X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/821221737967d65208a5831ed3e7349d5f1e7d7e..153d9724f6045a7f08d1e64352ff68993408fa01:/lib/cmdlib.py diff --git a/lib/cmdlib.py b/lib/cmdlib.py index 26057f3..04612d7 100644 --- a/lib/cmdlib.py +++ b/lib/cmdlib.py @@ -1,4 +1,4 @@ -#!/usr/bin/python +# # # Copyright (C) 2006, 2007 Google Inc. @@ -85,7 +85,7 @@ class LogicalUnit(object): " use 'gnt-cluster init' first.") if self.REQ_MASTER: master = sstore.GetMasterNode() - if master != socket.gethostname(): + if master != utils.HostInfo().name: raise errors.OpPrereqError("Commands must be run on the master" " node %s" % master) @@ -161,31 +161,55 @@ class NoHooksLU(LogicalUnit): This is a no-op, since we don't run hooks. """ - return + return {}, [], [] def _GetWantedNodes(lu, nodes): - """Returns list of checked and expanded nodes. + """Returns list of checked and expanded node names. Args: nodes: List of nodes (strings) or None for all """ - if nodes is not None and not isinstance(nodes, list): + if not isinstance(nodes, list): raise errors.OpPrereqError("Invalid argument type 'nodes'") if nodes: - wanted_nodes = [] + wanted = [] for name in nodes: - node = lu.cfg.GetNodeInfo(lu.cfg.ExpandNodeName(name)) + node = lu.cfg.ExpandNodeName(name) if node is None: raise errors.OpPrereqError("No such node name '%s'" % name) - wanted_nodes.append(node) + wanted.append(node) - return wanted_nodes else: - return [lu.cfg.GetNodeInfo(name) for name in lu.cfg.GetNodeList()] + wanted = lu.cfg.GetNodeList() + return utils.NiceSort(wanted) + + +def _GetWantedInstances(lu, instances): + """Returns list of checked and expanded instance names. + + Args: + instances: List of instances (strings) or None for all + + """ + if not isinstance(instances, list): + raise errors.OpPrereqError("Invalid argument type 'instances'") + + if instances: + wanted = [] + + for name in instances: + instance = lu.cfg.ExpandInstanceName(name) + if instance is None: + raise errors.OpPrereqError("No such instance name '%s'" % name) + wanted.append(instance) + + else: + wanted = lu.cfg.GetInstanceList() + return utils.NiceSort(wanted) def _CheckOutputFields(static, dynamic, selected): @@ -215,6 +239,7 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status, secondary_nodes: List of secondary nodes as strings """ env = { + "OP_TARGET": name, "INSTANCE_NAME": name, "INSTANCE_PRIMARY": primary_node, "INSTANCE_SECONDARIES": " ".join(secondary_nodes), @@ -451,24 +476,23 @@ def _InitSSHSetup(node): node: the name of this host as a fqdn """ - if os.path.exists('/root/.ssh/id_dsa'): - utils.CreateBackup('/root/.ssh/id_dsa') - if os.path.exists('/root/.ssh/id_dsa.pub'): - utils.CreateBackup('/root/.ssh/id_dsa.pub') + priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) - utils.RemoveFile('/root/.ssh/id_dsa') - utils.RemoveFile('/root/.ssh/id_dsa.pub') + 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", "/root/.ssh/id_dsa", + "-f", priv_key, "-q", "-N", ""]) if result.failed: raise errors.OpExecError("Could not generate ssh keypair, error %s" % result.output) - f = open('/root/.ssh/id_dsa.pub', 'r') + f = open(pub_key, 'r') try: - utils.AddAuthorizedKey('/root/.ssh/authorized_keys', f.read(8192)) + utils.AddAuthorizedKey(auth_keys, f.read(8192)) finally: f.close() @@ -504,6 +528,18 @@ def _InitGanetiServerSetup(ss): (result.cmd, result.exit_code, result.output)) +def _CheckInstanceBridgesExist(instance): + """Check that the brigdes needed by an instance exist. + + """ + # check bridges existance + brlist = [nic.bridge for nic in instance.nics] + if not rpc.call_bridges_exist(instance.primary_node, brlist): + raise errors.OpPrereqError("one or more target bridges %s does not" + " exist on destination node '%s'" % + (brlist, instance.primary_node)) + + class LUInitCluster(LogicalUnit): """Initialise the cluster. @@ -521,11 +557,8 @@ class LUInitCluster(LogicalUnit): ourselves in the post-run node list. """ - env = { - "CLUSTER": self.op.cluster_name, - "MASTER": self.hostname['hostname_full'], - } - return env, [], [self.hostname['hostname_full']] + env = {"OP_TARGET": self.op.cluster_name} + return env, [], [self.hostname.name] def CheckPrereq(self): """Verify that the passed name is a valid one. @@ -534,33 +567,32 @@ class LUInitCluster(LogicalUnit): if config.ConfigWriter.IsCluster(): raise errors.OpPrereqError("Cluster is already initialised") - hostname_local = socket.gethostname() - self.hostname = hostname = utils.LookupHostname(hostname_local) - if not hostname: - raise errors.OpPrereqError("Cannot resolve my own hostname ('%s')" % - hostname_local) + self.hostname = hostname = utils.HostInfo() - self.clustername = clustername = utils.LookupHostname(self.op.cluster_name) - if not clustername: - raise errors.OpPrereqError("Cannot resolve given cluster name ('%s')" - % self.op.cluster_name) + if hostname.ip.startswith("127."): + raise errors.OpPrereqError("This host's IP resolves to the private" + " range (%s). Please fix DNS or /etc/hosts." % + (hostname.ip,)) - result = utils.RunCmd(["fping", "-S127.0.0.1", "-q", hostname['ip']]) - if result.failed: + self.clustername = clustername = utils.HostInfo(self.op.cluster_name) + + if not utils.TcpPing(constants.LOCALHOST_IP_ADDRESS, hostname.ip, + constants.DEFAULT_NODED_PORT): raise errors.OpPrereqError("Inconsistency: this host's name resolves" " to %s,\nbut this ip address does not" " belong to this host." - " Aborting." % hostname['ip']) + " Aborting." % hostname.ip) 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']: - result = utils.RunCmd(["fping", "-S127.0.0.1", "-q", secondary_ip]) - if result.failed: - raise errors.OpPrereqError("You gave %s as secondary IP,\n" - "but it does not belong to this host." % - secondary_ip) + if (secondary_ip and + secondary_ip != hostname.ip and + (not utils.TcpPing(constants.LOCALHOST_IP_ADDRESS, secondary_ip, + constants.DEFAULT_NODED_PORT))): + raise errors.OpPrereqError("You gave %s as secondary IP,\n" + "but it does not belong to this host." % + secondary_ip) self.secondary_ip = secondary_ip # checks presence of the volume group given @@ -592,41 +624,36 @@ class LUInitCluster(LogicalUnit): hostname = self.hostname # set up the simple store - ss = ssconf.SimpleStore() + self.sstore = ss = ssconf.SimpleStore() ss.SetKey(ss.SS_HYPERVISOR, self.op.hypervisor_type) - ss.SetKey(ss.SS_MASTER_NODE, hostname['hostname_full']) - ss.SetKey(ss.SS_MASTER_IP, clustername['ip']) + 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['hostname']) + ss.SetKey(ss.SS_CLUSTER_NAME, clustername.name) # set up the inter-node password and certificate _InitGanetiServerSetup(ss) # start the master ip - rpc.call_node_start_master(hostname['hostname_full']) + rpc.call_node_start_master(hostname.name) # set up ssh config and /etc/hosts - f = open('/etc/ssh/ssh_host_rsa_key.pub', 'r') + f = open(constants.SSH_HOST_RSA_PUB, 'r') try: sshline = f.read() finally: f.close() sshkey = sshline.split(" ")[1] - _UpdateEtcHosts(hostname['hostname_full'], - hostname['ip'], - ) + _UpdateEtcHosts(hostname.name, hostname.ip) - _UpdateKnownHosts(hostname['hostname_full'], - hostname['ip'], - sshkey, - ) + _UpdateKnownHosts(hostname.name, hostname.ip, sshkey) - _InitSSHSetup(hostname['hostname']) + _InitSSHSetup(hostname.name) # init of cluster config file - cfgw = config.ConfigWriter() - cfgw.InitConfig(hostname['hostname'], hostname['ip'], self.secondary_ip, + 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) @@ -660,8 +687,9 @@ class LUDestroyCluster(NoHooksLU): """Destroys the cluster. """ - utils.CreateBackup('/root/.ssh/id_dsa') - utils.CreateBackup('/root/.ssh/id_dsa.pub') + priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) + utils.CreateBackup(priv_key) + utils.CreateBackup(pub_key) rpc.call_node_leave_cluster(self.sstore.GetMasterNode()) @@ -782,7 +810,7 @@ class LUVerifyCluster(NoHooksLU): (instance, node)) bad = True - return not bad + return bad def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn): """Verify if there are any unknown volumes in the cluster. @@ -909,6 +937,85 @@ class LUVerifyCluster(NoHooksLU): return int(bad) +class LURenameCluster(LogicalUnit): + """Rename the cluster. + + """ + HPATH = "cluster-rename" + HTYPE = constants.HTYPE_CLUSTER + _OP_REQP = ["name"] + + def BuildHooksEnv(self): + """Build hooks env. + + """ + env = { + "OP_TARGET": self.op.sstore.GetClusterName(), + "NEW_NAME": self.op.name, + } + mn = self.sstore.GetMasterNode() + return env, [mn], [mn] + + def CheckPrereq(self): + """Verify that the passed name is a valid one. + + """ + hostname = utils.HostInfo(self.op.name) + + new_name = hostname.name + self.ip = new_ip = hostname.ip + old_name = self.sstore.GetClusterName() + old_ip = self.sstore.GetMasterIP() + if new_name == old_name and new_ip == old_ip: + 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: + raise errors.OpPrereqError("The given cluster IP address (%s) is" + " reachable on the network. Aborting." % + new_ip) + + self.op.name = new_name + + def Exec(self, feedback_fn): + """Rename the cluster. + + """ + clustername = self.op.name + ip = self.ip + ss = self.sstore + + # shutdown the master IP + master = ss.GetMasterNode() + if not rpc.call_node_stop_master(master): + raise errors.OpExecError("Could not disable the master role") + + try: + # modify the sstore + ss.SetKey(ss.SS_MASTER_IP, ip) + ss.SetKey(ss.SS_CLUSTER_NAME, clustername) + + # Distribute updated ss config to all nodes + myself = self.cfg.GetNodeInfo(master) + dist_nodes = self.cfg.GetNodeList() + if myself.name in dist_nodes: + dist_nodes.remove(myself.name) + + logger.Debug("Copying updated ssconf data to all nodes") + for keyname in [ss.SS_CLUSTER_NAME, ss.SS_MASTER_IP]: + fname = ss.KeyToFilename(keyname) + result = rpc.call_upload_file(dist_nodes, fname) + for to_node in dist_nodes: + if not result[to_node]: + logger.Error("copy of file %s to node %s failed" % + (fname, to_node)) + finally: + if not rpc.call_node_start_master(master): + logger.Error("Could not re-enable the master role on the master,\n" + "please restart manually.") + + def _WaitForSync(cfgw, instance, oneshot=False, unlock=False): """Sleep and poll for an instance's disk to sync. @@ -1034,6 +1141,7 @@ class LURemoveNode(LogicalUnit): """ env = { + "OP_TARGET": self.op.node_name, "NODE_NAME": self.op.node_name, } all_nodes = self.cfg.GetNodeList() @@ -1094,7 +1202,7 @@ class LUQueryNodes(NoHooksLU): """Logical unit for querying nodes. """ - _OP_REQP = ["output_fields"] + _OP_REQP = ["output_fields", "names"] def CheckPrereq(self): """Check prerequisites. @@ -1103,21 +1211,24 @@ class LUQueryNodes(NoHooksLU): """ self.dynamic_fields = frozenset(["dtotal", "dfree", - "mtotal", "mnode", "mfree"]) + "mtotal", "mnode", "mfree", + "bootid"]) - _CheckOutputFields(static=["name", "pinst", "sinst", "pip", "sip"], + _CheckOutputFields(static=["name", "pinst_cnt", "sinst_cnt", + "pinst_list", "sinst_list", + "pip", "sip"], dynamic=self.dynamic_fields, selected=self.op.output_fields) + self.wanted = _GetWantedNodes(self, self.op.names) def Exec(self, feedback_fn): """Computes the list of nodes and their attributes. """ - nodenames = utils.NiceSort(self.cfg.GetNodeList()) + nodenames = self.wanted nodelist = [self.cfg.GetNodeInfo(name) for name in nodenames] - # begin data gathering if self.dynamic_fields.intersection(self.op.output_fields): @@ -1132,23 +1243,28 @@ class LUQueryNodes(NoHooksLU): "mfree": utils.TryConvert(int, nodeinfo['memory_free']), "dtotal": utils.TryConvert(int, nodeinfo['vg_size']), "dfree": utils.TryConvert(int, nodeinfo['vg_free']), + "bootid": nodeinfo['bootid'], } else: live_data[name] = {} else: live_data = dict.fromkeys(nodenames, {}) - node_to_primary = dict.fromkeys(nodenames, 0) - node_to_secondary = dict.fromkeys(nodenames, 0) + node_to_primary = dict([(name, set()) for name in nodenames]) + node_to_secondary = dict([(name, set()) for name in nodenames]) - if "pinst" in self.op.output_fields or "sinst" in self.op.output_fields: + inst_fields = frozenset(("pinst_cnt", "pinst_list", + "sinst_cnt", "sinst_list")) + if inst_fields & frozenset(self.op.output_fields): instancelist = self.cfg.GetInstanceList() - for instance in instancelist: - instanceinfo = self.cfg.GetInstanceInfo(instance) - node_to_primary[instanceinfo.primary_node] += 1 - for secnode in instanceinfo.secondary_nodes: - node_to_secondary[secnode] += 1 + for instance_name in instancelist: + inst = self.cfg.GetInstanceInfo(instance_name) + if inst.primary_node in node_to_primary: + node_to_primary[inst.primary_node].add(inst.name) + for secnode in inst.secondary_nodes: + if secnode in node_to_secondary: + node_to_secondary[secnode].add(inst.name) # end data gathering @@ -1158,19 +1274,22 @@ class LUQueryNodes(NoHooksLU): for field in self.op.output_fields: if field == "name": val = node.name - elif field == "pinst": - val = node_to_primary[node.name] - elif field == "sinst": - val = node_to_secondary[node.name] + elif field == "pinst_list": + val = list(node_to_primary[node.name]) + elif field == "sinst_list": + val = list(node_to_secondary[node.name]) + elif field == "pinst_cnt": + val = len(node_to_primary[node.name]) + elif field == "sinst_cnt": + val = len(node_to_secondary[node.name]) elif field == "pip": val = node.primary_ip elif field == "sip": val = node.secondary_ip elif field in self.dynamic_fields: - val = live_data[node.name].get(field, "?") + val = live_data[node.name].get(field, None) else: raise errors.ParameterError(field) - val = str(val) node_output.append(val) output.append(node_output) @@ -1200,7 +1319,7 @@ class LUQueryNodeVolumes(NoHooksLU): """Computes the list of nodes and their attributes. """ - nodenames = utils.NiceSort([node.name for node in self.nodes]) + nodenames = self.nodes volumes = rpc.call_node_volumes(nodenames) ilist = [self.cfg.GetInstanceInfo(iname) for iname @@ -1262,6 +1381,7 @@ class LUAddNode(LogicalUnit): """ env = { + "OP_TARGET": self.op.node_name, "NODE_NAME": self.op.node_name, "NODE_PIP": self.op.primary_ip, "NODE_SIP": self.op.secondary_ip, @@ -1284,12 +1404,10 @@ class LUAddNode(LogicalUnit): node_name = self.op.node_name cfg = self.cfg - dns_data = utils.LookupHostname(node_name) - if not dns_data: - raise errors.OpPrereqError("Node %s is not resolvable" % node_name) + dns_data = utils.HostInfo(node_name) - node = dns_data['hostname'] - primary_ip = self.op.primary_ip = dns_data['ip'] + node = dns_data.name + primary_ip = self.op.primary_ip = dns_data.ip secondary_ip = getattr(self.op, "secondary_ip", None) if secondary_ip is None: secondary_ip = primary_ip @@ -1324,17 +1442,18 @@ class LUAddNode(LogicalUnit): " new node doesn't have one") # checks reachablity - command = ["fping", "-q", primary_ip] - result = utils.RunCmd(command) - if result.failed: + if not utils.TcpPing(utils.HostInfo().name, + primary_ip, + constants.DEFAULT_NODED_PORT): raise errors.OpPrereqError("Node not reachable by ping") if not newbie_singlehomed: # check reachability from my secondary ip to newbie's secondary ip - command = ["fping", "-S%s" % myself.secondary_ip, "-q", secondary_ip] - result = utils.RunCmd(command) - if result.failed: - raise errors.OpPrereqError("Node secondary ip not reachable by ping") + if not utils.TcpPing(myself.secondary_ip, + secondary_ip, + constants.DEFAULT_NODED_PORT): + raise errors.OpPrereqError( + "Node secondary ip not reachable by TCP based ping to noded port") self.new_node = objects.Node(name=node, primary_ip=primary_ip, @@ -1401,10 +1520,11 @@ class LUAddNode(LogicalUnit): # setup ssh on node logger.Info("copy ssh key to node %s" % node) + priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) keyarray = [] - keyfiles = ["/etc/ssh/ssh_host_dsa_key", "/etc/ssh/ssh_host_dsa_key.pub", - "/etc/ssh/ssh_host_rsa_key", "/etc/ssh/ssh_host_rsa_key.pub", - "/root/.ssh/id_dsa", "/root/.ssh/id_dsa.pub"] + keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB, + constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB, + priv_key, pub_key] for i in keyfiles: f = open(i, 'r') @@ -1425,14 +1545,23 @@ class LUAddNode(LogicalUnit): self.cfg.GetHostKey()) if new_node.secondary_ip != new_node.primary_ip: - result = ssh.SSHCall(node, "root", - "fping -S 127.0.0.1 -q %s" % new_node.secondary_ip) - if result.failed: + if not rpc.call_node_tcp_ping(new_node.name, + constants.LOCALHOST_IP_ADDRESS, + new_node.secondary_ip, + constants.DEFAULT_NODED_PORT, + 10, False): raise errors.OpExecError("Node claims it doesn't have the" " secondary ip you gave (%s).\n" "Please fix and re-run this command." % new_node.secondary_ip) + success, msg = ssh.VerifyNodeHostname(node) + if not success: + raise errors.OpExecError("Node '%s' claims it has a different hostname" + " than the one the resolver gives: %s.\n" + "Please fix and re-run this command." % + (node, msg)) + # Distribute updated /etc/hosts and known_hosts to all nodes, # including the node just added myself = self.cfg.GetNodeInfo(self.sstore.GetMasterNode()) @@ -1476,6 +1605,7 @@ class LUMasterFailover(LogicalUnit): """ env = { + "OP_TARGET": self.new_master, "NEW_MASTER": self.new_master, "OLD_MASTER": self.old_master, } @@ -1487,8 +1617,7 @@ class LUMasterFailover(LogicalUnit): This checks that we are not already the master. """ - self.new_master = socket.gethostname() - + self.new_master = utils.HostInfo().name self.old_master = self.sstore.GetMasterNode() if self.old_master == self.new_master: @@ -1589,9 +1718,9 @@ class LUClusterCopyFile(NoHooksLU): """ filename = self.op.filename - myname = socket.gethostname() + myname = utils.HostInfo().name - for node in [node.name for node in self.nodes]: + for node in self.nodes: if node == myname: continue if not ssh.CopyFileToNode(node, filename): @@ -1637,8 +1766,8 @@ class LURunClusterCommand(NoHooksLU): """ data = [] for node in self.nodes: - result = ssh.SSHCall(node.name, "root", self.op.command) - data.append((node.name, result.output, result.exit_code)) + result = ssh.SSHCall(node, "root", self.op.command) + data.append((node, result.output, result.exit_code)) return data @@ -1707,6 +1836,12 @@ def _AssembleInstanceDisks(instance, cfg, ignore_secondaries=False): device_info.append((instance.primary_node, inst_disk.iv_name, master_result)) + # leave the disks configured for the primary node + # this is a workaround that would be fixed better by + # improving the logical/physical id handling + for disk in instance.disks: + cfg.SetDiskID(disk, instance.primary_node) + return disks_ok, device_info @@ -1817,11 +1952,7 @@ class LUStartupInstance(LogicalUnit): self.op.instance_name) # check bridges existance - brlist = [nic.bridge for nic in instance.nics] - if not rpc.call_bridges_exist(instance.primary_node, brlist): - raise errors.OpPrereqError("one or more target bridges %s does not" - " exist on destination node '%s'" % - (brlist, instance.primary_node)) + _CheckInstanceBridgesExist(instance) self.instance = instance self.op.instance_name = instance.name @@ -1859,6 +1990,82 @@ class LUStartupInstance(LogicalUnit): self.cfg.MarkInstanceUp(instance.name) +class LURebootInstance(LogicalUnit): + """Reboot an instance. + + """ + HPATH = "instance-reboot" + HTYPE = constants.HTYPE_INSTANCE + _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"] + + def BuildHooksEnv(self): + """Build hooks env. + + This runs on master, primary and secondary nodes of the instance. + + """ + env = { + "IGNORE_SECONDARIES": self.op.ignore_secondaries, + } + env.update(_BuildInstanceHookEnvByObject(self.instance)) + nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + + list(self.instance.secondary_nodes)) + 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) + + # check bridges existance + _CheckInstanceBridgesExist(instance) + + self.instance = instance + self.op.instance_name = instance.name + + def Exec(self, feedback_fn): + """Reboot the instance. + + """ + instance = self.instance + ignore_secondaries = self.op.ignore_secondaries + reboot_type = self.op.reboot_type + extra_args = getattr(self.op, "extra_args", "") + + node_current = instance.primary_node + + if reboot_type not in [constants.INSTANCE_REBOOT_SOFT, + constants.INSTANCE_REBOOT_HARD, + constants.INSTANCE_REBOOT_FULL]: + raise errors.ParameterError("reboot type not in [%s, %s, %s]" % + (constants.INSTANCE_REBOOT_SOFT, + constants.INSTANCE_REBOOT_HARD, + constants.INSTANCE_REBOOT_FULL)) + + if reboot_type in [constants.INSTANCE_REBOOT_SOFT, + constants.INSTANCE_REBOOT_HARD]: + if not rpc.call_instance_reboot(node_current, instance, + reboot_type, extra_args): + raise errors.OpExecError("Could not reboot instance") + else: + if not rpc.call_instance_shutdown(node_current, instance): + raise errors.OpExecError("could not shutdown instance for full reboot") + _ShutdownInstanceDisks(instance, self.cfg) + _StartInstanceDisks(self.cfg, instance, ignore_secondaries) + if not rpc.call_instance_start(node_current, instance, extra_args): + _ShutdownInstanceDisks(instance, self.cfg) + raise errors.OpExecError("Could not start instance for full reboot") + + self.cfg.MarkInstanceUp(instance.name) + + class LUShutdownInstance(LogicalUnit): """Shutdown an instance. @@ -1983,6 +2190,84 @@ class LUReinstallInstance(LogicalUnit): _ShutdownInstanceDisks(inst, self.cfg) +class LURenameInstance(LogicalUnit): + """Rename an instance. + + """ + HPATH = "instance-rename" + HTYPE = constants.HTYPE_INSTANCE + _OP_REQP = ["instance_name", "new_name"] + + def BuildHooksEnv(self): + """Build hooks env. + + This runs on master, primary and secondary nodes of the instance. + + """ + env = _BuildInstanceHookEnvByObject(self.instance) + env["INSTANCE_NEW_NAME"] = self.op.new_name + nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + + list(self.instance.secondary_nodes)) + return env, nl, nl + + def CheckPrereq(self): + """Check prerequisites. + + 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) + if instance.status != "down": + raise errors.OpPrereqError("Instance '%s' is marked to be up" % + self.op.instance_name) + remote_info = rpc.call_instance_info(instance.primary_node, instance.name) + if remote_info: + raise errors.OpPrereqError("Instance '%s' is running on the node %s" % + (self.op.instance_name, + instance.primary_node)) + self.instance = instance + + # new name verification + name_info = utils.HostInfo(self.op.new_name) + + self.op.new_name = new_name = name_info.name + if not getattr(self.op, "ignore_ip", False): + command = ["fping", "-q", name_info.ip] + result = utils.RunCmd(command) + if not result.failed: + raise errors.OpPrereqError("IP %s of instance %s already in use" % + (name_info.ip, new_name)) + + + def Exec(self, feedback_fn): + """Reinstall the instance. + + """ + inst = self.instance + old_name = inst.name + + self.cfg.RenameInstance(inst.name, self.op.new_name) + + # re-read the instance from the configuration after rename + inst = self.cfg.GetInstanceInfo(self.op.new_name) + + _StartInstanceDisks(self.cfg, inst, None) + try: + if not rpc.call_instance_run_rename(inst.primary_node, inst, old_name, + "sda", "sdb"): + msg = ("Could run OS rename script for instance %s\n" + "on node %s\n" + "(but the instance has been renamed in Ganeti)" % + (inst.name, inst.primary_node)) + logger.Error(msg) + finally: + _ShutdownInstanceDisks(inst, self.cfg) + + class LURemoveInstance(LogicalUnit): """Remove an instance. @@ -1998,8 +2283,7 @@ class LURemoveInstance(LogicalUnit): """ env = _BuildInstanceHookEnvByObject(self.instance) - nl = ([self.sstore.GetMasterNode(), self.instance.primary_node] + - list(self.instance.secondary_nodes)) + nl = [self.sstore.GetMasterNode()] return env, nl, nl def CheckPrereq(self): @@ -2024,12 +2308,19 @@ class LURemoveInstance(LogicalUnit): (instance.name, instance.primary_node)) if not rpc.call_instance_shutdown(instance.primary_node, instance): - raise errors.OpExecError("Could not shutdown instance %s on node %s" % - (instance.name, instance.primary_node)) + if self.op.ignore_failures: + feedback_fn("Warning: can't shutdown instance") + else: + raise errors.OpExecError("Could not shutdown instance %s on node %s" % + (instance.name, instance.primary_node)) logger.Info("removing block devices for instance %s" % instance.name) - _RemoveDisks(instance, self.cfg) + if not _RemoveDisks(instance, self.cfg): + if self.op.ignore_failures: + feedback_fn("Warning: can't remove instance's disks") + else: + raise errors.OpExecError("Can't remove instance's disks") logger.Info("removing instance %s out of cluster config" % instance.name) @@ -2040,7 +2331,7 @@ class LUQueryInstances(NoHooksLU): """Logical unit for querying instances. """ - _OP_REQP = ["output_fields"] + _OP_REQP = ["output_fields", "names"] def CheckPrereq(self): """Check prerequisites. @@ -2056,11 +2347,13 @@ class LUQueryInstances(NoHooksLU): dynamic=self.dynamic_fields, selected=self.op.output_fields) + self.wanted = _GetWantedInstances(self, self.op.names) + def Exec(self, feedback_fn): """Computes the list of nodes and their attributes. """ - instance_names = utils.NiceSort(self.cfg.GetInstanceList()) + instance_names = self.wanted instance_list = [self.cfg.GetInstanceInfo(iname) for iname in instance_names] @@ -2095,25 +2388,19 @@ class LUQueryInstances(NoHooksLU): elif field == "pnode": val = instance.primary_node elif field == "snodes": - val = ",".join(instance.secondary_nodes) or "-" + val = list(instance.secondary_nodes) elif field == "admin_state": - if instance.status == "down": - val = "no" - else: - val = "yes" + val = (instance.status != "down") elif field == "oper_state": if instance.primary_node in bad_nodes: - val = "(node down)" + val = None else: - if live_data.get(instance.name): - val = "running" - else: - val = "stopped" + val = bool(live_data.get(instance.name)) elif field == "admin_ram": val = instance.memory elif field == "oper_ram": if instance.primary_node in bad_nodes: - val = "(node down)" + val = None elif instance.name in live_data: val = live_data[instance.name].get("memory", "?") else: @@ -2129,12 +2416,11 @@ class LUQueryInstances(NoHooksLU): elif field == "sda_size" or field == "sdb_size": disk = instance.FindDisk(field[:3]) if disk is None: - val = "N/A" + val = None else: val = disk.size else: raise errors.ParameterError(field) - val = str(val) iout.append(val) output.append(iout) @@ -2174,8 +2460,17 @@ class LUFailoverInstance(LogicalUnit): raise errors.OpPrereqError("Instance '%s' not known" % self.op.instance_name) + if instance.disk_template not in constants.DTS_NET_MIRROR: + raise errors.OpPrereqError("Instance's disk layout is not" + " network mirrored, cannot failover.") + + secondary_nodes = instance.secondary_nodes + if not secondary_nodes: + raise errors.ProgrammerError("no secondary node but using " + "DT_REMOTE_RAID1 template") + # check memory requirements on the secondary node - target_node = instance.secondary_nodes[0] + target_node = secondary_nodes[0] nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName()) info = nodeinfo.get(target_node, None) if not info: @@ -2331,16 +2626,32 @@ def _GenerateMDDRBDBranch(cfg, primary, secondary, size, names): """ port = cfg.AllocatePort() vgname = cfg.GetVGName() - dev_data = objects.Disk(dev_type="lvm", size=size, + dev_data = objects.Disk(dev_type=constants.LD_LV, size=size, logical_id=(vgname, names[0])) - dev_meta = objects.Disk(dev_type="lvm", size=128, + dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128, logical_id=(vgname, names[1])) - drbd_dev = objects.Disk(dev_type="drbd", size=size, + 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. + + """ + 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_DRBD8, size=size, + logical_id = (primary, secondary, port), + children = [dev_data, dev_meta], + iv_name=iv_name) + return drbd_dev + def _GenerateDiskTemplate(cfg, template_name, instance_name, primary_node, secondary_nodes, disk_sz, swap_sz): @@ -2357,10 +2668,10 @@ def _GenerateDiskTemplate(cfg, template_name, raise errors.ProgrammerError("Wrong template configuration") names = _GenerateUniqueNames(cfg, [".sda", ".sdb"]) - sda_dev = objects.Disk(dev_type="lvm", size=disk_sz, + sda_dev = objects.Disk(dev_type=constants.LD_LV, size=disk_sz, logical_id=(vgname, names[0]), iv_name = "sda") - sdb_dev = objects.Disk(dev_type="lvm", size=swap_sz, + sdb_dev = objects.Disk(dev_type=constants.LD_LV, size=swap_sz, logical_id=(vgname, names[1]), iv_name = "sdb") disks = [sda_dev, sdb_dev] @@ -2371,22 +2682,22 @@ def _GenerateDiskTemplate(cfg, template_name, names = _GenerateUniqueNames(cfg, [".sda_m1", ".sda_m2", ".sdb_m1", ".sdb_m2"]) - sda_dev_m1 = objects.Disk(dev_type="lvm", size=disk_sz, + sda_dev_m1 = objects.Disk(dev_type=constants.LD_LV, size=disk_sz, logical_id=(vgname, names[0])) - sda_dev_m2 = objects.Disk(dev_type="lvm", size=disk_sz, + sda_dev_m2 = objects.Disk(dev_type=constants.LD_LV, size=disk_sz, logical_id=(vgname, names[1])) - md_sda_dev = objects.Disk(dev_type="md_raid1", iv_name = "sda", + md_sda_dev = objects.Disk(dev_type=constants.LD_MD_R1, iv_name = "sda", size=disk_sz, children = [sda_dev_m1, sda_dev_m2]) - sdb_dev_m1 = objects.Disk(dev_type="lvm", size=swap_sz, + sdb_dev_m1 = objects.Disk(dev_type=constants.LD_LV, size=swap_sz, logical_id=(vgname, names[2])) - sdb_dev_m2 = objects.Disk(dev_type="lvm", size=swap_sz, + sdb_dev_m2 = objects.Disk(dev_type=constants.LD_LV, size=swap_sz, logical_id=(vgname, names[3])) - md_sdb_dev = objects.Disk(dev_type="md_raid1", iv_name = "sdb", + md_sdb_dev = objects.Disk(dev_type=constants.LD_MD_R1, iv_name = "sdb", size=swap_sz, children = [sdb_dev_m1, sdb_dev_m2]) disks = [md_sda_dev, md_sdb_dev] - elif template_name == "remote_raid1": + elif template_name == constants.DT_REMOTE_RAID1: if len(secondary_nodes) != 1: raise errors.ProgrammerError("Wrong template configuration") remote_node = secondary_nodes[0] @@ -2394,13 +2705,24 @@ def _GenerateDiskTemplate(cfg, template_name, ".sdb_data", ".sdb_meta"]) drbd_sda_dev = _GenerateMDDRBDBranch(cfg, primary_node, remote_node, disk_sz, names[0:2]) - md_sda_dev = objects.Disk(dev_type="md_raid1", iv_name="sda", + md_sda_dev = objects.Disk(dev_type=constants.LD_MD_R1, iv_name="sda", children = [drbd_sda_dev], size=disk_sz) drbd_sdb_dev = _GenerateMDDRBDBranch(cfg, primary_node, remote_node, swap_sz, names[2:4]) - md_sdb_dev = objects.Disk(dev_type="md_raid1", iv_name="sdb", + md_sdb_dev = objects.Disk(dev_type=constants.LD_MD_R1, iv_name="sdb", children = [drbd_sdb_dev], size=swap_sz) disks = [md_sda_dev, md_sdb_dev] + elif template_name == constants.DT_DRBD8: + if len(secondary_nodes) != 1: + raise errors.ProgrammerError("Wrong template configuration") + remote_node = secondary_nodes[0] + names = _GenerateUniqueNames(cfg, [".sda_data", ".sda_meta", + ".sdb_data", ".sdb_meta"]) + drbd_sda_dev = _GenerateDRBD8Branch(cfg, primary_node, remote_node, + disk_sz, names[0:2], "sda") + drbd_sdb_dev = _GenerateDRBD8Branch(cfg, primary_node, remote_node, + swap_sz, names[2:4], "sdb") + disks = [drbd_sda_dev, drbd_sdb_dev] else: raise errors.ProgrammerError("Invalid disk template '%s'" % template_name) return disks @@ -2450,7 +2772,7 @@ def _RemoveDisks(instance, cfg): This abstracts away some work from `AddInstance()` and `RemoveInstance()`. Note that in case some of the devices couldn't - be remove, the removal will continue with the other ones (compare + be removed, the removal will continue with the other ones (compare with `_CreateDisks()`). Args: @@ -2482,7 +2804,7 @@ class LUCreateInstance(LogicalUnit): HTYPE = constants.HTYPE_INSTANCE _OP_REQP = ["instance_name", "mem_size", "disk_size", "pnode", "disk_template", "swap_size", "mode", "start", "vcpus", - "wait_for_sync"] + "wait_for_sync", "ip_check"] def BuildHooksEnv(self): """Build hooks env. @@ -2577,9 +2899,9 @@ class LUCreateInstance(LogicalUnit): if self.op.disk_template not in constants.DISK_TEMPLATES: raise errors.OpPrereqError("Invalid disk template name") - if self.op.disk_template == constants.DT_REMOTE_RAID1: + if self.op.disk_template in constants.DTS_NET_MIRROR: if getattr(self.op, "snode", None) is None: - raise errors.OpPrereqError("The 'remote_raid1' disk template needs" + raise errors.OpPrereqError("The networked disk templates need" " a mirror node") snode_name = self.cfg.ExpandNodeName(self.op.snode) @@ -2602,6 +2924,7 @@ class LUCreateInstance(LogicalUnit): constants.DT_LOCAL_RAID1: (self.op.disk_size + self.op.swap_size) * 2, # 256 MB are added for drbd metadata, 128MB for each drbd device constants.DT_REMOTE_RAID1: self.op.disk_size + self.op.swap_size + 256, + constants.DT_DRBD8: self.op.disk_size + self.op.swap_size + 256, } if self.op.disk_template not in req_size_dict: @@ -2627,12 +2950,9 @@ class LUCreateInstance(LogicalUnit): " primary node" % self.op.os_type) # instance verification - hostname1 = utils.LookupHostname(self.op.instance_name) - if not hostname1: - raise errors.OpPrereqError("Instance name '%s' not found in dns" % - self.op.instance_name) + hostname1 = utils.HostInfo(self.op.instance_name) - self.op.instance_name = instance_name = hostname1['hostname'] + self.op.instance_name = instance_name = hostname1.name instance_list = self.cfg.GetInstanceList() if instance_name in instance_list: raise errors.OpPrereqError("Instance '%s' is already in the cluster" % @@ -2642,7 +2962,7 @@ class LUCreateInstance(LogicalUnit): if ip is None or ip.lower() == "none": inst_ip = None elif ip.lower() == "auto": - inst_ip = hostname1['ip'] + inst_ip = hostname1.ip else: if not utils.IsValidIP(ip): raise errors.OpPrereqError("given IP address '%s' doesn't look" @@ -2650,11 +2970,15 @@ class LUCreateInstance(LogicalUnit): inst_ip = ip self.inst_ip = inst_ip - command = ["fping", "-q", hostname1['ip']] - result = utils.RunCmd(command) - if not result.failed: - raise errors.OpPrereqError("IP %s of instance %s already in use" % - (hostname1['ip'], instance_name)) + if self.op.start and not self.op.ip_check: + raise errors.OpPrereqError("Cannot ignore IP address conflicts when" + " adding an instance in start mode") + + if self.op.ip_check: + if utils.TcpPing(utils.HostInfo().name, hostname1.ip, + constants.DEFAULT_NODED_PORT): + raise errors.OpPrereqError("IP %s of instance %s already in use" % + (hostname1.ip, instance_name)) # bridge verification bridge = getattr(self.op, "bridge", None) @@ -2710,7 +3034,7 @@ class LUCreateInstance(LogicalUnit): if self.op.wait_for_sync: disk_abort = not _WaitForSync(self.cfg, iobj) - elif iobj.disk_template == "remote_raid1": + elif iobj.disk_template in constants.DTS_NET_MIRROR: # make sure the disks are not degraded (still sync-ing is ok) time.sleep(15) feedback_fn("* checking mirrors status") @@ -2899,8 +3223,8 @@ class LUAddMDDRBDComponent(LogicalUnit): # the device exists now # call the primary node to add the mirror to md logger.Info("adding new mirror component to md") - if not rpc.call_blockdev_addchild(instance.primary_node, - disk, new_drbd): + if not rpc.call_blockdev_addchildren(instance.primary_node, + disk, [new_drbd]): logger.Error("Can't add mirror compoment to md!") self.cfg.SetDiskID(new_drbd, remote_node) if not rpc.call_blockdev_remove(remote_node, new_drbd): @@ -2966,7 +3290,8 @@ class LURemoveMDDRBDComponent(LogicalUnit): raise errors.OpPrereqError("Can't find this device ('%s') in the" " instance." % self.op.disk_name) for child in disk.children: - if child.dev_type == "drbd" and child.logical_id[2] == self.op.disk_id: + if (child.dev_type == constants.LD_DRBD7 and + child.logical_id[2] == self.op.disk_id): break else: raise errors.OpPrereqError("Can't find the device with this port.") @@ -2991,8 +3316,8 @@ class LURemoveMDDRBDComponent(LogicalUnit): child = self.child logger.Info("remove mirror component") self.cfg.SetDiskID(disk, instance.primary_node) - if not rpc.call_blockdev_removechild(instance.primary_node, - disk, child): + if not rpc.call_blockdev_removechildren(instance.primary_node, + disk, [child]): raise errors.OpExecError("Can't remove child from mirror.") for node in child.logical_id[:2]: @@ -3072,7 +3397,6 @@ class LUReplaceDisks(LogicalUnit): # start of work remote_node = self.op.remote_node cfg = self.cfg - vgname = cfg.GetVGName() for dev in instance.disks: size = dev.size lv_names = [".%s_%s" % (dev.iv_name, suf) for suf in ["data", "meta"]] @@ -3103,8 +3427,8 @@ class LUReplaceDisks(LogicalUnit): # the device exists now # call the primary node to add the mirror to md logger.Info("adding new mirror component to md") - if not rpc.call_blockdev_addchild(instance.primary_node, dev, - new_drbd): + if not rpc.call_blockdev_addchildren(instance.primary_node, dev, + [new_drbd]): logger.Error("Can't add mirror compoment to md!") cfg.SetDiskID(new_drbd, remote_node) if not rpc.call_blockdev_remove(remote_node, new_drbd): @@ -3138,8 +3462,8 @@ class LUReplaceDisks(LogicalUnit): dev, child, new_drbd = iv_names[name] logger.Info("remove mirror %s component" % name) cfg.SetDiskID(dev, instance.primary_node) - if not rpc.call_blockdev_removechild(instance.primary_node, - dev, child): + if not rpc.call_blockdev_removechildren(instance.primary_node, + dev, [child]): logger.Error("Can't remove child from mirror, aborting" " *this device cleanup*.\nYou need to cleanup manually!!") continue @@ -3190,7 +3514,7 @@ class LUQueryInstanceData(NoHooksLU): """ self.cfg.SetDiskID(dev, instance.primary_node) dev_pstatus = rpc.call_blockdev_find(instance.primary_node, dev) - if dev.dev_type == "drbd": + if dev.dev_type in constants.LDS_DRBD: # we change the snode then (otherwise we use the one passed in) if dev.logical_id[0] == instance.primary_node: snode = dev.logical_id[1] @@ -3249,6 +3573,7 @@ class LUQueryInstanceData(NoHooksLU): "memory": instance.memory, "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics], "disks": disks, + "vcpus": instance.vcpus, } result[instance.name] = idict @@ -3256,37 +3581,6 @@ class LUQueryInstanceData(NoHooksLU): return result -class LUQueryNodeData(NoHooksLU): - """Logical unit for querying node data. - - """ - _OP_REQP = ["nodes"] - - def CheckPrereq(self): - """Check prerequisites. - - This only checks the optional node list against the existing names. - - """ - self.wanted_nodes = _GetWantedNodes(self, self.op.nodes) - - def Exec(self, feedback_fn): - """Compute and return the list of nodes. - - """ - ilist = [self.cfg.GetInstanceInfo(iname) for iname - in self.cfg.GetInstanceList()] - result = [] - for node in self.wanted_nodes: - result.append((node.name, node.primary_ip, node.secondary_ip, - [inst.name for inst in ilist - if inst.primary_node == node.name], - [inst.name for inst in ilist - if node.name in inst.secondary_nodes], - )) - return result - - class LUSetInstanceParms(LogicalUnit): """Modifies an instances's parameters. @@ -3409,7 +3703,7 @@ class LUQueryExports(NoHooksLU): that node. """ - return rpc.call_export_list([node.name for node in self.nodes]) + return rpc.call_export_list(self.nodes) class LUExportInstance(LogicalUnit): @@ -3482,7 +3776,7 @@ class LUExportInstance(LogicalUnit): logger.Error("could not snapshot block device %s on node %s" % (disk.logical_id[1], src_node)) else: - new_dev = objects.Disk(dev_type="lvm", size=disk.size, + new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size, logical_id=(vgname, new_dev_name), physical_id=(vgname, new_dev_name), iv_name=disk.iv_name) @@ -3546,7 +3840,7 @@ class TagsLU(NoHooksLU): self.op.name = name self.target = self.cfg.GetNodeInfo(name) elif self.op.kind == constants.TAG_INSTANCE: - name = self.cfg.ExpandInstanceName(name) + name = self.cfg.ExpandInstanceName(self.op.name) if name is None: raise errors.OpPrereqError("Invalid instance name (%s)" % (self.op.name,)) @@ -3570,11 +3864,11 @@ class LUGetTags(TagsLU): return self.target.GetTags() -class LUAddTag(TagsLU): +class LUAddTags(TagsLU): """Sets a tag on a given object. """ - _OP_REQP = ["kind", "name", "tag"] + _OP_REQP = ["kind", "name", "tags"] def CheckPrereq(self): """Check prerequisites. @@ -3583,14 +3877,16 @@ class LUAddTag(TagsLU): """ TagsLU.CheckPrereq(self) - objects.TaggableObject.ValidateTag(self.op.tag) + for tag in self.op.tags: + objects.TaggableObject.ValidateTag(tag) def Exec(self, feedback_fn): """Sets the tag. """ try: - self.target.AddTag(self.op.tag) + for tag in self.op.tags: + self.target.AddTag(tag) except errors.TagError, err: raise errors.OpExecError("Error while setting tag: %s" % str(err)) try: @@ -3601,11 +3897,11 @@ class LUAddTag(TagsLU): " aborted. Please retry.") -class LUDelTag(TagsLU): - """Delete a tag from a given object. +class LUDelTags(TagsLU): + """Delete a list of tags from a given object. """ - _OP_REQP = ["kind", "name", "tag"] + _OP_REQP = ["kind", "name", "tags"] def CheckPrereq(self): """Check prerequisites. @@ -3614,15 +3910,23 @@ class LUDelTag(TagsLU): """ TagsLU.CheckPrereq(self) - objects.TaggableObject.ValidateTag(self.op.tag) - if self.op.tag not in self.target.GetTags(): - raise errors.OpPrereqError("Tag not found") + for tag in self.op.tags: + objects.TaggableObject.ValidateTag(tag) + del_tags = frozenset(self.op.tags) + cur_tags = self.target.GetTags() + if not del_tags <= cur_tags: + diff_tags = del_tags - cur_tags + diff_names = ["'%s'" % tag for tag in diff_tags] + diff_names.sort() + raise errors.OpPrereqError("Tag(s) %s not found" % + (",".join(diff_names))) def Exec(self, feedback_fn): """Remove the tag from the object. """ - self.target.RemoveTag(self.op.tag) + for tag in self.op.tags: + self.target.RemoveTag(tag) try: self.cfg.Update(self.target) except errors.ConfigurationError: