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))
595 def _GetVgSpindlesInfo(name, excl_stor):
596 """Retrieves information about spindles in an LVM volume group.
600 @type excl_stor: bool
601 @param excl_stor: exclusive storage
603 @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
604 free spindles, total spindles respectively
608 (vg_free, vg_size) = bdev.LogicalVolume.GetVgSpindlesInfo(name)
619 def _GetHvInfo(name):
620 """Retrieves node information from a hypervisor.
622 The information returned depends on the hypervisor. Common items:
624 - vg_size is the size of the configured volume group in MiB
625 - vg_free is the free size of the volume group in MiB
626 - memory_dom0 is the memory allocated for domain0 in MiB
627 - memory_free is the currently available (free) ram in MiB
628 - memory_total is the total number of ram in MiB
629 - hv_version: the hypervisor version, if available
632 return hypervisor.GetHypervisor(name).GetNodeInfo()
635 def _GetNamedNodeInfo(names, fn):
636 """Calls C{fn} for all names in C{names} and returns a dictionary.
644 return map(fn, names)
647 def GetNodeInfo(storage_units, hv_names, excl_stor):
648 """Gives back a hash with different information about the node.
650 @type storage_units: list of pairs (string, string)
651 @param storage_units: List of pairs (storage unit, identifier) to ask for disk
652 space information. In case of lvm-vg, the identifier is
654 @type hv_names: list of string
655 @param hv_names: Names of the hypervisors to ask for node information
656 @type excl_stor: boolean
657 @param excl_stor: Whether exclusive_storage is active
658 @rtype: tuple; (string, None/dict, None/dict)
659 @return: Tuple containing boot ID, volume group information and hypervisor
663 bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
664 storage_info = _GetNamedNodeInfo(
666 (lambda storage_unit: _ApplyStorageInfoFunction(storage_unit[0],
669 hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo)
671 return (bootid, storage_info, hv_info)
674 # FIXME: implement storage reporting for all missing storage types.
675 _STORAGE_TYPE_INFO_FN = {
676 constants.ST_BLOCK: None,
677 constants.ST_DISKLESS: None,
678 constants.ST_EXT: None,
679 constants.ST_FILE: None,
680 constants.ST_LVM_PV: _GetVgSpindlesInfo,
681 constants.ST_LVM_VG: _GetVgInfo,
682 constants.ST_RADOS: None,
686 def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
687 """Looks up and applies the correct function to calculate free and total
688 storage for the given storage type.
690 @type storage_type: string
691 @param storage_type: the storage type for which the storage shall be reported.
692 @type storage_key: string
693 @param storage_key: identifier of a storage unit, e.g. the volume group name
694 of an LVM storage unit
696 @param args: various parameters that can be used for storage reporting. These
697 parameters and their semantics vary from storage type to storage type and
698 are just propagated in this function.
699 @return: the results of the application of the storage space function (see
700 _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
702 @raises NotImplementedError: for storage types who don't support space
705 fn = _STORAGE_TYPE_INFO_FN[storage_type]
707 return fn(storage_key, *args)
709 raise NotImplementedError
712 def _CheckExclusivePvs(pvi_list):
713 """Check that PVs are not shared among LVs
715 @type pvi_list: list of L{objects.LvmPvInfo} objects
716 @param pvi_list: information about the PVs
718 @rtype: list of tuples (string, list of strings)
719 @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
724 if len(pvi.lv_list) > 1:
725 res.append((pvi.name, pvi.lv_list))
729 def VerifyNode(what, cluster_name):
730 """Verify the status of the local node.
732 Based on the input L{what} parameter, various checks are done on the
735 If the I{filelist} key is present, this list of
736 files is checksummed and the file/checksum pairs are returned.
738 If the I{nodelist} key is present, we check that we have
739 connectivity via ssh with the target nodes (and check the hostname
742 If the I{node-net-test} key is present, we check that we have
743 connectivity to the given nodes via both primary IP and, if
744 applicable, secondary IPs.
747 @param what: a dictionary of things to check:
748 - filelist: list of files for which to compute checksums
749 - nodelist: list of nodes we should check ssh communication with
750 - node-net-test: list of nodes we should check node daemon port
752 - hypervisor: list with hypervisors to run the verify for
754 @return: a dictionary with the same keys as the input dict, and
755 values representing the result of the checks
759 my_name = netutils.Hostname.GetSysName()
760 port = netutils.GetDaemonPort(constants.NODED)
761 vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
763 if constants.NV_HYPERVISOR in what and vm_capable:
764 result[constants.NV_HYPERVISOR] = tmp = {}
765 for hv_name in what[constants.NV_HYPERVISOR]:
767 val = hypervisor.GetHypervisor(hv_name).Verify()
768 except errors.HypervisorError, err:
769 val = "Error while checking hypervisor: %s" % str(err)
772 if constants.NV_HVPARAMS in what and vm_capable:
773 result[constants.NV_HVPARAMS] = tmp = []
774 for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
776 logging.info("Validating hv %s, %s", hv_name, hvparms)
777 hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
778 except errors.HypervisorError, err:
779 tmp.append((source, hv_name, str(err)))
781 if constants.NV_FILELIST in what:
782 fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
783 what[constants.NV_FILELIST]))
784 result[constants.NV_FILELIST] = \
785 dict((vcluster.MakeVirtualPath(key), value)
786 for (key, value) in fingerprints.items())
788 if constants.NV_NODELIST in what:
789 (nodes, bynode) = what[constants.NV_NODELIST]
791 # Add nodes from other groups (different for each node)
793 nodes.extend(bynode[my_name])
798 random.shuffle(nodes)
800 # Try to contact all nodes
803 success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
807 result[constants.NV_NODELIST] = val
809 if constants.NV_NODENETTEST in what:
810 result[constants.NV_NODENETTEST] = tmp = {}
811 my_pip = my_sip = None
812 for name, pip, sip in what[constants.NV_NODENETTEST]:
818 tmp[my_name] = ("Can't find my own primary/secondary IP"
821 for name, pip, sip in what[constants.NV_NODENETTEST]:
823 if not netutils.TcpPing(pip, port, source=my_pip):
824 fail.append("primary")
826 if not netutils.TcpPing(sip, port, source=my_sip):
827 fail.append("secondary")
829 tmp[name] = ("failure using the %s interface(s)" %
832 if constants.NV_MASTERIP in what:
833 # FIXME: add checks on incoming data structures (here and in the
834 # rest of the function)
835 master_name, master_ip = what[constants.NV_MASTERIP]
836 if master_name == my_name:
837 source = constants.IP4_ADDRESS_LOCALHOST
840 result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
843 if constants.NV_USERSCRIPTS in what:
844 result[constants.NV_USERSCRIPTS] = \
845 [script for script in what[constants.NV_USERSCRIPTS]
846 if not utils.IsExecutable(script)]
848 if constants.NV_OOB_PATHS in what:
849 result[constants.NV_OOB_PATHS] = tmp = []
850 for path in what[constants.NV_OOB_PATHS]:
854 tmp.append("error stating out of band helper: %s" % err)
856 if stat.S_ISREG(st.st_mode):
857 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
860 tmp.append("out of band helper %s is not executable" % path)
862 tmp.append("out of band helper %s is not a file" % path)
864 if constants.NV_LVLIST in what and vm_capable:
866 val = GetVolumeList(utils.ListVolumeGroups().keys())
869 result[constants.NV_LVLIST] = val
871 if constants.NV_INSTANCELIST in what and vm_capable:
872 # GetInstanceList can fail
874 val = GetInstanceList(what[constants.NV_INSTANCELIST])
877 result[constants.NV_INSTANCELIST] = val
879 if constants.NV_VGLIST in what and vm_capable:
880 result[constants.NV_VGLIST] = utils.ListVolumeGroups()
882 if constants.NV_PVLIST in what and vm_capable:
883 check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
884 val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
885 filter_allocatable=False,
886 include_lvs=check_exclusive_pvs)
887 if check_exclusive_pvs:
888 result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
890 # Avoid sending useless data on the wire
892 result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
894 if constants.NV_VERSION in what:
895 result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
896 constants.RELEASE_VERSION)
898 if constants.NV_HVINFO in what and vm_capable:
899 hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
900 result[constants.NV_HVINFO] = hyper.GetNodeInfo()
902 if constants.NV_DRBDVERSION in what and vm_capable:
904 drbd_version = DRBD8.GetProcInfo().GetVersionString()
905 except errors.BlockDeviceError, err:
906 logging.warning("Can't get DRBD version", exc_info=True)
907 drbd_version = str(err)
908 result[constants.NV_DRBDVERSION] = drbd_version
910 if constants.NV_DRBDLIST in what and vm_capable:
912 used_minors = drbd.DRBD8.GetUsedDevs()
913 except errors.BlockDeviceError, err:
914 logging.warning("Can't get used minors list", exc_info=True)
915 used_minors = str(err)
916 result[constants.NV_DRBDLIST] = used_minors
918 if constants.NV_DRBDHELPER in what and vm_capable:
921 payload = drbd.DRBD8.GetUsermodeHelper()
922 except errors.BlockDeviceError, err:
923 logging.error("Can't get DRBD usermode helper: %s", str(err))
926 result[constants.NV_DRBDHELPER] = (status, payload)
928 if constants.NV_NODESETUP in what:
929 result[constants.NV_NODESETUP] = tmpr = []
930 if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
931 tmpr.append("The sysfs filesytem doesn't seem to be mounted"
932 " under /sys, missing required directories /sys/block"
933 " and /sys/class/net")
934 if (not os.path.isdir("/proc/sys") or
935 not os.path.isfile("/proc/sysrq-trigger")):
936 tmpr.append("The procfs filesystem doesn't seem to be mounted"
937 " under /proc, missing required directory /proc/sys and"
938 " the file /proc/sysrq-trigger")
940 if constants.NV_TIME in what:
941 result[constants.NV_TIME] = utils.SplitTime(time.time())
943 if constants.NV_OSLIST in what and vm_capable:
944 result[constants.NV_OSLIST] = DiagnoseOS()
946 if constants.NV_BRIDGES in what and vm_capable:
947 result[constants.NV_BRIDGES] = [bridge
948 for bridge in what[constants.NV_BRIDGES]
949 if not utils.BridgeExists(bridge)]
951 if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
952 result[constants.NV_FILE_STORAGE_PATHS] = \
953 bdev.ComputeWrongFileStoragePaths()
958 def GetBlockDevSizes(devices):
959 """Return the size of the given block devices
962 @param devices: list of block device nodes to query
965 dictionary of all block devices under /dev (key). The value is their
968 {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
974 for devpath in devices:
975 if not utils.IsBelowDir(DEV_PREFIX, devpath):
979 st = os.stat(devpath)
980 except EnvironmentError, err:
981 logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
984 if stat.S_ISBLK(st.st_mode):
985 result = utils.RunCmd(["blockdev", "--getsize64", devpath])
987 # We don't want to fail, just do not list this device as available
988 logging.warning("Cannot get size for block device %s", devpath)
991 size = int(result.stdout) / (1024 * 1024)
992 blockdevs[devpath] = size
996 def GetVolumeList(vg_names):
997 """Compute list of logical volumes and their size.
1000 @param vg_names: the volume groups whose LVs we should list, or
1001 empty for all volume groups
1004 dictionary of all partions (key) with value being a tuple of
1005 their size (in MiB), inactive and online status::
1007 {'xenvg/test1': ('20.06', True, True)}
1009 in case of errors, a string is returned with the error
1017 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1018 "--separator=%s" % sep,
1019 "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1021 _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1023 for line in result.stdout.splitlines():
1025 match = _LVSLINE_REGEX.match(line)
1027 logging.error("Invalid line returned from lvs output: '%s'", line)
1029 vg_name, name, size, attr = match.groups()
1030 inactive = attr[4] == "-"
1031 online = attr[5] == "o"
1032 virtual = attr[0] == "v"
1034 # we don't want to report such volumes as existing, since they
1035 # don't really hold data
1037 lvs[vg_name + "/" + name] = (size, inactive, online)
1042 def ListVolumeGroups():
1043 """List the volume groups and their size.
1046 @return: dictionary with keys volume name and values the
1050 return utils.ListVolumeGroups()
1054 """List all volumes on this node.
1058 A list of dictionaries, each having four keys:
1059 - name: the logical volume name,
1060 - size: the size of the logical volume
1061 - dev: the physical device on which the LV lives
1062 - vg: the volume group to which it belongs
1064 In case of errors, we return an empty list and log the
1067 Note that since a logical volume can live on multiple physical
1068 volumes, the resulting list might include a logical volume
1072 result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1074 "--options=lv_name,lv_size,devices,vg_name"])
1076 _Fail("Failed to list logical volumes, lvs output: %s",
1080 return dev.split("(")[0]
1082 def handle_dev(dev):
1083 return [parse_dev(x) for x in dev.split(",")]
1086 line = [v.strip() for v in line]
1087 return [{"name": line[0], "size": line[1],
1088 "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1091 for line in result.stdout.splitlines():
1092 if line.count("|") >= 3:
1093 all_devs.extend(map_line(line.split("|")))
1095 logging.warning("Strange line in the output from lvs: '%s'", line)
1099 def BridgesExist(bridges_list):
1100 """Check if a list of bridges exist on the current node.
1103 @return: C{True} if all of them exist, C{False} otherwise
1107 for bridge in bridges_list:
1108 if not utils.BridgeExists(bridge):
1109 missing.append(bridge)
1112 _Fail("Missing bridges %s", utils.CommaJoin(missing))
1115 def GetInstanceListForHypervisor(hname, hvparams=None,
1116 get_hv_fn=hypervisor.GetHypervisor):
1117 """Provides a list of instances of the given hypervisor.
1120 @param hname: name of the hypervisor
1121 @type hvparams: dict of strings
1122 @param hvparams: hypervisor parameters for the given hypervisor
1123 @type get_hv_fn: function
1124 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1125 name; optional parameter to increase testability
1128 @return: a list of all running instances on the current node
1129 - instance1.example.com
1130 - instance2.example.com
1135 hv = get_hv_fn(hname)
1136 names = hv.ListInstances(hvparams)
1137 results.extend(names)
1138 except errors.HypervisorError, err:
1139 _Fail("Error enumerating instances (hypervisor %s): %s",
1140 hname, err, exc=True)
1144 def GetInstanceList(hypervisor_list, all_hvparams=None,
1145 get_hv_fn=hypervisor.GetHypervisor):
1146 """Provides a list of instances.
1148 @type hypervisor_list: list
1149 @param hypervisor_list: the list of hypervisors to query information
1150 @type all_hvparams: dict of dict of strings
1151 @param all_hvparams: a dictionary mapping hypervisor types to respective
1152 cluster-wide hypervisor parameters
1153 @type get_hv_fn: function
1154 @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1155 name; optional parameter to increase testability
1158 @return: a list of all running instances on the current node
1159 - instance1.example.com
1160 - instance2.example.com
1164 for hname in hypervisor_list:
1166 if all_hvparams is not None:
1167 hvparams = all_hvparams[hname]
1168 results.extend(GetInstanceListForHypervisor(hname, hvparams,
1169 get_hv_fn=get_hv_fn))
1173 def GetInstanceInfo(instance, hname):
1174 """Gives back the information about an instance as a dictionary.
1176 @type instance: string
1177 @param instance: the instance name
1179 @param hname: the hypervisor type of the instance
1182 @return: dictionary with the following keys:
1183 - memory: memory size of instance (int)
1184 - state: xen state of instance (string)
1185 - time: cpu time of instance (float)
1186 - vcpus: the number of vcpus (int)
1191 iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
1192 if iinfo is not None:
1193 output["memory"] = iinfo[2]
1194 output["vcpus"] = iinfo[3]
1195 output["state"] = iinfo[4]
1196 output["time"] = iinfo[5]
1201 def GetInstanceMigratable(instance):
1202 """Computes whether an instance can be migrated.
1204 @type instance: L{objects.Instance}
1205 @param instance: object representing the instance to be checked.
1208 @return: tuple of (result, description) where:
1209 - result: whether the instance can be migrated or not
1210 - description: a description of the issue, if relevant
1213 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1214 iname = instance.name
1215 if iname not in hyper.ListInstances(instance.hvparams):
1216 _Fail("Instance %s is not running", iname)
1218 for idx in range(len(instance.disks)):
1219 link_name = _GetBlockDevSymlinkPath(iname, idx)
1220 if not os.path.islink(link_name):
1221 logging.warning("Instance %s is missing symlink %s for disk %d",
1222 iname, link_name, idx)
1225 def GetAllInstancesInfo(hypervisor_list):
1226 """Gather data about all instances.
1228 This is the equivalent of L{GetInstanceInfo}, except that it
1229 computes data for all instances at once, thus being faster if one
1230 needs data about more than one instance.
1232 @type hypervisor_list: list
1233 @param hypervisor_list: list of hypervisors to query for instance data
1236 @return: dictionary of instance: data, with data having the following keys:
1237 - memory: memory size of instance (int)
1238 - state: xen state of instance (string)
1239 - time: cpu time of instance (float)
1240 - vcpus: the number of vcpus
1245 for hname in hypervisor_list:
1246 iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
1248 for name, _, memory, vcpus, state, times in iinfo:
1256 # we only check static parameters, like memory and vcpus,
1257 # and not state and time which can change between the
1258 # invocations of the different hypervisors
1259 for key in "memory", "vcpus":
1260 if value[key] != output[name][key]:
1261 _Fail("Instance %s is running twice"
1262 " with different parameters", name)
1263 output[name] = value
1268 def _InstanceLogName(kind, os_name, instance, component):
1269 """Compute the OS log filename for a given instance and operation.
1271 The instance name and os name are passed in as strings since not all
1272 operations have these as part of an instance object.
1275 @param kind: the operation type (e.g. add, import, etc.)
1276 @type os_name: string
1277 @param os_name: the os name
1278 @type instance: string
1279 @param instance: the name of the instance being imported/added/etc.
1280 @type component: string or None
1281 @param component: the name of the component of the instance being
1285 # TODO: Use tempfile.mkstemp to create unique filename
1287 assert "/" not in component
1288 c_msg = "-%s" % component
1291 base = ("%s-%s-%s%s-%s.log" %
1292 (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1293 return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1296 def InstanceOsAdd(instance, reinstall, debug):
1297 """Add an OS to an instance.
1299 @type instance: L{objects.Instance}
1300 @param instance: Instance whose OS is to be installed
1301 @type reinstall: boolean
1302 @param reinstall: whether this is an instance reinstall
1303 @type debug: integer
1304 @param debug: debug level, passed to the OS scripts
1308 inst_os = OSFromDisk(instance.os)
1310 create_env = OSEnvironment(instance, inst_os, debug)
1312 create_env["INSTANCE_REINSTALL"] = "1"
1314 logfile = _InstanceLogName("add", instance.os, instance.name, None)
1316 result = utils.RunCmd([inst_os.create_script], env=create_env,
1317 cwd=inst_os.path, output=logfile, reset_env=True)
1319 logging.error("os create command '%s' returned error: %s, logfile: %s,"
1320 " output: %s", result.cmd, result.fail_reason, logfile,
1322 lines = [utils.SafeEncode(val)
1323 for val in utils.TailFile(logfile, lines=20)]
1324 _Fail("OS create script failed (%s), last lines in the"
1325 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1328 def RunRenameInstance(instance, old_name, debug):
1329 """Run the OS rename script for an instance.
1331 @type instance: L{objects.Instance}
1332 @param instance: Instance whose OS is to be installed
1333 @type old_name: string
1334 @param old_name: previous instance name
1335 @type debug: integer
1336 @param debug: debug level, passed to the OS scripts
1338 @return: the success of the operation
1341 inst_os = OSFromDisk(instance.os)
1343 rename_env = OSEnvironment(instance, inst_os, debug)
1344 rename_env["OLD_INSTANCE_NAME"] = old_name
1346 logfile = _InstanceLogName("rename", instance.os,
1347 "%s-%s" % (old_name, instance.name), None)
1349 result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1350 cwd=inst_os.path, output=logfile, reset_env=True)
1353 logging.error("os create command '%s' returned error: %s output: %s",
1354 result.cmd, result.fail_reason, result.output)
1355 lines = [utils.SafeEncode(val)
1356 for val in utils.TailFile(logfile, lines=20)]
1357 _Fail("OS rename script failed (%s), last lines in the"
1358 " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1361 def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1362 """Returns symlink path for block device.
1366 _dir = pathutils.DISK_LINKS_DIR
1368 return utils.PathJoin(_dir,
1370 (instance_name, constants.DISK_SEPARATOR, idx)))
1373 def _SymlinkBlockDev(instance_name, device_path, idx):
1374 """Set up symlinks to a instance's block device.
1376 This is an auxiliary function run when an instance is start (on the primary
1377 node) or when an instance is migrated (on the target node).
1380 @param instance_name: the name of the target instance
1381 @param device_path: path of the physical block device, on the node
1382 @param idx: the disk index
1383 @return: absolute path to the disk's symlink
1386 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1388 os.symlink(device_path, link_name)
1389 except OSError, err:
1390 if err.errno == errno.EEXIST:
1391 if (not os.path.islink(link_name) or
1392 os.readlink(link_name) != device_path):
1393 os.remove(link_name)
1394 os.symlink(device_path, link_name)
1401 def _RemoveBlockDevLinks(instance_name, disks):
1402 """Remove the block device symlinks belonging to the given instance.
1405 for idx, _ in enumerate(disks):
1406 link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1407 if os.path.islink(link_name):
1409 os.remove(link_name)
1411 logging.exception("Can't remove symlink '%s'", link_name)
1414 def _GatherAndLinkBlockDevs(instance):
1415 """Set up an instance's block device(s).
1417 This is run on the primary node at instance startup. The block
1418 devices must be already assembled.
1420 @type instance: L{objects.Instance}
1421 @param instance: the instance whose disks we shoul assemble
1423 @return: list of (disk_object, device_path)
1427 for idx, disk in enumerate(instance.disks):
1428 device = _RecursiveFindBD(disk)
1430 raise errors.BlockDeviceError("Block device '%s' is not set up." %
1434 link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1436 raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1439 block_devices.append((disk, link_name))
1441 return block_devices
1444 def StartInstance(instance, startup_paused, reason, store_reason=True):
1445 """Start an instance.
1447 @type instance: L{objects.Instance}
1448 @param instance: the instance object
1449 @type startup_paused: bool
1450 @param instance: pause instance at startup?
1451 @type reason: list of reasons
1452 @param reason: the reason trail for this startup
1453 @type store_reason: boolean
1454 @param store_reason: whether to store the shutdown reason trail on file
1458 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1461 if instance.name in running_instances:
1462 logging.info("Instance %s already running, not starting", instance.name)
1466 block_devices = _GatherAndLinkBlockDevs(instance)
1467 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1468 hyper.StartInstance(instance, block_devices, startup_paused)
1470 _StoreInstReasonTrail(instance.name, reason)
1471 except errors.BlockDeviceError, err:
1472 _Fail("Block device error: %s", err, exc=True)
1473 except errors.HypervisorError, err:
1474 _RemoveBlockDevLinks(instance.name, instance.disks)
1475 _Fail("Hypervisor error: %s", err, exc=True)
1478 def InstanceShutdown(instance, timeout, reason, store_reason=True):
1479 """Shut an instance down.
1481 @note: this functions uses polling with a hardcoded timeout.
1483 @type instance: L{objects.Instance}
1484 @param instance: the instance object
1485 @type timeout: integer
1486 @param timeout: maximum timeout for soft shutdown
1487 @type reason: list of reasons
1488 @param reason: the reason trail for this shutdown
1489 @type store_reason: boolean
1490 @param store_reason: whether to store the shutdown reason trail on file
1494 hv_name = instance.hypervisor
1495 hyper = hypervisor.GetHypervisor(hv_name)
1496 iname = instance.name
1498 if instance.name not in hyper.ListInstances(instance.hvparams):
1499 logging.info("Instance %s not running, doing nothing", iname)
1504 self.tried_once = False
1507 if iname not in hyper.ListInstances(instance.hvparams):
1511 hyper.StopInstance(instance, retry=self.tried_once)
1513 _StoreInstReasonTrail(instance.name, reason)
1514 except errors.HypervisorError, err:
1515 if iname not in hyper.ListInstances(instance.hvparams):
1516 # if the instance is no longer existing, consider this a
1517 # success and go to cleanup
1520 _Fail("Failed to stop instance %s: %s", iname, err)
1522 self.tried_once = True
1524 raise utils.RetryAgain()
1527 utils.Retry(_TryShutdown(), 5, timeout)
1528 except utils.RetryTimeout:
1529 # the shutdown did not succeed
1530 logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1533 hyper.StopInstance(instance, force=True)
1534 except errors.HypervisorError, err:
1535 if iname in hyper.ListInstances(instance.hvparams):
1536 # only raise an error if the instance still exists, otherwise
1537 # the error could simply be "instance ... unknown"!
1538 _Fail("Failed to force stop instance %s: %s", iname, err)
1542 if iname in hyper.ListInstances(instance.hvparams):
1543 _Fail("Could not shutdown instance %s even by destroy", iname)
1546 hyper.CleanupInstance(instance.name)
1547 except errors.HypervisorError, err:
1548 logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1550 _RemoveBlockDevLinks(iname, instance.disks)
1553 def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1554 """Reboot an instance.
1556 @type instance: L{objects.Instance}
1557 @param instance: the instance object to reboot
1558 @type reboot_type: str
1559 @param reboot_type: the type of reboot, one the following
1561 - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1562 instance OS, do not recreate the VM
1563 - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1564 restart the VM (at the hypervisor level)
1565 - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1566 not accepted here, since that mode is handled differently, in
1567 cmdlib, and translates into full stop and start of the
1568 instance (instead of a call_instance_reboot RPC)
1569 @type shutdown_timeout: integer
1570 @param shutdown_timeout: maximum timeout for soft shutdown
1571 @type reason: list of reasons
1572 @param reason: the reason trail for this reboot
1576 running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1579 if instance.name not in running_instances:
1580 _Fail("Cannot reboot instance %s that is not running", instance.name)
1582 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1583 if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1585 hyper.RebootInstance(instance)
1586 except errors.HypervisorError, err:
1587 _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1588 elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1590 InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1591 result = StartInstance(instance, False, reason, store_reason=False)
1592 _StoreInstReasonTrail(instance.name, reason)
1594 except errors.HypervisorError, err:
1595 _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1597 _Fail("Invalid reboot_type received: %s", reboot_type)
1600 def InstanceBalloonMemory(instance, memory):
1601 """Resize an instance's memory.
1603 @type instance: L{objects.Instance}
1604 @param instance: the instance object
1606 @param memory: new memory amount in MB
1610 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1611 running = hyper.ListInstances(instance.hvparams)
1612 if instance.name not in running:
1613 logging.info("Instance %s is not running, cannot balloon", instance.name)
1616 hyper.BalloonInstanceMemory(instance, memory)
1617 except errors.HypervisorError, err:
1618 _Fail("Failed to balloon instance memory: %s", err, exc=True)
1621 def MigrationInfo(instance):
1622 """Gather information about an instance to be migrated.
1624 @type instance: L{objects.Instance}
1625 @param instance: the instance definition
1628 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1630 info = hyper.MigrationInfo(instance)
1631 except errors.HypervisorError, err:
1632 _Fail("Failed to fetch migration information: %s", err, exc=True)
1636 def AcceptInstance(instance, info, target):
1637 """Prepare the node to accept an instance.
1639 @type instance: L{objects.Instance}
1640 @param instance: the instance definition
1641 @type info: string/data (opaque)
1642 @param info: migration information, from the source node
1643 @type target: string
1644 @param target: target host (usually ip), on this node
1647 # TODO: why is this required only for DTS_EXT_MIRROR?
1648 if instance.disk_template in constants.DTS_EXT_MIRROR:
1649 # Create the symlinks, as the disks are not active
1652 _GatherAndLinkBlockDevs(instance)
1653 except errors.BlockDeviceError, err:
1654 _Fail("Block device error: %s", err, exc=True)
1656 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1658 hyper.AcceptInstance(instance, info, target)
1659 except errors.HypervisorError, err:
1660 if instance.disk_template in constants.DTS_EXT_MIRROR:
1661 _RemoveBlockDevLinks(instance.name, instance.disks)
1662 _Fail("Failed to accept instance: %s", err, exc=True)
1665 def FinalizeMigrationDst(instance, info, success):
1666 """Finalize any preparation to accept an instance.
1668 @type instance: L{objects.Instance}
1669 @param instance: the instance definition
1670 @type info: string/data (opaque)
1671 @param info: migration information, from the source node
1672 @type success: boolean
1673 @param success: whether the migration was a success or a failure
1676 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1678 hyper.FinalizeMigrationDst(instance, info, success)
1679 except errors.HypervisorError, err:
1680 _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1683 def MigrateInstance(instance, target, live):
1684 """Migrates an instance to another node.
1686 @type instance: L{objects.Instance}
1687 @param instance: the instance definition
1688 @type target: string
1689 @param target: the target node name
1691 @param live: whether the migration should be done live or not (the
1692 interpretation of this parameter is left to the hypervisor)
1693 @raise RPCFail: if migration fails for some reason
1696 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1699 hyper.MigrateInstance(instance, target, live)
1700 except errors.HypervisorError, err:
1701 _Fail("Failed to migrate instance: %s", err, exc=True)
1704 def FinalizeMigrationSource(instance, success, live):
1705 """Finalize the instance migration on the source node.
1707 @type instance: L{objects.Instance}
1708 @param instance: the instance definition of the migrated instance
1710 @param success: whether the migration succeeded or not
1712 @param live: whether the user requested a live migration or not
1713 @raise RPCFail: If the execution fails for some reason
1716 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1719 hyper.FinalizeMigrationSource(instance, success, live)
1720 except Exception, err: # pylint: disable=W0703
1721 _Fail("Failed to finalize the migration on the source node: %s", err,
1725 def GetMigrationStatus(instance):
1726 """Get the migration status
1728 @type instance: L{objects.Instance}
1729 @param instance: the instance that is being migrated
1730 @rtype: L{objects.MigrationStatus}
1731 @return: the status of the current migration (one of
1732 L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1733 progress info that can be retrieved from the hypervisor
1734 @raise RPCFail: If the migration status cannot be retrieved
1737 hyper = hypervisor.GetHypervisor(instance.hypervisor)
1739 return hyper.GetMigrationStatus(instance)
1740 except Exception, err: # pylint: disable=W0703
1741 _Fail("Failed to get migration status: %s", err, exc=True)
1744 def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1745 """Creates a block device for an instance.
1747 @type disk: L{objects.Disk}
1748 @param disk: the object describing the disk we should create
1750 @param size: the size of the physical underlying device, in MiB
1752 @param owner: the name of the instance for which disk is created,
1753 used for device cache data
1754 @type on_primary: boolean
1755 @param on_primary: indicates if it is the primary node or not
1757 @param info: string that will be sent to the physical device
1758 creation, used for example to set (LVM) tags on LVs
1759 @type excl_stor: boolean
1760 @param excl_stor: Whether exclusive_storage is active
1762 @return: the new unique_id of the device (this can sometime be
1763 computed only after creation), or None. On secondary nodes,
1764 it's not required to return anything.
1767 # TODO: remove the obsolete "size" argument
1768 # pylint: disable=W0613
1771 for child in disk.children:
1773 crdev = _RecursiveAssembleBD(child, owner, on_primary)
1774 except errors.BlockDeviceError, err:
1775 _Fail("Can't assemble device %s: %s", child, err)
1776 if on_primary or disk.AssembleOnSecondary():
1777 # we need the children open in case the device itself has to
1780 # pylint: disable=E1103
1782 except errors.BlockDeviceError, err:
1783 _Fail("Can't make child '%s' read-write: %s", child, err)
1787 device = bdev.Create(disk, clist, excl_stor)
1788 except errors.BlockDeviceError, err:
1789 _Fail("Can't create block device: %s", err)
1791 if on_primary or disk.AssembleOnSecondary():
1794 except errors.BlockDeviceError, err:
1795 _Fail("Can't assemble device after creation, unusual event: %s", err)
1796 if on_primary or disk.OpenOnSecondary():
1798 device.Open(force=True)
1799 except errors.BlockDeviceError, err:
1800 _Fail("Can't make device r/w after creation, unusual event: %s", err)
1801 DevCacheManager.UpdateCache(device.dev_path, owner,
1802 on_primary, disk.iv_name)
1804 device.SetInfo(info)
1806 return device.unique_id
1809 def _WipeDevice(path, offset, size):
1810 """This function actually wipes the device.
1812 @param path: The path to the device to wipe
1813 @param offset: The offset in MiB in the file
1814 @param size: The size in MiB to write
1817 # Internal sizes are always in Mebibytes; if the following "dd" command
1818 # should use a different block size the offset and size given to this
1819 # function must be adjusted accordingly before being passed to "dd".
1820 block_size = 1024 * 1024
1822 cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1823 "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1825 result = utils.RunCmd(cmd)
1828 _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1829 result.fail_reason, result.output)
1832 def BlockdevWipe(disk, offset, size):
1833 """Wipes a block device.
1835 @type disk: L{objects.Disk}
1836 @param disk: the disk object we want to wipe
1838 @param offset: The offset in MiB in the file
1840 @param size: The size in MiB to write
1844 rdev = _RecursiveFindBD(disk)
1845 except errors.BlockDeviceError:
1849 _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1851 # Do cross verify some of the parameters
1853 _Fail("Negative offset")
1855 _Fail("Negative size")
1856 if offset > rdev.size:
1857 _Fail("Offset is bigger than device size")
1858 if (offset + size) > rdev.size:
1859 _Fail("The provided offset and size to wipe is bigger than device size")
1861 _WipeDevice(rdev.dev_path, offset, size)
1864 def BlockdevPauseResumeSync(disks, pause):
1865 """Pause or resume the sync of the block device.
1867 @type disks: list of L{objects.Disk}
1868 @param disks: the disks object we want to pause/resume
1870 @param pause: Wheater to pause or resume
1876 rdev = _RecursiveFindBD(disk)
1877 except errors.BlockDeviceError:
1881 success.append((False, ("Cannot change sync for device %s:"
1882 " device not found" % disk.iv_name)))
1885 result = rdev.PauseResumeSync(pause)
1888 success.append((result, None))
1894 success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1899 def BlockdevRemove(disk):
1900 """Remove a block device.
1902 @note: This is intended to be called recursively.
1904 @type disk: L{objects.Disk}
1905 @param disk: the disk object we should remove
1907 @return: the success of the operation
1912 rdev = _RecursiveFindBD(disk)
1913 except errors.BlockDeviceError, err:
1914 # probably can't attach
1915 logging.info("Can't attach to device %s in remove", disk)
1917 if rdev is not None:
1918 r_path = rdev.dev_path
1921 except errors.BlockDeviceError, err:
1922 msgs.append(str(err))
1924 DevCacheManager.RemoveCache(r_path)
1927 for child in disk.children:
1929 BlockdevRemove(child)
1930 except RPCFail, err:
1931 msgs.append(str(err))
1934 _Fail("; ".join(msgs))
1937 def _RecursiveAssembleBD(disk, owner, as_primary):
1938 """Activate a block device for an instance.
1940 This is run on the primary and secondary nodes for an instance.
1942 @note: this function is called recursively.
1944 @type disk: L{objects.Disk}
1945 @param disk: the disk we try to assemble
1947 @param owner: the name of the instance which owns the disk
1948 @type as_primary: boolean
1949 @param as_primary: if we should make the block device
1952 @return: the assembled device or None (in case no device
1954 @raise errors.BlockDeviceError: in case there is an error
1955 during the activation of the children or the device
1961 mcn = disk.ChildrenNeeded()
1963 mcn = 0 # max number of Nones allowed
1965 mcn = len(disk.children) - mcn # max number of Nones
1966 for chld_disk in disk.children:
1968 cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1969 except errors.BlockDeviceError, err:
1970 if children.count(None) >= mcn:
1973 logging.error("Error in child activation (but continuing): %s",
1975 children.append(cdev)
1977 if as_primary or disk.AssembleOnSecondary():
1978 r_dev = bdev.Assemble(disk, children)
1980 if as_primary or disk.OpenOnSecondary():
1982 DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1983 as_primary, disk.iv_name)
1990 def BlockdevAssemble(disk, owner, as_primary, idx):
1991 """Activate a block device for an instance.
1993 This is a wrapper over _RecursiveAssembleBD.
1995 @rtype: str or boolean
1996 @return: a C{/dev/...} path for primary nodes, and
1997 C{True} for secondary nodes
2001 result = _RecursiveAssembleBD(disk, owner, as_primary)
2002 if isinstance(result, BlockDev):
2003 # pylint: disable=E1103
2004 result = result.dev_path
2006 _SymlinkBlockDev(owner, result, idx)
2007 except errors.BlockDeviceError, err:
2008 _Fail("Error while assembling disk: %s", err, exc=True)
2009 except OSError, err:
2010 _Fail("Error while symlinking disk: %s", err, exc=True)
2015 def BlockdevShutdown(disk):
2016 """Shut down a block device.
2018 First, if the device is assembled (Attach() is successful), then
2019 the device is shutdown. Then the children of the device are
2022 This function is called recursively. Note that we don't cache the
2023 children or such, as oppossed to assemble, shutdown of different
2024 devices doesn't require that the upper device was active.
2026 @type disk: L{objects.Disk}
2027 @param disk: the description of the disk we should
2033 r_dev = _RecursiveFindBD(disk)
2034 if r_dev is not None:
2035 r_path = r_dev.dev_path
2038 DevCacheManager.RemoveCache(r_path)
2039 except errors.BlockDeviceError, err:
2040 msgs.append(str(err))
2043 for child in disk.children:
2045 BlockdevShutdown(child)
2046 except RPCFail, err:
2047 msgs.append(str(err))
2050 _Fail("; ".join(msgs))
2053 def BlockdevAddchildren(parent_cdev, new_cdevs):
2054 """Extend a mirrored block device.
2056 @type parent_cdev: L{objects.Disk}
2057 @param parent_cdev: the disk to which we should add children
2058 @type new_cdevs: list of L{objects.Disk}
2059 @param new_cdevs: the list of children which we should add
2063 parent_bdev = _RecursiveFindBD(parent_cdev)
2064 if parent_bdev is None:
2065 _Fail("Can't find parent device '%s' in add children", parent_cdev)
2066 new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2067 if new_bdevs.count(None) > 0:
2068 _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2069 parent_bdev.AddChildren(new_bdevs)
2072 def BlockdevRemovechildren(parent_cdev, new_cdevs):
2073 """Shrink a mirrored block device.
2075 @type parent_cdev: L{objects.Disk}
2076 @param parent_cdev: the disk from which we should remove children
2077 @type new_cdevs: list of L{objects.Disk}
2078 @param new_cdevs: the list of children which we should remove
2082 parent_bdev = _RecursiveFindBD(parent_cdev)
2083 if parent_bdev is None:
2084 _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2086 for disk in new_cdevs:
2087 rpath = disk.StaticDevPath()
2089 bd = _RecursiveFindBD(disk)
2091 _Fail("Can't find device %s while removing children", disk)
2093 devs.append(bd.dev_path)
2095 if not utils.IsNormAbsPath(rpath):
2096 _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2098 parent_bdev.RemoveChildren(devs)
2101 def BlockdevGetmirrorstatus(disks):
2102 """Get the mirroring status of a list of devices.
2104 @type disks: list of L{objects.Disk}
2105 @param disks: the list of disks which we should query
2107 @return: List of L{objects.BlockDevStatus}, one for each disk
2108 @raise errors.BlockDeviceError: if any of the disks cannot be
2114 rbd = _RecursiveFindBD(dsk)
2116 _Fail("Can't find device %s", dsk)
2118 stats.append(rbd.CombinedSyncStatus())
2123 def BlockdevGetmirrorstatusMulti(disks):
2124 """Get the mirroring status of a list of devices.
2126 @type disks: list of L{objects.Disk}
2127 @param disks: the list of disks which we should query
2129 @return: List of tuples, (bool, status), one for each disk; bool denotes
2130 success/failure, status is L{objects.BlockDevStatus} on success, string
2137 rbd = _RecursiveFindBD(disk)
2139 result.append((False, "Can't find device %s" % disk))
2142 status = rbd.CombinedSyncStatus()
2143 except errors.BlockDeviceError, err:
2144 logging.exception("Error while getting disk status")
2145 result.append((False, str(err)))
2147 result.append((True, status))
2149 assert len(disks) == len(result)
2154 def _RecursiveFindBD(disk):
2155 """Check if a device is activated.
2157 If so, return information about the real device.
2159 @type disk: L{objects.Disk}
2160 @param disk: the disk object we need to find
2162 @return: None if the device can't be found,
2163 otherwise the device instance
2168 for chdisk in disk.children:
2169 children.append(_RecursiveFindBD(chdisk))
2171 return bdev.FindDevice(disk, children)
2174 def _OpenRealBD(disk):
2175 """Opens the underlying block device of a disk.
2177 @type disk: L{objects.Disk}
2178 @param disk: the disk object we want to open
2181 real_disk = _RecursiveFindBD(disk)
2182 if real_disk is None:
2183 _Fail("Block device '%s' is not set up", disk)
2190 def BlockdevFind(disk):
2191 """Check if a device is activated.
2193 If it is, return information about the real device.
2195 @type disk: L{objects.Disk}
2196 @param disk: the disk to find
2197 @rtype: None or objects.BlockDevStatus
2198 @return: None if the disk cannot be found, otherwise a the current
2203 rbd = _RecursiveFindBD(disk)
2204 except errors.BlockDeviceError, err:
2205 _Fail("Failed to find device: %s", err, exc=True)
2210 return rbd.GetSyncStatus()
2213 def BlockdevGetdimensions(disks):
2214 """Computes the size of the given disks.
2216 If a disk is not found, returns None instead.
2218 @type disks: list of L{objects.Disk}
2219 @param disks: the list of disk to compute the size for
2221 @return: list with elements None if the disk cannot be found,
2222 otherwise the pair (size, spindles), where spindles is None if the
2223 device doesn't support that
2229 rbd = _RecursiveFindBD(cf)
2230 except errors.BlockDeviceError:
2236 result.append(rbd.GetActualDimensions())
2240 def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2241 """Export a block device to a remote node.
2243 @type disk: L{objects.Disk}
2244 @param disk: the description of the disk to export
2245 @type dest_node: str
2246 @param dest_node: the destination node to export to
2247 @type dest_path: str
2248 @param dest_path: the destination path on the target node
2249 @type cluster_name: str
2250 @param cluster_name: the cluster name, needed for SSH hostalias
2254 real_disk = _OpenRealBD(disk)
2256 # the block size on the read dd is 1MiB to match our units
2257 expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2258 "dd if=%s bs=1048576 count=%s",
2259 real_disk.dev_path, str(disk.size))
2261 # we set here a smaller block size as, due to ssh buffering, more
2262 # than 64-128k will mostly ignored; we use nocreat to fail if the
2263 # device is not already there or we pass a wrong path; we use
2264 # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2265 # to not buffer too much memory; this means that at best, we flush
2266 # every 64k, which will not be very fast
2267 destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2268 " oflag=dsync", dest_path)
2270 remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2271 constants.SSH_LOGIN_USER,
2274 # all commands have been checked, so we're safe to combine them
2275 command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2277 result = utils.RunCmd(["bash", "-c", command])
2280 _Fail("Disk copy command '%s' returned error: %s"
2281 " output: %s", command, result.fail_reason, result.output)
2284 def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2285 """Write a file to the filesystem.
2287 This allows the master to overwrite(!) a file. It will only perform
2288 the operation if the file belongs to a list of configuration files.
2290 @type file_name: str
2291 @param file_name: the target file name
2293 @param data: the new contents of the file
2295 @param mode: the mode to give the file (can be None)
2297 @param uid: the owner of the file
2299 @param gid: the group of the file
2301 @param atime: the atime to set on the file (can be None)
2303 @param mtime: the mtime to set on the file (can be None)
2307 file_name = vcluster.LocalizeVirtualPath(file_name)
2309 if not os.path.isabs(file_name):
2310 _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2312 if file_name not in _ALLOWED_UPLOAD_FILES:
2313 _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2316 raw_data = _Decompress(data)
2318 if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2319 _Fail("Invalid username/groupname type")
2321 getents = runtime.GetEnts()
2322 uid = getents.LookupUser(uid)
2323 gid = getents.LookupGroup(gid)
2325 utils.SafeWriteFile(file_name, None,
2326 data=raw_data, mode=mode, uid=uid, gid=gid,
2327 atime=atime, mtime=mtime)
2330 def RunOob(oob_program, command, node, timeout):
2331 """Executes oob_program with given command on given node.
2333 @param oob_program: The path to the executable oob_program
2334 @param command: The command to invoke on oob_program
2335 @param node: The node given as an argument to the program
2336 @param timeout: Timeout after which we kill the oob program
2339 @raise RPCFail: If execution fails for some reason
2342 result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2345 _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2346 result.fail_reason, result.output)
2348 return result.stdout
2351 def _OSOndiskAPIVersion(os_dir):
2352 """Compute and return the API version of a given OS.
2354 This function will try to read the API version of the OS residing in
2355 the 'os_dir' directory.
2358 @param os_dir: the directory in which we should look for the OS
2360 @return: tuple (status, data) with status denoting the validity and
2361 data holding either the vaid versions or an error message
2364 api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2367 st = os.stat(api_file)
2368 except EnvironmentError, err:
2369 return False, ("Required file '%s' not found under path %s: %s" %
2370 (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2372 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2373 return False, ("File '%s' in %s is not a regular file" %
2374 (constants.OS_API_FILE, os_dir))
2377 api_versions = utils.ReadFile(api_file).splitlines()
2378 except EnvironmentError, err:
2379 return False, ("Error while reading the API version file at %s: %s" %
2380 (api_file, utils.ErrnoOrStr(err)))
2383 api_versions = [int(version.strip()) for version in api_versions]
2384 except (TypeError, ValueError), err:
2385 return False, ("API version(s) can't be converted to integer: %s" %
2388 return True, api_versions
2391 def DiagnoseOS(top_dirs=None):
2392 """Compute the validity for all OSes.
2394 @type top_dirs: list
2395 @param top_dirs: the list of directories in which to
2396 search (if not given defaults to
2397 L{pathutils.OS_SEARCH_PATH})
2398 @rtype: list of L{objects.OS}
2399 @return: a list of tuples (name, path, status, diagnose, variants,
2400 parameters, api_version) for all (potential) OSes under all
2401 search paths, where:
2402 - name is the (potential) OS name
2403 - path is the full path to the OS
2404 - status True/False is the validity of the OS
2405 - diagnose is the error message for an invalid OS, otherwise empty
2406 - variants is a list of supported OS variants, if any
2407 - parameters is a list of (name, help) parameters, if any
2408 - api_version is a list of support OS API versions
2411 if top_dirs is None:
2412 top_dirs = pathutils.OS_SEARCH_PATH
2415 for dir_name in top_dirs:
2416 if os.path.isdir(dir_name):
2418 f_names = utils.ListVisibleFiles(dir_name)
2419 except EnvironmentError, err:
2420 logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2422 for name in f_names:
2423 os_path = utils.PathJoin(dir_name, name)
2424 status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2427 variants = os_inst.supported_variants
2428 parameters = os_inst.supported_parameters
2429 api_versions = os_inst.api_versions
2432 variants = parameters = api_versions = []
2433 result.append((name, os_path, status, diagnose, variants,
2434 parameters, api_versions))
2439 def _TryOSFromDisk(name, base_dir=None):
2440 """Create an OS instance from disk.
2442 This function will return an OS instance if the given name is a
2445 @type base_dir: string
2446 @keyword base_dir: Base directory containing OS installations.
2447 Defaults to a search in all the OS_SEARCH_PATH dirs.
2449 @return: success and either the OS instance if we find a valid one,
2453 if base_dir is None:
2454 os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2456 os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2459 return False, "Directory for OS %s not found in search path" % name
2461 status, api_versions = _OSOndiskAPIVersion(os_dir)
2464 return status, api_versions
2466 if not constants.OS_API_VERSIONS.intersection(api_versions):
2467 return False, ("API version mismatch for path '%s': found %s, want %s." %
2468 (os_dir, api_versions, constants.OS_API_VERSIONS))
2470 # OS Files dictionary, we will populate it with the absolute path
2471 # names; if the value is True, then it is a required file, otherwise
2473 os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2475 if max(api_versions) >= constants.OS_API_V15:
2476 os_files[constants.OS_VARIANTS_FILE] = False
2478 if max(api_versions) >= constants.OS_API_V20:
2479 os_files[constants.OS_PARAMETERS_FILE] = True
2481 del os_files[constants.OS_SCRIPT_VERIFY]
2483 for (filename, required) in os_files.items():
2484 os_files[filename] = utils.PathJoin(os_dir, filename)
2487 st = os.stat(os_files[filename])
2488 except EnvironmentError, err:
2489 if err.errno == errno.ENOENT and not required:
2490 del os_files[filename]
2492 return False, ("File '%s' under path '%s' is missing (%s)" %
2493 (filename, os_dir, utils.ErrnoOrStr(err)))
2495 if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2496 return False, ("File '%s' under path '%s' is not a regular file" %
2499 if filename in constants.OS_SCRIPTS:
2500 if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2501 return False, ("File '%s' under path '%s' is not executable" %
2505 if constants.OS_VARIANTS_FILE in os_files:
2506 variants_file = os_files[constants.OS_VARIANTS_FILE]
2509 utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2510 except EnvironmentError, err:
2511 # we accept missing files, but not other errors
2512 if err.errno != errno.ENOENT:
2513 return False, ("Error while reading the OS variants file at %s: %s" %
2514 (variants_file, utils.ErrnoOrStr(err)))
2517 if constants.OS_PARAMETERS_FILE in os_files:
2518 parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2520 parameters = utils.ReadFile(parameters_file).splitlines()
2521 except EnvironmentError, err:
2522 return False, ("Error while reading the OS parameters file at %s: %s" %
2523 (parameters_file, utils.ErrnoOrStr(err)))
2524 parameters = [v.split(None, 1) for v in parameters]
2526 os_obj = objects.OS(name=name, path=os_dir,
2527 create_script=os_files[constants.OS_SCRIPT_CREATE],
2528 export_script=os_files[constants.OS_SCRIPT_EXPORT],
2529 import_script=os_files[constants.OS_SCRIPT_IMPORT],
2530 rename_script=os_files[constants.OS_SCRIPT_RENAME],
2531 verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2533 supported_variants=variants,
2534 supported_parameters=parameters,
2535 api_versions=api_versions)
2539 def OSFromDisk(name, base_dir=None):
2540 """Create an OS instance from disk.
2542 This function will return an OS instance if the given name is a
2543 valid OS name. Otherwise, it will raise an appropriate
2544 L{RPCFail} exception, detailing why this is not a valid OS.
2546 This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2547 an exception but returns true/false status data.
2549 @type base_dir: string
2550 @keyword base_dir: Base directory containing OS installations.
2551 Defaults to a search in all the OS_SEARCH_PATH dirs.
2552 @rtype: L{objects.OS}
2553 @return: the OS instance if we find a valid one
2554 @raise RPCFail: if we don't find a valid OS
2557 name_only = objects.OS.GetName(name)
2558 status, payload = _TryOSFromDisk(name_only, base_dir)
2566 def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2567 """Calculate the basic environment for an os script.
2570 @param os_name: full operating system name (including variant)
2571 @type inst_os: L{objects.OS}
2572 @param inst_os: operating system for which the environment is being built
2573 @type os_params: dict
2574 @param os_params: the OS parameters
2575 @type debug: integer
2576 @param debug: debug level (0 or 1, for OS Api 10)
2578 @return: dict of environment variables
2579 @raise errors.BlockDeviceError: if the block device
2585 max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2586 result["OS_API_VERSION"] = "%d" % api_version
2587 result["OS_NAME"] = inst_os.name
2588 result["DEBUG_LEVEL"] = "%d" % debug
2591 if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2592 variant = objects.OS.GetVariant(os_name)
2594 variant = inst_os.supported_variants[0]
2597 result["OS_VARIANT"] = variant
2600 for pname, pvalue in os_params.items():
2601 result["OSP_%s" % pname.upper()] = pvalue
2603 # Set a default path otherwise programs called by OS scripts (or
2604 # even hooks called from OS scripts) might break, and we don't want
2605 # to have each script require setting a PATH variable
2606 result["PATH"] = constants.HOOKS_PATH
2611 def OSEnvironment(instance, inst_os, debug=0):
2612 """Calculate the environment for an os script.
2614 @type instance: L{objects.Instance}
2615 @param instance: target instance for the os script run
2616 @type inst_os: L{objects.OS}
2617 @param inst_os: operating system for which the environment is being built
2618 @type debug: integer
2619 @param debug: debug level (0 or 1, for OS Api 10)
2621 @return: dict of environment variables
2622 @raise errors.BlockDeviceError: if the block device
2626 result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2628 for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2629 result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2631 result["HYPERVISOR"] = instance.hypervisor
2632 result["DISK_COUNT"] = "%d" % len(instance.disks)
2633 result["NIC_COUNT"] = "%d" % len(instance.nics)
2634 result["INSTANCE_SECONDARY_NODES"] = \
2635 ("%s" % " ".join(instance.secondary_nodes))
2638 for idx, disk in enumerate(instance.disks):
2639 real_disk = _OpenRealBD(disk)
2640 result["DISK_%d_PATH" % idx] = real_disk.dev_path
2641 result["DISK_%d_ACCESS" % idx] = disk.mode
2642 if constants.HV_DISK_TYPE in instance.hvparams:
2643 result["DISK_%d_FRONTEND_TYPE" % idx] = \
2644 instance.hvparams[constants.HV_DISK_TYPE]
2645 if disk.dev_type in constants.LDS_BLOCK:
2646 result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2647 elif disk.dev_type == constants.LD_FILE:
2648 result["DISK_%d_BACKEND_TYPE" % idx] = \
2649 "file:%s" % disk.physical_id[0]
2652 for idx, nic in enumerate(instance.nics):
2653 result["NIC_%d_MAC" % idx] = nic.mac
2655 result["NIC_%d_IP" % idx] = nic.ip
2656 result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2657 if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2658 result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2659 if nic.nicparams[constants.NIC_LINK]:
2660 result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2662 nobj = objects.Network.FromDict(nic.netinfo)
2663 result.update(nobj.HooksDict("NIC_%d_" % idx))
2664 if constants.HV_NIC_TYPE in instance.hvparams:
2665 result["NIC_%d_FRONTEND_TYPE" % idx] = \
2666 instance.hvparams[constants.HV_NIC_TYPE]
2669 for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2670 for key, value in source.items():
2671 result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2676 def DiagnoseExtStorage(top_dirs=None):
2677 """Compute the validity for all ExtStorage Providers.
2679 @type top_dirs: list
2680 @param top_dirs: the list of directories in which to
2681 search (if not given defaults to
2682 L{pathutils.ES_SEARCH_PATH})
2683 @rtype: list of L{objects.ExtStorage}
2684 @return: a list of tuples (name, path, status, diagnose, parameters)
2685 for all (potential) ExtStorage Providers under all
2686 search paths, where:
2687 - name is the (potential) ExtStorage Provider
2688 - path is the full path to the ExtStorage Provider
2689 - status True/False is the validity of the ExtStorage Provider
2690 - diagnose is the error message for an invalid ExtStorage Provider,
2692 - parameters is a list of (name, help) parameters, if any
2695 if top_dirs is None:
2696 top_dirs = pathutils.ES_SEARCH_PATH
2699 for dir_name in top_dirs:
2700 if os.path.isdir(dir_name):
2702 f_names = utils.ListVisibleFiles(dir_name)
2703 except EnvironmentError, err:
2704 logging.exception("Can't list the ExtStorage directory %s: %s",
2707 for name in f_names:
2708 es_path = utils.PathJoin(dir_name, name)
2709 status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2712 parameters = es_inst.supported_parameters
2716 result.append((name, es_path, status, diagnose, parameters))
2721 def BlockdevGrow(disk, amount, dryrun, backingstore):
2722 """Grow a stack of block devices.
2724 This function is called recursively, with the childrens being the
2725 first ones to resize.
2727 @type disk: L{objects.Disk}
2728 @param disk: the disk to be grown
2729 @type amount: integer
2730 @param amount: the amount (in mebibytes) to grow with
2731 @type dryrun: boolean
2732 @param dryrun: whether to execute the operation in simulation mode
2733 only, without actually increasing the size
2734 @param backingstore: whether to execute the operation on backing storage
2735 only, or on "logical" storage only; e.g. DRBD is logical storage,
2736 whereas LVM, file, RBD are backing storage
2737 @rtype: (status, result)
2738 @return: a tuple with the status of the operation (True/False), and
2739 the errors message if status is False
2742 r_dev = _RecursiveFindBD(disk)
2744 _Fail("Cannot find block device %s", disk)
2747 r_dev.Grow(amount, dryrun, backingstore)
2748 except errors.BlockDeviceError, err:
2749 _Fail("Failed to grow block device: %s", err, exc=True)
2752 def BlockdevSnapshot(disk):
2753 """Create a snapshot copy of a block device.
2755 This function is called recursively, and the snapshot is actually created
2756 just for the leaf lvm backend device.
2758 @type disk: L{objects.Disk}
2759 @param disk: the disk to be snapshotted
2761 @return: snapshot disk ID as (vg, lv)
2764 if disk.dev_type == constants.LD_DRBD8:
2765 if not disk.children:
2766 _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2768 return BlockdevSnapshot(disk.children[0])
2769 elif disk.dev_type == constants.LD_LV:
2770 r_dev = _RecursiveFindBD(disk)
2771 if r_dev is not None:
2772 # FIXME: choose a saner value for the snapshot size
2773 # let's stay on the safe side and ask for the full size, for now
2774 return r_dev.Snapshot(disk.size)
2776 _Fail("Cannot find block device %s", disk)
2778 _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2779 disk.unique_id, disk.dev_type)
2782 def BlockdevSetInfo(disk, info):
2783 """Sets 'metadata' information on block devices.
2785 This function sets 'info' metadata on block devices. Initial
2786 information is set at device creation; this function should be used
2787 for example after renames.
2789 @type disk: L{objects.Disk}
2790 @param disk: the disk to be grown
2792 @param info: new 'info' metadata
2793 @rtype: (status, result)
2794 @return: a tuple with the status of the operation (True/False), and
2795 the errors message if status is False
2798 r_dev = _RecursiveFindBD(disk)
2800 _Fail("Cannot find block device %s", disk)
2804 except errors.BlockDeviceError, err:
2805 _Fail("Failed to set information on block device: %s", err, exc=True)
2808 def FinalizeExport(instance, snap_disks):
2809 """Write out the export configuration information.
2811 @type instance: L{objects.Instance}
2812 @param instance: the instance which we export, used for
2813 saving configuration
2814 @type snap_disks: list of L{objects.Disk}
2815 @param snap_disks: list of snapshot block devices, which
2816 will be used to get the actual name of the dump file
2821 destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2822 finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2824 config = objects.SerializableConfigParser()
2826 config.add_section(constants.INISECT_EXP)
2827 config.set(constants.INISECT_EXP, "version", "0")
2828 config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2829 config.set(constants.INISECT_EXP, "source", instance.primary_node)
2830 config.set(constants.INISECT_EXP, "os", instance.os)
2831 config.set(constants.INISECT_EXP, "compression", "none")
2833 config.add_section(constants.INISECT_INS)
2834 config.set(constants.INISECT_INS, "name", instance.name)
2835 config.set(constants.INISECT_INS, "maxmem", "%d" %
2836 instance.beparams[constants.BE_MAXMEM])
2837 config.set(constants.INISECT_INS, "minmem", "%d" %
2838 instance.beparams[constants.BE_MINMEM])
2839 # "memory" is deprecated, but useful for exporting to old ganeti versions
2840 config.set(constants.INISECT_INS, "memory", "%d" %
2841 instance.beparams[constants.BE_MAXMEM])
2842 config.set(constants.INISECT_INS, "vcpus", "%d" %
2843 instance.beparams[constants.BE_VCPUS])
2844 config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2845 config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2846 config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2849 for nic_count, nic in enumerate(instance.nics):
2851 config.set(constants.INISECT_INS, "nic%d_mac" %
2852 nic_count, "%s" % nic.mac)
2853 config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2854 config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2856 for param in constants.NICS_PARAMETER_TYPES:
2857 config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2858 "%s" % nic.nicparams.get(param, None))
2859 # TODO: redundant: on load can read nics until it doesn't exist
2860 config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2863 for disk_count, disk in enumerate(snap_disks):
2866 config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2867 ("%s" % disk.iv_name))
2868 config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2869 ("%s" % disk.physical_id[1]))
2870 config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2873 config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2875 # New-style hypervisor/backend parameters
2877 config.add_section(constants.INISECT_HYP)
2878 for name, value in instance.hvparams.items():
2879 if name not in constants.HVC_GLOBALS:
2880 config.set(constants.INISECT_HYP, name, str(value))
2882 config.add_section(constants.INISECT_BEP)
2883 for name, value in instance.beparams.items():
2884 config.set(constants.INISECT_BEP, name, str(value))
2886 config.add_section(constants.INISECT_OSP)
2887 for name, value in instance.osparams.items():
2888 config.set(constants.INISECT_OSP, name, str(value))
2890 utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2891 data=config.Dumps())
2892 shutil.rmtree(finaldestdir, ignore_errors=True)
2893 shutil.move(destdir, finaldestdir)
2896 def ExportInfo(dest):
2897 """Get export configuration information.
2900 @param dest: directory containing the export
2902 @rtype: L{objects.SerializableConfigParser}
2903 @return: a serializable config file containing the
2907 cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2909 config = objects.SerializableConfigParser()
2912 if (not config.has_section(constants.INISECT_EXP) or
2913 not config.has_section(constants.INISECT_INS)):
2914 _Fail("Export info file doesn't have the required fields")
2916 return config.Dumps()
2920 """Return a list of exports currently available on this machine.
2923 @return: list of the exports
2926 if os.path.isdir(pathutils.EXPORT_DIR):
2927 return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
2929 _Fail("No exports directory")
2932 def RemoveExport(export):
2933 """Remove an existing export from the node.
2936 @param export: the name of the export to remove
2940 target = utils.PathJoin(pathutils.EXPORT_DIR, export)
2943 shutil.rmtree(target)
2944 except EnvironmentError, err:
2945 _Fail("Error while removing the export: %s", err, exc=True)
2948 def BlockdevRename(devlist):
2949 """Rename a list of block devices.
2951 @type devlist: list of tuples
2952 @param devlist: list of tuples of the form (disk,
2953 new_logical_id, new_physical_id); disk is an
2954 L{objects.Disk} object describing the current disk,
2955 and new logical_id/physical_id is the name we
2958 @return: True if all renames succeeded, False otherwise
2963 for disk, unique_id in devlist:
2964 dev = _RecursiveFindBD(disk)
2966 msgs.append("Can't find device %s in rename" % str(disk))
2970 old_rpath = dev.dev_path
2971 dev.Rename(unique_id)
2972 new_rpath = dev.dev_path
2973 if old_rpath != new_rpath:
2974 DevCacheManager.RemoveCache(old_rpath)
2975 # FIXME: we should add the new cache information here, like:
2976 # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2977 # but we don't have the owner here - maybe parse from existing
2978 # cache? for now, we only lose lvm data when we rename, which
2979 # is less critical than DRBD or MD
2980 except errors.BlockDeviceError, err:
2981 msgs.append("Can't rename device '%s' to '%s': %s" %
2982 (dev, unique_id, err))
2983 logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2986 _Fail("; ".join(msgs))
2989 def _TransformFileStorageDir(fs_dir):
2990 """Checks whether given file_storage_dir is valid.
2992 Checks wheter the given fs_dir is within the cluster-wide default
2993 file_storage_dir or the shared_file_storage_dir, which are stored in
2994 SimpleStore. Only paths under those directories are allowed.
2997 @param fs_dir: the path to check
2999 @return: the normalized path if valid, None otherwise
3002 if not (constants.ENABLE_FILE_STORAGE or
3003 constants.ENABLE_SHARED_FILE_STORAGE):
3004 _Fail("File storage disabled at configure time")
3006 bdev.CheckFileStoragePath(fs_dir)
3008 return os.path.normpath(fs_dir)
3011 def CreateFileStorageDir(file_storage_dir):
3012 """Create file storage directory.
3014 @type file_storage_dir: str
3015 @param file_storage_dir: directory to create
3018 @return: tuple with first element a boolean indicating wheter dir
3019 creation was successful or not
3022 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3023 if os.path.exists(file_storage_dir):
3024 if not os.path.isdir(file_storage_dir):
3025 _Fail("Specified storage dir '%s' is not a directory",
3029 os.makedirs(file_storage_dir, 0750)
3030 except OSError, err:
3031 _Fail("Cannot create file storage directory '%s': %s",
3032 file_storage_dir, err, exc=True)
3035 def RemoveFileStorageDir(file_storage_dir):
3036 """Remove file storage directory.
3038 Remove it only if it's empty. If not log an error and return.
3040 @type file_storage_dir: str
3041 @param file_storage_dir: the directory we should cleanup
3042 @rtype: tuple (success,)
3043 @return: tuple of one element, C{success}, denoting
3044 whether the operation was successful
3047 file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3048 if os.path.exists(file_storage_dir):
3049 if not os.path.isdir(file_storage_dir):
3050 _Fail("Specified Storage directory '%s' is not a directory",
3052 # deletes dir only if empty, otherwise we want to fail the rpc call
3054 os.rmdir(file_storage_dir)
3055 except OSError, err:
3056 _Fail("Cannot remove file storage directory '%s': %s",
3057 file_storage_dir, err)
3060 def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3061 """Rename the file storage directory.
3063 @type old_file_storage_dir: str
3064 @param old_file_storage_dir: the current path
3065 @type new_file_storage_dir: str
3066 @param new_file_storage_dir: the name we should rename to
3067 @rtype: tuple (success,)
3068 @return: tuple of one element, C{success}, denoting
3069 whether the operation was successful
3072 old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3073 new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3074 if not os.path.exists(new_file_storage_dir):
3075 if os.path.isdir(old_file_storage_dir):
3077 os.rename(old_file_storage_dir, new_file_storage_dir)
3078 except OSError, err:
3079 _Fail("Cannot rename '%s' to '%s': %s",
3080 old_file_storage_dir, new_file_storage_dir, err)
3082 _Fail("Specified storage dir '%s' is not a directory",
3083 old_file_storage_dir)
3085 if os.path.exists(old_file_storage_dir):
3086 _Fail("Cannot rename '%s' to '%s': both locations exist",
3087 old_file_storage_dir, new_file_storage_dir)
3090 def _EnsureJobQueueFile(file_name):
3091 """Checks whether the given filename is in the queue directory.
3093 @type file_name: str
3094 @param file_name: the file name we should check
3096 @raises RPCFail: if the file is not valid
3099 if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3100 _Fail("Passed job queue file '%s' does not belong to"
3101 " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3104 def JobQueueUpdate(file_name, content):
3105 """Updates a file in the queue directory.
3107 This is just a wrapper over L{utils.io.WriteFile}, with proper
3110 @type file_name: str
3111 @param file_name: the job file name
3113 @param content: the new job contents
3115 @return: the success of the operation
3118 file_name = vcluster.LocalizeVirtualPath(file_name)
3120 _EnsureJobQueueFile(file_name)
3121 getents = runtime.GetEnts()
3123 # Write and replace the file atomically
3124 utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3125 gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3128 def JobQueueRename(old, new):
3129 """Renames a job queue file.
3131 This is just a wrapper over os.rename with proper checking.
3134 @param old: the old (actual) file name
3136 @param new: the desired file name
3138 @return: the success of the operation and payload
3141 old = vcluster.LocalizeVirtualPath(old)
3142 new = vcluster.LocalizeVirtualPath(new)
3144 _EnsureJobQueueFile(old)
3145 _EnsureJobQueueFile(new)
3147 getents = runtime.GetEnts()
3149 utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3150 dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3153 def BlockdevClose(instance_name, disks):
3154 """Closes the given block devices.
3156 This means they will be switched to secondary mode (in case of
3159 @param instance_name: if the argument is not empty, the symlinks
3160 of this instance will be removed
3161 @type disks: list of L{objects.Disk}
3162 @param disks: the list of disks to be closed
3163 @rtype: tuple (success, message)
3164 @return: a tuple of success and message, where success
3165 indicates the succes of the operation, and message
3166 which will contain the error details in case we
3172 rd = _RecursiveFindBD(cf)
3174 _Fail("Can't find device %s", cf)
3181 except errors.BlockDeviceError, err:
3182 msg.append(str(err))
3184 _Fail("Can't make devices secondary: %s", ",".join(msg))
3187 _RemoveBlockDevLinks(instance_name, disks)
3190 def ValidateHVParams(hvname, hvparams):
3191 """Validates the given hypervisor parameters.
3193 @type hvname: string
3194 @param hvname: the hypervisor name
3195 @type hvparams: dict
3196 @param hvparams: the hypervisor parameters to be validated
3201 hv_type = hypervisor.GetHypervisor(hvname)
3202 hv_type.ValidateParameters(hvparams)
3203 except errors.HypervisorError, err:
3204 _Fail(str(err), log=False)
3207 def _CheckOSPList(os_obj, parameters):
3208 """Check whether a list of parameters is supported by the OS.
3210 @type os_obj: L{objects.OS}
3211 @param os_obj: OS object to check
3212 @type parameters: list
3213 @param parameters: the list of parameters to check
3216 supported = [v[0] for v in os_obj.supported_parameters]
3217 delta = frozenset(parameters).difference(supported)
3219 _Fail("The following parameters are not supported"
3220 " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3223 def ValidateOS(required, osname, checks, osparams):
3224 """Validate the given OS' parameters.
3226 @type required: boolean
3227 @param required: whether absence of the OS should translate into
3229 @type osname: string
3230 @param osname: the OS to be validated
3232 @param checks: list of the checks to run (currently only 'parameters')
3233 @type osparams: dict
3234 @param osparams: dictionary with OS parameters
3236 @return: True if the validation passed, or False if the OS was not
3237 found and L{required} was false
3240 if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3241 _Fail("Unknown checks required for OS %s: %s", osname,
3242 set(checks).difference(constants.OS_VALIDATE_CALLS))
3244 name_only = objects.OS.GetName(osname)
3245 status, tbv = _TryOSFromDisk(name_only, None)
3253 if max(tbv.api_versions) < constants.OS_API_V20:
3256 if constants.OS_VALIDATE_PARAMETERS in checks:
3257 _CheckOSPList(tbv, osparams.keys())
3259 validate_env = OSCoreEnv(osname, tbv, osparams)
3260 result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3261 cwd=tbv.path, reset_env=True)
3263 logging.error("os validate command '%s' returned error: %s output: %s",
3264 result.cmd, result.fail_reason, result.output)
3265 _Fail("OS validation script failed (%s), output: %s",
3266 result.fail_reason, result.output, log=False)
3272 """Demotes the current node from master candidate role.
3275 # try to ensure we're not the master by mistake
3276 master, myself = ssconf.GetMasterAndMyself()
3277 if master == myself:
3278 _Fail("ssconf status shows I'm the master node, will not demote")
3280 result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3281 if not result.failed:
3282 _Fail("The master daemon is running, will not demote")
3285 if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3286 utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3287 except EnvironmentError, err:
3288 if err.errno != errno.ENOENT:
3289 _Fail("Error while backing up cluster file: %s", err, exc=True)
3291 utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3294 def _GetX509Filenames(cryptodir, name):
3295 """Returns the full paths for the private key and certificate.
3298 return (utils.PathJoin(cryptodir, name),
3299 utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3300 utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3303 def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3304 """Creates a new X509 certificate for SSL/TLS.
3307 @param validity: Validity in seconds
3308 @rtype: tuple; (string, string)
3309 @return: Certificate name and public part
3312 (key_pem, cert_pem) = \
3313 utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3314 min(validity, _MAX_SSL_CERT_VALIDITY))
3316 cert_dir = tempfile.mkdtemp(dir=cryptodir,
3317 prefix="x509-%s-" % utils.TimestampForFilename())
3319 name = os.path.basename(cert_dir)
3320 assert len(name) > 5
3322 (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3324 utils.WriteFile(key_file, mode=0400, data=key_pem)
3325 utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3327 # Never return private key as it shouldn't leave the node
3328 return (name, cert_pem)
3330 shutil.rmtree(cert_dir, ignore_errors=True)
3334 def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3335 """Removes a X509 certificate.
3338 @param name: Certificate name
3341 (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3343 utils.RemoveFile(key_file)
3344 utils.RemoveFile(cert_file)
3348 except EnvironmentError, err:
3349 _Fail("Cannot remove certificate directory '%s': %s",
3353 def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3354 """Returns the command for the requested input/output.
3356 @type instance: L{objects.Instance}
3357 @param instance: The instance object
3358 @param mode: Import/export mode
3359 @param ieio: Input/output type
3360 @param ieargs: Input/output arguments
3363 assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3370 if ieio == constants.IEIO_FILE:
3371 (filename, ) = ieargs
3373 if not utils.IsNormAbsPath(filename):
3374 _Fail("Path '%s' is not normalized or absolute", filename)
3376 real_filename = os.path.realpath(filename)
3377 directory = os.path.dirname(real_filename)
3379 if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3380 _Fail("File '%s' is not under exports directory '%s': %s",
3381 filename, pathutils.EXPORT_DIR, real_filename)
3384 utils.Makedirs(directory, mode=0750)
3386 quoted_filename = utils.ShellQuote(filename)
3388 if mode == constants.IEM_IMPORT:
3389 suffix = "> %s" % quoted_filename
3390 elif mode == constants.IEM_EXPORT:
3391 suffix = "< %s" % quoted_filename
3393 # Retrieve file size
3395 st = os.stat(filename)
3396 except EnvironmentError, err:
3397 logging.error("Can't stat(2) %s: %s", filename, err)
3399 exp_size = utils.BytesToMebibyte(st.st_size)
3401 elif ieio == constants.IEIO_RAW_DISK:
3404 real_disk = _OpenRealBD(disk)
3406 if mode == constants.IEM_IMPORT:
3407 # we set here a smaller block size as, due to transport buffering, more
3408 # than 64-128k will mostly ignored; we use nocreat to fail if the device
3409 # is not already there or we pass a wrong path; we use notrunc to no
3410 # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3411 # much memory; this means that at best, we flush every 64k, which will
3413 suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3414 " bs=%s oflag=dsync"),
3418 elif mode == constants.IEM_EXPORT:
3419 # the block size on the read dd is 1MiB to match our units
3420 prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3422 str(1024 * 1024), # 1 MB
3424 exp_size = disk.size
3426 elif ieio == constants.IEIO_SCRIPT:
3427 (disk, disk_index, ) = ieargs
3429 assert isinstance(disk_index, (int, long))
3431 real_disk = _OpenRealBD(disk)
3433 inst_os = OSFromDisk(instance.os)
3434 env = OSEnvironment(instance, inst_os)
3436 if mode == constants.IEM_IMPORT:
3437 env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3438 env["IMPORT_INDEX"] = str(disk_index)
3439 script = inst_os.import_script
3441 elif mode == constants.IEM_EXPORT:
3442 env["EXPORT_DEVICE"] = real_disk.dev_path
3443 env["EXPORT_INDEX"] = str(disk_index)
3444 script = inst_os.export_script
3446 # TODO: Pass special environment only to script
3447 script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3449 if mode == constants.IEM_IMPORT:
3450 suffix = "| %s" % script_cmd
3452 elif mode == constants.IEM_EXPORT:
3453 prefix = "%s |" % script_cmd
3455 # Let script predict size
3456 exp_size = constants.IE_CUSTOM_SIZE
3459 _Fail("Invalid %s I/O mode %r", mode, ieio)
3461 return (env, prefix, suffix, exp_size)
3464 def _CreateImportExportStatusDir(prefix):
3465 """Creates status directory for import/export.
3468 return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3470 (prefix, utils.TimestampForFilename())))
3473 def StartImportExportDaemon(mode, opts, host, port, instance, component,
3475 """Starts an import or export daemon.
3477 @param mode: Import/output mode
3478 @type opts: L{objects.ImportExportOptions}
3479 @param opts: Daemon options
3481 @param host: Remote host for export (None for import)
3483 @param port: Remote port for export (None for import)
3484 @type instance: L{objects.Instance}
3485 @param instance: Instance object
3486 @type component: string
3487 @param component: which part of the instance is transferred now,
3489 @param ieio: Input/output type
3490 @param ieioargs: Input/output arguments
3493 if mode == constants.IEM_IMPORT:
3496 if not (host is None and port is None):
3497 _Fail("Can not specify host or port on import")
3499 elif mode == constants.IEM_EXPORT:
3502 if host is None or port is None:
3503 _Fail("Host and port must be specified for an export")
3506 _Fail("Invalid mode %r", mode)
3508 if (opts.key_name is None) ^ (opts.ca_pem is None):
3509 _Fail("Cluster certificate can only be used for both key and CA")
3511 (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3512 _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3514 if opts.key_name is None:
3516 key_path = pathutils.NODED_CERT_FILE
3517 cert_path = pathutils.NODED_CERT_FILE
3518 assert opts.ca_pem is None
3520 (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3522 assert opts.ca_pem is not None
3524 for i in [key_path, cert_path]:
3525 if not os.path.exists(i):
3526 _Fail("File '%s' does not exist" % i)
3528 status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3530 status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3531 pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3532 ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3534 if opts.ca_pem is None:
3536 ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3541 utils.WriteFile(ca_file, data=ca, mode=0400)
3544 pathutils.IMPORT_EXPORT_DAEMON,
3546 "--key=%s" % key_path,
3547 "--cert=%s" % cert_path,
3548 "--ca=%s" % ca_file,
3552 cmd.append("--host=%s" % host)
3555 cmd.append("--port=%s" % port)
3558 cmd.append("--ipv6")
3560 cmd.append("--ipv4")
3563 cmd.append("--compress=%s" % opts.compress)
3566 cmd.append("--magic=%s" % opts.magic)
3568 if exp_size is not None:
3569 cmd.append("--expected-size=%s" % exp_size)
3572 cmd.append("--cmd-prefix=%s" % cmd_prefix)
3575 cmd.append("--cmd-suffix=%s" % cmd_suffix)
3577 if mode == constants.IEM_EXPORT:
3578 # Retry connection a few times when connecting to remote peer
3579 cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3580 cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3581 elif opts.connect_timeout is not None:
3582 assert mode == constants.IEM_IMPORT
3583 # Overall timeout for establishing connection while listening
3584 cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3586 logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3588 # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3589 # support for receiving a file descriptor for output
3590 utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3593 # The import/export name is simply the status directory name
3594 return os.path.basename(status_dir)
3597 shutil.rmtree(status_dir, ignore_errors=True)
3601 def GetImportExportStatus(names):
3602 """Returns import/export daemon status.
3604 @type names: sequence
3605 @param names: List of names
3606 @rtype: List of dicts
3607 @return: Returns a list of the state of each named import/export or None if a
3608 status couldn't be read
3614 status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3618 data = utils.ReadFile(status_file)
3619 except EnvironmentError, err:
3620 if err.errno != errno.ENOENT:
3628 result.append(serializer.LoadJson(data))
3633 def AbortImportExport(name):
3634 """Sends SIGTERM to a running import/export daemon.
3637 logging.info("Abort import/export %s", name)
3639 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3640 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3643 logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3645 utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3648 def CleanupImportExport(name):
3649 """Cleanup after an import or export.
3651 If the import/export daemon is still running it's killed. Afterwards the
3652 whole status directory is removed.
3655 logging.info("Finalizing import/export %s", name)
3657 status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3659 pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3662 logging.info("Import/export %s is still running with PID %s",
3664 utils.KillProcess(pid, waitpid=False)
3666 shutil.rmtree(status_dir, ignore_errors=True)
3669 def _FindDisks(nodes_ip, disks):
3670 """Sets the physical ID on disks and returns the block devices.
3673 # set the correct physical ID
3674 my_name = netutils.Hostname.GetSysName()
3676 cf.SetPhysicalID(my_name, nodes_ip)
3681 rd = _RecursiveFindBD(cf)
3683 _Fail("Can't find device %s", cf)
3688 def DrbdDisconnectNet(nodes_ip, disks):
3689 """Disconnects the network on a list of drbd devices.
3692 bdevs = _FindDisks(nodes_ip, disks)
3698 except errors.BlockDeviceError, err:
3699 _Fail("Can't change network configuration to standalone mode: %s",
3703 def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3704 """Attaches the network on a list of drbd devices.
3707 bdevs = _FindDisks(nodes_ip, disks)
3710 for idx, rd in enumerate(bdevs):
3712 _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3713 except EnvironmentError, err:
3714 _Fail("Can't create symlink: %s", err)
3715 # reconnect disks, switch to new master configuration and if
3716 # needed primary mode
3719 rd.AttachNet(multimaster)
3720 except errors.BlockDeviceError, err:
3721 _Fail("Can't change network configuration: %s", err)
3723 # wait until the disks are connected; we need to retry the re-attach
3724 # if the device becomes standalone, as this might happen if the one
3725 # node disconnects and reconnects in a different mode before the
3726 # other node reconnects; in this case, one or both of the nodes will
3727 # decide it has wrong configuration and switch to standalone
3730 all_connected = True
3733 stats = rd.GetProcStatus()
3735 all_connected = (all_connected and
3736 (stats.is_connected or stats.is_in_resync))
3738 if stats.is_standalone:
3739 # peer had different config info and this node became
3740 # standalone, even though this should not happen with the
3741 # new staged way of changing disk configs
3743 rd.AttachNet(multimaster)
3744 except errors.BlockDeviceError, err:
3745 _Fail("Can't change network configuration: %s", err)
3747 if not all_connected:
3748 raise utils.RetryAgain()
3751 # Start with a delay of 100 miliseconds and go up to 5 seconds
3752 utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3753 except utils.RetryTimeout:
3754 _Fail("Timeout in disk reconnecting")
3757 # change to primary mode
3761 except errors.BlockDeviceError, err:
3762 _Fail("Can't change to primary mode: %s", err)
3765 def DrbdWaitSync(nodes_ip, disks):
3766 """Wait until DRBDs have synchronized.
3770 stats = rd.GetProcStatus()
3771 if not (stats.is_connected or stats.is_in_resync):
3772 raise utils.RetryAgain()
3775 bdevs = _FindDisks(nodes_ip, disks)
3781 # poll each second for 15 seconds
3782 stats = utils.Retry(_helper, 1, 15, args=[rd])
3783 except utils.RetryTimeout:
3784 stats = rd.GetProcStatus()
3786 if not (stats.is_connected or stats.is_in_resync):
3787 _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3788 alldone = alldone and (not stats.is_in_resync)
3789 if stats.sync_percent is not None:
3790 min_resync = min(min_resync, stats.sync_percent)
3792 return (alldone, min_resync)
3795 def GetDrbdUsermodeHelper():
3796 """Returns DRBD usermode helper currently configured.
3800 return drbd.DRBD8.GetUsermodeHelper()
3801 except errors.BlockDeviceError, err:
3805 def PowercycleNode(hypervisor_type):
3806 """Hard-powercycle the node.
3808 Because we need to return first, and schedule the powercycle in the
3809 background, we won't be able to report failures nicely.
3812 hyper = hypervisor.GetHypervisor(hypervisor_type)
3816 # if we can't fork, we'll pretend that we're in the child process
3819 return "Reboot scheduled in 5 seconds"
3820 # ensure the child is running on ram
3823 except Exception: # pylint: disable=W0703
3826 hyper.PowercycleNode()
3829 def _VerifyRestrictedCmdName(cmd):
3830 """Verifies a restricted command name.
3833 @param cmd: Command name
3834 @rtype: tuple; (boolean, string or None)
3835 @return: The tuple's first element is the status; if C{False}, the second
3836 element is an error message string, otherwise it's C{None}
3840 return (False, "Missing command name")
3842 if os.path.basename(cmd) != cmd:
3843 return (False, "Invalid command name")
3845 if not constants.EXT_PLUGIN_MASK.match(cmd):
3846 return (False, "Command name contains forbidden characters")
3851 def _CommonRestrictedCmdCheck(path, owner):
3852 """Common checks for restricted command file system directories and files.
3855 @param path: Path to check
3856 @param owner: C{None} or tuple containing UID and GID
3857 @rtype: tuple; (boolean, string or C{os.stat} result)
3858 @return: The tuple's first element is the status; if C{False}, the second
3859 element is an error message string, otherwise it's the result of C{os.stat}
3863 # Default to root as owner
3868 except EnvironmentError, err:
3869 return (False, "Can't stat(2) '%s': %s" % (path, err))
3871 if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
3872 return (False, "Permissions on '%s' are too permissive" % path)
3874 if (st.st_uid, st.st_gid) != owner:
3875 (owner_uid, owner_gid) = owner
3876 return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
3881 def _VerifyRestrictedCmdDirectory(path, _owner=None):
3882 """Verifies restricted command directory.
3885 @param path: Path to check
3886 @rtype: tuple; (boolean, string or None)
3887 @return: The tuple's first element is the status; if C{False}, the second
3888 element is an error message string, otherwise it's C{None}
3891 (status, value) = _CommonRestrictedCmdCheck(path, _owner)
3894 return (False, value)
3896 if not stat.S_ISDIR(value.st_mode):
3897 return (False, "Path '%s' is not a directory" % path)
3902 def _VerifyRestrictedCmd(path, cmd, _owner=None):
3903 """Verifies a whole restricted command and returns its executable filename.
3906 @param path: Directory containing restricted commands
3908 @param cmd: Command name
3909 @rtype: tuple; (boolean, string)
3910 @return: The tuple's first element is the status; if C{False}, the second
3911 element is an error message string, otherwise the second element is the
3912 absolute path to the executable
3915 executable = utils.PathJoin(path, cmd)
3917 (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
3922 if not utils.IsExecutable(executable):
3923 return (False, "access(2) thinks '%s' can't be executed" % executable)
3925 return (True, executable)
3928 def _PrepareRestrictedCmd(path, cmd,
3929 _verify_dir=_VerifyRestrictedCmdDirectory,
3930 _verify_name=_VerifyRestrictedCmdName,
3931 _verify_cmd=_VerifyRestrictedCmd):
3932 """Performs a number of tests on a restricted command.
3935 @param path: Directory containing restricted commands
3937 @param cmd: Command name
3938 @return: Same as L{_VerifyRestrictedCmd}
3941 # Verify the directory first
3942 (status, msg) = _verify_dir(path)
3944 # Check command if everything was alright
3945 (status, msg) = _verify_name(cmd)
3950 # Check actual executable
3951 return _verify_cmd(path, cmd)
3954 def RunRestrictedCmd(cmd,
3955 _lock_timeout=_RCMD_LOCK_TIMEOUT,
3956 _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
3957 _path=pathutils.RESTRICTED_COMMANDS_DIR,
3958 _sleep_fn=time.sleep,
3959 _prepare_fn=_PrepareRestrictedCmd,
3960 _runcmd_fn=utils.RunCmd,
3961 _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
3962 """Executes a restricted command after performing strict tests.
3965 @param cmd: Command name
3967 @return: Command output
3968 @raise RPCFail: In case of an error
3971 logging.info("Preparing to run restricted command '%s'", cmd)
3974 _Fail("Restricted commands disabled at configure time")
3980 lock = utils.FileLock.Open(_lock_file)
3981 lock.Exclusive(blocking=True, timeout=_lock_timeout)
3983 (status, value) = _prepare_fn(_path, cmd)
3986 cmdresult = _runcmd_fn([value], env={}, reset_env=True,
3987 postfork_fn=lambda _: lock.Unlock())
3989 logging.error(value)
3990 except Exception: # pylint: disable=W0703
3991 # Keep original error in log
3992 logging.exception("Caught exception")
3994 if cmdresult is None:
3995 logging.info("Sleeping for %0.1f seconds before returning",
3996 _RCMD_INVALID_DELAY)
3997 _sleep_fn(_RCMD_INVALID_DELAY)
3999 # Do not include original error message in returned error
4000 _Fail("Executing command '%s' failed" % cmd)
4001 elif cmdresult.failed or cmdresult.fail_reason:
4002 _Fail("Restricted command '%s' failed: %s; output: %s",
4003 cmd, cmdresult.fail_reason, cmdresult.output)
4005 return cmdresult.output
4007 if lock is not None:
4008 # Release lock at last
4013 def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4014 """Creates or removes the watcher pause file.
4016 @type until: None or number
4017 @param until: Unix timestamp saying until when the watcher shouldn't run
4021 logging.info("Received request to no longer pause watcher")
4022 utils.RemoveFile(_filename)
4024 logging.info("Received request to pause watcher until %s", until)
4026 if not ht.TNumber(until):
4027 _Fail("Duration must be numeric")
4029 utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4032 class HooksRunner(object):
4035 This class is instantiated on the node side (ganeti-noded) and not
4039 def __init__(self, hooks_base_dir=None):
4040 """Constructor for hooks runner.
4042 @type hooks_base_dir: str or None
4043 @param hooks_base_dir: if not None, this overrides the
4044 L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4047 if hooks_base_dir is None:
4048 hooks_base_dir = pathutils.HOOKS_BASE_DIR
4049 # yeah, _BASE_DIR is not valid for attributes, we use it like a
4051 self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4053 def RunLocalHooks(self, node_list, hpath, phase, env):
4054 """Check that the hooks will be run only locally and then run them.
4057 assert len(node_list) == 1
4059 _, myself = ssconf.GetMasterAndMyself()
4060 assert node == myself
4062 results = self.RunHooks(hpath, phase, env)
4064 # Return values in the form expected by HooksMaster
4065 return {node: (None, False, results)}
4067 def RunHooks(self, hpath, phase, env):
4068 """Run the scripts in the hooks directory.
4071 @param hpath: the path to the hooks directory which
4074 @param phase: either L{constants.HOOKS_PHASE_PRE} or
4075 L{constants.HOOKS_PHASE_POST}
4077 @param env: dictionary with the environment for the hook
4079 @return: list of 3-element tuples:
4081 - script result, either L{constants.HKR_SUCCESS} or
4082 L{constants.HKR_FAIL}
4083 - output of the script
4085 @raise errors.ProgrammerError: for invalid input
4089 if phase == constants.HOOKS_PHASE_PRE:
4091 elif phase == constants.HOOKS_PHASE_POST:
4094 _Fail("Unknown hooks phase '%s'", phase)
4096 subdir = "%s-%s.d" % (hpath, suffix)
4097 dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4101 if not os.path.isdir(dir_name):
4102 # for non-existing/non-dirs, we simply exit instead of logging a
4103 # warning at every operation
4106 runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4108 for (relname, relstatus, runresult) in runparts_results:
4109 if relstatus == constants.RUNPARTS_SKIP:
4110 rrval = constants.HKR_SKIP
4112 elif relstatus == constants.RUNPARTS_ERR:
4113 rrval = constants.HKR_FAIL
4114 output = "Hook script execution error: %s" % runresult
4115 elif relstatus == constants.RUNPARTS_RUN:
4116 if runresult.failed:
4117 rrval = constants.HKR_FAIL
4119 rrval = constants.HKR_SUCCESS
4120 output = utils.SafeEncode(runresult.output.strip())
4121 results.append(("%s/%s" % (subdir, relname), rrval, output))
4126 class IAllocatorRunner(object):
4127 """IAllocator runner.
4129 This class is instantiated on the node side (ganeti-noded) and not on
4134 def Run(name, idata):
4135 """Run an iallocator script.
4138 @param name: the iallocator script name
4140 @param idata: the allocator input data
4143 @return: two element tuple of:
4145 - either error message or stdout of allocator (for success)
4148 alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4150 if alloc_script is None:
4151 _Fail("iallocator module '%s' not found in the search path", name)
4153 fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4157 result = utils.RunCmd([alloc_script, fin_name])
4159 _Fail("iallocator module '%s' failed: %s, output '%s'",
4160 name, result.fail_reason, result.output)
4164 return result.stdout
4167 class DevCacheManager(object):
4168 """Simple class for managing a cache of block device information.
4171 _DEV_PREFIX = "/dev/"
4172 _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4175 def _ConvertPath(cls, dev_path):
4176 """Converts a /dev/name path to the cache file name.
4178 This replaces slashes with underscores and strips the /dev
4179 prefix. It then returns the full path to the cache file.
4182 @param dev_path: the C{/dev/} path name
4184 @return: the converted path name
4187 if dev_path.startswith(cls._DEV_PREFIX):
4188 dev_path = dev_path[len(cls._DEV_PREFIX):]
4189 dev_path = dev_path.replace("/", "_")
4190 fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4194 def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4195 """Updates the cache information for a given device.
4198 @param dev_path: the pathname of the device
4200 @param owner: the owner (instance name) of the device
4201 @type on_primary: bool
4202 @param on_primary: whether this is the primary
4205 @param iv_name: the instance-visible name of the
4206 device, as in objects.Disk.iv_name
4211 if dev_path is None:
4212 logging.error("DevCacheManager.UpdateCache got a None dev_path")
4214 fpath = cls._ConvertPath(dev_path)
4220 iv_name = "not_visible"
4221 fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4223 utils.WriteFile(fpath, data=fdata)
4224 except EnvironmentError, err:
4225 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4228 def RemoveCache(cls, dev_path):
4229 """Remove data for a dev_path.
4231 This is just a wrapper over L{utils.io.RemoveFile} with a converted
4232 path name and logging.
4235 @param dev_path: the pathname of the device
4240 if dev_path is None:
4241 logging.error("DevCacheManager.RemoveCache got a None dev_path")
4243 fpath = cls._ConvertPath(dev_path)
4245 utils.RemoveFile(fpath)
4246 except EnvironmentError, err:
4247 logging.exception("Can't update bdev cache for %s: %s", dev_path, err)