X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/ea8ac9c99ca6b59db76bd7f549eb059613aa81b4..8a53b55f3a83f3bdf3b231f90e766fc15ec51895:/lib/bootstrap.py diff --git a/lib/bootstrap.py b/lib/bootstrap.py index 5a07895..d18e68b 100644 --- a/lib/bootstrap.py +++ b/lib/bootstrap.py @@ -42,6 +42,14 @@ from ganeti import hypervisor from ganeti import bdev from ganeti import netutils from ganeti import backend +from ganeti import luxi + + +# ec_id for InitConfig's temporary reservation manager +_INITCONF_ECID = "initconfig-ecid" + +#: After how many seconds daemon must be responsive +_DAEMON_READY_TIMEOUT = 10.0 def _InitSSHSetup(): @@ -178,10 +186,30 @@ def _WaitForNodeDaemon(node_name): raise utils.RetryAgain() try: - utils.Retry(_CheckNodeDaemon, 1.0, 10.0) + utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT) except utils.RetryTimeout: raise errors.OpExecError("Node daemon on %s didn't answer queries within" - " 10 seconds" % node_name) + " %s seconds" % (node_name, _DAEMON_READY_TIMEOUT)) + + +def _WaitForMasterDaemon(): + """Wait for master daemon to become responsive. + + """ + def _CheckMasterDaemon(): + try: + cl = luxi.Client() + (cluster_name, ) = cl.QueryConfigValues(["cluster_name"]) + except Exception: + raise utils.RetryAgain() + + logging.debug("Received cluster name %s from master", cluster_name) + + try: + utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT) + except utils.RetryTimeout: + raise errors.OpExecError("Master daemon didn't answer queries within" + " %s seconds" % _DAEMON_READY_TIMEOUT) def _InitFileStorage(file_storage_dir): @@ -216,15 +244,14 @@ def _InitFileStorage(file_storage_dir): return file_storage_dir -#pylint: disable-msg=R0913 -def InitCluster(cluster_name, mac_prefix, +def InitCluster(cluster_name, mac_prefix, # pylint: disable-msg=R0913 master_netdev, file_storage_dir, candidate_pool_size, secondary_ip=None, vg_name=None, beparams=None, - nicparams=None, hvparams=None, enabled_hypervisors=None, - modify_etc_hosts=True, modify_ssh_setup=True, - maintain_node_health=False, drbd_helper=None, - uid_pool=None, default_iallocator=None, - primary_ip_version=None): + nicparams=None, ndparams=None, hvparams=None, + enabled_hypervisors=None, modify_etc_hosts=True, + modify_ssh_setup=True, maintain_node_health=False, + drbd_helper=None, uid_pool=None, default_iallocator=None, + primary_ip_version=None, prealloc_wipe_disks=False): """Initialise the cluster. @type candidate_pool_size: int @@ -337,6 +364,11 @@ def InitCluster(cluster_name, mac_prefix, utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES) objects.NIC.CheckParameterSyntax(nicparams) + if ndparams is not None: + utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES) + else: + ndparams = dict(constants.NDC_DEFAULTS) + # hvparams is a mapping of hypervisor->hvparams dict for hv_name, hv_params in hvparams.iteritems(): utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) @@ -380,6 +412,7 @@ def InitCluster(cluster_name, mac_prefix, enabled_hypervisors=enabled_hypervisors, beparams={constants.PP_DEFAULT: beparams}, nicparams={constants.PP_DEFAULT: nicparams}, + ndparams=ndparams, hvparams=hvparams, candidate_pool_size=candidate_pool_size, modify_etc_hosts=modify_etc_hosts, @@ -387,11 +420,11 @@ def InitCluster(cluster_name, mac_prefix, uid_pool=uid_pool, ctime=now, mtime=now, - uuid=utils.NewUUID(), maintain_node_health=maintain_node_health, drbd_usermode_helper=drbd_helper, default_iallocator=default_iallocator, primary_ip_family=ipcls.family, + prealloc_wipe_disks=prealloc_wipe_disks, ) master_node_config = objects.Node(name=hostname.name, primary_ip=hostname.ip, @@ -399,6 +432,7 @@ def InitCluster(cluster_name, mac_prefix, serial_no=1, master_candidate=True, offline=False, drained=False, + ctime=now, mtime=now, ) InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) cfg = config.ConfigWriter(offline=True) @@ -409,10 +443,14 @@ def InitCluster(cluster_name, mac_prefix, # set up the inter-node password and certificate _InitGanetiServerSetup(hostname.name) - # start the master ip - # TODO: Review rpc call from bootstrap - # TODO: Warn on failed start master - rpc.RpcRunner.call_node_start_master(hostname.name, True, False) + logging.debug("Starting daemons") + result = utils.RunCmd([constants.DAEMON_UTIL, "start-all"]) + if result.failed: + raise errors.OpExecError("Could not start daemons, command %s" + " had exitcode %s and error %s" % + (result.cmd, result.exit_code, result.output)) + + _WaitForMasterDaemon() def InitConfig(version, cluster_config, master_node_config, @@ -432,13 +470,26 @@ def InitConfig(version, cluster_config, master_node_config, @param cfg_file: configuration file path """ + uuid_generator = config.TemporaryReservationManager() + cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID, + _INITCONF_ECID) + master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID, + _INITCONF_ECID) nodes = { master_node_config.name: master_node_config, } - + default_nodegroup = objects.NodeGroup( + uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID), + name=constants.INITIAL_NODE_GROUP_NAME, + members=[master_node_config.name], + ) + nodegroups = { + default_nodegroup.uuid: default_nodegroup, + } now = time.time() config_data = objects.ConfigData(version=version, cluster=cluster_config, + nodegroups=nodegroups, nodes=nodes, instances={}, serial_no=1, @@ -482,27 +533,7 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check): """ family = ssconf.SimpleStore().GetPrimaryIPFamily() sshrunner = ssh.SshRunner(cluster_name, - ipv6=family==netutils.IP6Address.family) - - noded_cert = utils.ReadFile(constants.NODED_CERT_FILE) - rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE) - confd_hmac_key = utils.ReadFile(constants.CONFD_HMAC_KEY) - - # 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. HMAC keys are hexadecimal strings, - # so the same restrictions apply. - for content in (noded_cert, rapi_cert, confd_hmac_key): - if re.search('^!EOF\.', content, re.MULTILINE): - raise errors.OpExecError("invalid SSL certificate or HMAC key") - - if not noded_cert.endswith("\n"): - noded_cert += "\n" - if not rapi_cert.endswith("\n"): - rapi_cert += "\n" - if not confd_hmac_key.endswith("\n"): - confd_hmac_key += "\n" + ipv6=(family == netutils.IP6Address.family)) bind_address = constants.IP4_ADDRESS_ANY if family == netutils.IP6Address.family: @@ -512,26 +543,16 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check): # 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 - # TODO: Could this command exceed a shell's maximum command length? - mycommand = ("umask 077 && " - "cat > '%s' << '!EOF.' && \n" - "%s!EOF.\n" - "cat > '%s' << '!EOF.' && \n" - "%s!EOF.\n" - "cat > '%s' << '!EOF.' && \n" - "%s!EOF.\n" - "chmod 0400 %s %s %s && " - "%s start %s -b '%s'" % - (constants.NODED_CERT_FILE, noded_cert, - constants.RAPI_CERT_FILE, rapi_cert, - constants.CONFD_HMAC_KEY, confd_hmac_key, - constants.NODED_CERT_FILE, constants.RAPI_CERT_FILE, - constants.CONFD_HMAC_KEY, - constants.DAEMON_UTIL, constants.NODED, bind_address)) + sshrunner.CopyFileToNode(node, constants.NODED_CERT_FILE) + sshrunner.CopyFileToNode(node, constants.RAPI_CERT_FILE) + sshrunner.CopyFileToNode(node, constants.CONFD_HMAC_KEY) + mycommand = ("%s stop-all; %s start %s -b %s" % + (constants.DAEMON_UTIL, constants.DAEMON_UTIL, constants.NODED, + utils.ShellQuote(bind_address))) result = sshrunner.Run(node, 'root', mycommand, batch=False, ask_key=ssh_key_check, - use_cluster_key=False, + use_cluster_key=True, strict_host_check=ssh_key_check) if result.failed: raise errors.OpExecError("Remote command on node %s, error: %s," @@ -595,12 +616,36 @@ def MasterFailover(no_voting=False): logging.info("Setting master to %s, old master: %s", new_master, old_master) + try: + # instantiate a real config writer, as we now know we have the + # configuration data + cfg = config.ConfigWriter(accept_foreign=True) + + cluster_info = cfg.GetClusterInfo() + cluster_info.master_node = new_master + # this will also regenerate the ssconf files, since we updated the + # cluster info + cfg.Update(cluster_info, logging.error) + except errors.ConfigurationError, err: + logging.error("Error while trying to set the new master: %s", + str(err)) + return 1 + + # if cfg.Update worked, then it means the old master daemon won't be + # able now to write its own config file (we rely on locking in both + # backend.UploadFile() and ConfigWriter._Write(); hence the next + # step is to kill the old master + + logging.info("Stopping the master daemon on node %s", old_master) + result = rpc.RpcRunner.call_node_stop_master(old_master, True) msg = result.fail_msg if msg: logging.error("Could not disable the master role on the old master" " %s, please disable manually: %s", old_master, msg) + logging.info("Checking master IP non-reachability...") + master_ip = sstore.GetMasterIP() total_timeout = 30 # Here we have a phase where no master should be running @@ -615,15 +660,7 @@ def MasterFailover(no_voting=False): " continuing but activating the master on the current" " node will probably fail", total_timeout) - # instantiate a real config writer, as we now know we have the - # configuration data - cfg = config.ConfigWriter() - - cluster_info = cfg.GetClusterInfo() - cluster_info.master_node = new_master - # this will also regenerate the ssconf files, since we updated the - # cluster info - cfg.Update(cluster_info, logging.error) + logging.info("Starting the master daemons on the new master") result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting) msg = result.fail_msg @@ -632,6 +669,7 @@ def MasterFailover(no_voting=False): " %s, please check: %s", new_master, msg) rcode = 1 + logging.info("Master failed over from %s to %s", old_master, new_master) return rcode