X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/c8a0948f2e864c310a6e2a52c03fd8df4033237a..fc490dbe206e4c89b75d0f3e96ff4418242bac34:/lib/cmdlib.py diff --git a/lib/cmdlib.py b/lib/cmdlib.py index 3cb1156..2e26cfb 100644 --- a/lib/cmdlib.py +++ b/lib/cmdlib.py @@ -163,8 +163,16 @@ class NoHooksLU(LogicalUnit): return {}, [], [] +def _AddHostToEtcHosts(hostname): + """Wrapper around utils.SetEtcHostsEntry. + + """ + hi = utils.HostInfo(name=hostname) + utils.SetEtcHostsEntry(constants.ETC_HOSTS, hi.ip, hi.name, [hi.ShortName()]) + + def _RemoveHostFromEtcHosts(hostname): - """Wrapper around utils.RemoteEtcHostsEntry. + """Wrapper around utils.RemoveEtcHostsEntry. """ hi = utils.HostInfo(name=hostname) @@ -259,11 +267,12 @@ def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status, if nics: nic_count = len(nics) - for idx, (ip, bridge) in enumerate(nics): + for idx, (ip, bridge, mac) in enumerate(nics): if ip is None: ip = "" env["INSTANCE_NIC%d_IP" % idx] = ip env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge + env["INSTANCE_NIC%d_HWADDR" % idx] = mac else: nic_count = 0 @@ -287,7 +296,7 @@ def _BuildInstanceHookEnvByObject(instance, override=None): 'status': instance.os, 'memory': instance.memory, 'vcpus': instance.vcpus, - 'nics': [(nic.ip, nic.bridge) for nic in instance.nics], + 'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics], } if override: args.update(override) @@ -491,17 +500,23 @@ class LUInitCluster(LogicalUnit): 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 /etc/hosts." % - (hostname.ip,)) + " range (%s). Please fix DNS or %s." % + (hostname.ip, constants.ETC_HOSTS)) self.clustername = clustername = utils.HostInfo(self.op.cluster_name) - if not utils.TcpPing(constants.LOCALHOST_IP_ADDRESS, hostname.ip, - constants.DEFAULT_NODED_PORT): + 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." @@ -512,10 +527,10 @@ class LUInitCluster(LogicalUnit): raise errors.OpPrereqError("Invalid secondary ip given") 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." % + (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 @@ -530,7 +545,7 @@ class LUInitCluster(LogicalUnit): raise errors.OpPrereqError("Invalid mac prefix given '%s'" % self.op.mac_prefix) - if self.op.hypervisor_type not in hypervisor.VALID_HTYPES: + if self.op.hypervisor_type not in constants.HYPER_TYPES: raise errors.OpPrereqError("Invalid hypervisor type given '%s'" % self.op.hypervisor_type) @@ -542,8 +557,8 @@ class LUInitCluster(LogicalUnit): 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) + raise errors.OpPrereqError("Init.d script '%s' missing or not" + " executable." % constants.NODE_INITD_SCRIPT) def Exec(self, feedback_fn): """Initialize the cluster. @@ -574,10 +589,7 @@ class LUInitCluster(LogicalUnit): f.close() sshkey = sshline.split(" ")[1] - hi = utils.HostInfo(name=hostname.name) - utils.AddEtcHostsEntry(constants.ETC_HOSTS, hostname.name, hi.ip) - utils.AddEtcHostsEntry(constants.ETC_HOSTS, hi.ShortName(), hi.ip) - del hi + _AddHostToEtcHosts(hostname.name) _UpdateKnownHosts(hostname.name, hostname.ip, sshkey) @@ -620,11 +632,12 @@ class LUDestroyCluster(NoHooksLU): """ master = self.sstore.GetMasterNode() + if not rpc.call_node_stop_master(master): + raise errors.OpExecError("Could not disable the master role") priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS) utils.CreateBackup(priv_key) utils.CreateBackup(pub_key) rpc.call_node_leave_cluster(master) - _RemoveHostFromEtcHosts(master) class LUVerifyCluster(NoHooksLU): @@ -792,7 +805,8 @@ class LUVerifyCluster(NoHooksLU): """ bad = False feedback_fn("* Verifying global settings") - self.cfg.VerifyConfig() + for msg in self.cfg.VerifyConfig(): + feedback_fn(" - ERROR: %s" % msg) vg_name = self.cfg.GetVGName() nodelist = utils.NiceSort(self.cfg.GetNodeList()) @@ -829,12 +843,17 @@ class LUVerifyCluster(NoHooksLU): # node_volume volumeinfo = all_volumeinfo[node] - if type(volumeinfo) != dict: + if isinstance(volumeinfo, basestring): + feedback_fn(" - ERROR: LVM problem on node %s: %s" % + (node, volumeinfo[-400:].encode('string_escape'))) + bad = True + node_volume[node] = {} + elif not isinstance(volumeinfo, dict): feedback_fn(" - ERROR: connection to %s failed" % (node,)) bad = True continue - - node_volume[node] = volumeinfo + else: + node_volume[node] = volumeinfo # node_instance nodeinstance = all_instanceinfo[node] @@ -870,6 +889,78 @@ class LUVerifyCluster(NoHooksLU): return int(bad) +class LUVerifyDisks(NoHooksLU): + """Verifies the cluster disks status. + + """ + _OP_REQP = [] + + def CheckPrereq(self): + """Check prerequisites. + + This has no prerequisites. + + """ + pass + + def Exec(self, feedback_fn): + """Verify integrity of cluster disks. + + """ + result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {} + + vg_name = self.cfg.GetVGName() + nodes = utils.NiceSort(self.cfg.GetNodeList()) + instances = [self.cfg.GetInstanceInfo(name) + for name in self.cfg.GetInstanceList()] + + nv_dict = {} + for inst in instances: + inst_lvs = {} + if (inst.status != "up" or + inst.disk_template not in constants.DTS_NET_MIRROR): + continue + inst.MapLVsByNode(inst_lvs) + # transform { iname: {node: [vol,],},} to {(node, vol): iname} + for node, vol_list in inst_lvs.iteritems(): + for vol in vol_list: + nv_dict[(node, vol)] = inst + + if not nv_dict: + return result + + node_lvs = rpc.call_volume_list(nodes, vg_name) + + to_act = set() + for node in nodes: + # node_volume + lvs = node_lvs[node] + + if isinstance(lvs, basestring): + logger.Info("error enumerating LVs on node %s: %s" % (node, lvs)) + res_nlvm[node] = lvs + elif not isinstance(lvs, dict): + logger.Info("connection to node %s failed or invalid data returned" % + (node,)) + res_nodes.append(node) + continue + + for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems(): + inst = nv_dict.pop((node, lv_name), None) + if (not lv_online and inst is not None + and inst.name not in res_instances): + res_instances.append(inst.name) + + # any leftover items in nv_dict are missing LVs, let's arrange the + # data better + for key, inst in nv_dict.iteritems(): + if inst.name not in res_missing: + res_missing[inst.name] = [] + res_missing[inst.name].append(key) + + return result + + class LURenameCluster(LogicalUnit): """Rename the cluster. @@ -883,7 +974,7 @@ class LURenameCluster(LogicalUnit): """ env = { - "OP_TARGET": self.op.sstore.GetClusterName(), + "OP_TARGET": self.sstore.GetClusterName(), "NEW_NAME": self.op.name, } mn = self.sstore.GetMasterNode() @@ -945,8 +1036,8 @@ class LURenameCluster(LogicalUnit): (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.") + logger.Error("Could not re-enable the master role on the master," + " please restart manually.") def _WaitForSync(cfgw, instance, proc, oneshot=False, unlock=False): @@ -1386,23 +1477,25 @@ class LUAddNode(LogicalUnit): " new node doesn't have one") # checks reachablity - if not utils.TcpPing(utils.HostInfo().name, - primary_ip, - constants.DEFAULT_NODED_PORT): + if not utils.TcpPing(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 - 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") + if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, + source=myself.secondary_ip): + 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, 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. @@ -1484,10 +1577,7 @@ 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 - hi = utils.HostInfo(name=new_node.name) - utils.AddEtcHostsEntry(constants.ETC_HOSTS, new_node.name, hi.ip) - utils.AddEtcHostsEntry(constants.ETC_HOSTS, hi.ShortName(), hi.ip) - del hi + _AddHostToEtcHosts(new_node.name) _UpdateKnownHosts(new_node.name, new_node.primary_ip, self.cfg.GetHostKey()) @@ -1498,16 +1588,15 @@ class LUAddNode(LogicalUnit): 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) + raise errors.OpExecError("Node claims it doesn't have the secondary ip" + " you gave (%s). 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." % + " than the one the resolver gives: %s." + " Please fix and re-run this command." % (node, msg)) # Distribute updated /etc/hosts and known_hosts to all nodes, @@ -1518,7 +1607,7 @@ class LUAddNode(LogicalUnit): dist_nodes.remove(myself.name) logger.Debug("Copying hosts and known_hosts to all nodes") - for fname in ("/etc/hosts", constants.SSH_KNOWN_HOSTS_FILE): + for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE): result = rpc.call_upload_file(dist_nodes, fname) for to_node in dist_nodes: if not result[to_node]: @@ -1526,6 +1615,8 @@ class LUAddNode(LogicalUnit): (fname, to_node)) to_copy = ss.GetFileList() + if self.sstore.GetHypervisorType() == constants.HT_XEN_HVM31: + to_copy.append(constants.VNC_PASSWORD_FILE) for fname in to_copy: if not ssh.CopyFileToNode(node, fname): logger.Error("could not copy file %s to node %s" % (fname, node)) @@ -1570,8 +1661,8 @@ class LUMasterFailover(LogicalUnit): 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.\n" - "%s is already the master" % + " where you want the new master to be." + " %s is already the master" % self.old_master) def Exec(self, feedback_fn): @@ -1600,8 +1691,8 @@ class LUMasterFailover(LogicalUnit): 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,\n" - "please fix manually.") + feedback_fn("Error in activating the master IP on the new master," + " please fix manually.") @@ -1768,22 +1859,41 @@ def _AssembleInstanceDisks(instance, cfg, ignore_secondaries=False): """ device_info = [] disks_ok = True + iname = instance.name + # With the two passes mechanism we try to reduce the window of + # opportunity for the race condition of switching DRBD to primary + # before handshaking occured, but we do not eliminate it + + # The proper fix would be to wait (with some limits) until the + # connection has been made and drbd transitions from WFConnection + # into any other network-connected state (Connected, SyncTarget, + # SyncSource, etc.) + + # 1st pass, assemble on all nodes in secondary mode for inst_disk in instance.disks: - master_result = None for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): cfg.SetDiskID(node_disk, node) - is_primary = node == instance.primary_node - result = rpc.call_blockdev_assemble(node, node_disk, - instance.name, is_primary) + result = rpc.call_blockdev_assemble(node, node_disk, iname, False) if not result: - logger.Error("could not prepare block device %s on node %s (is_pri" - "mary=%s)" % (inst_disk.iv_name, node, is_primary)) - if is_primary or not ignore_secondaries: + logger.Error("could not prepare block device %s on node %s" + " (is_primary=False, pass=1)" % (inst_disk.iv_name, node)) + if not ignore_secondaries: disks_ok = False - if is_primary: - master_result = result - device_info.append((instance.primary_node, inst_disk.iv_name, - master_result)) + + # FIXME: race condition on drbd migration to primary + + # 2nd pass, do only the primary node + for inst_disk in instance.disks: + for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node): + if node != instance.primary_node: + continue + cfg.SetDiskID(node_disk, node) + result = rpc.call_blockdev_assemble(node, node_disk, iname, True) + if not result: + logger.Error("could not prepare block device %s on node %s" + " (is_primary=True, pass=2)" % (inst_disk.iv_name, node)) + disks_ok = False + device_info.append((instance.primary_node, inst_disk.iv_name, result)) # leave the disks configured for the primary node # this is a workaround that would be fixed better by @@ -1866,6 +1976,36 @@ def _ShutdownInstanceDisks(instance, cfg, ignore_primary=False): return result +def _CheckNodeFreeMemory(cfg, node, reason, requested): + """Checks if a node has enough free memory. + + This function check if a given node has the needed amount of free + memory. In case the node has less memory or we cannot get the + information from the node, this function raise an OpPrereqError + exception. + + Args: + - cfg: a ConfigWriter instance + - node: the node name + - reason: string to use in the error message + - requested: the amount of memory in MiB + + """ + nodeinfo = rpc.call_node_info([node], cfg.GetVGName()) + if not nodeinfo or not isinstance(nodeinfo, dict): + raise errors.OpPrereqError("Could not contact node %s for resource" + " information" % (node,)) + + free_mem = nodeinfo[node].get('memory_free') + if not isinstance(free_mem, int): + raise errors.OpPrereqError("Can't compute free memory on node %s, result" + " was '%s'" % (node, free_mem)) + if requested > free_mem: + raise errors.OpPrereqError("Not enough memory on node %s for %s:" + " needed %s MiB, available %s MiB" % + (node, reason, requested, free_mem)) + + class LUStartupInstance(LogicalUnit): """Starts an instance. @@ -1903,6 +2043,10 @@ class LUStartupInstance(LogicalUnit): # check bridges existance _CheckInstanceBridgesExist(instance) + _CheckNodeFreeMemory(self.cfg, instance.primary_node, + "starting instance %s" % instance.name, + instance.memory) + self.instance = instance self.op.instance_name = instance.name @@ -1916,20 +2060,6 @@ class LUStartupInstance(LogicalUnit): node_current = instance.primary_node - nodeinfo = rpc.call_node_info([node_current], self.cfg.GetVGName()) - if not nodeinfo: - raise errors.OpExecError("Could not contact node %s for infos" % - (node_current)) - - freememory = nodeinfo[node_current]['memory_free'] - memory = instance.memory - if memory > freememory: - raise errors.OpExecError("Not enough memory to start instance" - " %s on node %s" - " needed %s MiB, available %s MiB" % - (instance.name, node_current, memory, - freememory)) - _StartInstanceDisks(self.cfg, instance, force) if not rpc.call_instance_start(node_current, instance, extra_args): @@ -2132,8 +2262,8 @@ class LUReinstallInstance(LogicalUnit): try: feedback_fn("Running the instance OS create scripts...") if not rpc.call_instance_os_add(inst.primary_node, inst, "sda", "sdb"): - raise errors.OpExecError("Could not install OS for instance %s " - "on node %s" % + raise errors.OpExecError("Could not install OS for instance %s" + " on node %s" % (inst.name, inst.primary_node)) finally: _ShutdownInstanceDisks(inst, self.cfg) @@ -2184,6 +2314,11 @@ class LURenameInstance(LogicalUnit): name_info = utils.HostInfo(self.op.new_name) self.op.new_name = new_name = name_info.name + instance_list = self.cfg.GetInstanceList() + if new_name in instance_list: + raise errors.OpPrereqError("Instance '%s' is already in the cluster" % + instance_name) + if not getattr(self.op, "ignore_ip", False): command = ["fping", "-q", name_info.ip] result = utils.RunCmd(command) @@ -2208,9 +2343,8 @@ class LURenameInstance(LogicalUnit): 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)" % + msg = ("Could run OS rename script for instance %s on node %s (but the" + " instance has been renamed in Ganeti)" % (inst.name, inst.primary_node)) logger.Error(msg) finally: @@ -2288,11 +2422,11 @@ class LUQueryInstances(NoHooksLU): This checks that the fields required are valid output fields. """ - self.dynamic_fields = frozenset(["oper_state", "oper_ram"]) + self.dynamic_fields = frozenset(["oper_state", "oper_ram", "status"]) _CheckOutputFields(static=["name", "os", "pnode", "snodes", "admin_state", "admin_ram", "disk_template", "ip", "mac", "bridge", - "sda_size", "sdb_size"], + "sda_size", "sdb_size", "vcpus"], dynamic=self.dynamic_fields, selected=self.op.output_fields) @@ -2345,6 +2479,21 @@ class LUQueryInstances(NoHooksLU): val = None else: val = bool(live_data.get(instance.name)) + elif field == "status": + if instance.primary_node in bad_nodes: + val = "ERROR_nodedown" + else: + running = bool(live_data.get(instance.name)) + if running: + if instance.status != "down": + val = "running" + else: + val = "ERROR_up" + else: + if instance.status != "down": + val = "ERROR_down" + else: + val = "ADMIN_down" elif field == "admin_ram": val = instance.memory elif field == "oper_ram": @@ -2368,6 +2517,8 @@ class LUQueryInstances(NoHooksLU): val = None else: val = disk.size + elif field == "vcpus": + val = instance.vcpus else: raise errors.ParameterError(field) iout.append(val) @@ -2418,18 +2569,10 @@ class LUFailoverInstance(LogicalUnit): raise errors.ProgrammerError("no secondary node but using " "DT_REMOTE_RAID1 template") - # check memory requirements on the secondary node target_node = secondary_nodes[0] - nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName()) - info = nodeinfo.get(target_node, None) - if not info: - raise errors.OpPrereqError("Cannot get current information" - " from node '%s'" % nodeinfo) - if instance.memory > info['memory_free']: - raise errors.OpPrereqError("Not enough memory on target node %s." - " %d MB available, %d MB required" % - (target_node, info['memory_free'], - instance.memory)) + # check memory requirements on the secondary node + _CheckNodeFreeMemory(self.cfg, target_node, "failing over instance %s" % + instance.name, instance.memory) # check bridge existance brlist = [nic.bridge for nic in instance.nics] @@ -2460,21 +2603,6 @@ class LUFailoverInstance(LogicalUnit): raise errors.OpExecError("Disk %s is degraded on target node," " aborting failover." % dev.iv_name) - feedback_fn("* checking target node resource availability") - nodeinfo = rpc.call_node_info([target_node], self.cfg.GetVGName()) - - if not nodeinfo: - raise errors.OpExecError("Could not contact target node %s." % - target_node) - - free_memory = int(nodeinfo[target_node]['memory_free']) - memory = instance.memory - if memory > free_memory: - raise errors.OpExecError("Not enough memory to create instance %s on" - " node %s. needed %s MiB, available %s MiB" % - (instance.name, target_node, memory, - free_memory)) - feedback_fn("* shutting down instance on source node") logger.Info("Shutting down instance %s on node %s" % (instance.name, source_node)) @@ -2617,9 +2745,9 @@ def _GenerateDiskTemplate(cfg, template_name, #TODO: compute space requirements vgname = cfg.GetVGName() - if template_name == "diskless": + if template_name == constants.DT_DISKLESS: disks = [] - elif template_name == "plain": + elif template_name == constants.DT_PLAIN: if len(secondary_nodes) != 0: raise errors.ProgrammerError("Wrong template configuration") @@ -2631,7 +2759,7 @@ def _GenerateDiskTemplate(cfg, template_name, logical_id=(vgname, names[1]), iv_name = "sdb") disks = [sda_dev, sdb_dev] - elif template_name == "local_raid1": + elif template_name == constants.DT_LOCAL_RAID1: if len(secondary_nodes) != 0: raise errors.ProgrammerError("Wrong template configuration") @@ -2761,7 +2889,7 @@ 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", "ip_check"] + "wait_for_sync", "ip_check", "mac"] def BuildHooksEnv(self): """Build hooks env. @@ -2787,7 +2915,7 @@ class LUCreateInstance(LogicalUnit): os_type=self.op.os_type, memory=self.op.mem_size, vcpus=self.op.vcpus, - nics=[(self.inst_ip, self.op.bridge)], + nics=[(self.inst_ip, self.op.bridge, self.op.mac)], )) nl = ([self.sstore.GetMasterNode(), self.op.pnode] + @@ -2799,6 +2927,10 @@ class LUCreateInstance(LogicalUnit): """Check prerequisites. """ + for attr in ["kernel_path", "initrd_path", "hvm_boot_order"]: + if not hasattr(self.op, attr): + setattr(self.op, attr, None) + if self.op.mode not in (constants.INSTANCE_CREATE, constants.INSTANCE_IMPORT): raise errors.OpPrereqError("Invalid instance creation mode '%s'" % @@ -2870,13 +3002,9 @@ class LUCreateInstance(LogicalUnit): " the primary node.") self.secondaries.append(snode_name) - # Check lv size requirements - nodenames = [pnode.name] + self.secondaries - nodeinfo = rpc.call_node_info(nodenames, self.cfg.GetVGName()) - # Required free disk space as a function of disk and swap space req_size_dict = { - constants.DT_DISKLESS: 0, + constants.DT_DISKLESS: None, constants.DT_PLAIN: self.op.disk_size + self.op.swap_size, 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 @@ -2890,15 +3018,23 @@ class LUCreateInstance(LogicalUnit): req_size = req_size_dict[self.op.disk_template] - for node in nodenames: - info = nodeinfo.get(node, None) - if not info: - raise errors.OpPrereqError("Cannot get current information" - " from node '%s'" % nodeinfo) - if req_size > info['vg_free']: - raise errors.OpPrereqError("Not enough disk space on target node %s." - " %d MB available, %d MB required" % - (node, info['vg_free'], req_size)) + # Check lv size requirements + if req_size is not None: + nodenames = [pnode.name] + self.secondaries + 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'" % nodeinfo) + 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 req_size > info['vg_free']: + raise errors.OpPrereqError("Not enough disk space on target node %s." + " %d MB available, %d MB required" % + (node, info['vg_free'], req_size)) # os verification os_obj = rpc.call_os_get(pnode.name, self.op.os_type) @@ -2906,6 +3042,9 @@ class LUCreateInstance(LogicalUnit): raise errors.OpPrereqError("OS '%s' not in supported os list for" " primary node" % self.op.os_type) + if self.op.kernel_path == constants.VALUE_NONE: + raise errors.OpPrereqError("Can't set instance kernel to none") + # instance verification hostname1 = utils.HostInfo(self.op.instance_name) @@ -2932,11 +3071,16 @@ class LUCreateInstance(LogicalUnit): " adding an instance in start mode") if self.op.ip_check: - if utils.TcpPing(utils.HostInfo().name, hostname1.ip, - constants.DEFAULT_NODED_PORT): + if utils.TcpPing(hostname1.ip, constants.DEFAULT_NODED_PORT): raise errors.OpPrereqError("IP %s of instance %s already in use" % (hostname1.ip, instance_name)) + # MAC address verification + if self.op.mac != "auto": + if not utils.IsValidMac(self.op.mac.lower()): + raise errors.OpPrereqError("invalid MAC address specified: %s" % + self.op.mac) + # bridge verification bridge = getattr(self.op, "bridge", None) if bridge is None: @@ -2949,6 +3093,12 @@ class LUCreateInstance(LogicalUnit): " destination node '%s'" % (self.op.bridge, pnode.name)) + # boot order verification + if self.op.hvm_boot_order is not None: + if len(self.op.hvm_boot_order.strip("acdn")) != 0: + raise errors.OpPrereqError("invalid boot order specified," + " must be one or more of [acdn]") + if self.op.start: self.instance_status = 'up' else: @@ -2961,10 +3111,21 @@ class LUCreateInstance(LogicalUnit): instance = self.op.instance_name pnode_name = self.pnode.name - nic = objects.NIC(bridge=self.op.bridge, mac=self.cfg.GenerateMAC()) + if self.op.mac == "auto": + mac_address = self.cfg.GenerateMAC() + else: + mac_address = self.op.mac + + nic = objects.NIC(bridge=self.op.bridge, mac=mac_address) if self.inst_ip is not None: nic.ip = self.inst_ip + ht_kind = self.sstore.GetHypervisorType() + if ht_kind in constants.HTS_REQ_PORT: + network_port = self.cfg.AllocatePort() + else: + network_port = None + disks = _GenerateDiskTemplate(self.cfg, self.op.disk_template, instance, pnode_name, @@ -2978,6 +3139,10 @@ class LUCreateInstance(LogicalUnit): nics=[nic], disks=disks, disk_template=self.op.disk_template, status=self.instance_status, + network_port=network_port, + kernel_path=self.op.kernel_path, + initrd_path=self.op.initrd_path, + hvm_boot_order=self.op.hvm_boot_order, ) feedback_fn("* creating instance disks...") @@ -3077,7 +3242,7 @@ class LUConnectConsole(NoHooksLU): logger.Debug("connecting to console of %s on %s" % (instance.name, node)) hyper = hypervisor.GetHypervisor() - console_cmd = hyper.GetShellCommandForConsole(instance.name) + console_cmd = hyper.GetShellCommandForConsole(instance) # build ssh cmdline argv = ["ssh", "-q", "-t"] argv.extend(ssh.KNOWN_HOSTS_OPTS) @@ -3142,10 +3307,9 @@ class LUAddMDDRBDComponent(LogicalUnit): raise errors.OpPrereqError("Can't find this device ('%s') in the" " instance." % self.op.disk_name) if len(disk.children) > 1: - raise errors.OpPrereqError("The device already has two slave" - " devices.\n" - "This would create a 3-disk raid1" - " which we don't allow.") + raise errors.OpPrereqError("The device already has two slave devices." + " This would create a 3-disk raid1 which we" + " don't allow.") self.disk = disk def Exec(self, feedback_fn): @@ -3329,6 +3493,7 @@ class LUReplaceDisks(LogicalUnit): 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.DTS_NET_MIRROR: raise errors.OpPrereqError("Instance's disk layout is not" @@ -3367,13 +3532,18 @@ class LUReplaceDisks(LogicalUnit): raise errors.OpPrereqError("Template 'remote_raid1' only allows all" " disks replacement, not individual ones") if instance.disk_template == constants.DT_DRBD8: + if (self.op.mode == constants.REPLACE_DISK_ALL and + remote_node is not None): + # switch to replace secondary mode + self.op.mode = constants.REPLACE_DISK_SEC + if self.op.mode == constants.REPLACE_DISK_ALL: - raise errors.OpPrereqError("Template 'drbd8' only allows primary or" + raise errors.OpPrereqError("Template 'drbd' only allows primary or" " secondary disk replacement, not" " both at once") elif self.op.mode == constants.REPLACE_DISK_PRI: if remote_node is not None: - raise errors.OpPrereqError("Template 'drbd8' does not allow changing" + raise errors.OpPrereqError("Template 'drbd' does not allow changing" " the secondary while doing a primary" " node disk replacement") self.tgt_node = instance.primary_node @@ -3417,9 +3587,8 @@ class LUReplaceDisks(LogicalUnit): if not _CreateBlockDevOnSecondary(cfg, remote_node, instance, new_drbd, False, _GetInstanceInfoText(instance)): - raise errors.OpExecError("Failed to create new component on" - " secondary node %s\n" - "Full abort, cleanup manually!" % + raise errors.OpExecError("Failed to create new component on secondary" + " node %s. Full abort, cleanup manually!" % remote_node) logger.Info("adding new mirror component on primary") @@ -3430,8 +3599,8 @@ class LUReplaceDisks(LogicalUnit): # remove secondary dev cfg.SetDiskID(new_drbd, remote_node) rpc.call_blockdev_remove(remote_node, new_drbd) - raise errors.OpExecError("Failed to create volume on primary!\n" - "Full abort, cleanup manually!!") + raise errors.OpExecError("Failed to create volume on primary!" + " Full abort, cleanup manually!!") # the device exists now # call the primary node to add the mirror to md @@ -3590,7 +3759,7 @@ class LUReplaceDisks(LogicalUnit): # ok, we created the new LVs, so now we know we have the needed # storage; as such, we proceed on the target node to rename # old_lv to _old, and new_lv to old_lv; note that we rename LVs - # using the assumption than logical_id == physical_id (which in + # using the assumption that logical_id == physical_id (which in # turn is the unique_id on that node) # FIXME(iustin): use a better name for the replaced LVs @@ -3626,7 +3795,7 @@ class LUReplaceDisks(LogicalUnit): if not rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs): for new_lv in new_lvs: if not rpc.call_blockdev_remove(tgt_node, new_lv): - warning("Can't rollback device %s", "manually cleanup unused" + warning("Can't rollback device %s", hint="manually cleanup unused" " logical volumes") raise errors.OpExecError("Can't add local storage to drbd") @@ -3655,7 +3824,7 @@ class LUReplaceDisks(LogicalUnit): for lv in old_lvs: cfg.SetDiskID(lv, tgt_node) if not rpc.call_blockdev_remove(tgt_node, lv): - warning("Can't remove old LV", "manually remove unused LVs") + warning("Can't remove old LV", hint="manually remove unused LVs") continue def _ExecD8Secondary(self, feedback_fn): @@ -3758,23 +3927,47 @@ class LUReplaceDisks(LogicalUnit): cfg.SetDiskID(dev, old_node) if not rpc.call_blockdev_shutdown(old_node, dev): warning("Failed to shutdown drbd for %s on old node" % dev.iv_name, - "Please cleanup this device manuall as soon as possible") + hint="Please cleanup this device manually as soon as possible") - # we have new storage, we 'rename' the network on the primary - info("switching primary drbd for %s to new secondary node" % dev.iv_name) + info("detaching primary drbds from the network (=> standalone)") + done = 0 + for dev in instance.disks: cfg.SetDiskID(dev, pri_node) - # rename to the ip of the new node - new_uid = list(dev.physical_id) - new_uid[2] = self.remote_node_info.secondary_ip - rlist = [(dev, tuple(new_uid))] - if not rpc.call_blockdev_rename(pri_node, rlist): - raise errors.OpExecError("Can't detach & re-attach drbd %s on node" - " %s from %s to %s" % - (dev.iv_name, pri_node, old_node, new_node)) - dev.logical_id = (pri_node, new_node, dev.logical_id[2]) + # set the physical (unique in bdev terms) id to None, meaning + # detach from network + dev.physical_id = (None,) * len(dev.physical_id) + # and 'find' the device, which will 'fix' it to match the + # standalone state + if rpc.call_blockdev_find(pri_node, dev): + done += 1 + else: + warning("Failed to detach drbd %s from network, unusual case" % + dev.iv_name) + + if not done: + # no detaches succeeded (very unlikely) + raise errors.OpExecError("Can't detach at least one DRBD from old node") + + # if we managed to detach at least one, we update all the disks of + # the instance to point to the new secondary + info("updating instance configuration") + for dev in instance.disks: + dev.logical_id = (pri_node, new_node) + dev.logical_id[2:] cfg.SetDiskID(dev, pri_node) - cfg.Update(instance) + cfg.Update(instance) + # and now perform the drbd attach + info("attaching primary drbds to new secondary (standalone => connected)") + failures = [] + for dev in instance.disks: + info("attaching primary drbd for %s to new secondary node" % dev.iv_name) + # since the attach is smart, it's enough to 'find' the device, + # it will automatically activate the network, if the physical_id + # is correct + cfg.SetDiskID(dev, pri_node) + if not rpc.call_blockdev_find(pri_node, dev): + warning("can't attach drbd %s to new secondary!" % dev.iv_name, + "please do a gnt-instance info to see the status of disks") # this can fail as the old devices are degraded and _WaitForSync # does a combined result over all disks, so we don't check its @@ -3796,7 +3989,7 @@ class LUReplaceDisks(LogicalUnit): cfg.SetDiskID(lv, old_node) if not rpc.call_blockdev_remove(old_node, lv): warning("Can't remove LV on old secondary", - "Cleanup stale volumes by hand") + hint="Cleanup stale volumes by hand") def Exec(self, feedback_fn): """Execute disk replacement. @@ -3838,7 +4031,7 @@ class LUQueryInstanceData(NoHooksLU): instance = self.cfg.GetInstanceInfo(self.cfg.ExpandInstanceName(name)) if instance is None: raise errors.OpPrereqError("No such instance name '%s'" % name) - self.wanted_instances.append(instance) + self.wanted_instances.append(instance) else: self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name in self.cfg.GetInstanceList()] @@ -3910,7 +4103,11 @@ class LUQueryInstanceData(NoHooksLU): "memory": instance.memory, "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics], "disks": disks, + "network_port": instance.network_port, "vcpus": instance.vcpus, + "kernel_path": instance.kernel_path, + "initrd_path": instance.initrd_path, + "hvm_boot_order": instance.hvm_boot_order, } result[instance.name] = idict @@ -3937,7 +4134,7 @@ class LUSetInstanceParms(LogicalUnit): args['memory'] = self.mem if self.vcpus: args['vcpus'] = self.vcpus - if self.do_ip or self.do_bridge: + if self.do_ip or self.do_bridge or self.mac: if self.do_ip: ip = self.ip else: @@ -3946,7 +4143,11 @@ class LUSetInstanceParms(LogicalUnit): bridge = self.bridge else: bridge = self.instance.nics[0].bridge - args['nics'] = [(ip, bridge)] + if self.mac: + mac = self.mac + else: + mac = self.instance.nics[0].mac + args['nics'] = [(ip, bridge, mac)] env = _BuildInstanceHookEnvByObject(self.instance, override=args) nl = [self.sstore.GetMasterNode(), self.instance.primary_node] + list(self.instance.secondary_nodes) @@ -3961,8 +4162,14 @@ class LUSetInstanceParms(LogicalUnit): self.mem = getattr(self.op, "mem", None) self.vcpus = getattr(self.op, "vcpus", None) self.ip = getattr(self.op, "ip", None) + self.mac = getattr(self.op, "mac", None) self.bridge = getattr(self.op, "bridge", None) - if [self.mem, self.vcpus, self.ip, self.bridge].count(None) == 4: + self.kernel_path = getattr(self.op, "kernel_path", None) + self.initrd_path = getattr(self.op, "initrd_path", None) + self.hvm_boot_order = getattr(self.op, "hvm_boot_order", None) + all_parms = [self.mem, self.vcpus, self.ip, self.bridge, self.mac, + self.kernel_path, self.initrd_path, self.hvm_boot_order] + if all_parms.count(None) == len(all_parms): raise errors.OpPrereqError("No changes submitted") if self.mem is not None: try: @@ -3984,6 +4191,42 @@ class LUSetInstanceParms(LogicalUnit): else: self.do_ip = False self.do_bridge = (self.bridge is not None) + if self.mac is not None: + if self.cfg.IsMacInUse(self.mac): + raise errors.OpPrereqError('MAC address %s already in use in cluster' % + self.mac) + if not utils.IsValidMac(self.mac): + raise errors.OpPrereqError('Invalid MAC address %s' % self.mac) + + if self.kernel_path is not None: + self.do_kernel_path = True + if self.kernel_path == constants.VALUE_NONE: + raise errors.OpPrereqError("Can't set instance to no kernel") + + if self.kernel_path != constants.VALUE_DEFAULT: + if not os.path.isabs(self.kernel_path): + raise errors.OpPrereqError("The kernel path must be an absolute" + " filename") + else: + self.do_kernel_path = False + + if self.initrd_path is not None: + self.do_initrd_path = True + if self.initrd_path not in (constants.VALUE_NONE, + constants.VALUE_DEFAULT): + if not os.path.isabs(self.initrd_path): + raise errors.OpPrereqError("The initrd path must be an absolute" + " filename") + else: + self.do_initrd_path = False + + # boot order verification + if self.hvm_boot_order is not None: + if self.hvm_boot_order != constants.VALUE_DEFAULT: + if len(self.hvm_boot_order.strip("acdn")) != 0: + raise errors.OpPrereqError("invalid boot order specified," + " must be one or more of [acdn]" + " or 'default'") instance = self.cfg.GetInstanceInfo( self.cfg.ExpandInstanceName(self.op.instance_name)) @@ -4013,6 +4256,21 @@ class LUSetInstanceParms(LogicalUnit): if self.bridge: instance.nics[0].bridge = self.bridge result.append(("bridge", self.bridge)) + if self.mac: + instance.nics[0].mac = self.mac + result.append(("mac", self.mac)) + if self.do_kernel_path: + instance.kernel_path = self.kernel_path + result.append(("kernel_path", self.kernel_path)) + if self.do_initrd_path: + instance.initrd_path = self.initrd_path + result.append(("initrd_path", self.initrd_path)) + if self.hvm_boot_order: + if self.hvm_boot_order == constants.VALUE_DEFAULT: + instance.hvm_boot_order = None + else: + instance.hvm_boot_order = self.hvm_boot_order + result.append(("hvm_boot_order", self.hvm_boot_order)) self.cfg.AddInstance(instance) @@ -4306,3 +4564,39 @@ class LUDelTags(TagsLU): raise errors.OpRetryError("There has been a modification to the" " 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 + time. + + """ + _OP_REQP = ["duration", "on_master", "on_nodes"] + + def CheckPrereq(self): + """Check prerequisites. + + This checks that we have a good list of nodes and/or the duration + is valid. + + """ + + if self.op.on_nodes: + self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes) + + def Exec(self, feedback_fn): + """Do the actual sleep. + + """ + if self.op.on_master: + if not utils.TestDelay(self.op.duration): + raise errors.OpExecError("Error during master delay test") + if self.op.on_nodes: + result = rpc.call_test_delay(self.op.on_nodes, self.op.duration) + if not result: + raise errors.OpExecError("Complete failure from rpc call") + for node, node_result in result.items(): + if not node_result: + raise errors.OpExecError("Failure during rpc call to node %s," + " result: %s" % (node, node_result))