X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/31684f840bb8b321e958db75c6c8cd96a606ca28..905108aba23c253f17ad1a4f2bf4883b8108bda5:/lib/bootstrap.py diff --git a/lib/bootstrap.py b/lib/bootstrap.py index 889fa71..4bc9ebf 100644 --- a/lib/bootstrap.py +++ b/lib/bootstrap.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007, 2008, 2010 Google Inc. +# Copyright (C) 2006, 2007, 2008, 2010, 2011 Google Inc. # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by @@ -42,10 +42,16 @@ from ganeti import hypervisor from ganeti import bdev from ganeti import netutils from ganeti import backend +from ganeti import luxi +from ganeti import jstore + # 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(): """Setup the SSH configuration for the cluster. @@ -181,10 +187,30 @@ 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): @@ -195,14 +221,11 @@ def _InitFileStorage(file_storage_dir): time) or the normalized path to the storage directory """ - if not constants.ENABLE_FILE_STORAGE: - return "" - file_storage_dir = os.path.normpath(file_storage_dir) if not os.path.isabs(file_storage_dir): - raise errors.OpPrereqError("The file storage directory you passed is" - " not an absolute path.", errors.ECODE_INVAL) + raise errors.OpPrereqError("File storage directory '%s' is not an absolute" + " path" % file_storage_dir, errors.ECODE_INVAL) if not os.path.exists(file_storage_dir): try: @@ -219,15 +242,14 @@ def _InitFileStorage(file_storage_dir): return file_storage_dir -#pylint: disable-msg=R0913 -def InitCluster(cluster_name, mac_prefix, - 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): +def InitCluster(cluster_name, mac_prefix, # pylint: disable-msg=R0913 + master_netdev, file_storage_dir, shared_file_storage_dir, + candidate_pool_size, secondary_ip=None, vg_name=None, + beparams=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 @@ -321,7 +343,15 @@ def InitCluster(cluster_name, mac_prefix, curr_helper), errors.ECODE_INVAL) - file_storage_dir = _InitFileStorage(file_storage_dir) + if constants.ENABLE_FILE_STORAGE: + file_storage_dir = _InitFileStorage(file_storage_dir) + else: + file_storage_dir = "" + + if constants.ENABLE_SHARED_FILE_STORAGE: + shared_file_storage_dir = _InitFileStorage(shared_file_storage_dir) + else: + shared_file_storage_dir = "" if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix, @@ -340,6 +370,11 @@ 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) @@ -364,6 +399,12 @@ def InitCluster(cluster_name, mac_prefix, raise errors.OpPrereqError("Invalid default iallocator script '%s'" " specified" % default_iallocator, errors.ECODE_INVAL) + elif constants.HTOOLS: + # htools was enabled at build-time, we default to it + if utils.FindFile(constants.IALLOC_HAIL, + constants.IALLOCATOR_SEARCH_PATH, + os.path.isfile): + default_iallocator = constants.IALLOC_HAIL now = time.time() @@ -380,9 +421,11 @@ def InitCluster(cluster_name, mac_prefix, master_netdev=master_netdev, cluster_name=clustername.name, file_storage_dir=file_storage_dir, + shared_file_storage_dir=shared_file_storage_dir, 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, @@ -394,6 +437,7 @@ def InitCluster(cluster_name, mac_prefix, 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, @@ -401,6 +445,7 @@ 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) @@ -411,10 +456,14 @@ 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, @@ -444,7 +493,7 @@ def InitConfig(version, cluster_config, master_node_config, } default_nodegroup = objects.NodeGroup( uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID), - name="default", + name=constants.INITIAL_NODE_GROUP_NAME, members=[master_node_config.name], ) nodegroups = { @@ -497,27 +546,7 @@ 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: @@ -530,10 +559,9 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check): 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, - bind_address)) + 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, @@ -601,12 +629,36 @@ 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 @@ -621,15 +673,11 @@ 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() + if jstore.CheckDrainFlag(): + logging.info("Undraining job queue") + jstore.SetDrainFlag(False) - 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 @@ -638,6 +686,7 @@ 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