append(bridge)
1219 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1222 def GetInstanceListForHypervisor(hname, hvparams=None,
1223 get_hv_fn=hypervisor.GetHypervisor):
1224 """Provides a list of instances of the given hypervisor.
1227 @param hname: name of the hypervisor
1228 @type hvparams: dict of strings
1229 @param hvparams: hypervisor parameters for the given hypervisor
1230 @type get_hv_fn: function
1231 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1232 name; optional parameter to increase testability
1235 @return: a list of all running instances on the current node
1236 - instance1.example.com
1237 - instance2.example.com
1242 hv = get_hv_fn(hname)
1243 names = hv.ListInstances(hvparams=hvparams)
1244 results.extend(names)
1245 except errors.HypervisorError, err:
1246 _Fail("Error enumerating instances (hypervisor %s): %s",
1247 hname, err, exc=True)
1251 def GetInstanceList(hypervisor_list, all_hvparams=None,
1252 get_hv_fn=hypervisor.GetHypervisor):
1253 """Provides a list of instances.
1255 @type hypervisor_list: list
1256 @param hypervisor_list: the list of hypervisors to query information
1257 @type all_hvparams: dict of dict of strings
1258 @param all_hvparams: a dictionary mapping hypervisor types to respective
1259 cluster-wide hypervisor parameters
1260 @type get_hv_fn: function
1261 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1262 name; optional parameter to increase testability
1265 @return: a list of all running instances on the current node
1266 - instance1.example.com
1267 - instance2.example.com
1271 for hname in hypervisor_list:
1272 hvparams = all_hvparams[hname]
1273 results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1274 get_hv_fn=get_hv_fn))
1278 def GetInstanceInfo(instance, hname, hvparams=None):
1279 """Gives back the information about an instance as a dictionary.
1281 @type instance: string
1282 @param instance: the instance name
1284 @param hname: the hypervisor type of the instance
1285 @type hvparams: dict of strings
1286 @param hvparams: the instance's hvparams
1289 @return: dictionary with the following keys:
1290 - memory: memory size of instance (int)
1291 - state: xen state of instance (string)
1292 - time: cpu time of instance (float)
1293 - vcpus: the number of vcpus (int)
1298 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1300 if iinfo is not None:
1301 output["memory"] = iinfo[2]
1302 output["vcpus"] = iinfo[3]
1303 output["state"] = iinfo[4]
1304 output["time"] = iinfo[5]
1309 def GetInstanceMigratable(instance):
1310 """Computes whether an instance can be migrated.
1312 @type instance: L{objects.Instance}
1313 @param instance: object representing the instance to be checked.
1316 @return: tuple of (result, description) where:
1317 - result: whether the instance can be migrated or not
1318 - description: a description of the issue, if relevant
1321 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1322 iname = instance.name
1323 if iname not in hyper.ListInstances(instance.hvparams):
1324 _Fail("Instance %s is not running", iname)
1326 for idx in range(len(instance.disks)):
1327 link_name = _GetBlockDevSymlinkPath(iname, idx)
1328 if not os.path.islink(link_name):
1329 logging.warning("Instance %s is missing symlink %s for disk %d",
1330 iname, link_name, idx)
1333 def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1334 """Gather data about all instances.
1336 This is the equivalent of L{GetInstanceInfo}, except that it
1337 computes data for all instances at once, thus being faster if one
1338 needs data about more than one instance.
1340 @type hypervisor_list: list
1341 @param hypervisor_list: list of hypervisors to query for instance data
1342 @type all_hvparams: dict of dict of strings
1343 @param all_hvparams: mapping of hypervisor names to hvparams
1346 @return: dictionary of instance: data, with data having the following keys:
1347 - memory: memory size of instance (int)
1348 - state: xen state of instance (string)
1349 - time: cpu time of instance (float)
1350 - vcpus: the number of vcpus
1355 for hname in hypervisor_list:
1356 hvparams = all_hvparams[hname]
1357 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1359 for name, _, memory, vcpus, state, times in iinfo:
1367 # we only check static parameters, like memory and vcpus,
1368 # and not state and time which can change between the
1369 # invocations of the different hypervisors
1370 for key in "memory", "vcpus":
1371 if value[key] != output[name][key]:
1372 _Fail("Instance %s is running twice"
1373 " with different parameters", name)
1374 output[name] = value
1379 def _InstanceLogName(kind, os_name, instance, component):
1380 """Compute the OS log filename for a given instance and operation.
1382 The instance name and os name are passed in as strings since not all
1383 operations have these as part of an instance object.
1386 @param kind: the operation type (e.g. add, import, etc.)
1387 @type os_name: string
1388 @param os_name: the os name
1389 @type instance: string
1390 @param instance: the name of the instance being imported/added/etc.
1391 @type component: string or None
1392 @param component: the name of the component of the instance being
1396 # TODO: Use tempfile.mkstemp to create unique filename
1398 assert "/" not in component
1399 c_msg = "-%s" % component
1402 base = ("%s-%s-%s%s-%s.log" %
1403 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1404 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1407 def InstanceOsAdd(instance, reinstall, debug):
1408 """Add an OS to an instance.
1410 @type instance: L{objects.Instance}
1411 @param instance: Instance whose OS is to be installed
1412 @type reinstall: boolean
1413 @param reinstall: whether this is an instance reinstall
1414 @type debug: integer
1415 @param debug: debug level, passed to the OS scripts
1419 inst_os = OSFromDisk(instance.os)
1421 create_env = OSEnvironment(instance, inst_os, debug)
1423 create_env["INSTANCE_REINSTALL"] = "1"
1425 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1427 result = utils.RunCmd([inst_os.create_script], env=create_env,
1428 cwd=inst_os.path, output=logfile, reset_env=True)
1430 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1431 " output: %s", result.cmd, result.fail_reason, logfile,
1433 lines = [utils.SafeEncode(val)
1434 for val in utils.TailFile(logfile, lines=20)]
1435 _Fail("OS create script failed (%s), last lines in the"
1436 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1439 def RunRenameInstance(instance, old_name, debug):
1440 """Run the OS rename script for an instance.
1442 @type instance: L{objects.Instance}
1443 @param instance: Instance whose OS is to be installed
1444 @type old_name: string
1445 @param old_name: previous instance name
1446 @type debug: integer
1447 @param debug: debug level, passed to the OS scripts
1449 @return: the success of the operation
1452 inst_os = OSFromDisk(instance.os)
1454 rename_env = OSEnvironment(instance, inst_os, debug)
1455 rename_env["OLD_INSTANCE_NAME"] = old_name
1457 logfile = _InstanceLogName("rename", instance.os,
1458 "%s-%s" % (old_name, instance.name), None)
1460 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1461 cwd=inst_os.path, output=logfile, reset_env=True)
1464 logging.error("os create command '%s' returned error: %s output: %s",
1465 result.cmd, result.fail_reason, result.output)
1466 lines = [utils.SafeEncode(val)
1467 for val in utils.TailFile(logfile, lines=20)]
1468 _Fail("OS rename script failed (%s), last lines in the"
1469 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1472 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1473 """Returns symlink path for block device.
1477 _dir = pathutils.DISK_LINKS_DIR
1479 return utils.PathJoin(_dir,
1481 (instance_name, constants.DISK_SEPARATOR, idx)))
1484 def _SymlinkBlockDev(instance_name, device_path, idx):
1485 """Set up symlinks to a instance's block device.
1487 This is an auxiliary function run when an instance is start (on the primary
1488 node) or when an instance is migrated (on the target node).
1491 @param instance_name: the name of the target instance
1492 @param device_path: path of the physical block device, on the node
1493 @param idx: the disk index
1494 @return: absolute path to the disk's symlink
1497 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1499 os.symlink(device_path, link_name)
1500 except OSError, err:
1501 if err.errno == errno.EEXIST:
1502 if (not os.path.islink(link_name) or
1503 os.readlink(link_name) != device_path):
1504 os.remove(link_name)
1505 os.symlink(device_path, link_name)
1512 def _RemoveBlockDevLinks(instance_name, disks):
1513 """Remove the block device symlinks belonging to the given instance.
1516 for idx, _ in enumerate(disks):
1517 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1518 if os.path.islink(link_name):
1520 os.remove(link_name)
1522 logging.exception("Can't remove symlink '%s'", link_name)
1525 def _GatherAndLinkBlockDevs(instance):
1526 """Set up an instance's block device(s).
1528 This is run on the primary node at instance startup. The block
1529 devices must be already assembled.
1531 @type instance: L{objects.Instance}
1532 @param instance: the instance whose disks we shoul assemble
1534 @return: list of (disk_object, device_path)
1538 for idx, disk in enumerate(instance.disks):
1539 device = _RecursiveFindBD(disk)
1541 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1545 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1547 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1550 block_devices.append((disk, link_name))
1552 return block_devices
1555 def StartInstance(instance, startup_paused, reason, store_reason=True):
1556 """Start an instance.
1558 @type instance: L{objects.Instance}
1559 @param instance: the instance object
1560 @type startup_paused: bool
1561 @param instance: pause instance at startup?
1562 @type reason: list of reasons
1563 @param reason: the reason trail for this startup
1564 @type store_reason: boolean
1565 @param store_reason: whether to store the shutdown reason trail on file
1569 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1572 if instance.name in running_instances:
1573 logging.info("Instance %s already running, not starting", instance.name)
1577 block_devices = _GatherAndLinkBlockDevs(instance)
1578 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1579 hyper.StartInstance(instance, block_devices, startup_paused)
1581 _StoreInstReasonTrail(instance.name, reason)
1582 except errors.BlockDeviceError, err:
1583 _Fail("Block device error: %s", err, exc=True)
1584 except errors.HypervisorError, err:
1585 _RemoveBlockDevLinks(instance.name, instance.disks)
1586 _Fail("Hypervisor error: %s", err, exc=True)
1589 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1590 """Shut an instance down.
1592 @note: this functions uses polling with a hardcoded timeout.
1594 @type instance: L{objects.Instance}
1595 @param instance: the instance object
1596 @type timeout: integer
1597 @param timeout: maximum timeout for soft shutdown
1598 @type reason: list of reasons
1599 @param reason: the reason trail for this shutdown
1600 @type store_reason: boolean
1601 @param store_reason: whether to store the shutdown reason trail on file
1605 hv_name = instance.hypervisor
1606 hyper = hypervisor.GetHypervisor(hv_name)
1607 iname = instance.name
1609 if instance.name not in hyper.ListInstances(instance.hvparams):
1610 logging.info("Instance %s not running, doing nothing", iname)
1615 self.tried_once = False
1618 if iname not in hyper.ListInstances(instance.hvparams):
1622 hyper.StopInstance(instance, retry=self.tried_once)
1624 _StoreInstReasonTrail(instance.name, reason)
1625 except errors.HypervisorError, err:
1626 if iname not in hyper.ListInstances(instance.hvparams):
1627 # if the instance is no longer existing, consider this a
1628 # success and go to cleanup
1631 _Fail("Failed to stop instance %s: %s", iname, err)
1633 self.tried_once = True
1635 raise utils.RetryAgain()
1638 utils.Retry(_TryShutdown(), 5, timeout)
1639 except utils.RetryTimeout:
1640 # the shutdown did not succeed
1641 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1644 hyper.StopInstance(instance, force=True)
1645 except errors.HypervisorError, err:
1646 if iname in hyper.ListInstances(instance.hvparams):
1647 # only raise an error if the instance still exists, otherwise
1648 # the error could simply be "instance ... unknown"!
1649 _Fail("Failed to force stop instance %s: %s", iname, err)
1653 if iname in hyper.ListInstances(instance.hvparams):
1654 _Fail("Could not shutdown instance %s even by destroy", iname)
1657 hyper.CleanupInstance(instance.name)
1658 except errors.HypervisorError, err:
1659 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1661 _RemoveBlockDevLinks(iname, instance.disks)
1664 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1665 """Reboot an instance.
1667 @type instance: L{objects.Instance}
1668 @param instance: the instance object to reboot
1669 @type reboot_type: str
1670 @param reboot_type: the type of reboot, one the following
1672 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1673 instance OS, do not recreate the VM
1674 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1675 restart the VM (at the hypervisor level)
1676 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1677 not accepted here, since that mode is handled differently, in
1678 cmdlib, and translates into full stop and start of the
1679 instance (instead of a call_instance_reboot RPC)
1680 @type shutdown_timeout: integer
1681 @param shutdown_timeout: maximum timeout for soft shutdown
1682 @type reason: list of reasons
1683 @param reason: the reason trail for this reboot
1687 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1690 if instance.name not in running_instances:
1691 _Fail("Cannot reboot instance %s that is not running", instance.name)
1693 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1694 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1696 hyper.RebootInstance(instance)
1697 except errors.HypervisorError, err:
1698 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1699 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1701 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1702 result = StartInstance(instance, False, reason, store_reason=False)
1703 _StoreInstReasonTrail(instance.name, reason)
1705 except errors.HypervisorError, err:
1706 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1708 _Fail("Invalid reboot_type received: %s", reboot_type)
1711 def InstanceBalloonMemory(instance, memory):
1712 """Resize an instance's memory.
1714 @type instance: L{objects.Instance}
1715 @param instance: the instance object
1717 @param memory: new memory amount in MB
1721 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1722 running = hyper.ListInstances(instance.hvparams)
1723 if instance.name not in running:
1724 logging.info("Instance %s is not running, cannot balloon", instance.name)
1727 hyper.BalloonInstanceMemory(instance, memory)
1728 except errors.HypervisorError, err:
1729 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1732 def MigrationInfo(instance):
1733 """Gather information about an instance to be migrated.
1735 @type instance: L{objects.Instance}
1736 @param instance: the instance definition
1739 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1741 info = hyper.MigrationInfo(instance)
1742 except errors.HypervisorError, err:
1743 _Fail("Failed to fetch migration information: %s", err, exc=True)
1747 def AcceptInstance(instance, info, target):
1748 """Prepare the node to accept an instance.
1750 @type instance: L{objects.Instance}
1751 @param instance: the instance definition
1752 @type info: string/data (opaque)
1753 @param info: migration information, from the source node
1754 @type target: string
1755 @param target: target host (usually ip), on this node
1758 # TODO: why is this required only for DTS_EXT_MIRROR?
1759 if instance.disk_template in constants.DTS_EXT_MIRROR:
1760 # Create the symlinks, as the disks are not active
1763 _GatherAndLinkBlockDevs(instance)
1764 except errors.BlockDeviceError, err:
1765 _Fail("Block device error: %s", err, exc=True)
1767 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1769 hyper.AcceptInstance(instance, info, target)
1770 except errors.HypervisorError, err:
1771 if instance.disk_template in constants.DTS_EXT_MIRROR:
1772 _RemoveBlockDevLinks(instance.name, instance.disks)
1773 _Fail("Failed to accept instance: %s", err, exc=True)
1776 def FinalizeMigrationDst(instance, info, success):
1777 """Finalize any preparation to accept an instance.
1779 @type instance: L{objects.Instance}
1780 @param instance: the instance definition
1781 @type info: string/data (opaque)
1782 @param info: migration information, from the source node
1783 @type success: boolean
1784 @param success: whether the migration was a success or a failure
1787 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1789 hyper.FinalizeMigrationDst(instance, info, success)
1790 except errors.HypervisorError, err:
1791 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1794 def MigrateInstance(cluster_name, instance, target, live):
1795 """Migrates an instance to another node.
1797 @type cluster_name: string
1798 @param cluster_name: name of the cluster
1799 @type instance: L{objects.Instance}
1800 @param instance: the instance definition
1801 @type target: string
1802 @param target: the target node name
1804 @param live: whether the migration should be done live or not (the
1805 interpretation of this parameter is left to the hypervisor)
1806 @raise RPCFail: if migration fails for some reason
1809 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1812 hyper.MigrateInstance(cluster_name, instance, target, live)
1813 except errors.HypervisorError, err:
1814 _Fail("Failed to migrate instance: %s", err, exc=True)
1817 def FinalizeMigrationSource(instance, success, live):
1818 """Finalize the instance migration on the source node.
1820 @type instance: L{objects.Instance}
1821 @param instance: the instance definition of the migrated instance
1823 @param success: whether the migration succeeded or not
1825 @param live: whether the user requested a live migration or not
1826 @raise RPCFail: If the execution fails for some reason
1829 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1832 hyper.FinalizeMigrationSource(instance, success, live)
1833 except Exception, err: # pylint: disable=W0703
1834 _Fail("Failed to finalize the migration on the source node: %s", err,
1838 def GetMigrationStatus(instance):
1839 """Get the migration status
1841 @type instance: L{objects.Instance}
1842 @param instance: the instance that is being migrated
1843 @rtype: L{objects.MigrationStatus}
1844 @return: the status of the current migration (one of
1845 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1846 progress info that can be retrieved from the hypervisor
1847 @raise RPCFail: If the migration status cannot be retrieved
1850 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1852 return hyper.GetMigrationStatus(instance)
1853 except Exception, err: # pylint: disable=W0703
1854 _Fail("Failed to get migration status: %s", err, exc=True)
1857 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1858 """Creates a block device for an instance.
1860 @type disk: L{objects.Disk}
1861 @param disk: the object describing the disk we should create
1863 @param size: the size of the physical underlying device, in MiB
1865 @param owner: the name of the instance for which disk is created,
1866 used for device cache data
1867 @type on_primary: boolean
1868 @param on_primary: indicates if it is the primary node or not
1870 @param info: string that will be sent to the physical device
1871 creation, used for example to set (LVM) tags on LVs
1872 @type excl_stor: boolean
1873 @param excl_stor: Whether exclusive_storage is active
1875 @return: the new unique_id of the device (this can sometime be
1876 computed only after creation), or None. On secondary nodes,
1877 it's not required to return anything.
1880 # TODO: remove the obsolete "size" argument
1881 # pylint: disable=W0613
1884 for child in disk.children:
1886 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1887 except errors.BlockDeviceError, err:
1888 _Fail("Can't assemble device %s: %s", child, err)
1889 if on_primary or disk.AssembleOnSecondary():
1890 # we need the children open in case the device itself has to
1893 # pylint: disable=E1103
1895 except errors.BlockDeviceError, err:
1896 _Fail("Can't make child '%s' read-write: %s", child, err)
1900 device = bdev.Create(disk, clist, excl_stor)
1901 except errors.BlockDeviceError, err:
1902 _Fail("Can't create block device: %s", err)
1904 if on_primary or disk.AssembleOnSecondary():
1907 except errors.BlockDeviceError, err:
1908 _Fail("Can't assemble device after creation, unusual event: %s", err)
1909 if on_primary or disk.OpenOnSecondary():
1911 device.Open(force=True)
1912 except errors.BlockDeviceError, err:
1913 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1914 DevCacheManager.UpdateCache(device.dev_path, owner,
1915 on_primary, disk.iv_name)
1917 device.SetInfo(info)
1919 return device.unique_id
1922 def _WipeDevice(path, offset, size):
1923 """This function actually wipes the device.
1925 @param path: The path to the device to wipe
1926 @param offset: The offset in MiB in the file
1927 @param size: The size in MiB to write
1930 # Internal sizes are always in Mebibytes; if the following "dd" command
1931 # should use a different block size the offset and size given to this
1932 # function must be adjusted accordingly before being passed to "dd".
1933 block_size = 1024 * 1024
1935 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1936 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1938 result = utils.RunCmd(cmd)
1941 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1942 result.fail_reason, result.output)
1945 def BlockdevWipe(disk, offset, size):
1946 """Wipes a block device.
1948 @type disk: L{objects.Disk}
1949 @param disk: the disk object we want to wipe
1951 @param offset: The offset in MiB in the file
1953 @param size: The size in MiB to write
1957 rdev = _RecursiveFindBD(disk)
1958 except errors.BlockDeviceError:
1962 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1964 # Do cross verify some of the parameters
1966 _Fail("Negative offset")
1968 _Fail("Negative size")
1969 if offset > rdev.size:
1970 _Fail("Offset is bigger than device size")
1971 if (offset + size) > rdev.size:
1972 _Fail("The provided offset and size to wipe is bigger than device size")
1974 _WipeDevice(rdev.dev_path, offset, size)
1977 def BlockdevPauseResumeSync(disks, pause):
1978 """Pause or resume the sync of the block device.
1980 @type disks: list of L{objects.Disk}
1981 @param disks: the disks object we want to pause/resume
1983 @param pause: Wheater to pause or resume
1989 rdev = _RecursiveFindBD(disk)
1990 except errors.BlockDeviceError:
1994 success.append((False, ("Cannot change sync for device %s:"
1995 " device not found" % disk.iv_name)))
1998 result = rdev.PauseResumeSync(pause)
2001 success.append((result, None))
2007 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2012 def BlockdevRemove(disk):
2013 """Remove a block device.
2015 @note: This is intended to be called recursively.
2017 @type disk: L{objects.Disk}
2018 @param disk: the disk object we should remove
2020 @return: the success of the operation
2025 rdev = _RecursiveFindBD(disk)
2026 except errors.BlockDeviceError, err:
2027 # probably can't attach
2028 logging.info("Can't attach to device %s in remove", disk)
2030 if rdev is not None:
2031 r_path = rdev.dev_path
2034 except errors.BlockDeviceError, err:
2035 msgs.append(str(err))
2037 DevCacheManager.RemoveCache(r_path)
2040 for child in disk.children:
2042 BlockdevRemove(child)
2043 except RPCFail, err:
2044 msgs.append(str(err))
2047 _Fail("; ".join(msgs))
2050 def _RecursiveAssembleBD(disk, owner, as_primary):
2051 """Activate a block device for an instance.
2053 This is run on the primary and secondary nodes for an instance.
2055 @note: this function is called recursively.
2057 @type disk: L{objects.Disk}
2058 @param disk: the disk we try to assemble
2060 @param owner: the name of the instance which owns the disk
2061 @type as_primary: boolean
2062 @param as_primary: if we should make the block device
2065 @return: the assembled device or None (in case no device
2067 @raise errors.BlockDeviceError: in case there is an error
2068 during the activation of the children or the device
2074 mcn = disk.ChildrenNeeded()
2076 mcn = 0 # max number of Nones allowed
2078 mcn = len(disk.children) - mcn # max number of Nones
2079 for chld_disk in disk.children:
2081 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2082 except errors.BlockDeviceError, err:
2083 if children.count(None) >= mcn:
2086 logging.error("Error in child activation (but continuing): %s",
2088 children.append(cdev)
2090 if as_primary or disk.AssembleOnSecondary():
2091 r_dev = bdev.Assemble(disk, children)
2093 if as_primary or disk.OpenOnSecondary():
2095 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2096 as_primary, disk.iv_name)
2103 def BlockdevAssemble(disk, owner, as_primary, idx):
2104 """Activate a block device for an instance.
2106 This is a wrapper over _RecursiveAssembleBD.
2108 @rtype: str or boolean
2109 @return: a C{/dev/...} path for primary nodes, and
2110 C{True} for secondary nodes
2114 result = _RecursiveAssembleBD(disk, owner, as_primary)
2115 if isinstance(result, BlockDev):
2116 # pylint: disable=E1103
2117 result = result.dev_path
2119 _SymlinkBlockDev(owner, result, idx)
2120 except errors.BlockDeviceError, err:
2121 _Fail("Error while assembling disk: %s", err, exc=True)
2122 except OSError, err:
2123 _Fail("Error while symlinking disk: %s", err, exc=True)
2128 def BlockdevShutdown(disk):
2129 """Shut down a block device.
2131 First, if the device is assembled (Attach() is successful), then
2132 the device is shutdown. Then the children of the device are
2135 This function is called recursively. Note that we don't cache the
2136 children or such, as oppossed to assemble, shutdown of different
2137 devices doesn't require that the upper device was active.
2139 @type disk: L{objects.Disk}
2140 @param disk: the description of the disk we should
2146 r_dev = _RecursiveFindBD(disk)
2147 if r_dev is not None:
2148 r_path = r_dev.dev_path
2151 DevCacheManager.RemoveCache(r_path)
2152 except errors.BlockDeviceError, err:
2153 msgs.append(str(err))
2156 for child in disk.children:
2158 BlockdevShutdown(child)
2159 except RPCFail, err:
2160 msgs.append(str(err))
2163 _Fail("; ".join(msgs))
2166 def BlockdevAddchildren(parent_cdev, new_cdevs):
2167 """Extend a mirrored block device.
2169 @type parent_cdev: L{objects.Disk}
2170 @param parent_cdev: the disk to which we should add children
2171 @type new_cdevs: list of L{objects.Disk}
2172 @param new_cdevs: the list of children which we should add
2176 parent_bdev = _RecursiveFindBD(parent_cdev)
2177 if parent_bdev is None:
2178 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2179 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2180 if new_bdevs.count(None) > 0:
2181 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2182 parent_bdev.AddChildren(new_bdevs)
2185 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2186 """Shrink a mirrored block device.
2188 @type parent_cdev: L{objects.Disk}
2189 @param parent_cdev: the disk from which we should remove children
2190 @type new_cdevs: list of L{objects.Disk}
2191 @param new_cdevs: the list of children which we should remove
2195 parent_bdev = _RecursiveFindBD(parent_cdev)
2196 if parent_bdev is None:
2197 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2199 for disk in new_cdevs:
2200 rpath = disk.StaticDevPath()
2202 bd = _RecursiveFindBD(disk)
2204 _Fail("Can't find device %s while removing children", disk)
2206 devs.append(bd.dev_path)
2208 if not utils.IsNormAbsPath(rpath):
2209 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2211 parent_bdev.RemoveChildren(devs)
2214 def BlockdevGetmirrorstatus(disks):
2215 """Get the mirroring status of a list of devices.
2217 @type disks: list of L{objects.Disk}
2218 @param disks: the list of disks which we should query
2220 @return: List of L{objects.BlockDevStatus}, one for each disk
2221 @raise errors.BlockDeviceError: if any of the disks cannot be
2227 rbd = _RecursiveFindBD(dsk)
2229 _Fail("Can't find device %s", dsk)
2231 stats.append(rbd.CombinedSyncStatus())
2236 def BlockdevGetmirrorstatusMulti(disks):
2237 """Get the mirroring status of a list of devices.
2239 @type disks: list of L{objects.Disk}
2240 @param disks: the list of disks which we should query
2242 @return: List of tuples, (bool, status), one for each disk; bool denotes
2243 success/failure, status is L{objects.BlockDevStatus} on success, string
2250 rbd = _RecursiveFindBD(disk)
2252 result.append((False, "Can't find device %s" % disk))
2255 status = rbd.CombinedSyncStatus()
2256 except errors.BlockDeviceError, err:
2257 logging.exception("Error while getting disk status")
2258 result.append((False, str(err)))
2260 result.append((True, status))
2262 assert len(disks) == len(result)
2267 def _RecursiveFindBD(disk):
2268 """Check if a device is activated.
2270 If so, return information about the real device.
2272 @type disk: L{objects.Disk}
2273 @param disk: the disk object we need to find
2275 @return: None if the device can't be found,
2276 otherwise the device instance
2281 for chdisk in disk.children:
2282 children.append(_RecursiveFindBD(chdisk))
2284 return bdev.FindDevice(disk, children)
2287 def _OpenRealBD(disk):
2288 """Opens the underlying block device of a disk.
2290 @type disk: L{objects.Disk}
2291 @param disk: the disk object we want to open
2294 real_disk = _RecursiveFindBD(disk)
2295 if real_disk is None:
2296 _Fail("Block device '%s' is not set up", disk)
2303 def BlockdevFind(disk):
2304 """Check if a device is activated.
2306 If it is, return information about the real device.
2308 @type disk: L{objects.Disk}
2309 @param disk: the disk to find
2310 @rtype: None or objects.BlockDevStatus
2311 @return: None if the disk cannot be found, otherwise a the current
2316 rbd = _RecursiveFindBD(disk)
2317 except errors.BlockDeviceError, err:
2318 _Fail("Failed to find device: %s", err, exc=True)
2323 return rbd.GetSyncStatus()
2326 def BlockdevGetdimensions(disks):
2327 """Computes the size of the given disks.
2329 If a disk is not found, returns None instead.
2331 @type disks: list of L{objects.Disk}
2332 @param disks: the list of disk to compute the size for
2334 @return: list with elements None if the disk cannot be found,
2335 otherwise the pair (size, spindles), where spindles is None if the
2336 device doesn't support that
2342 rbd = _RecursiveFindBD(cf)
2343 except errors.BlockDeviceError:
2349 result.append(rbd.GetActualDimensions())
2353 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2354 """Export a block device to a remote node.
2356 @type disk: L{objects.Disk}
2357 @param disk: the description of the disk to export
2358 @type dest_node: str
2359 @param dest_node: the destination node to export to
2360 @type dest_path: str
2361 @param dest_path: the destination path on the target node
2362 @type cluster_name: str
2363 @param cluster_name: the cluster name, needed for SSH hostalias
2367 real_disk = _OpenRealBD(disk)
2369 # the block size on the read dd is 1MiB to match our units
2370 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2371 "dd if=%s bs=1048576 count=%s",
2372 real_disk.dev_path, str(disk.size))
2374 # we set here a smaller block size as, due to ssh buffering, more
2375 # than 64-128k will mostly ignored; we use nocreat to fail if the
2376 # device is not already there or we pass a wrong path; we use
2377 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2378 # to not buffer too much memory; this means that at best, we flush
2379 # every 64k, which will not be very fast
2380 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2381 " oflag=dsync", dest_path)
2383 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2384 constants.SSH_LOGIN_USER,
2387 # all commands have been checked, so we're safe to combine them
2388 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2390 result = utils.RunCmd(["bash", "-c", command])
2393 _Fail("Disk copy command '%s' returned error: %s"
2394 " output: %s", command, result.fail_reason, result.output)
2397 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2398 """Write a file to the filesystem.
2400 This allows the master to overwrite(!) a file. It will only perform
2401 the operation if the file belongs to a list of configuration files.
2403 @type file_name: str
2404 @param file_name: the target file name
2406 @param data: the new contents of the file
2408 @param mode: the mode to give the file (can be None)
2410 @param uid: the owner of the file
2412 @param gid: the group of the file
2414 @param atime: the atime to set on the file (can be None)
2416 @param mtime: the mtime to set on the file (can be None)
2420 file_name = vcluster.LocalizeVirtualPath(file_name)
2422 if not os.path.isabs(file_name):
2423 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2425 if file_name not in _ALLOWED_UPLOAD_FILES:
2426 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2429 raw_data = _Decompress(data)
2431 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2432 _Fail("Invalid username/groupname type")
2434 getents = runtime.GetEnts()
2435 uid = getents.LookupUser(uid)
2436 gid = getents.LookupGroup(gid)
2438 utils.SafeWriteFile(file_name, None,
2439 data=raw_data, mode=mode, uid=uid, gid=gid,
2440 atime=atime, mtime=mtime)
2443 def RunOob(oob_program, command, node, timeout):
2444 """Executes oob_program with given command on given node.
2446 @param oob_program: The path to the executable oob_program
2447 @param command: The command to invoke on oob_program
2448 @param node: The node given as an argument to the program
2449 @param timeout: Timeout after which we kill the oob program
2452 @raise RPCFail: If execution fails for some reason
2455 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2458 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2459 result.fail_reason, result.output)
2461 return result.stdout
2464 def _OSOndiskAPIVersion(os_dir):
2465 """Compute and return the API version of a given OS.
2467 This function will try to read the API version of the OS residing in
2468 the 'os_dir' directory.
2471 @param os_dir: the directory in which we should look for the OS
2473 @return: tuple (status, data) with status denoting the validity and
2474 data holding either the vaid versions or an error message
2477 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2480 st = os.stat(api_file)
2481 except EnvironmentError, err:
2482 return False, ("Required file '%s' not found under path %s: %s" %
2483 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2485 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2486 return False, ("File '%s' in %s is not a regular file" %
2487 (constants.OS_API_FILE, os_dir))
2490 api_versions = utils.ReadFile(api_file).splitlines()
2491 except EnvironmentError, err:
2492 return False, ("Error while reading the API version file at %s: %s" %
2493 (api_file, utils.ErrnoOrStr(err)))
2496 api_versions = [int(version.strip()) for version in api_versions]
2497 except (TypeError, ValueError), err:
2498 return False, ("API version(s) can't be converted to integer: %s" %
2501 return True, api_versions
2504 def DiagnoseOS(top_dirs=None):
2505 """Compute the validity for all OSes.
2507 @type top_dirs: list
2508 @param top_dirs: the list of directories in which to
2509 search (if not given defaults to
2510 L{pathutils.OS_SEARCH_PATH})
2511 @rtype: list of L{objects.OS}
2512 @return: a list of tuples (name, path, status, diagnose, variants,
2513 parameters, api_version) for all (potential) OSes under all
2514 search paths, where:
2515 - name is the (potential) OS name
2516 - path is the full path to the OS
2517 - status True/False is the validity of the OS
2518 - diagnose is the error message for an invalid OS, otherwise empty
2519 - variants is a list of supported OS variants, if any
2520 - parameters is a list of (name, help) parameters, if any
2521 - api_version is a list of support OS API versions
2524 if top_dirs is None:
2525 top_dirs = pathutils.OS_SEARCH_PATH
2528 for dir_name in top_dirs:
2529 if os.path.isdir(dir_name):
2531 f_names = utils.ListVisibleFiles(dir_name)
2532 except EnvironmentError, err:
2533 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2535 for name in f_names:
2536 os_path = utils.PathJoin(dir_name, name)
2537 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2540 variants = os_inst.supported_variants
2541 parameters = os_inst.supported_parameters
2542 api_versions = os_inst.api_versions
2545 variants = parameters = api_versions = []
2546 result.append((name, os_path, status, diagnose, variants,
2547 parameters, api_versions))
2552 def _TryOSFromDisk(name, base_dir=None):
2553 """Create an OS instance from disk.
2555 This function will return an OS instance if the given name is a
2558 @type base_dir: string
2559 @keyword base_dir: Base directory containing OS installations.
2560 Defaults to a search in all the OS_SEARCH_PATH dirs.
2562 @return: success and either the OS instance if we find a valid one,
2566 if base_dir is None:
2567 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2569 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2572 return False, "Directory for OS %s not found in search path" % name
2574 status, api_versions = _OSOndiskAPIVersion(os_dir)
2577 return status, api_versions
2579 if not constants.OS_API_VERSIONS.intersection(api_versions):
2580 return False, ("API version mismatch for path '%s': found %s, want %s." %
2581 (os_dir, api_versions, constants.OS_API_VERSIONS))
2583 # OS Files dictionary, we will populate it with the absolute path
2584 # names; if the value is True, then it is a required file, otherwise
2586 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2588 if max(api_versions) >= constants.OS_API_V15:
2589 os_files[constants.OS_VARIANTS_FILE] = False
2591 if max(api_versions) >= constants.OS_API_V20:
2592 os_files[constants.OS_PARAMETERS_FILE] = True
2594 del os_files[constants.OS_SCRIPT_VERIFY]
2596 for (filename, required) in os_files.items():
2597 os_files[filename] = utils.PathJoin(os_dir, filename)
2600 st = os.stat(os_files[filename])
2601 except EnvironmentError, err:
2602 if err.errno == errno.ENOENT and not required:
2603 del os_files[filename]
2605 return False, ("File '%s' under path '%s' is missing (%s)" %
2606 (filename, os_dir, utils.ErrnoOrStr(err)))
2608 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2609 return False, ("File '%s' under path '%s' is not a regular file" %
2612 if filename in constants.OS_SCRIPTS:
2613 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2614 return False, ("File '%s' under path '%s' is not executable" %
2618 if constants.OS_VARIANTS_FILE in os_files:
2619 variants_file = os_files[constants.OS_VARIANTS_FILE]
2622 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2623 except EnvironmentError, err:
2624 # we accept missing files, but not other errors
2625 if err.errno != errno.ENOENT:
2626 return False, ("Error while reading the OS variants file at %s: %s" %
2627 (variants_file, utils.ErrnoOrStr(err)))
2630 if constants.OS_PARAMETERS_FILE in os_files:
2631 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2633 parameters = utils.ReadFile(parameters_file).splitlines()
2634 except EnvironmentError, err:
2635 return False, ("Error while reading the OS parameters file at %s: %s" %
2636 (parameters_file, utils.ErrnoOrStr(err)))
2637 parameters = [v.split(None, 1) for v in parameters]
2639 os_obj = objects.OS(name=name, path=os_dir,
2640 create_script=os_files[constants.OS_SCRIPT_CREATE],
2641 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2642 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2643 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2644 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2646 supported_variants=variants,
2647 supported_parameters=parameters,
2648 api_versions=api_versions)
2652 def OSFromDisk(name, base_dir=None):
2653 """Create an OS instance from disk.
2655 This function will return an OS instance if the given name is a
2656 valid OS name. Otherwise, it will raise an appropriate
2657 L{RPCFail} exception, detailing why this is not a valid OS.
2659 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2660 an exception but returns true/false status data.
2662 @type base_dir: string
2663 @keyword base_dir: Base directory containing OS installations.
2664 Defaults to a search in all the OS_SEARCH_PATH dirs.
2665 @rtype: L{objects.OS}
2666 @return: the OS instance if we find a valid one
2667 @raise RPCFail: if we don't find a valid OS
2670 name_only = objects.OS.GetName(name)
2671 status, payload = _TryOSFromDisk(name_only, base_dir)
2679 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2680 """Calculate the basic environment for an os script.
2683 @param os_name: full operating system name (including variant)
2684 @type inst_os: L{objects.OS}
2685 @param inst_os: operating system for which the environment is being built
2686 @type os_params: dict
2687 @param os_params: the OS parameters
2688 @type debug: integer
2689 @param debug: debug level (0 or 1, for OS Api 10)
2691 @return: dict of environment variables
2692 @raise errors.BlockDeviceError: if the block device
2698 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2699 result["OS_API_VERSION"] = "%d" % api_version
2700 result["OS_NAME"] = inst_os.name
2701 result["DEBUG_LEVEL"] = "%d" % debug
2704 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2705 variant = objects.OS.GetVariant(os_name)
2707 variant = inst_os.supported_variants[0]
2710 result["OS_VARIANT"] = variant
2713 for pname, pvalue in os_params.items():
2714 result["OSP_%s" % pname.upper()] = pvalue
2716 # Set a default path otherwise programs called by OS scripts (or
2717 # even hooks called from OS scripts) might break, and we don't want
2718 # to have each script require setting a PATH variable
2719 result["PATH"] = constants.HOOKS_PATH
2724 def OSEnvironment(instance, inst_os, debug=0):
2725 """Calculate the environment for an os script.
2727 @type instance: L{objects.Instance}
2728 @param instance: target instance for the os script run
2729 @type inst_os: L{objects.OS}
2730 @param inst_os: operating system for which the environment is being built
2731 @type debug: integer
2732 @param debug: debug level (0 or 1, for OS Api 10)
2734 @return: dict of environment variables
2735 @raise errors.BlockDeviceError: if the block device
2739 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2741 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2742 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2744 result["HYPERVISOR"] = instance.hypervisor
2745 result["DISK_COUNT"] = "%d" % len(instance.disks)
2746 result["NIC_COUNT"] = "%d" % len(instance.nics)
2747 result["INSTANCE_SECONDARY_NODES"] = \
2748 ("%s" % " ".join(instance.secondary_nodes))
2751 for idx, disk in enumerate(instance.disks):
2752 real_disk = _OpenRealBD(disk)
2753 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2754 result["DISK_%d_ACCESS" % idx] = disk.mode
2755 if constants.HV_DISK_TYPE in instance.hvparams:
2756 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2757 instance.hvparams[constants.HV_DISK_TYPE]
2758 if disk.dev_type in constants.LDS_BLOCK:
2759 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2760 elif disk.dev_type == constants.LD_FILE:
2761 result["DISK_%d_BACKEND_TYPE" % idx] = \
2762 "file:%s" % disk.physical_id[0]
2765 for idx, nic in enumerate(instance.nics):
2766 result["NIC_%d_MAC" % idx] = nic.mac
2768 result["NIC_%d_IP" % idx] = nic.ip
2769 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2770 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2771 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2772 if nic.nicparams[constants.NIC_LINK]:
2773 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2775 nobj = objects.Network.FromDict(nic.netinfo)
2776 result.update(nobj.HooksDict("NIC_%d_" % idx))
2777 if constants.HV_NIC_TYPE in instance.hvparams:
2778 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2779 instance.hvparams[constants.HV_NIC_TYPE]
2782 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2783 for key, value in source.items():
2784 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2789 def DiagnoseExtStorage(top_dirs=None):
2790 """Compute the validity for all ExtStorage Providers.
2792 @type top_dirs: list
2793 @param top_dirs: the list of directories in which to
2794 search (if not given defaults to
2795 L{pathutils.ES_SEARCH_PATH})
2796 @rtype: list of L{objects.ExtStorage}
2797 @return: a list of tuples (name, path, status, diagnose, parameters)
2798 for all (potential) ExtStorage Providers under all
2799 search paths, where:
2800 - name is the (potential) ExtStorage Provider
2801 - path is the full path to the ExtStorage Provider
2802 - status True/False is the validity of the ExtStorage Provider
2803 - diagnose is the error message for an invalid ExtStorage Provider,
2805 - parameters is a list of (name, help) parameters, if any
2808 if top_dirs is None:
2809 top_dirs = pathutils.ES_SEARCH_PATH
2812 for dir_name in top_dirs:
2813 if os.path.isdir(dir_name):
2815 f_names = utils.ListVisibleFiles(dir_name)
2816 except EnvironmentError, err:
2817 logging.exception("Can't list the ExtStorage directory %s: %s",
2820 for name in f_names:
2821 es_path = utils.PathJoin(dir_name, name)
2822 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2825 parameters = es_inst.supported_parameters
2829 result.append((name, es_path, status, diagnose, parameters))
2834 def BlockdevGrow(disk, amount, dryrun, backingstore):
2835 """Grow a stack of block devices.
2837 This function is called recursively, with the childrens being the
2838 first ones to resize.
2840 @type disk: L{objects.Disk}
2841 @param disk: the disk to be grown
2842 @type amount: integer
2843 @param amount: the amount (in mebibytes) to grow with
2844 @type dryrun: boolean
2845 @param dryrun: whether to execute the operation in simulation mode
2846 only, without actually increasing the size
2847 @param backingstore: whether to execute the operation on backing storage
2848 only, or on "logical" storage only; e.g. DRBD is logical storage,
2849 whereas LVM, file, RBD are backing storage
2850 @rtype: (status, result)
2851 @return: a tuple with the status of the operation (True/False), and
2852 the errors message if status is False
2855 r_dev = _RecursiveFindBD(disk)
2857 _Fail("Cannot find block device %s", disk)
2860 r_dev.Grow(amount, dryrun, backingstore)
2861 except errors.BlockDeviceError, err:
2862 _Fail("Failed to grow block device: %s", err, exc=True)
2865 def BlockdevSnapshot(disk):
2866 """Create a snapshot copy of a block device.
2868 This function is called recursively, and the snapshot is actually created
2869 just for the leaf lvm backend device.
2871 @type disk: L{objects.Disk}
2872 @param disk: the disk to be snapshotted
2874 @return: snapshot disk ID as (vg, lv)
2877 if disk.dev_type == constants.LD_DRBD8:
2878 if not disk.children:
2879 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2881 return BlockdevSnapshot(disk.children[0])
2882 elif disk.dev_type == constants.LD_LV:
2883 r_dev = _RecursiveFindBD(disk)
2884 if r_dev is not None:
2885 # FIXME: choose a saner value for the snapshot size
2886 # let's stay on the safe side and ask for the full size, for now
2887 return r_dev.Snapshot(disk.size)
2889 _Fail("Cannot find block device %s", disk)
2891 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2892 disk.unique_id, disk.dev_type)
2895 def BlockdevSetInfo(disk, info):
2896 """Sets 'metadata' information on block devices.
2898 This function sets 'info' metadata on block devices. Initial
2899 information is set at device creation; this function should be used
2900 for example after renames.
2902 @type disk: L{objects.Disk}
2903 @param disk: the disk to be grown
2905 @param info: new 'info' metadata
2906 @rtype: (status, result)
2907 @return: a tuple with the status of the operation (True/False), and
2908 the errors message if status is False
2911 r_dev = _RecursiveFindBD(disk)
2913 _Fail("Cannot find block device %s", disk)
2917 except errors.BlockDeviceError, err:
2918 _Fail("Failed to set information on block device: %s", err, exc=True)
2921 def FinalizeExport(instance, snap_disks):
2922 """Write out the export configuration information.
2924 @type instance: L{objects.Instance}
2925 @param instance: the instance which we export, used for
2926 saving configuration
2927 @type snap_disks: list of L{objects.Disk}
2928 @param snap_disks: list of snapshot block devices, which
2929 will be used to get the actual name of the dump file
2934 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2935 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2937 config = objects.SerializableConfigParser()
2939 config.add_section(constants.INISECT_EXP)
2940 config.set(constants.INISECT_EXP, "version", "0")
2941 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2942 config.set(constants.INISECT_EXP, "source", instance.primary_node)
2943 config.set(constants.INISECT_EXP, "os", instance.os)
2944 config.set(constants.INISECT_EXP, "compression", "none")
2946 config.add_section(constants.INISECT_INS)
2947 config.set(constants.INISECT_INS, "name", instance.name)
2948 config.set(constants.INISECT_INS, "maxmem", "%d" %
2949 instance.beparams[constants.BE_MAXMEM])
2950 config.set(constants.INISECT_INS, "minmem", "%d" %
2951 instance.beparams[constants.BE_MINMEM])
2952 # "memory" is deprecated, but useful for exporting to old ganeti versions
2953 config.set(constants.INISECT_INS, "memory", "%d" %
2954 instance.beparams[constants.BE_MAXMEM])
2955 config.set(constants.INISECT_INS, "vcpus", "%d" %
2956 instance.beparams[constants.BE_VCPUS])
2957 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2958 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2959 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2962 for nic_count, nic in enumerate(instance.nics):
2964 config.set(constants.INISECT_INS, "nic%d_mac" %
2965 nic_count, "%s" % nic.mac)
2966 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2967 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2969 for param in constants.NICS_PARAMETER_TYPES:
2970 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2971 "%s" % nic.nicparams.get(param, None))
2972 # TODO: redundant: on load can read nics until it doesn't exist
2973 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2976 for disk_count, disk in enumerate(snap_disks):
2979 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2980 ("%s" % disk.iv_name))
2981 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2982 ("%s" % disk.physical_id[1]))
2983 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2986 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2988 # New-style hypervisor/backend parameters
2990 config.add_section(constants.INISECT_HYP)
2991 for name, value in instance.hvparams.items():
2992 if name not in constants.HVC_GLOBALS:
2993 config.set(constants.INISECT_HYP, name, str(value))
2995 config.add_section(constants.INISECT_BEP)
2996 for name, value in instance.beparams.items():
2997 config.set(constants.INISECT_BEP, name, str(value))
2999 config.add_section(constants.INISECT_OSP)
3000 for name, value in instance.osparams.items():
3001 config.set(constants.INISECT_OSP, name, str(value))
3003 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3004 data=config.Dumps())
3005 shutil.rmtree(finaldestdir, ignore_errors=True)
3006 shutil.move(destdir, finaldestdir)
3009 def ExportInfo(dest):
3010 """Get export configuration information.
3013 @param dest: directory containing the export
3015 @rtype: L{objects.SerializableConfigParser}
3016 @return: a serializable config file containing the
3020 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3022 config = objects.SerializableConfigParser()
3025 if (not config.has_section(constants.INISECT_EXP) or
3026 not config.has_section(constants.INISECT_INS)):
3027 _Fail("Export info file doesn't have the required fields")
3029 return config.Dumps()
3033 """Return a list of exports currently available on this machine.
3036 @return: list of the exports
3039 if os.path.isdir(pathutils.EXPORT_DIR):
3040 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3042 _Fail("No exports directory")
3045 def RemoveExport(export):
3046 """Remove an existing export from the node.
3049 @param export: the name of the export to remove
3053 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3056 shutil.rmtree(target)
3057 except EnvironmentError, err:
3058 _Fail("Error while removing the export: %s", err, exc=True)
3061 def BlockdevRename(devlist):
3062 """Rename a list of block devices.
3064 @type devlist: list of tuples
3065 @param devlist: list of tuples of the form (disk,
3066 new_logical_id, new_physical_id); disk is an
3067 L{objects.Disk} object describing the current disk,
3068 and new logical_id/physical_id is the name we
3071 @return: True if all renames succeeded, False otherwise
3076 for disk, unique_id in devlist:
3077 dev = _RecursiveFindBD(disk)
3079 msgs.append("Can't find device %s in rename" % str(disk))
3083 old_rpath = dev.dev_path
3084 dev.Rename(unique_id)
3085 new_rpath = dev.dev_path
3086 if old_rpath != new_rpath:
3087 DevCacheManager.RemoveCache(old_rpath)
3088 # FIXME: we should add the new cache information here, like:
3089 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3090 # but we don't have the owner here - maybe parse from existing
3091 # cache? for now, we only lose lvm data when we rename, which
3092 # is less critical than DRBD or MD
3093 except errors.BlockDeviceError, err:
3094 msgs.append("Can't rename device '%s' to '%s': %s" %
3095 (dev, unique_id, err))
3096 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3099 _Fail("; ".join(msgs))
3102 def _TransformFileStorageDir(fs_dir):
3103 """Checks whether given file_storage_dir is valid.
3105 Checks wheter the given fs_dir is within the cluster-wide default
3106 file_storage_dir or the shared_file_storage_dir, which are stored in
3107 SimpleStore. Only paths under those directories are allowed.
3110 @param fs_dir: the path to check
3112 @return: the normalized path if valid, None otherwise
3115 if not (constants.ENABLE_FILE_STORAGE or
3116 constants.ENABLE_SHARED_FILE_STORAGE):
3117 _Fail("File storage disabled at configure time")
3119 bdev.CheckFileStoragePath(fs_dir)
3121 return os.path.normpath(fs_dir)
3124 def CreateFileStorageDir(file_storage_dir):
3125 """Create file storage directory.
3127 @type file_storage_dir: str
3128 @param file_storage_dir: directory to create
3131 @return: tuple with first element a boolean indicating wheter dir
3132 creation was successful or not
3135 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3136 if os.path.exists(file_storage_dir):
3137 if not os.path.isdir(file_storage_dir):
3138 _Fail("Specified storage dir '%s' is not a directory",
3142 os.makedirs(file_storage_dir, 0750)
3143 except OSError, err:
3144 _Fail("Cannot create file storage directory '%s': %s",
3145 file_storage_dir, err, exc=True)
3148 def RemoveFileStorageDir(file_storage_dir):
3149 """Remove file storage directory.
3151 Remove it only if it's empty. If not log an error and return.
3153 @type file_storage_dir: str
3154 @param file_storage_dir: the directory we should cleanup
3155 @rtype: tuple (success,)
3156 @return: tuple of one element, C{success}, denoting
3157 whether the operation was successful
3160 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3161 if os.path.exists(file_storage_dir):
3162 if not os.path.isdir(file_storage_dir):
3163 _Fail("Specified Storage directory '%s' is not a directory",
3165 # deletes dir only if empty, otherwise we want to fail the rpc call
3167 os.rmdir(file_storage_dir)
3168 except OSError, err:
3169 _Fail("Cannot remove file storage directory '%s': %s",
3170 file_storage_dir, err)
3173 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3174 """Rename the file storage directory.
3176 @type old_file_storage_dir: str
3177 @param old_file_storage_dir: the current path
3178 @type new_file_storage_dir: str
3179 @param new_file_storage_dir: the name we should rename to
3180 @rtype: tuple (success,)
3181 @return: tuple of one element, C{success}, denoting
3182 whether the operation was successful
3185 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3186 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3187 if not os.path.exists(new_file_storage_dir):
3188 if os.path.isdir(old_file_storage_dir):
3190 os.rename(old_file_storage_dir, new_file_storage_dir)
3191 except OSError, err:
3192 _Fail("Cannot rename '%s' to '%s': %s",
3193 old_file_storage_dir, new_file_storage_dir, err)
3195 _Fail("Specified storage dir '%s' is not a directory",
3196 old_file_storage_dir)
3198 if os.path.exists(old_file_storage_dir):
3199 _Fail("Cannot rename '%s' to '%s': both locations exist",
3200 old_file_storage_dir, new_file_storage_dir)
3203 def _EnsureJobQueueFile(file_name):
3204 """Checks whether the given filename is in the queue directory.
3206 @type file_name: str
3207 @param file_name: the file name we should check
3209 @raises RPCFail: if the file is not valid
3212 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3213 _Fail("Passed job queue file '%s' does not belong to"
3214 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3217 def JobQueueUpdate(file_name, content):
3218 """Updates a file in the queue directory.
3220 This is just a wrapper over L{utils.io.WriteFile}, with proper
3223 @type file_name: str
3224 @param file_name: the job file name
3226 @param content: the new job contents
3228 @return: the success of the operation
3231 file_name = vcluster.LocalizeVirtualPath(file_name)
3233 _EnsureJobQueueFile(file_name)
3234 getents = runtime.GetEnts()
3236 # Write and replace the file atomically
3237 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3238 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3241 def JobQueueRename(old, new):
3242 """Renames a job queue file.
3244 This is just a wrapper over os.rename with proper checking.
3247 @param old: the old (actual) file name
3249 @param new: the desired file name
3251 @return: the success of the operation and payload
3254 old = vcluster.LocalizeVirtualPath(old)
3255 new = vcluster.LocalizeVirtualPath(new)
3257 _EnsureJobQueueFile(old)
3258 _EnsureJobQueueFile(new)
3260 getents = runtime.GetEnts()
3262 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3263 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3266 def BlockdevClose(instance_name, disks):
3267 """Closes the given block devices.
3269 This means they will be switched to secondary mode (in case of
3272 @param instance_name: if the argument is not empty, the symlinks
3273 of this instance will be removed
3274 @type disks: list of L{objects.Disk}
3275 @param disks: the list of disks to be closed
3276 @rtype: tuple (success, message)
3277 @return: a tuple of success and message, where success
3278 indicates the succes of the operation, and message
3279 which will contain the error details in case we
3285 rd = _RecursiveFindBD(cf)
3287 _Fail("Can't find device %s", cf)
3294 except errors.BlockDeviceError, err:
3295 msg.append(str(err))
3297 _Fail("Can't make devices secondary: %s", ",".join(msg))
3300 _RemoveBlockDevLinks(instance_name, disks)
3303 def ValidateHVParams(hvname, hvparams):
3304 """Validates the given hypervisor parameters.
3306 @type hvname: string
3307 @param hvname: the hypervisor name
3308 @type hvparams: dict
3309 @param hvparams: the hypervisor parameters to be validated
3314 hv_type = hypervisor.GetHypervisor(hvname)
3315 hv_type.ValidateParameters(hvparams)
3316 except errors.HypervisorError, err:
3317 _Fail(str(err), log=False)
3320 def _CheckOSPList(os_obj, parameters):
3321 """Check whether a list of parameters is supported by the OS.
3323 @type os_obj: L{objects.OS}
3324 @param os_obj: OS object to check
3325 @type parameters: list
3326 @param parameters: the list of parameters to check
3329 supported = [v[0] for v in os_obj.supported_parameters]
3330 delta = frozenset(parameters).difference(supported)
3332 _Fail("The following parameters are not supported"
3333 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3336 def ValidateOS(required, osname, checks, osparams):
3337 """Validate the given OS' parameters.
3339 @type required: boolean
3340 @param required: whether absence of the OS should translate into
3342 @type osname: string
3343 @param osname: the OS to be validated
3345 @param checks: list of the checks to run (currently only 'parameters')
3346 @type osparams: dict
3347 @param osparams: dictionary with OS parameters
3349 @return: True if the validation passed, or False if the OS was not
3350 found and L{required} was false
3353 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3354 _Fail("Unknown checks required for OS %s: %s", osname,
3355 set(checks).difference(constants.OS_VALIDATE_CALLS))
3357 name_only = objects.OS.GetName(osname)
3358 status, tbv = _TryOSFromDisk(name_only, None)
3366 if max(tbv.api_versions) < constants.OS_API_V20:
3369 if constants.OS_VALIDATE_PARAMETERS in checks:
3370 _CheckOSPList(tbv, osparams.keys())
3372 validate_env = OSCoreEnv(osname, tbv, osparams)
3373 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3374 cwd=tbv.path, reset_env=True)
3376 logging.error("os validate command '%s' returned error: %s output: %s",
3377 result.cmd, result.fail_reason, result.output)
3378 _Fail("OS validation script failed (%s), output: %s",
3379 result.fail_reason, result.output, log=False)
3385 """Demotes the current node from master candidate role.
3388 # try to ensure we're not the master by mistake
3389 master, myself = ssconf.GetMasterAndMyself()
3390 if master == myself:
3391 _Fail("ssconf status shows I'm the master node, will not demote")
3393 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3394 if not result.failed:
3395 _Fail("The master daemon is running, will not demote")
3398 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3399 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3400 except EnvironmentError, err:
3401 if err.errno != errno.ENOENT:
3402 _Fail("Error while backing up cluster file: %s", err, exc=True)
3404 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3407 def _GetX509Filenames(cryptodir, name):
3408 """Returns the full paths for the private key and certificate.
3411 return (utils.PathJoin(cryptodir, name),
3412 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3413 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3416 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3417 """Creates a new X509 certificate for SSL/TLS.
3420 @param validity: Validity in seconds
3421 @rtype: tuple; (string, string)
3422 @return: Certificate name and public part
3425 (key_pem, cert_pem) = \
3426 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3427 min(validity, _MAX_SSL_CERT_VALIDITY))
3429 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3430 prefix="x509-%s-" % utils.TimestampForFilename())
3432 name = os.path.basename(cert_dir)
3433 assert len(name) > 5
3435 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3437 utils.WriteFile(key_file, mode=0400, data=key_pem)
3438 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3440 # Never return private key as it shouldn't leave the node
3441 return (name, cert_pem)
3443 shutil.rmtree(cert_dir, ignore_errors=True)
3447 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3448 """Removes a X509 certificate.
3451 @param name: Certificate name
3454 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3456 utils.RemoveFile(key_file)
3457 utils.RemoveFile(cert_file)
3461 except EnvironmentError, err:
3462 _Fail("Cannot remove certificate directory '%s': %s",
3466 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3467 """Returns the command for the requested input/output.
3469 @type instance: L{objects.Instance}
3470 @param instance: The instance object
3471 @param mode: Import/export mode
3472 @param ieio: Input/output type
3473 @param ieargs: Input/output arguments
3476 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3483 if ieio == constants.IEIO_FILE:
3484 (filename, ) = ieargs
3486 if not utils.IsNormAbsPath(filename):
3487 _Fail("Path '%s' is not normalized or absolute", filename)
3489 real_filename = os.path.realpath(filename)
3490 directory = os.path.dirname(real_filename)
3492 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3493 _Fail("File '%s' is not under exports directory '%s': %s",
3494 filename, pathutils.EXPORT_DIR, real_filename)
3497 utils.Makedirs(directory, mode=0750)
3499 quoted_filename = utils.ShellQuote(filename)
3501 if mode == constants.IEM_IMPORT:
3502 suffix = "> %s" % quoted_filename
3503 elif mode == constants.IEM_EXPORT:
3504 suffix = "< %s" % quoted_filename
3506 # Retrieve file size
3508 st = os.stat(filename)
3509 except EnvironmentError, err:
3510 logging.error("Can't stat(2) %s: %s", filename, err)
3512 exp_size = utils.BytesToMebibyte(st.st_size)
3514 elif ieio == constants.IEIO_RAW_DISK:
3517 real_disk = _OpenRealBD(disk)
3519 if mode == constants.IEM_IMPORT:
3520 # we set here a smaller block size as, due to transport buffering, more
3521 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3522 # is not already there or we pass a wrong path; we use notrunc to no
3523 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3524 # much memory; this means that at best, we flush every 64k, which will
3526 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3527 " bs=%s oflag=dsync"),
3531 elif mode == constants.IEM_EXPORT:
3532 # the block size on the read dd is 1MiB to match our units
3533 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3535 str(1024 * 1024), # 1 MB
3537 exp_size = disk.size
3539 elif ieio == constants.IEIO_SCRIPT:
3540 (disk, disk_index, ) = ieargs
3542 assert isinstance(disk_index, (int, long))
3544 real_disk = _OpenRealBD(disk)
3546 inst_os = OSFromDisk(instance.os)
3547 env = OSEnvironment(instance, inst_os)
3549 if mode == constants.IEM_IMPORT:
3550 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3551 env["IMPORT_INDEX"] = str(disk_index)
3552 script = inst_os.import_script
3554 elif mode == constants.IEM_EXPORT:
3555 env["EXPORT_DEVICE"] = real_disk.dev_path
3556 env["EXPORT_INDEX"] = str(disk_index)
3557 script = inst_os.export_script
3559 # TODO: Pass special environment only to script
3560 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3562 if mode == constants.IEM_IMPORT:
3563 suffix = "| %s" % script_cmd
3565 elif mode == constants.IEM_EXPORT:
3566 prefix = "%s |" % script_cmd
3568 # Let script predict size
3569 exp_size = constants.IE_CUSTOM_SIZE
3572 _Fail("Invalid %s I/O mode %r", mode, ieio)
3574 return (env, prefix, suffix, exp_size)
3577 def _CreateImportExportStatusDir(prefix):
3578 """Creates status directory for import/export.
3581 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3583 (prefix, utils.TimestampForFilename())))
3586 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3588 """Starts an import or export daemon.
3590 @param mode: Import/output mode
3591 @type opts: L{objects.ImportExportOptions}
3592 @param opts: Daemon options
3594 @param host: Remote host for export (None for import)
3596 @param port: Remote port for export (None for import)
3597 @type instance: L{objects.Instance}
3598 @param instance: Instance object
3599 @type component: string
3600 @param component: which part of the instance is transferred now,
3602 @param ieio: Input/output type
3603 @param ieioargs: Input/output arguments
3606 if mode == constants.IEM_IMPORT:
3609 if not (host is None and port is None):
3610 _Fail("Can not specify host or port on import")
3612 elif mode == constants.IEM_EXPORT:
3615 if host is None or port is None:
3616 _Fail("Host and port must be specified for an export")
3619 _Fail("Invalid mode %r", mode)
3621 if (opts.key_name is None) ^ (opts.ca_pem is None):
3622 _Fail("Cluster certificate can only be used for both key and CA")
3624 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3625 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3627 if opts.key_name is None:
3629 key_path = pathutils.NODED_CERT_FILE
3630 cert_path = pathutils.NODED_CERT_FILE
3631 assert opts.ca_pem is None
3633 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3635 assert opts.ca_pem is not None
3637 for i in [key_path, cert_path]:
3638 if not os.path.exists(i):
3639 _Fail("File '%s' does not exist" % i)
3641 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3643 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3644 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3645 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3647 if opts.ca_pem is None:
3649 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3654 utils.WriteFile(ca_file, data=ca, mode=0400)
3657 pathutils.IMPORT_EXPORT_DAEMON,
3659 "--key=%s" % key_path,
3660 "--cert=%s" % cert_path,
3661 "--ca=%s" % ca_file,
3665 cmd.append("--host=%s" % host)
3668 cmd.append("--port=%s" % port)
3671 cmd.append("--ipv6")
3673 cmd.append("--ipv4")
3676 cmd.append("--compress=%s" % opts.compress)
3679 cmd.append("--magic=%s" % opts.magic)
3681 if exp_size is not None:
3682 cmd.append("--expected-size=%s" % exp_size)
3685 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3688 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3690 if mode == constants.IEM_EXPORT:
3691 # Retry connection a few times when connecting to remote peer
3692 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3693 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3694 elif opts.connect_timeout is not None:
3695 assert mode == constants.IEM_IMPORT
3696 # Overall timeout for establishing connection while listening
3697 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3699 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3701 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3702 # support for receiving a file descriptor for output
3703 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3706 # The import/export name is simply the status directory name
3707 return os.path.basename(status_dir)
3710 shutil.rmtree(status_dir, ignore_errors=True)
3714 def GetImportExportStatus(names):
3715 """Returns import/export daemon status.
3717 @type names: sequence
3718 @param names: List of names
3719 @rtype: List of dicts
3720 @return: Returns a list of the state of each named import/export or None if a
3721 status couldn't be read
3727 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3731 data = utils.ReadFile(status_file)
3732 except EnvironmentError, err:
3733 if err.errno != errno.ENOENT:
3741 result.append(serializer.LoadJson(data))
3746 def AbortImportExport(name):
3747 """Sends SIGTERM to a running import/export daemon.
3750 logging.info("Abort import/export %s", name)
3752 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3753 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3756 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3758 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3761 def CleanupImportExport(name):
3762 """Cleanup after an import or export.
3764 If the import/export daemon is still running it's killed. Afterwards the
3765 whole status directory is removed.
3768 logging.info("Finalizing import/export %s", name)
3770 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3772 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3775 logging.info("Import/export %s is still running with PID %s",
3777 utils.KillProcess(pid, waitpid=False)
3779 shutil.rmtree(status_dir, ignore_errors=True)
3782 def _FindDisks(target_node_uuid, nodes_ip, disks):
3783 """Sets the physical ID on disks and returns the block devices.
3786 # set the correct physical ID
3788 cf.SetPhysicalID(target_node_uuid, nodes_ip)
3793 rd = _RecursiveFindBD(cf)
3795 _Fail("Can't find device %s", cf)
3800 def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
3801 """Disconnects the network on a list of drbd devices.
3804 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3810 except errors.BlockDeviceError, err:
3811 _Fail("Can't change network configuration to standalone mode: %s",
3815 def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
3817 """Attaches the network on a list of drbd devices.
3820 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3823 for idx, rd in enumerate(bdevs):
3825 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3826 except EnvironmentError, err:
3827 _Fail("Can't create symlink: %s", err)
3828 # reconnect disks, switch to new master configuration and if
3829 # needed primary mode
3832 rd.AttachNet(multimaster)
3833 except errors.BlockDeviceError, err:
3834 _Fail("Can't change network configuration: %s", err)
3836 # wait until the disks are connected; we need to retry the re-attach
3837 # if the device becomes standalone, as this might happen if the one
3838 # node disconnects and reconnects in a different mode before the
3839 # other node reconnects; in this case, one or both of the nodes will
3840 # decide it has wrong configuration and switch to standalone
3843 all_connected = True
3846 stats = rd.GetProcStatus()
3848 all_connected = (all_connected and
3849 (stats.is_connected or stats.is_in_resync))
3851 if stats.is_standalone:
3852 # peer had different config info and this node became
3853 # standalone, even though this should not happen with the
3854 # new staged way of changing disk configs
3856 rd.AttachNet(multimaster)
3857 except errors.BlockDeviceError, err:
3858 _Fail("Can't change network configuration: %s", err)
3860 if not all_connected:
3861 raise utils.RetryAgain()
3864 # Start with a delay of 100 miliseconds and go up to 5 seconds
3865 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3866 except utils.RetryTimeout:
3867 _Fail("Timeout in disk reconnecting")
3870 # change to primary mode
3874 except errors.BlockDeviceError, err:
3875 _Fail("Can't change to primary mode: %s", err)
3878 def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
3879 """Wait until DRBDs have synchronized.
3883 stats = rd.GetProcStatus()
3884 if not (stats.is_connected or stats.is_in_resync):
3885 raise utils.RetryAgain()
3888 bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3894 # poll each second for 15 seconds
3895 stats = utils.Retry(_helper, 1, 15, args=[rd])
3896 except utils.RetryTimeout:
3897 stats = rd.GetProcStatus()
3899 if not (stats.is_connected or stats.is_in_resync):
3900 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3901 alldone = alldone and (not stats.is_in_resync)
3902 if stats.sync_percent is not None:
3903 min_resync = min(min_resync, stats.sync_percent)
3905 return (alldone, min_resync)
3908 def GetDrbdUsermodeHelper():
3909 """Returns DRBD usermode helper currently configured.
3913 return drbd.DRBD8.GetUsermodeHelper()
3914 except errors.BlockDeviceError, err:
3918 def PowercycleNode(hypervisor_type, hvparams=None):
3919 """Hard-powercycle the node.
3921 Because we need to return first, and schedule the powercycle in the
3922 background, we won't be able to report failures nicely.
3925 hyper = hypervisor.GetHypervisor(hypervisor_type)
3929 # if we can't fork, we'll pretend that we're in the child process
3932 return "Reboot scheduled in 5 seconds"
3933 # ensure the child is running on ram
3936 except Exception: # pylint: disable=W0703
3939 hyper.PowercycleNode(hvparams=hvparams)
3942 def _VerifyRestrictedCmdName(cmd):
3943 """Verifies a restricted command name.
3946 @param cmd: Command name
3947 @rtype: tuple; (boolean, string or None)
3948 @return: The tuple's first element is the status; if C{False}, the second
3949 element is an error message string, otherwise it's C{None}
3953 return (False, "Missing command name")
3955 if os.path.basename(cmd) != cmd:
3956 return (False, "Invalid command name")
3958 if not constants.EXT_PLUGIN_MASK.match(cmd):
3959 return (False, "Command name contains forbidden characters")
3964 def _CommonRestrictedCmdCheck(path, owner):
3965 """Common checks for restricted command file system directories and files.
3968 @param path: Path to check
3969 @param owner: C{None} or tuple containing UID and GID
3970 @rtype: tuple; (boolean, string or C{os.stat} result)
3971 @return: The tuple's first element is the status; if C{False}, the second
3972 element is an error message string, otherwise it's the result of C{os.stat}
3976 # Default to root as owner
3981 except EnvironmentError, err:
3982 return (False, "Can't stat(2) '%s': %s" % (path, err))
3984 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
3985 return (False, "Permissions on '%s' are too permissive" % path)
3987 if (st.st_uid, st.st_gid) != owner:
3988 (owner_uid, owner_gid) = owner
3989 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
3994 def _VerifyRestrictedCmdDirectory(path, _owner=None):
3995 """Verifies restricted command directory.
3998 @param path: Path to check
3999 @rtype: tuple; (boolean, string or None)
4000 @return: The tuple's first element is the status; if C{False}, the second
4001 element is an error message string, otherwise it's C{None}
4004 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4007 return (False, value)
4009 if not stat.S_ISDIR(value.st_mode):
4010 return (False, "Path '%s' is not a directory" % path)
4015 def _VerifyRestrictedCmd(path, cmd, _owner=None):
4016 """Verifies a whole restricted command and returns its executable filename.
4019 @param path: Directory containing restricted commands
4021 @param cmd: Command name
4022 @rtype: tuple; (boolean, string)
4023 @return: The tuple's first element is the status; if C{False}, the second
4024 element is an error message string, otherwise the second element is the
4025 absolute path to the executable
4028 executable = utils.PathJoin(path, cmd)
4030 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4035 if not utils.IsExecutable(executable):
4036 return (False, "access(2) thinks '%s' can't be executed" % executable)
4038 return (True, executable)
4041 def _PrepareRestrictedCmd(path, cmd,
4042 _verify_dir=_VerifyRestrictedCmdDirectory,
4043 _verify_name=_VerifyRestrictedCmdName,
4044 _verify_cmd=_VerifyRestrictedCmd):
4045 """Performs a number of tests on a restricted command.
4048 @param path: Directory containing restricted commands
4050 @param cmd: Command name
4051 @return: Same as L{_VerifyRestrictedCmd}
4054 # Verify the directory first
4055 (status, msg) = _verify_dir(path)
4057 # Check command if everything was alright
4058 (status, msg) = _verify_name(cmd)
4063 # Check actual executable
4064 return _verify_cmd(path, cmd)
4067 def RunRestrictedCmd(cmd,
4068 _lock_timeout=_RCMD_LOCK_TIMEOUT,
4069 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4070 _path=pathutils.RESTRICTED_COMMANDS_DIR,
4071 _sleep_fn=time.sleep,
4072 _prepare_fn=_PrepareRestrictedCmd,
4073 _runcmd_fn=utils.RunCmd,
4074 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4075 """Executes a restricted command after performing strict tests.
4078 @param cmd: Command name
4080 @return: Command output
4081 @raise RPCFail: In case of an error
4084 logging.info("Preparing to run restricted command '%s'", cmd)
4087 _Fail("Restricted commands disabled at configure time")
4093 lock = utils.FileLock.Open(_lock_file)
4094 lock.Exclusive(blocking=True, timeout=_lock_timeout)
4096 (status, value) = _prepare_fn(_path, cmd)
4099 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4100 postfork_fn=lambda _: lock.Unlock())
4102 logging.error(value)
4103 except Exception: # pylint: disable=W0703
4104 # Keep original error in log
4105 logging.exception("Caught exception")
4107 if cmdresult is None:
4108 logging.info("Sleeping for %0.1f seconds before returning",
4109 _RCMD_INVALID_DELAY)
4110 _sleep_fn(_RCMD_INVALID_DELAY)
4112 # Do not include original error message in returned error
4113 _Fail("Executing command '%s' failed" % cmd)
4114 elif cmdresult.failed or cmdresult.fail_reason:
4115 _Fail("Restricted command '%s' failed: %s; output: %s",
4116 cmd, cmdresult.fail_reason, cmdresult.output)
4118 return cmdresult.output
4120 if lock is not None:
4121 # Release lock at last
4126 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4127 """Creates or removes the watcher pause file.
4129 @type until: None or number
4130 @param until: Unix timestamp saying until when the watcher shouldn't run
4134 logging.info("Received request to no longer pause watcher")
4135 utils.RemoveFile(_filename)
4137 logging.info("Received request to pause watcher until %s", until)
4139 if not ht.TNumber(until):
4140 _Fail("Duration must be numeric")
4142 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4145 class HooksRunner(object):
4148 This class is instantiated on the node side (ganeti-noded) and not
4152 def __init__(self, hooks_base_dir=None):
4153 """Constructor for hooks runner.
4155 @type hooks_base_dir: str or None
4156 @param hooks_base_dir: if not None, this overrides the
4157 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4160 if hooks_base_dir is None:
4161 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4162 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4164 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4166 def RunLocalHooks(self, node_list, hpath, phase, env):
4167 """Check that the hooks will be run only locally and then run them.
4170 assert len(node_list) == 1
4172 _, myself = ssconf.GetMasterAndMyself()
4173 assert node == myself
4175 results = self.RunHooks(hpath, phase, env)
4177 # Return values in the form expected by HooksMaster
4178 return {node: (None, False, results)}
4180 def RunHooks(self, hpath, phase, env):
4181 """Run the scripts in the hooks directory.
4184 @param hpath: the path to the hooks directory which
4187 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4188 L{constants.HOOKS_PHASE_POST}
4190 @param env: dictionary with the environment for the hook
4192 @return: list of 3-element tuples:
4194 - script result, either L{constants.HKR_SUCCESS} or
4195 L{constants.HKR_FAIL}
4196 - output of the script
4198 @raise errors.ProgrammerError: for invalid input
4202 if phase == constants.HOOKS_PHASE_PRE:
4204 elif phase == constants.HOOKS_PHASE_POST:
4207 _Fail("Unknown hooks phase '%s'", phase)
4209 subdir = "%s-%s.d" % (hpath, suffix)
4210 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4214 if not os.path.isdir(dir_name):
4215 # for non-existing/non-dirs, we simply exit instead of logging a
4216 # warning at every operation
4219 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4221 for (relname, relstatus, runresult) in runparts_results:
4222 if relstatus == constants.RUNPARTS_SKIP:
4223 rrval = constants.HKR_SKIP
4225 elif relstatus == constants.RUNPARTS_ERR:
4226 rrval = constants.HKR_FAIL
4227 output = "Hook script execution error: %s" % runresult
4228 elif relstatus == constants.RUNPARTS_RUN:
4229 if runresult.failed:
4230 rrval = constants.HKR_FAIL
4232 rrval = constants.HKR_SUCCESS
4233 output = utils.SafeEncode(runresult.output.strip())
4234 results.append(("%s/%s" % (subdir, relname), rrval, output))
4239 class IAllocatorRunner(object):
4240 """IAllocator runner.
4242 This class is instantiated on the node side (ganeti-noded) and not on
4247 def Run(name, idata):
4248 """Run an iallocator script.
4251 @param name: the iallocator script name
4253 @param idata: the allocator input data
4256 @return: two element tuple of:
4258 - either error message or stdout of allocator (for success)
4261 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4263 if alloc_script is None:
4264 _Fail("iallocator module '%s' not found in the search path", name)
4266 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4270 result = utils.RunCmd([alloc_script, fin_name])
4272 _Fail("iallocator module '%s' failed: %s, output '%s'",
4273 name, result.fail_reason, result.output)
4277 return result.stdout
4280 class DevCacheManager(object):
4281 """Simple class for managing a cache of block device information.
4284 _DEV_PREFIX = "/dev/"
4285 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4288 def _ConvertPath(cls, dev_path):
4289 """Converts a /dev/name path to the cache file name.
4291 This replaces slashes with underscores and strips the /dev
4292 prefix. It then returns the full path to the cache file.
4295 @param dev_path: the C{/dev/} path name
4297 @return: the converted path name
4300 if dev_path.startswith(cls._DEV_PREFIX):
4301 dev_path = dev_path[len(cls._DEV_PREFIX):]
4302 dev_path = dev_path.replace("/", "_")
4303 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4307 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4308 """Updates the cache information for a given device.
4311 @param dev_path: the pathname of the device
4313 @param owner: the owner (instance name) of the device
4314 @type on_primary: bool
4315 @param on_primary: whether this is the primary
4318 @param iv_name: the instance-visible name of the
4319 device, as in objects.Disk.iv_name
4324 if dev_path is None:
4325 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4327 fpath = cls._ConvertPath(dev_path)
4333 iv_name = "not_visible"
4334 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4336 utils.WriteFile(fpath, data=fdata)
4337 except EnvironmentError, err:
4338 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4341 def RemoveCache(cls, dev_path):
4342 """Remove data for a dev_path.
4344 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4345 path name and logging.
4348 @param dev_path: the pathname of the device
4353 if dev_path is None:
4354 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4356 fpath = cls._ConvertPath(dev_path)
4358 utils.RemoveFile(fpath)
4359 except EnvironmentError, err:
4360 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)