X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/c6d58a2b8290c951c1fa173d5f0f0a901376cbc6..97efde459ccd6770faab6c2a81db7a9ee60104d3:/lib/cmdlib.py diff --git a/lib/cmdlib.py b/lib/cmdlib.py index 86f6da2..dee811d 100644 --- a/lib/cmdlib.py +++ b/lib/cmdlib.py @@ -37,11 +37,10 @@ from ganeti import logger from ganeti import utils from ganeti import errors from ganeti import hypervisor -from ganeti import config +from ganeti import locking from ganeti import constants from ganeti import objects from ganeti import opcodes -from ganeti import ssconf from ganeti import serializer @@ -49,21 +48,27 @@ class LogicalUnit(object): """Logical Unit base class. Subclasses must follow these rules: - - implement CheckPrereq which also fills in the opcode instance - with all the fields (even if as None) + - implement ExpandNames + - implement CheckPrereq - implement Exec - implement BuildHooksEnv - redefine HPATH and HTYPE - - optionally redefine their run requirements (REQ_MASTER); note that all - commands require root permissions + - optionally redefine their run requirements: + REQ_MASTER: the LU needs to run on the master node + REQ_WSSTORE: the LU needs a writable SimpleStore + REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively + + Note that all commands require root permissions. """ HPATH = None HTYPE = None _OP_REQP = [] REQ_MASTER = True + REQ_WSSTORE = False + REQ_BGL = True - def __init__(self, processor, op, cfg, sstore): + def __init__(self, processor, op, context, sstore): """Constructor for LogicalUnit. This needs to be overriden in derived classes in order to check op @@ -72,8 +77,13 @@ class LogicalUnit(object): """ self.proc = processor self.op = op - self.cfg = cfg + self.cfg = context.cfg self.sstore = sstore + self.context = context + self.needed_locks = None + self.share_locks = dict(((i, 0) for i in locking.LEVELS)) + # Used to force good behavior when calling helper functions + self.recalculate_locks = {} self.__ssh = None for attr_name in self._OP_REQP: @@ -82,7 +92,7 @@ class LogicalUnit(object): raise errors.OpPrereqError("Required parameter '%s' missing" % attr_name) - if not cfg.IsCluster(): + if not self.cfg.IsCluster(): raise errors.OpPrereqError("Cluster not initialized yet," " use 'gnt-cluster init' first.") if self.REQ_MASTER: @@ -101,6 +111,68 @@ class LogicalUnit(object): ssh = property(fget=__GetSSH) + def ExpandNames(self): + """Expand names for this LU. + + This method is called before starting to execute the opcode, and it should + update all the parameters of the opcode to their canonical form (e.g. a + short node name must be fully expanded after this method has successfully + completed). This way locking, hooks, logging, ecc. can work correctly. + + LUs which implement this method must also populate the self.needed_locks + member, as a dict with lock levels as keys, and a list of needed lock names + as values. Rules: + - Use an empty dict if you don't need any lock + - If you don't need any lock at a particular level omit that level + - Don't put anything for the BGL level + - If you want all locks at a level use None as a value + (this reflects what LockSet does, and will be replaced before + CheckPrereq with the full list of nodes that have been locked) + + If you need to share locks (rather than acquire them exclusively) at one + level you can modify self.share_locks, setting a true value (usually 1) for + that level. By default locks are not shared. + + Examples: + # Acquire all nodes and one instance + self.needed_locks = { + locking.LEVEL_NODE: None, + locking.LEVEL_INSTANCES: ['instance1.example.tld'], + } + # Acquire just two nodes + self.needed_locks = { + locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'], + } + # Acquire no locks + self.needed_locks = {} # No, you can't leave it to the default value None + + """ + # The implementation of this method is mandatory only if the new LU is + # concurrent, so that old LUs don't need to be changed all at the same + # time. + if self.REQ_BGL: + self.needed_locks = {} # Exclusive LUs don't need locks. + else: + raise NotImplementedError + + def DeclareLocks(self, level): + """Declare LU locking needs for a level + + While most LUs can just declare their locking needs at ExpandNames time, + sometimes there's the need to calculate some locks after having acquired + the ones before. This function is called just before acquiring locks at a + particular level, but after acquiring the ones at lower levels, and permits + such calculations. It can be used to modify self.needed_locks, and by + default it does nothing. + + This function is only called if you have something already set in + self.needed_locks for the level. + + @param level: Locking level which is going to be locked + @type level: member of ganeti.locking.LEVELS + + """ + def CheckPrereq(self): """Check prerequisites for this LU. @@ -113,9 +185,7 @@ class LogicalUnit(object): not fulfilled. Its return value is ignored. This method should also update all the parameters of the opcode to - their canonical form; e.g. a short node name must be fully - expanded after this method has successfully completed (so that - hooks, logging, etc. work correctly). + their canonical form if it hasn't been done by ExpandNames before. """ raise NotImplementedError @@ -170,33 +240,75 @@ class LogicalUnit(object): """ return lu_result + def _ExpandAndLockInstance(self): + """Helper function to expand and lock an instance. -class NoHooksLU(LogicalUnit): - """Simple LU which runs no hooks. + Many LUs that work on an instance take its name in self.op.instance_name + and need to expand it and then declare the expanded name for locking. This + function does it, and then updates self.op.instance_name to the expanded + name. It also initializes needed_locks as a dict, if this hasn't been done + before. - This LU is intended as a parent for other LogicalUnits which will - run no hooks, in order to reduce duplicate code. + """ + if self.needed_locks is None: + self.needed_locks = {} + else: + assert locking.LEVEL_INSTANCE not in self.needed_locks, \ + "_ExpandAndLockInstance called with instance-level locks set" + expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name) + if expanded_name is None: + raise errors.OpPrereqError("Instance '%s' not known" % + self.op.instance_name) + self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name + self.op.instance_name = expanded_name - """ - HPATH = None - HTYPE = None + def _LockInstancesNodes(self): + """Helper function to declare instances' nodes for locking. + This function should be called after locking one or more instances to lock + their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE] + with all primary or secondary nodes for instances already locked and + present in self.needed_locks[locking.LEVEL_INSTANCE]. -def _AddHostToEtcHosts(hostname): - """Wrapper around utils.SetEtcHostsEntry. + It should be called from DeclareLocks, and for safety only works if + self.recalculate_locks[locking.LEVEL_NODE] is set. - """ - hi = utils.HostInfo(name=hostname) - utils.SetEtcHostsEntry(constants.ETC_HOSTS, hi.ip, hi.name, [hi.ShortName()]) + In the future it may grow parameters to just lock some instance's nodes, or + to just lock primaries or secondary nodes, if needed. + + If should be called in DeclareLocks in a way similar to: + + if level == locking.LEVEL_NODE: + self._LockInstancesNodes() + + """ + assert locking.LEVEL_NODE in self.recalculate_locks, \ + "_LockInstancesNodes helper function called with no nodes to recalculate" + + # TODO: check if we're really been called with the instance locks held + + # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the + # future we might want to have different behaviors depending on the value + # of self.recalculate_locks[locking.LEVEL_NODE] + wanted_nodes = [] + for instance_name in self.needed_locks[locking.LEVEL_INSTANCE]: + instance = self.context.cfg.GetInstanceInfo(instance_name) + wanted_nodes.append(instance.primary_node) + wanted_nodes.extend(instance.secondary_nodes) + self.needed_locks[locking.LEVEL_NODE] = wanted_nodes + + del self.recalculate_locks[locking.LEVEL_NODE] -def _RemoveHostFromEtcHosts(hostname): - """Wrapper around utils.RemoveEtcHostsEntry. +class NoHooksLU(LogicalUnit): + """Simple LU which runs no hooks. + + This LU is intended as a parent for other LogicalUnits which will + run no hooks, in order to reduce duplicate code. """ - hi = utils.HostInfo(name=hostname) - utils.RemoveEtcHostsEntry(constants.ETC_HOSTS, hi.name) - utils.RemoveEtcHostsEntry(constants.ETC_HOSTS, hi.ShortName()) + HPATH = None + HTYPE = None def _GetWantedNodes(lu, nodes): @@ -322,85 +434,6 @@ def _BuildInstanceHookEnvByObject(instance, override=None): return _BuildInstanceHookEnv(**args) -def _HasValidVG(vglist, vgname): - """Checks if the volume group list is valid. - - A non-None return value means there's an error, and the return value - is the error message. - - """ - vgsize = vglist.get(vgname, None) - if vgsize is None: - return "volume group '%s' missing" % vgname - elif vgsize < 20480: - return ("volume group '%s' too small (20480MiB required, %dMib found)" % - (vgname, vgsize)) - return None - - -def _InitSSHSetup(node): - """Setup the SSH configuration for the cluster. - - - This generates a dsa keypair for root, adds the pub key to the - permitted hosts and adds the hostkey to its own known hosts. - - Args: - node: the name of this host as a fqdn - - """ - priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) - - for name in priv_key, pub_key: - if os.path.exists(name): - utils.CreateBackup(name) - utils.RemoveFile(name) - - result = utils.RunCmd(["ssh-keygen", "-t", "dsa", - "-f", priv_key, - "-q", "-N", ""]) - if result.failed: - raise errors.OpExecError("Could not generate ssh keypair, error %s" % - result.output) - - f = open(pub_key, 'r') - try: - utils.AddAuthorizedKey(auth_keys, f.read(8192)) - finally: - f.close() - - -def _InitGanetiServerSetup(ss): - """Setup the necessary configuration for the initial node daemon. - - This creates the nodepass file containing the shared password for - the cluster and also generates the SSL certificate. - - """ - # Create pseudo random password - randpass = sha.new(os.urandom(64)).hexdigest() - # and write it into sstore - ss.SetKey(ss.SS_NODED_PASS, randpass) - - result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024", - "-days", str(365*5), "-nodes", "-x509", - "-keyout", constants.SSL_CERT_FILE, - "-out", constants.SSL_CERT_FILE, "-batch"]) - if result.failed: - raise errors.OpExecError("could not generate server ssl cert, command" - " %s had exitcode %s and error message %s" % - (result.cmd, result.exit_code, result.output)) - - os.chmod(constants.SSL_CERT_FILE, 0400) - - result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"]) - - if result.failed: - raise errors.OpExecError("Could not start the node daemon, command %s" - " had exitcode %s and error %s" % - (result.cmd, result.exit_code, result.output)) - - def _CheckInstanceBridgesExist(instance): """Check that the brigdes needed by an instance exist. @@ -413,161 +446,6 @@ def _CheckInstanceBridgesExist(instance): (brlist, instance.primary_node)) -class LUInitCluster(LogicalUnit): - """Initialise the cluster. - - """ - HPATH = "cluster-init" - HTYPE = constants.HTYPE_CLUSTER - _OP_REQP = ["cluster_name", "hypervisor_type", "mac_prefix", - "def_bridge", "master_netdev", "file_storage_dir"] - REQ_CLUSTER = False - - def BuildHooksEnv(self): - """Build hooks env. - - Notes: Since we don't require a cluster, we must manually add - ourselves in the post-run node list. - - """ - env = {"OP_TARGET": self.op.cluster_name} - return env, [], [self.hostname.name] - - def CheckPrereq(self): - """Verify that the passed name is a valid one. - - """ - if config.ConfigWriter.IsCluster(): - raise errors.OpPrereqError("Cluster is already initialised") - - if self.op.hypervisor_type == constants.HT_XEN_HVM31: - if not os.path.exists(constants.VNC_PASSWORD_FILE): - raise errors.OpPrereqError("Please prepare the cluster VNC" - "password file %s" % - constants.VNC_PASSWORD_FILE) - - self.hostname = hostname = utils.HostInfo() - - if hostname.ip.startswith("127."): - raise errors.OpPrereqError("This host's IP resolves to the private" - " range (%s). Please fix DNS or %s." % - (hostname.ip, constants.ETC_HOSTS)) - - if not utils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT, - source=constants.LOCALHOST_IP_ADDRESS): - raise errors.OpPrereqError("Inconsistency: this host's name resolves" - " to %s,\nbut this ip address does not" - " belong to this host." - " Aborting." % hostname.ip) - - self.clustername = clustername = utils.HostInfo(self.op.cluster_name) - - if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, - timeout=5): - raise errors.OpPrereqError("Cluster IP already active. Aborting.") - - secondary_ip = getattr(self.op, "secondary_ip", None) - if secondary_ip and not utils.IsValidIP(secondary_ip): - raise errors.OpPrereqError("Invalid secondary ip given") - if (secondary_ip and - secondary_ip != hostname.ip and - (not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, - source=constants.LOCALHOST_IP_ADDRESS))): - raise errors.OpPrereqError("You gave %s as secondary IP," - " but it does not belong to this host." % - secondary_ip) - self.secondary_ip = secondary_ip - - if not hasattr(self.op, "vg_name"): - self.op.vg_name = None - # if vg_name not None, checks if volume group is valid - if self.op.vg_name: - vgstatus = _HasValidVG(utils.ListVolumeGroups(), self.op.vg_name) - if vgstatus: - raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" - " you are not using lvm" % vgstatus) - - self.op.file_storage_dir = os.path.normpath(self.op.file_storage_dir) - - if not os.path.isabs(self.op.file_storage_dir): - raise errors.OpPrereqError("The file storage directory you have is" - " not an absolute path.") - - if not os.path.exists(self.op.file_storage_dir): - try: - os.makedirs(self.op.file_storage_dir, 0750) - except OSError, err: - raise errors.OpPrereqError("Cannot create file storage directory" - " '%s': %s" % - (self.op.file_storage_dir, err)) - - if not os.path.isdir(self.op.file_storage_dir): - raise errors.OpPrereqError("The file storage directory '%s' is not" - " a directory." % self.op.file_storage_dir) - - if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", - self.op.mac_prefix): - raise errors.OpPrereqError("Invalid mac prefix given '%s'" % - self.op.mac_prefix) - - if self.op.hypervisor_type not in constants.HYPER_TYPES: - raise errors.OpPrereqError("Invalid hypervisor type given '%s'" % - self.op.hypervisor_type) - - result = utils.RunCmd(["ip", "link", "show", "dev", self.op.master_netdev]) - if result.failed: - raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % - (self.op.master_netdev, - result.output.strip())) - - if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and - os.access(constants.NODE_INITD_SCRIPT, os.X_OK)): - raise errors.OpPrereqError("Init.d script '%s' missing or not" - " executable." % constants.NODE_INITD_SCRIPT) - - def Exec(self, feedback_fn): - """Initialize the cluster. - - """ - clustername = self.clustername - hostname = self.hostname - - # set up the simple store - self.sstore = ss = ssconf.SimpleStore() - ss.SetKey(ss.SS_HYPERVISOR, self.op.hypervisor_type) - ss.SetKey(ss.SS_MASTER_NODE, hostname.name) - ss.SetKey(ss.SS_MASTER_IP, clustername.ip) - ss.SetKey(ss.SS_MASTER_NETDEV, self.op.master_netdev) - ss.SetKey(ss.SS_CLUSTER_NAME, clustername.name) - ss.SetKey(ss.SS_FILE_STORAGE_DIR, self.op.file_storage_dir) - ss.SetKey(ss.SS_CONFIG_VERSION, constants.CONFIG_VERSION) - - # set up the inter-node password and certificate - _InitGanetiServerSetup(ss) - - # start the master ip - rpc.call_node_start_master(hostname.name) - - # set up ssh config and /etc/hosts - f = open(constants.SSH_HOST_RSA_PUB, 'r') - try: - sshline = f.read() - finally: - f.close() - sshkey = sshline.split(" ")[1] - - _AddHostToEtcHosts(hostname.name) - _InitSSHSetup(hostname.name) - - # init of cluster config file - self.cfg = cfgw = config.ConfigWriter() - cfgw.InitConfig(hostname.name, hostname.ip, self.secondary_ip, - sshkey, self.op.mac_prefix, - self.op.vg_name, self.op.def_bridge) - - ssh.WriteKnownHostsFile(cfgw, ss, constants.SSH_KNOWN_HOSTS_FILE) - - class LUDestroyCluster(NoHooksLU): """Logical unit for destroying the cluster. @@ -598,7 +476,7 @@ class LUDestroyCluster(NoHooksLU): """ master = self.sstore.GetMasterNode() - if not rpc.call_node_stop_master(master): + if not rpc.call_node_stop_master(master, False): raise errors.OpExecError("Could not disable the master role") priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) utils.CreateBackup(priv_key) @@ -649,7 +527,8 @@ class LUVerifyCluster(LogicalUnit): (node,)) bad = True else: - vgstatus = _HasValidVG(vglist, self.cfg.GetVGName()) + vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(), + constants.MIN_VG_SIZE) if vgstatus: feedback_fn(" - ERROR: %s on node %s" % (vgstatus, node)) bad = True @@ -989,7 +868,8 @@ class LUVerifyCluster(LogicalUnit): lu_result: previous Exec result """ - # We only really run POST phase hooks, and are only interested in their results + # We only really run POST phase hooks, and are only interested in + # their results if phase == constants.HOOKS_PHASE_POST: # Used to change hooks' output to proper indentation indent_re = re.compile('^', re.M) @@ -1099,6 +979,7 @@ class LURenameCluster(LogicalUnit): HPATH = "cluster-rename" HTYPE = constants.HTYPE_CLUSTER _OP_REQP = ["name"] + REQ_WSSTORE = True def BuildHooksEnv(self): """Build hooks env. @@ -1125,8 +1006,7 @@ class LURenameCluster(LogicalUnit): raise errors.OpPrereqError("Neither the name nor the IP address of the" " cluster has changed") if new_ip != old_ip: - result = utils.RunCmd(["fping", "-q", new_ip]) - if not result.failed: + if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT): raise errors.OpPrereqError("The given cluster IP address (%s) is" " reachable on the network. Aborting." % new_ip) @@ -1143,7 +1023,7 @@ class LURenameCluster(LogicalUnit): # shutdown the master IP master = ss.GetMasterNode() - if not rpc.call_node_stop_master(master): + if not rpc.call_node_stop_master(master, False): raise errors.OpExecError("Could not disable the master role") try: @@ -1166,7 +1046,7 @@ class LURenameCluster(LogicalUnit): logger.Error("copy of file %s to node %s failed" % (fname, to_node)) finally: - if not rpc.call_node_start_master(master): + if not rpc.call_node_start_master(master, False): logger.Error("Could not re-enable the master role on the master," " please restart manually.") @@ -1228,7 +1108,8 @@ class LUSetClusterParams(LogicalUnit): node_list = self.cfg.GetNodeList() vglist = rpc.call_vg_list(node_list) for node in node_list: - vgstatus = _HasValidVG(vglist[node], self.op.vg_name) + vgstatus = utils.CheckVolumeGroupSize(vglist[node], self.op.vg_name, + constants.MIN_VG_SIZE) if vgstatus: raise errors.OpPrereqError("Error on node '%s': %s" % (node, vgstatus)) @@ -1295,15 +1176,7 @@ def _WaitForSync(cfgw, instance, proc, oneshot=False, unlock=False): if done or oneshot: break - if unlock: - #utils.Unlock('cmd') - pass - try: - time.sleep(min(60, max_time)) - finally: - if unlock: - #utils.Lock('cmd') - pass + time.sleep(min(60, max_time)) if done: proc.LogInfo("Instance %s's disks are in sync." % instance.name) @@ -1431,7 +1304,7 @@ class LURemoveNode(LogicalUnit): """Build hooks env. This doesn't run on the target node in the pre phase as a failed - node would not allows itself to run. + node would then be impossible to remove. """ env = { @@ -1485,13 +1358,13 @@ class LURemoveNode(LogicalUnit): rpc.call_node_leave_cluster(node.name) - self.ssh.Run(node.name, 'root', "%s stop" % constants.NODE_INITD_SCRIPT) - logger.Info("Removing node %s from config" % node.name) self.cfg.RemoveNode(node.name) + # Remove the node from the Ganeti Lock Manager + self.context.glm.remove(locking.LEVEL_NODE, node.name) - _RemoveHostFromEtcHosts(node.name) + utils.RemoveHostFromEtcHosts(node.name) class LUQueryNodes(NoHooksLU): @@ -1515,7 +1388,7 @@ class LUQueryNodes(NoHooksLU): _CheckOutputFields(static=["name", "pinst_cnt", "sinst_cnt", "pinst_list", "sinst_list", - "pip", "sip"], + "pip", "sip", "tags"], dynamic=self.dynamic_fields, selected=self.op.output_fields) @@ -1586,6 +1459,8 @@ class LUQueryNodes(NoHooksLU): val = node.primary_ip elif field == "sip": val = node.secondary_ip + elif field == "tags": + val = list(node.GetTags()) elif field in self.dynamic_fields: val = live_data[node.name].get(field, None) else: @@ -1767,11 +1642,6 @@ class LUAddNode(LogicalUnit): primary_ip=primary_ip, secondary_ip=secondary_ip) - if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31: - if not os.path.exists(constants.VNC_PASSWORD_FILE): - raise errors.OpPrereqError("Cluster VNC password file %s missing" % - constants.VNC_PASSWORD_FILE) - def Exec(self, feedback_fn): """Adds the new node to the cluster. @@ -1779,46 +1649,7 @@ class LUAddNode(LogicalUnit): new_node = self.new_node node = new_node.name - # set up inter-node password and certificate and restarts the node daemon - gntpass = self.sstore.GetNodeDaemonPassword() - if not re.match('^[a-zA-Z0-9.]{1,64}$', gntpass): - raise errors.OpExecError("ganeti password corruption detected") - f = open(constants.SSL_CERT_FILE) - try: - gntpem = f.read(8192) - finally: - f.close() - # in the base64 pem encoding, neither '!' nor '.' are valid chars, - # so we use this to detect an invalid certificate; as long as the - # cert doesn't contain this, the here-document will be correctly - # parsed by the shell sequence below - if re.search('^!EOF\.', gntpem, re.MULTILINE): - raise errors.OpExecError("invalid PEM encoding in the SSL certificate") - if not gntpem.endswith("\n"): - raise errors.OpExecError("PEM must end with newline") - logger.Info("copy cluster pass to %s and starting the node daemon" % node) - - # and then connect with ssh to set password and start ganeti-noded - # note that all the below variables are sanitized at this point, - # either by being constants or by the checks above - ss = self.sstore - mycommand = ("umask 077 && " - "echo '%s' > '%s' && " - "cat > '%s' << '!EOF.' && \n" - "%s!EOF.\n%s restart" % - (gntpass, ss.KeyToFilename(ss.SS_NODED_PASS), - constants.SSL_CERT_FILE, gntpem, - constants.NODE_INITD_SCRIPT)) - - result = self.ssh.Run(node, 'root', mycommand, batch=False, ask_key=True) - if result.failed: - raise errors.OpExecError("Remote command on node %s, error: %s," - " output: %s" % - (node, result.fail_reason, result.output)) - # check connectivity - time.sleep(4) - result = rpc.call_version([node])[node] if result: if constants.PROTOCOL_VERSION == result: @@ -1853,7 +1684,7 @@ class LUAddNode(LogicalUnit): raise errors.OpExecError("Cannot transfer ssh keys to the new node") # Add node to our /etc/hosts, and add key to known_hosts - _AddHostToEtcHosts(new_node.name) + utils.AddHostToEtcHosts(new_node.name) if new_node.secondary_ip != new_node.primary_ip: if not rpc.call_node_tcp_ping(new_node.name, @@ -1865,12 +1696,22 @@ class LUAddNode(LogicalUnit): " you gave (%s). Please fix and re-run this" " command." % new_node.secondary_ip) - success, msg = self.ssh.VerifyNodeHostname(node) - if not success: - raise errors.OpExecError("Node '%s' claims it has a different hostname" - " than the one the resolver gives: %s." - " Please fix and re-run this command." % - (node, msg)) + node_verify_list = [self.sstore.GetMasterNode()] + node_verify_param = { + 'nodelist': [node], + # TODO: do a node-net-test as well? + } + + result = rpc.call_node_verify(node_verify_list, node_verify_param) + for verifier in node_verify_list: + if not result[verifier]: + raise errors.OpExecError("Cannot communicate with %s's node daemon" + " for remote verification" % verifier) + if result[verifier]['nodelist']: + for failed in result[verifier]['nodelist']: + feedback_fn("ssh/hostname verification failed %s -> %s" % + (verifier, result[verifier]['nodelist'][failed])) + raise errors.OpExecError("ssh/hostname verification failed.") # Distribute updated /etc/hosts and known_hosts to all nodes, # including the node just added @@ -1889,87 +1730,19 @@ class LUAddNode(LogicalUnit): logger.Error("copy of file %s to node %s failed" % (fname, to_node)) - to_copy = ss.GetFileList() + to_copy = self.sstore.GetFileList() if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31: to_copy.append(constants.VNC_PASSWORD_FILE) for fname in to_copy: - if not self.ssh.CopyFileToNode(node, fname): + result = rpc.call_upload_file([node], fname) + if not result[node]: logger.Error("could not copy file %s to node %s" % (fname, node)) if not self.op.readd: logger.Info("adding node %s to cluster.conf" % node) self.cfg.AddNode(new_node) - - -class LUMasterFailover(LogicalUnit): - """Failover the master node to the current node. - - This is a special LU in that it must run on a non-master node. - - """ - HPATH = "master-failover" - HTYPE = constants.HTYPE_CLUSTER - REQ_MASTER = False - _OP_REQP = [] - - def BuildHooksEnv(self): - """Build hooks env. - - This will run on the new master only in the pre phase, and on all - the nodes in the post phase. - - """ - env = { - "OP_TARGET": self.new_master, - "NEW_MASTER": self.new_master, - "OLD_MASTER": self.old_master, - } - return env, [self.new_master], self.cfg.GetNodeList() - - def CheckPrereq(self): - """Check prerequisites. - - This checks that we are not already the master. - - """ - self.new_master = utils.HostInfo().name - self.old_master = self.sstore.GetMasterNode() - - if self.old_master == self.new_master: - raise errors.OpPrereqError("This commands must be run on the node" - " where you want the new master to be." - " %s is already the master" % - self.old_master) - - def Exec(self, feedback_fn): - """Failover the master node. - - This command, when run on a non-master node, will cause the current - master to cease being master, and the non-master to become new - master. - - """ - #TODO: do not rely on gethostname returning the FQDN - logger.Info("setting master to %s, old master: %s" % - (self.new_master, self.old_master)) - - if not rpc.call_node_stop_master(self.old_master): - logger.Error("could disable the master role on the old master" - " %s, please disable manually" % self.old_master) - - ss = self.sstore - ss.SetKey(ss.SS_MASTER_NODE, self.new_master) - if not rpc.call_upload_file(self.cfg.GetNodeList(), - ss.KeyToFilename(ss.SS_MASTER_NODE)): - logger.Error("could not distribute the new simple store master file" - " to the other nodes, please check.") - - if not rpc.call_node_start_master(self.new_master): - logger.Error("could not start the master role on the new master" - " %s, please check" % self.new_master) - feedback_fn("Error in activating the master IP on the new master," - " please fix manually.") - + # Add the new node to the Ganeti Lock Manager + self.context.glm.add(locking.LEVEL_NODE, node) class LUQueryClusterInfo(NoHooksLU): @@ -1978,6 +1751,10 @@ class LUQueryClusterInfo(NoHooksLU): """ _OP_REQP = [] REQ_MASTER = False + REQ_BGL = False + + def ExpandNames(self): + self.needed_locks = {} def CheckPrereq(self): """No prerequsites needed for this LU. @@ -2004,50 +1781,15 @@ class LUQueryClusterInfo(NoHooksLU): return result -class LUClusterCopyFile(NoHooksLU): - """Copy file to cluster. - - """ - _OP_REQP = ["nodes", "filename"] - - def CheckPrereq(self): - """Check prerequisites. - - It should check that the named file exists and that the given list - of nodes is valid. - - """ - if not os.path.exists(self.op.filename): - raise errors.OpPrereqError("No such filename '%s'" % self.op.filename) - - self.nodes = _GetWantedNodes(self, self.op.nodes) - - def Exec(self, feedback_fn): - """Copy a file from master to some nodes. - - Args: - opts - class with options as members - args - list containing a single element, the file name - Opts used: - nodes - list containing the name of target nodes; if empty, all nodes - - """ - filename = self.op.filename - - myname = utils.HostInfo().name - - for node in self.nodes: - if node == myname: - continue - if not self.ssh.CopyFileToNode(node, filename): - logger.Error("Copy of file %s to node %s failed" % (filename, node)) - - class LUDumpClusterConfig(NoHooksLU): """Return a text-representation of the cluster-config. """ _OP_REQP = [] + REQ_BGL = False + + def ExpandNames(self): + self.needed_locks = {} def CheckPrereq(self): """No prerequisites. @@ -2062,38 +1804,6 @@ class LUDumpClusterConfig(NoHooksLU): return self.cfg.DumpConfig() -class LURunClusterCommand(NoHooksLU): - """Run a command on some nodes. - - """ - _OP_REQP = ["command", "nodes"] - - def CheckPrereq(self): - """Check prerequisites. - - It checks that the given list of nodes is valid. - - """ - self.nodes = _GetWantedNodes(self, self.op.nodes) - - def Exec(self, feedback_fn): - """Run a command on some nodes. - - """ - # put the master at the end of the nodes list - master_node = self.sstore.GetMasterNode() - if master_node in self.nodes: - self.nodes.remove(master_node) - self.nodes.append(master_node) - - data = [] - for node in self.nodes: - result = self.ssh.Run(node, "root", self.op.command) - data.append((node, result.output, result.exit_code)) - - return data - - class LUActivateInstanceDisks(NoHooksLU): """Bring up an instance's disks. @@ -2296,6 +2006,16 @@ class LUStartupInstance(LogicalUnit): HPATH = "instance-start" HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name", "force"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() + self.needed_locks[locking.LEVEL_NODE] = [] + self.recalculate_locks[locking.LEVEL_NODE] = 'replace' + + def DeclareLocks(self, level): + if level == locking.LEVEL_NODE: + self._LockInstancesNodes() def BuildHooksEnv(self): """Build hooks env. @@ -2317,11 +2037,9 @@ class LUStartupInstance(LogicalUnit): This checks that the instance is in the cluster. """ - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("Instance '%s' not known" % - self.op.instance_name) + self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert self.instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name # check bridges existance _CheckInstanceBridgesExist(instance) @@ -2330,9 +2048,6 @@ class LUStartupInstance(LogicalUnit): "starting instance %s" % instance.name, instance.memory) - self.instance = instance - self.op.instance_name = instance.name - def Exec(self, feedback_fn): """Start the instance. @@ -2359,6 +2074,16 @@ class LURebootInstance(LogicalUnit): HPATH = "instance-reboot" HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() + self.needed_locks[locking.LEVEL_NODE] = [] + self.recalculate_locks[locking.LEVEL_NODE] = 'replace' + + def DeclareLocks(self, level): + if level == locking.LEVEL_NODE: + self._LockInstancesNodes() def BuildHooksEnv(self): """Build hooks env. @@ -2380,18 +2105,13 @@ class LURebootInstance(LogicalUnit): This checks that the instance is in the cluster. """ - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("Instance '%s' not known" % - self.op.instance_name) + self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert self.instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name # check bridges existance _CheckInstanceBridgesExist(instance) - self.instance = instance - self.op.instance_name = instance.name - def Exec(self, feedback_fn): """Reboot the instance. @@ -2435,6 +2155,16 @@ class LUShutdownInstance(LogicalUnit): HPATH = "instance-stop" HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() + self.needed_locks[locking.LEVEL_NODE] = [] + self.recalculate_locks[locking.LEVEL_NODE] = 'replace' + + def DeclareLocks(self, level): + if level == locking.LEVEL_NODE: + self._LockInstancesNodes() def BuildHooksEnv(self): """Build hooks env. @@ -2453,12 +2183,9 @@ class LUShutdownInstance(LogicalUnit): This checks that the instance is in the cluster. """ - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("Instance '%s' not known" % - self.op.instance_name) - self.instance = instance + self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert self.instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name def Exec(self, feedback_fn): """Shutdown the instance. @@ -2480,6 +2207,16 @@ class LUReinstallInstance(LogicalUnit): HPATH = "instance-reinstall" HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() + self.needed_locks[locking.LEVEL_NODE] = [] + self.recalculate_locks[locking.LEVEL_NODE] = 'replace' + + def DeclareLocks(self, level): + if level == locking.LEVEL_NODE: + self._LockInstancesNodes() def BuildHooksEnv(self): """Build hooks env. @@ -2498,11 +2235,10 @@ class LUReinstallInstance(LogicalUnit): This checks that the instance is in the cluster and is not running. """ - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("Instance '%s' not known" % - self.op.instance_name) + instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name + if instance.disk_template == constants.DT_DISKLESS: raise errors.OpPrereqError("Instance '%s' has no disks" % self.op.instance_name) @@ -2603,9 +2339,7 @@ class LURenameInstance(LogicalUnit): new_name) if not getattr(self.op, "ignore_ip", False): - command = ["fping", "-q", name_info.ip] - result = utils.RunCmd(command) - if not result.failed: + if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT): raise errors.OpPrereqError("IP %s of instance %s already in use" % (name_info.ip, new_name)) @@ -2621,6 +2355,9 @@ class LURenameInstance(LogicalUnit): old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1]) self.cfg.RenameInstance(inst.name, self.op.new_name) + # Change the instance lock. This is definitely safe while we hold the BGL + self.context.glm.remove(locking.LEVEL_INSTANCE, inst.name) + self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name) # re-read the instance from the configuration after rename inst = self.cfg.GetInstanceInfo(self.op.new_name) @@ -2713,6 +2450,8 @@ class LURemoveInstance(LogicalUnit): logger.Info("removing instance %s out of cluster config" % instance.name) self.cfg.RemoveInstance(instance.name) + # Remove the new instance from the Ganeti Lock Manager + self.context.glm.remove(locking.LEVEL_INSTANCE, instance.name) class LUQueryInstances(NoHooksLU): @@ -2731,7 +2470,7 @@ class LUQueryInstances(NoHooksLU): _CheckOutputFields(static=["name", "os", "pnode", "snodes", "admin_state", "admin_ram", "disk_template", "ip", "mac", "bridge", - "sda_size", "sdb_size", "vcpus"], + "sda_size", "sdb_size", "vcpus", "tags"], dynamic=self.dynamic_fields, selected=self.op.output_fields) @@ -2824,6 +2563,8 @@ class LUQueryInstances(NoHooksLU): val = disk.size elif field == "vcpus": val = instance.vcpus + elif field == "tags": + val = list(instance.GetTags()) else: raise errors.ParameterError(field) iout.append(val) @@ -2839,6 +2580,16 @@ class LUFailoverInstance(LogicalUnit): HPATH = "instance-failover" HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name", "ignore_consistency"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() + self.needed_locks[locking.LEVEL_NODE] = [] + self.recalculate_locks[locking.LEVEL_NODE] = 'replace' + + def DeclareLocks(self, level): + if level == locking.LEVEL_NODE: + self._LockInstancesNodes() def BuildHooksEnv(self): """Build hooks env. @@ -2859,11 +2610,9 @@ class LUFailoverInstance(LogicalUnit): This checks that the instance is in the cluster. """ - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("Instance '%s' not known" % - self.op.instance_name) + self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert self.instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name if instance.disk_template not in constants.DTS_NET_MIRROR: raise errors.OpPrereqError("Instance's disk layout is not" @@ -2886,8 +2635,6 @@ class LUFailoverInstance(LogicalUnit): " exist on destination node '%s'" % (brlist, target_node)) - self.instance = instance - def Exec(self, feedback_fn): """Failover an instance. @@ -2927,7 +2674,7 @@ class LUFailoverInstance(LogicalUnit): instance.primary_node = target_node # distribute new instance config to the other nodes - self.cfg.AddInstance(instance) + self.cfg.Update(instance) # Only start the instance if it's marked as up if instance.status == "up": @@ -3011,22 +2758,6 @@ def _GenerateUniqueNames(cfg, exts): return results -def _GenerateMDDRBDBranch(cfg, primary, secondary, size, names): - """Generate a drbd device complete with its children. - - """ - port = cfg.AllocatePort() - vgname = cfg.GetVGName() - dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, - logical_id=(vgname, names[0])) - dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128, - logical_id=(vgname, names[1])) - drbd_dev = objects.Disk(dev_type=constants.LD_DRBD7, size=size, - logical_id = (primary, secondary, port), - children = [dev_data, dev_meta]) - return drbd_dev - - def _GenerateDRBD8Branch(cfg, primary, secondary, size, names, iv_name): """Generate a drbd8 device complete with its children. @@ -3459,7 +3190,7 @@ class LUCreateInstance(LogicalUnit): info = nodeinfo.get(node, None) if not info: raise errors.OpPrereqError("Cannot get current information" - " from node '%s'" % nodeinfo) + " from node '%s'" % node) vg_free = info.get('vg_free', None) if not isinstance(vg_free, int): raise errors.OpPrereqError("Can't compute free disk space on" @@ -3585,6 +3316,8 @@ class LUCreateInstance(LogicalUnit): feedback_fn("adding instance %s to cluster config" % instance) self.cfg.AddInstance(iobj) + # Add the new instance to the Ganeti Lock Manager + self.context.glm.add(locking.LEVEL_INSTANCE, instance) if self.op.wait_for_sync: disk_abort = not _WaitForSync(self.cfg, iobj, self.proc) @@ -3599,6 +3332,8 @@ class LUCreateInstance(LogicalUnit): if disk_abort: _RemoveDisks(iobj, self.cfg) self.cfg.RemoveInstance(iobj.name) + # Remove the new instance from the Ganeti Lock Manager + self.context.glm.remove(locking.LEVEL_INSTANCE, iobj.name) raise errors.OpExecError("There are some degraded disks for" " this instance") @@ -3643,6 +3378,10 @@ class LUConnectConsole(NoHooksLU): """ _OP_REQP = ["instance_name"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() def CheckPrereq(self): """Check prerequisites. @@ -3650,12 +3389,9 @@ class LUConnectConsole(NoHooksLU): This checks that the instance is in the cluster. """ - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("Instance '%s' not known" % - self.op.instance_name) - self.instance = instance + self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert self.instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name def Exec(self, feedback_fn): """Connect to the console of an instance @@ -4156,6 +3892,12 @@ class LUReplaceDisks(LogicalUnit): """ instance = self.instance + + # Activate the instance disks if we're replacing them on a down instance + if instance.status == "down": + op = opcodes.OpActivateInstanceDisks(instance_name=instance.name) + self.proc.ChainOpCode(op) + if instance.disk_template == constants.DT_DRBD8: if self.op.remote_node is None: fn = self._ExecD8DiskOnly @@ -4163,7 +3905,97 @@ class LUReplaceDisks(LogicalUnit): fn = self._ExecD8Secondary else: raise errors.ProgrammerError("Unhandled disk replacement case") - return fn(feedback_fn) + + ret = fn(feedback_fn) + + # Deactivate the instance disks if we're replacing them on a down instance + if instance.status == "down": + op = opcodes.OpDeactivateInstanceDisks(instance_name=instance.name) + self.proc.ChainOpCode(op) + + return ret + + +class LUGrowDisk(LogicalUnit): + """Grow a disk of an instance. + + """ + HPATH = "disk-grow" + HTYPE = constants.HTYPE_INSTANCE + _OP_REQP = ["instance_name", "disk", "amount"] + + def BuildHooksEnv(self): + """Build hooks env. + + This runs on the master, the primary and all the secondaries. + + """ + env = { + "DISK": self.op.disk, + "AMOUNT": self.op.amount, + } + env.update(_BuildInstanceHookEnvByObject(self.instance)) + nl = [ + self.sstore.GetMasterNode(), + self.instance.primary_node, + ] + return env, nl, nl + + def CheckPrereq(self): + """Check prerequisites. + + This checks that the instance is in the cluster. + + """ + instance = self.cfg.GetInstanceInfo( + self.cfg.ExpandInstanceName(self.op.instance_name)) + if instance is None: + raise errors.OpPrereqError("Instance '%s' not known" % + self.op.instance_name) + self.instance = instance + self.op.instance_name = instance.name + + if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8): + raise errors.OpPrereqError("Instance's disk layout does not support" + " growing.") + + if instance.FindDisk(self.op.disk) is None: + raise errors.OpPrereqError("Disk '%s' not found for instance '%s'" % + (self.op.disk, instance.name)) + + nodenames = [instance.primary_node] + list(instance.secondary_nodes) + nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName()) + for node in nodenames: + info = nodeinfo.get(node, None) + if not info: + raise errors.OpPrereqError("Cannot get current information" + " from node '%s'" % node) + vg_free = info.get('vg_free', None) + if not isinstance(vg_free, int): + raise errors.OpPrereqError("Can't compute free disk space on" + " node %s" % node) + if self.op.amount > info['vg_free']: + raise errors.OpPrereqError("Not enough disk space on target node %s:" + " %d MiB available, %d MiB required" % + (node, info['vg_free'], self.op.amount)) + + def Exec(self, feedback_fn): + """Execute disk grow. + + """ + instance = self.instance + disk = instance.FindDisk(self.op.disk) + for node in (instance.secondary_nodes + (instance.primary_node,)): + self.cfg.SetDiskID(disk, node) + result = rpc.call_blockdev_grow(node, disk, self.op.amount) + if not result or not isinstance(result, tuple) or len(result) != 2: + raise errors.OpExecError("grow request failed to node %s" % node) + elif not result[0]: + raise errors.OpExecError("grow request failed to node %s: %s" % + (node, result[1])) + disk.RecordGrow(self.op.amount) + self.cfg.Update(instance) + return class LUQueryInstanceData(NoHooksLU): @@ -4289,6 +4121,10 @@ class LUSetInstanceParams(LogicalUnit): HPATH = "instance-modify" HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name"] + REQ_BGL = False + + def ExpandNames(self): + self._ExpandAndLockInstance() def BuildHooksEnv(self): """Build hooks env. @@ -4326,6 +4162,9 @@ class LUSetInstanceParams(LogicalUnit): This only checks the instance list against the existing names. """ + # FIXME: all the parameters could be checked before, in ExpandNames, or in + # a separate CheckArguments function, if we implement one, so the operation + # can be aborted without waiting for any lock, should it have an error... self.mem = getattr(self.op, "mem", None) self.vcpus = getattr(self.op, "vcpus", None) self.ip = getattr(self.op, "ip", None) @@ -4420,13 +4259,9 @@ class LUSetInstanceParams(LogicalUnit): " like a valid IP address" % self.op.vnc_bind_address) - instance = self.cfg.GetInstanceInfo( - self.cfg.ExpandInstanceName(self.op.instance_name)) - if instance is None: - raise errors.OpPrereqError("No such instance name '%s'" % - self.op.instance_name) - self.op.instance_name = instance.name - self.instance = instance + self.instance = self.cfg.GetInstanceInfo(self.op.instance_name) + assert self.instance is not None, \ + "Cannot retrieve locked instance %s" % self.op.instance_name return def Exec(self, feedback_fn): @@ -4476,7 +4311,7 @@ class LUSetInstanceParams(LogicalUnit): instance.vnc_bind_address = self.vnc_bind_address result.append(("vnc_bind_address", self.vnc_bind_address)) - self.cfg.AddInstance(instance) + self.cfg.Update(instance) return result @@ -4565,8 +4400,8 @@ class LUExportInstance(LogicalUnit): if self.op.shutdown: # shutdown the instance, but not the disks if not rpc.call_instance_shutdown(src_node, instance): - raise errors.OpExecError("Could not shutdown instance %s on node %s" % - (instance.name, src_node)) + raise errors.OpExecError("Could not shutdown instance %s on node %s" % + (instance.name, src_node)) vgname = self.cfg.GetVGName() @@ -4704,7 +4539,7 @@ class LUGetTags(TagsLU): """Returns the tag list. """ - return self.target.GetTags() + return list(self.target.GetTags()) class LUSearchTags(NoHooksLU): @@ -4813,25 +4648,35 @@ class LUDelTags(TagsLU): " config file and the operation has been" " aborted. Please retry.") + class LUTestDelay(NoHooksLU): """Sleep for a specified amount of time. - This LU sleeps on the master and/or nodes for a specified amoutn of + This LU sleeps on the master and/or nodes for a specified amount of time. """ _OP_REQP = ["duration", "on_master", "on_nodes"] + REQ_BGL = False - def CheckPrereq(self): - """Check prerequisites. + def ExpandNames(self): + """Expand names and set required locks. - This checks that we have a good list of nodes and/or the duration - is valid. + This expands the node list, if any. """ - + self.needed_locks = {} if self.op.on_nodes: + # _GetWantedNodes can be used here, but is not always appropriate to use + # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for + # more information. self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes) + self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes + + def CheckPrereq(self): + """Check prerequisites. + + """ def Exec(self, feedback_fn): """Do the actual sleep. @@ -5089,9 +4934,8 @@ class IAllocator(object): if rcode == constants.IARUN_NOTFOUND: raise errors.OpExecError("Can't find allocator '%s'" % name) elif rcode == constants.IARUN_FAILURE: - raise errors.OpExecError("Instance allocator call failed: %s," - " output: %s" % - (fail, stdout+stderr)) + raise errors.OpExecError("Instance allocator call failed: %s," + " output: %s" % (fail, stdout+stderr)) self.out_text = stdout if validate: self._ValidateResult()