X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/878c42aed39f1ae9e89730b9d2cc23c3693c6729..8e00f36a0eff4a16f06922b92fcbaee204af1757:/lib/backend.py diff --git a/lib/backend.py b/lib/backend.py index 59484b8..abe6b39 100644 --- a/lib/backend.py +++ b/lib/backend.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012 Google Inc. +# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc. # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by @@ -64,10 +64,11 @@ from ganeti import mcpu from ganeti import compat from ganeti import pathutils from ganeti import vcluster +from ganeti import ht _BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id" -_ALLOWED_CLEAN_DIRS = frozenset([ +_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([ pathutils.DATA_DIR, pathutils.JOB_QUEUE_ARCHIVE_DIR, pathutils.QUEUE_DIR, @@ -87,15 +88,15 @@ _LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$") _MASTER_START = "start" _MASTER_STOP = "stop" -#: Maximum file permissions for remote command directory and executables +#: Maximum file permissions for restricted command directory and executables _RCMD_MAX_MODE = (stat.S_IRWXU | stat.S_IRGRP | stat.S_IXGRP | stat.S_IROTH | stat.S_IXOTH) -#: Delay before returning an error for remote commands +#: Delay before returning an error for restricted commands _RCMD_INVALID_DELAY = 10 -#: How long to wait to acquire lock for remote commands (shorter than +#: How long to wait to acquire lock for restricted commands (shorter than #: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many #: command requests arrive _RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8 @@ -109,6 +110,34 @@ class RPCFail(Exception): """ +def _GetInstReasonFilename(instance_name): + """Path of the file containing the reason of the instance status change. + + @type instance_name: string + @param instance_name: The name of the instance + @rtype: string + @return: The path of the file + + """ + return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name) + + +def _StoreInstReasonTrail(instance_name, trail): + """Serialize a reason trail related to an instance change of state to file. + + The exact location of the file depends on the name of the instance and on + the configuration of the Ganeti cluster defined at deploy time. + + @type instance_name: string + @param instance_name: The name of the instance + @rtype: None + + """ + json = serializer.DumpJson(trail) + filename = _GetInstReasonFilename(instance_name) + utils.WriteFile(filename, data=json) + + def _Fail(msg, *args, **kwargs): """Log an error and the raise an RPCFail exception. @@ -357,8 +386,8 @@ def _RunMasterSetupScript(master_params, action, use_external_mip_script): result = utils.RunCmd([setup_script, action], env=env, reset_env=True) if result.failed: - _Fail("Failed to %s the master IP. Script return value: %s" % - (action, result.exit_code), log=True) + _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" % + (action, result.exit_code, result.output), log=True) @RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup", @@ -540,12 +569,12 @@ def LeaveCluster(modify_ssh_setup): raise errors.QuitGanetiException(True, "Shutdown scheduled") -def _GetVgInfo(name): +def _GetVgInfo(name, excl_stor): """Retrieves information about a LVM volume group. """ # TODO: GetVGInfo supports returning information for multiple VGs at once - vginfo = bdev.LogicalVolume.GetVGInfo([name]) + vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor) if vginfo: vg_free = int(round(vginfo[0][0], 0)) vg_size = int(round(vginfo[0][1], 0)) @@ -588,25 +617,44 @@ def _GetNamedNodeInfo(names, fn): return map(fn, names) -def GetNodeInfo(vg_names, hv_names): +def GetNodeInfo(vg_names, hv_names, excl_stor): """Gives back a hash with different information about the node. @type vg_names: list of string @param vg_names: Names of the volume groups to ask for disk space information @type hv_names: list of string @param hv_names: Names of the hypervisors to ask for node information + @type excl_stor: boolean + @param excl_stor: Whether exclusive_storage is active @rtype: tuple; (string, None/dict, None/dict) @return: Tuple containing boot ID, volume group information and hypervisor information """ bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n") - vg_info = _GetNamedNodeInfo(vg_names, _GetVgInfo) + vg_info = _GetNamedNodeInfo(vg_names, (lambda vg: _GetVgInfo(vg, excl_stor))) hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo) return (bootid, vg_info, hv_info) +def _CheckExclusivePvs(pvi_list): + """Check that PVs are not shared among LVs + + @type pvi_list: list of L{objects.LvmPvInfo} objects + @param pvi_list: information about the PVs + + @rtype: list of tuples (string, list of strings) + @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...]) + + """ + res = [] + for pvi in pvi_list: + if len(pvi.lv_list) > 1: + res.append((pvi.name, pvi.lv_list)) + return res + + def VerifyNode(what, cluster_name): """Verify the status of the local node. @@ -761,9 +809,16 @@ def VerifyNode(what, cluster_name): result[constants.NV_VGLIST] = utils.ListVolumeGroups() if constants.NV_PVLIST in what and vm_capable: - result[constants.NV_PVLIST] = \ - bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST], - filter_allocatable=False) + check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what + val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST], + filter_allocatable=False, + include_lvs=check_exclusive_pvs) + if check_exclusive_pvs: + result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val) + for pvi in val: + # Avoid sending useless data on the wire + pvi.lv_list = [] + result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val) if constants.NV_VERSION in what: result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION, @@ -1190,9 +1245,16 @@ def RunRenameInstance(instance, old_name, debug): " log file:\n%s", result.fail_reason, "\n".join(lines), log=False) -def _GetBlockDevSymlinkPath(instance_name, idx): - return utils.PathJoin(pathutils.DISK_LINKS_DIR, "%s%s%d" % - (instance_name, constants.DISK_SEPARATOR, idx)) +def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None): + """Returns symlink path for block device. + + """ + if _dir is None: + _dir = pathutils.DISK_LINKS_DIR + + return utils.PathJoin(_dir, + ("%s%s%s" % + (instance_name, constants.DISK_SEPARATOR, idx))) def _SymlinkBlockDev(instance_name, device_path, idx): @@ -1397,7 +1459,8 @@ def InstanceReboot(instance, reboot_type, shutdown_timeout): elif reboot_type == constants.INSTANCE_REBOOT_HARD: try: InstanceShutdown(instance, shutdown_timeout) - return StartInstance(instance, False) + result = StartInstance(instance, False) + return result except errors.HypervisorError, err: _Fail("Failed to hard reboot instance %s: %s", instance.name, err) else: @@ -1548,7 +1611,7 @@ def GetMigrationStatus(instance): _Fail("Failed to get migration status: %s", err, exc=True) -def BlockdevCreate(disk, size, owner, on_primary, info): +def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor): """Creates a block device for an instance. @type disk: L{objects.Disk} @@ -1563,6 +1626,8 @@ def BlockdevCreate(disk, size, owner, on_primary, info): @type info: string @param info: string that will be sent to the physical device creation, used for example to set (LVM) tags on LVs + @type excl_stor: boolean + @param excl_stor: Whether exclusive_storage is active @return: the new unique_id of the device (this can sometime be computed only after creation), or None. On secondary nodes, @@ -1589,7 +1654,7 @@ def BlockdevCreate(disk, size, owner, on_primary, info): clist.append(crdev) try: - device = bdev.Create(disk, clist) + device = bdev.Create(disk, clist, excl_stor) except errors.BlockDeviceError, err: _Fail("Can't create block device: %s", err) @@ -2462,8 +2527,9 @@ def OSEnvironment(instance, inst_os, debug=0): result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK] if nic.nicparams[constants.NIC_LINK]: result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK] - if nic.network: - result["NIC_%d_NETWORK" % idx] = nic.network + if nic.netinfo: + nobj = objects.Network.FromDict(nic.netinfo) + result.update(nobj.HooksDict("NIC_%d_" % idx)) if constants.HV_NIC_TYPE in instance.hvparams: result["NIC_%d_FRONTEND_TYPE" % idx] = \ instance.hvparams[constants.HV_NIC_TYPE] @@ -2476,6 +2542,51 @@ def OSEnvironment(instance, inst_os, debug=0): return result +def DiagnoseExtStorage(top_dirs=None): + """Compute the validity for all ExtStorage Providers. + + @type top_dirs: list + @param top_dirs: the list of directories in which to + search (if not given defaults to + L{pathutils.ES_SEARCH_PATH}) + @rtype: list of L{objects.ExtStorage} + @return: a list of tuples (name, path, status, diagnose, parameters) + for all (potential) ExtStorage Providers under all + search paths, where: + - name is the (potential) ExtStorage Provider + - path is the full path to the ExtStorage Provider + - status True/False is the validity of the ExtStorage Provider + - diagnose is the error message for an invalid ExtStorage Provider, + otherwise empty + - parameters is a list of (name, help) parameters, if any + + """ + if top_dirs is None: + top_dirs = pathutils.ES_SEARCH_PATH + + result = [] + for dir_name in top_dirs: + if os.path.isdir(dir_name): + try: + f_names = utils.ListVisibleFiles(dir_name) + except EnvironmentError, err: + logging.exception("Can't list the ExtStorage directory %s: %s", + dir_name, err) + break + for name in f_names: + es_path = utils.PathJoin(dir_name, name) + status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name) + if status: + diagnose = "" + parameters = es_inst.supported_parameters + else: + diagnose = es_inst + parameters = [] + result.append((name, es_path, status, diagnose, parameters)) + + return result + + def BlockdevGrow(disk, amount, dryrun, backingstore): """Grow a stack of block devices. @@ -2609,6 +2720,8 @@ def FinalizeExport(instance, snap_disks): config.set(constants.INISECT_INS, "nic%d_mac" % nic_count, "%s" % nic.mac) config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip) + config.set(constants.INISECT_INS, "nic%d_network" % nic_count, + "%s" % nic.network) for param in constants.NICS_PARAMETER_TYPES: config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param), "%s" % nic.nicparams.get(param, None)) @@ -2878,7 +2991,7 @@ def JobQueueUpdate(file_name, content): # Write and replace the file atomically utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid, - gid=getents.masterd_gid) + gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS) def JobQueueRename(old, new): @@ -2902,8 +3015,8 @@ def JobQueueRename(old, new): getents = runtime.GetEnts() - utils.RenameFile(old, new, mkdir=True, mkdir_mode=0700, - dir_uid=getents.masterd_uid, dir_gid=getents.masterd_gid) + utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750, + dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid) def BlockdevClose(instance_name, disks): @@ -3582,8 +3695,8 @@ def PowercycleNode(hypervisor_type): hyper.PowercycleNode() -def _VerifyRemoteCommandName(cmd): - """Verifies a remote command name. +def _VerifyRestrictedCmdName(cmd): + """Verifies a restricted command name. @type cmd: string @param cmd: Command name @@ -3604,8 +3717,8 @@ def _VerifyRemoteCommandName(cmd): return (True, None) -def _CommonRemoteCommandCheck(path, owner): - """Common checks for remote command file system directories and files. +def _CommonRestrictedCmdCheck(path, owner): + """Common checks for restricted command file system directories and files. @type path: string @param path: Path to check @@ -3634,8 +3747,8 @@ def _CommonRemoteCommandCheck(path, owner): return (True, st) -def _VerifyRemoteCommandDirectory(path, _owner=None): - """Verifies remote command directory. +def _VerifyRestrictedCmdDirectory(path, _owner=None): + """Verifies restricted command directory. @type path: string @param path: Path to check @@ -3644,7 +3757,7 @@ def _VerifyRemoteCommandDirectory(path, _owner=None): element is an error message string, otherwise it's C{None} """ - (status, value) = _CommonRemoteCommandCheck(path, _owner) + (status, value) = _CommonRestrictedCmdCheck(path, _owner) if not status: return (False, value) @@ -3655,11 +3768,11 @@ def _VerifyRemoteCommandDirectory(path, _owner=None): return (True, None) -def _VerifyRemoteCommand(path, cmd, _owner=None): - """Verifies a whole remote command and returns its executable filename. +def _VerifyRestrictedCmd(path, cmd, _owner=None): + """Verifies a whole restricted command and returns its executable filename. @type path: string - @param path: Directory containing remote commands + @param path: Directory containing restricted commands @type cmd: string @param cmd: Command name @rtype: tuple; (boolean, string) @@ -3670,7 +3783,7 @@ def _VerifyRemoteCommand(path, cmd, _owner=None): """ executable = utils.PathJoin(path, cmd) - (status, msg) = _CommonRemoteCommandCheck(executable, _owner) + (status, msg) = _CommonRestrictedCmdCheck(executable, _owner) if not status: return (False, msg) @@ -3681,17 +3794,17 @@ def _VerifyRemoteCommand(path, cmd, _owner=None): return (True, executable) -def _PrepareRemoteCommand(path, cmd, - _verify_dir=_VerifyRemoteCommandDirectory, - _verify_name=_VerifyRemoteCommandName, - _verify_cmd=_VerifyRemoteCommand): - """Performs a number of tests on a remote command. +def _PrepareRestrictedCmd(path, cmd, + _verify_dir=_VerifyRestrictedCmdDirectory, + _verify_name=_VerifyRestrictedCmdName, + _verify_cmd=_VerifyRestrictedCmd): + """Performs a number of tests on a restricted command. @type path: string - @param path: Directory containing remote commands + @param path: Directory containing restricted commands @type cmd: string @param cmd: Command name - @return: Same as L{_VerifyRemoteCommand} + @return: Same as L{_VerifyRestrictedCmd} """ # Verify the directory first @@ -3707,15 +3820,15 @@ def _PrepareRemoteCommand(path, cmd, return _verify_cmd(path, cmd) -def RunRemoteCommand(cmd, +def RunRestrictedCmd(cmd, _lock_timeout=_RCMD_LOCK_TIMEOUT, _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE, _path=pathutils.RESTRICTED_COMMANDS_DIR, _sleep_fn=time.sleep, - _prepare_fn=_PrepareRemoteCommand, + _prepare_fn=_PrepareRestrictedCmd, _runcmd_fn=utils.RunCmd, _enabled=constants.ENABLE_RESTRICTED_COMMANDS): - """Executes a remote command after performing strict tests. + """Executes a restricted command after performing strict tests. @type cmd: string @param cmd: Command name @@ -3724,10 +3837,10 @@ def RunRemoteCommand(cmd, @raise RPCFail: In case of an error """ - logging.info("Preparing to run remote command '%s'", cmd) + logging.info("Preparing to run restricted command '%s'", cmd) if not _enabled: - _Fail("Remote commands disabled at configure time") + _Fail("Restricted commands disabled at configure time") lock = None try: @@ -3755,7 +3868,7 @@ def RunRemoteCommand(cmd, # Do not include original error message in returned error _Fail("Executing command '%s' failed" % cmd) elif cmdresult.failed or cmdresult.fail_reason: - _Fail("Remote command '%s' failed: %s; output: %s", + _Fail("Restricted command '%s' failed: %s; output: %s", cmd, cmdresult.fail_reason, cmdresult.output) else: return cmdresult.output @@ -3766,6 +3879,25 @@ def RunRemoteCommand(cmd, lock = None +def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE): + """Creates or removes the watcher pause file. + + @type until: None or number + @param until: Unix timestamp saying until when the watcher shouldn't run + + """ + if until is None: + logging.info("Received request to no longer pause watcher") + utils.RemoveFile(_filename) + else: + logging.info("Received request to pause watcher until %s", until) + + if not ht.TNumber(until): + _Fail("Duration must be numeric") + + utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644) + + class HooksRunner(object): """Hook runner.