X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/a7359d915896a9be2fe9b4adac08265e64f1c236..2492231f7381ce33f2164322f799c84b2d7cceef:/lib/config.py diff --git a/lib/config.py b/lib/config.py index c72ca62..7d81b37 100644 --- a/lib/config.py +++ b/lib/config.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007, 2008, 2009, 2010 Google Inc. +# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 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 @@ -88,8 +88,8 @@ class TemporaryReservationManager: def Reserve(self, ec_id, resource): if self.Reserved(resource): - raise errors.ReservationError("Duplicate reservation for resource: %s." % - (resource)) + raise errors.ReservationError("Duplicate reservation for resource '%s'" + % str(resource)) if ec_id not in self._ec_reserved: self._ec_reserved[ec_id] = set([resource]) else: @@ -125,6 +125,13 @@ class TemporaryReservationManager: return new_resource +def _MatchNameComponentIgnoreCase(short_name, names): + """Wrapper around L{utils.text.MatchNameComponent}. + + """ + return utils.MatchNameComponent(short_name, names, case_sensitive=False) + + class ConfigWriter: """The interface to the cluster configuration. @@ -179,6 +186,18 @@ class ConfigWriter: return mac @locking.ssynchronized(_config_lock, shared=1) + def GetNdParams(self, node): + """Get the node params populated with cluster defaults. + + @type node: L{object.Node} + @param node: The node we want to know the params for + @return: A dict with the filled in node params + + """ + nodegroup = self._UnlockedGetNodeGroup(node.group) + return self._config_data.cluster.FillND(node, nodegroup) + + @locking.ssynchronized(_config_lock, shared=1) def GenerateMAC(self, ec_id): """Generate a MAC for an instance. @@ -200,7 +219,7 @@ class ConfigWriter: if mac in all_macs: raise errors.ReservationError("mac already in use") else: - self._temporary_macs.Reserve(mac, ec_id) + self._temporary_macs.Reserve(ec_id, mac) @locking.ssynchronized(_config_lock, shared=1) def ReserveLV(self, lv_name, ec_id): @@ -214,7 +233,7 @@ class ConfigWriter: if lv_name in all_lvs: raise errors.ReservationError("LV already in use") else: - self._temporary_lvs.Reserve(lv_name, ec_id) + self._temporary_lvs.Reserve(ec_id, lv_name) @locking.ssynchronized(_config_lock, shared=1) def GenerateDRBDSecret(self, ec_id): @@ -355,29 +374,52 @@ class ConfigWriter: configuration errors """ + # pylint: disable-msg=R0914 result = [] seen_macs = [] ports = {} data = self._config_data + cluster = data.cluster seen_lids = [] seen_pids = [] # global cluster checks - if not data.cluster.enabled_hypervisors: + if not cluster.enabled_hypervisors: result.append("enabled hypervisors list doesn't have any entries") - invalid_hvs = set(data.cluster.enabled_hypervisors) - constants.HYPER_TYPES + invalid_hvs = set(cluster.enabled_hypervisors) - constants.HYPER_TYPES if invalid_hvs: result.append("enabled hypervisors contains invalid entries: %s" % invalid_hvs) - missing_hvp = (set(data.cluster.enabled_hypervisors) - - set(data.cluster.hvparams.keys())) + missing_hvp = (set(cluster.enabled_hypervisors) - + set(cluster.hvparams.keys())) if missing_hvp: result.append("hypervisor parameters missing for the enabled" " hypervisor(s) %s" % utils.CommaJoin(missing_hvp)) - if data.cluster.master_node not in data.nodes: + if cluster.master_node not in data.nodes: result.append("cluster has invalid primary node '%s'" % - data.cluster.master_node) + cluster.master_node) + + def _helper(owner, attr, value, template): + try: + utils.ForceDictType(value, template) + except errors.GenericError, err: + result.append("%s has invalid %s: %s" % (owner, attr, err)) + + def _helper_nic(owner, params): + try: + objects.NIC.CheckParameterSyntax(params) + except errors.ConfigurationError, err: + result.append("%s has invalid nicparams: %s" % (owner, err)) + + # check cluster parameters + _helper("cluster", "beparams", cluster.SimpleFillBE({}), + constants.BES_PARAMETER_TYPES) + _helper("cluster", "nicparams", cluster.SimpleFillNIC({}), + constants.NICS_PARAMETER_TYPES) + _helper_nic("cluster", cluster.SimpleFillNIC({})) + _helper("cluster", "ndparams", cluster.SimpleFillND({}), + constants.NDS_PARAMETER_TYPES) # per-instance checks for instance_name in data.instances: @@ -398,6 +440,17 @@ class ConfigWriter: (instance_name, idx, nic.mac)) else: seen_macs.append(nic.mac) + if nic.nicparams: + filled = cluster.SimpleFillNIC(nic.nicparams) + owner = "instance %s nic %d" % (instance.name, idx) + _helper(owner, "nicparams", + filled, constants.NICS_PARAMETER_TYPES) + _helper_nic(owner, filled) + + # parameter checks + if instance.beparams: + _helper("instance %s" % instance.name, "beparams", + cluster.FillBE(instance), constants.BES_PARAMETER_TYPES) # gather the drbd ports for duplicate checks for dsk in instance.disks: @@ -420,7 +473,7 @@ class ConfigWriter: result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids)) # cluster-wide pool of free ports - for free_port in data.cluster.tcpudp_port_pool: + for free_port in cluster.tcpudp_port_pool: if free_port not in ports: ports[free_port] = [] ports[free_port].append(("cluster", "port marked as free")) @@ -436,11 +489,11 @@ class ConfigWriter: # highest used tcp port check if keys: - if keys[-1] > data.cluster.highest_used_port: + if keys[-1] > cluster.highest_used_port: result.append("Highest used port mismatch, saved %s, computed %s" % - (data.cluster.highest_used_port, keys[-1])) + (cluster.highest_used_port, keys[-1])) - if not data.nodes[data.cluster.master_node].master_candidate: + if not data.nodes[cluster.master_node].master_candidate: result.append("Master node is not a master candidate") # master candidate checks @@ -459,21 +512,33 @@ class ConfigWriter: " drain=%s, offline=%s" % (node.name, node.master_candidate, node.drained, node.offline)) + if node.group not in data.nodegroups: + result.append("Node '%s' has invalid group '%s'" % + (node.name, node.group)) + else: + _helper("node %s" % node.name, "ndparams", + cluster.FillND(node, data.nodegroups[node.group]), + constants.NDS_PARAMETER_TYPES) # nodegroups checks nodegroups_names = set() for nodegroup_uuid in data.nodegroups: nodegroup = data.nodegroups[nodegroup_uuid] if nodegroup.uuid != nodegroup_uuid: - result.append("nodegroup '%s' (uuid: '%s') indexed by wrong uuid '%s'" + result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'" % (nodegroup.name, nodegroup.uuid, nodegroup_uuid)) if utils.UUID_RE.match(nodegroup.name.lower()): - result.append("nodegroup '%s' (uuid: '%s') has uuid-like name" % + result.append("node group '%s' (uuid: '%s') has uuid-like name" % (nodegroup.name, nodegroup.uuid)) if nodegroup.name in nodegroups_names: - result.append("duplicate nodegroup name '%s'" % nodegroup.name) + result.append("duplicate node group name '%s'" % nodegroup.name) else: nodegroups_names.add(nodegroup.name) + if nodegroup.ndparams: + _helper("group %s" % nodegroup.name, "ndparams", + cluster.SimpleFillND(nodegroup.ndparams), + constants.NDS_PARAMETER_TYPES) + # drbd minors check _, duplicates = self._UnlockedComputeDRBDMap() @@ -482,13 +547,13 @@ class ConfigWriter: " %s and %s" % (minor, node, instance_a, instance_b)) # IP checks - default_nicparams = data.cluster.nicparams[constants.PP_DEFAULT] + default_nicparams = cluster.nicparams[constants.PP_DEFAULT] ips = {} def _AddIpAddress(ip, name): ips.setdefault(ip, []).append(name) - _AddIpAddress(data.cluster.master_ip, "cluster_ip") + _AddIpAddress(cluster.master_ip, "cluster_ip") for node in data.nodes.values(): _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name) @@ -820,6 +885,13 @@ class ConfigWriter: return self._config_data.cluster.file_storage_dir @locking.ssynchronized(_config_lock, shared=1) + def GetSharedFileStorageDir(self): + """Get the shared file storage dir for this cluster. + + """ + return self._config_data.cluster.shared_file_storage_dir + + @locking.ssynchronized(_config_lock, shared=1) def GetHypervisorType(self): """Get the hypervisor type for this cluster. @@ -852,8 +924,76 @@ class ConfigWriter: """ return self._config_data.cluster.primary_ip_family - @locking.ssynchronized(_config_lock, shared=1) - def LookupNodeGroup(self, target): + @locking.ssynchronized(_config_lock) + def AddNodeGroup(self, group, ec_id, check_uuid=True): + """Add a node group to the configuration. + + This method calls group.UpgradeConfig() to fill any missing attributes + according to their default values. + + @type group: L{objects.NodeGroup} + @param group: the NodeGroup object to add + @type ec_id: string + @param ec_id: unique id for the job to use when creating a missing UUID + @type check_uuid: bool + @param check_uuid: add an UUID to the group if it doesn't have one or, if + it does, ensure that it does not exist in the + configuration already + + """ + self._UnlockedAddNodeGroup(group, ec_id, check_uuid) + self._WriteConfig() + + def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid): + """Add a node group to the configuration. + + """ + logging.info("Adding node group %s to configuration", group.name) + + # Some code might need to add a node group with a pre-populated UUID + # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass + # the "does this UUID" exist already check. + if check_uuid: + self._EnsureUUID(group, ec_id) + + try: + existing_uuid = self._UnlockedLookupNodeGroup(group.name) + except errors.OpPrereqError: + pass + else: + raise errors.OpPrereqError("Desired group name '%s' already exists as a" + " node group (UUID: %s)" % + (group.name, existing_uuid), + errors.ECODE_EXISTS) + + group.serial_no = 1 + group.ctime = group.mtime = time.time() + group.UpgradeConfig() + + self._config_data.nodegroups[group.uuid] = group + self._config_data.cluster.serial_no += 1 + + @locking.ssynchronized(_config_lock) + def RemoveNodeGroup(self, group_uuid): + """Remove a node group from the configuration. + + @type group_uuid: string + @param group_uuid: the UUID of the node group to remove + + """ + logging.info("Removing node group %s from configuration", group_uuid) + + if group_uuid not in self._config_data.nodegroups: + raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid) + + assert len(self._config_data.nodegroups) != 1, \ + "Group '%s' is the only group, cannot be removed" % group_uuid + + del self._config_data.nodegroups[group_uuid] + self._config_data.cluster.serial_no += 1 + self._WriteConfig() + + def _UnlockedLookupNodeGroup(self, target): """Lookup a node group's UUID. @type target: string or None @@ -865,7 +1005,7 @@ class ConfigWriter: """ if target is None: if len(self._config_data.nodegroups) != 1: - raise errors.OpPrereqError("More than one nodegroup exists. Target" + raise errors.OpPrereqError("More than one node group exists. Target" " group must be specified explicitely.") else: return self._config_data.nodegroups.keys()[0] @@ -874,10 +1014,24 @@ class ConfigWriter: for nodegroup in self._config_data.nodegroups.values(): if nodegroup.name == target: return nodegroup.uuid - raise errors.OpPrereqError("Nodegroup '%s' not found", target) + raise errors.OpPrereqError("Node group '%s' not found" % target, + errors.ECODE_NOENT) @locking.ssynchronized(_config_lock, shared=1) - def GetNodeGroup(self, uuid): + def LookupNodeGroup(self, target): + """Lookup a node group's UUID. + + This function is just a wrapper over L{_UnlockedLookupNodeGroup}. + + @type target: string or None + @param target: group name or UUID or None to look for the default + @rtype: string + @return: nodegroup UUID + + """ + return self._UnlockedLookupNodeGroup(target) + + def _UnlockedGetNodeGroup(self, uuid): """Lookup a node group. @type uuid: string @@ -892,6 +1046,18 @@ class ConfigWriter: return self._config_data.nodegroups[uuid] @locking.ssynchronized(_config_lock, shared=1) + def GetNodeGroup(self, uuid): + """Lookup a node group. + + @type uuid: string + @param uuid: group UUID + @rtype: L{objects.NodeGroup} or None + @return: nodegroup object, or None if not found + + """ + return self._UnlockedGetNodeGroup(uuid) + + @locking.ssynchronized(_config_lock, shared=1) def GetAllNodeGroupsInfo(self): """Get the configuration of all node groups. @@ -905,6 +1071,17 @@ class ConfigWriter: """ return self._config_data.nodegroups.keys() + @locking.ssynchronized(_config_lock, shared=1) + def GetNodeGroupMembersByNodes(self, nodes): + """Get nodes which are member in the same nodegroups as the given nodes. + + """ + ngfn = lambda node_name: self._UnlockedGetNodeInfo(node_name).group + return frozenset(member_name + for node_name in nodes + for member_name in + self._UnlockedGetNodeGroup(ngfn(node_name)).members) + @locking.ssynchronized(_config_lock) def AddInstance(self, instance, ec_id): """Add an instance to the config. @@ -1005,10 +1182,11 @@ class ConfigWriter: if disk.dev_type == constants.LD_FILE: # rename the file paths in logical and physical id file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1])) + disk_fname = "disk%s" % disk.iv_name.split("/")[1] disk.physical_id = disk.logical_id = (disk.logical_id[0], utils.PathJoin(file_storage_dir, inst.name, - disk.iv_name)) + disk_fname)) # Force update of ssconf files self._config_data.cluster.serial_no += 1 @@ -1041,14 +1219,12 @@ class ConfigWriter: """ return self._UnlockedGetInstanceList() - @locking.ssynchronized(_config_lock, shared=1) def ExpandInstanceName(self, short_name): """Attempt to expand an incomplete instance name. """ - return utils.MatchNameComponent(short_name, - self._config_data.instances.keys(), - case_sensitive=False) + # Locking is done in L{ConfigWriter.GetInstanceList} + return _MatchNameComponentIgnoreCase(short_name, self.GetInstanceList()) def _UnlockedGetInstanceInfo(self, instance_name): """Returns information about an instance. @@ -1078,6 +1254,25 @@ class ConfigWriter: return self._UnlockedGetInstanceInfo(instance_name) @locking.ssynchronized(_config_lock, shared=1) + def GetInstanceNodeGroups(self, instance_name, primary_only=False): + """Returns set of node group UUIDs for instance's nodes. + + @rtype: frozenset + + """ + instance = self._UnlockedGetInstanceInfo(instance_name) + if not instance: + raise errors.ConfigurationError("Unknown instance '%s'" % instance_name) + + if primary_only: + nodes = [instance.primary_node] + else: + nodes = instance.all_nodes + + return frozenset(self._UnlockedGetNodeInfo(node_name).group + for node_name in nodes) + + @locking.ssynchronized(_config_lock, shared=1) def GetAllInstancesInfo(self): """Get the configuration of all instances. @@ -1124,14 +1319,12 @@ class ConfigWriter: self._config_data.cluster.serial_no += 1 self._WriteConfig() - @locking.ssynchronized(_config_lock, shared=1) def ExpandNodeName(self, short_name): - """Attempt to expand an incomplete instance name. + """Attempt to expand an incomplete node name. """ - return utils.MatchNameComponent(short_name, - self._config_data.nodes.keys(), - case_sensitive=False) + # Locking is done in L{ConfigWriter.GetNodeList} + return _MatchNameComponentIgnoreCase(short_name, self.GetNodeList()) def _UnlockedGetNodeInfo(self, node_name): """Get the configuration of a node, as stored in the config. @@ -1183,6 +1376,26 @@ class ConfigWriter: sec.append(inst.name) return (pri, sec) + @locking.ssynchronized(_config_lock, shared=1) + def GetNodeGroupInstances(self, uuid, primary_only=False): + """Get the instances of a node group. + + @param uuid: Node group UUID + @param primary_only: Whether to only consider primary nodes + @rtype: frozenset + @return: List of instance names in node group + + """ + if primary_only: + nodes_fn = lambda inst: [inst.primary_node] + else: + nodes_fn = lambda inst: inst.all_nodes + + return frozenset(inst.name + for inst in self._config_data.instances.values() + for node_name in nodes_fn(inst) + if self._UnlockedGetNodeInfo(node_name).group == uuid) + def _UnlockedGetNodeList(self): """Return the list of nodes which are in the configuration. @@ -1217,6 +1430,15 @@ class ConfigWriter: return self._UnlockedGetOnlineNodeList() @locking.ssynchronized(_config_lock, shared=1) + def GetVmCapableNodeList(self): + """Return the list of nodes which are not vm capable. + + """ + all_nodes = [self._UnlockedGetNodeInfo(node) + for node in self._UnlockedGetNodeList()] + return [node.name for node in all_nodes if node.vm_capable] + + @locking.ssynchronized(_config_lock, shared=1) def GetNonVmCapableNodeList(self): """Return the list of nodes which are not vm capable. @@ -1238,6 +1460,17 @@ class ConfigWriter: for node in self._UnlockedGetNodeList()]) return my_dict + @locking.ssynchronized(_config_lock, shared=1) + def GetNodeGroupsFromNodes(self, nodes): + """Returns groups for a list of nodes. + + @type nodes: list of string + @param nodes: List of node names + @rtype: frozenset + + """ + return frozenset(self._UnlockedGetNodeInfo(name).group for name in nodes) + def _UnlockedGetMasterCandidateStats(self, exceptions=None): """Get the number of current and maximum desired and possible candidates. @@ -1410,13 +1643,10 @@ class ConfigWriter: item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID) modified = True if not self._config_data.nodegroups: - default_nodegroup_uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID) - default_nodegroup = objects.NodeGroup( - uuid=default_nodegroup_uuid, - name="default", - members=[], - ) - self._config_data.nodegroups[default_nodegroup_uuid] = default_nodegroup + default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME + default_nodegroup = objects.NodeGroup(name=default_nodegroup_name, + members=[]) + self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True) modified = True for node in self._config_data.nodes.values(): if not node.group: @@ -1575,10 +1805,11 @@ class ConfigWriter: self._config_data.nodegroups.values()] nodegroups_data = fn(utils.NiceSort(nodegroups)) - return { + ssconf_values = { constants.SS_CLUSTER_NAME: cluster.cluster_name, constants.SS_CLUSTER_TAGS: cluster_tags, constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir, + constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir, constants.SS_MASTER_CANDIDATES: mc_data, constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data, constants.SS_MASTER_IP: cluster.master_ip, @@ -1597,6 +1828,13 @@ class ConfigWriter: constants.SS_UID_POOL: uid_pool, constants.SS_NODEGROUPS: nodegroups_data, } + bad_values = [(k, v) for k, v in ssconf_values.items() + if not isinstance(v, (str, basestring))] + if bad_values: + err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values) + raise errors.ConfigurationError("Some ssconf key(s) have non-string" + " values: %s" % err) + return ssconf_values @locking.ssynchronized(_config_lock, shared=1) def GetSsconfValues(self): @@ -1688,6 +1926,8 @@ class ConfigWriter: update_serial = True elif isinstance(target, objects.Instance): test = target in self._config_data.instances.values() + elif isinstance(target, objects.NodeGroup): + test = target in self._config_data.nodegroups.values() else: raise errors.ProgrammerError("Invalid object type (%s) passed to" " ConfigWriter.Update" % type(target))