X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/4f358f284490332404560f05dd92b920cb8535df..bc5d0215:/lib/bootstrap.py diff --git a/lib/bootstrap.py b/lib/bootstrap.py index 24f9a0a..62d59ce 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. @@ -82,10 +88,14 @@ def GenerateHmacKey(file_name): backup=True) -def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key, - new_cds, rapi_cert_pem=None, cds=None, +def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert, + new_confd_hmac_key, new_cds, + rapi_cert_pem=None, spice_cert_pem=None, + spice_cacert_pem=None, cds=None, nodecert_file=constants.NODED_CERT_FILE, rapicert_file=constants.RAPI_CERT_FILE, + spicecert_file=constants.SPICE_CERT_FILE, + spicecacert_file=constants.SPICE_CACERT_FILE, hmackey_file=constants.CONFD_HMAC_KEY, cds_file=constants.CLUSTER_DOMAIN_SECRET_FILE): """Updates the cluster certificates, keys and secrets. @@ -94,18 +104,29 @@ def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key, @param new_cluster_cert: Whether to generate a new cluster certificate @type new_rapi_cert: bool @param new_rapi_cert: Whether to generate a new RAPI certificate + @type new_spice_cert: bool + @param new_spice_cert: Whether to generate a new SPICE certificate @type new_confd_hmac_key: bool @param new_confd_hmac_key: Whether to generate a new HMAC key @type new_cds: bool @param new_cds: Whether to generate a new cluster domain secret @type rapi_cert_pem: string @param rapi_cert_pem: New RAPI certificate in PEM format + @type spice_cert_pem: string + @param spice_cert_pem: New SPICE certificate in PEM format + @type spice_cacert_pem: string + @param spice_cacert_pem: Certificate of the CA that signed the SPICE + certificate, in PEM format @type cds: string @param cds: New cluster domain secret @type nodecert_file: string @param nodecert_file: optional override of the node cert file path @type rapicert_file: string @param rapicert_file: optional override of the rapi cert file path + @type spicecert_file: string + @param spicecert_file: optional override of the spice cert file path + @type spicecacert_file: string + @param spicecacert_file: optional override of the spice CA cert file path @type hmackey_file: string @param hmackey_file: optional override of the hmac key file path @@ -139,6 +160,31 @@ def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key, logging.debug("Generating new RAPI certificate at %s", rapicert_file) utils.GenerateSelfSignedSslCert(rapicert_file) + # SPICE + spice_cert_exists = os.path.exists(spicecert_file) + spice_cacert_exists = os.path.exists(spicecacert_file) + if spice_cert_pem: + # spice_cert_pem implies also spice_cacert_pem + logging.debug("Writing SPICE certificate at %s", spicecert_file) + utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True) + logging.debug("Writing SPICE CA certificate at %s", spicecacert_file) + utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True) + elif new_spice_cert or not spice_cert_exists: + if spice_cert_exists: + utils.CreateBackup(spicecert_file) + if spice_cacert_exists: + utils.CreateBackup(spicecacert_file) + + logging.debug("Generating new self-signed SPICE certificate at %s", + spicecert_file) + (_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file) + + # Self-signed certificate -> the public certificate is also the CA public + # certificate + logging.debug("Writing the public certificate to %s", + spicecert_file) + utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem) + # Cluster domain secret if cds: logging.debug("Writing cluster domain secret to %s", cds_file) @@ -160,7 +206,7 @@ def _InitGanetiServerSetup(master_name): """ # Generate cluster secrets - GenerateClusterCrypto(True, False, False, False) + GenerateClusterCrypto(True, False, False, False, False) result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED]) if result.failed: @@ -176,15 +222,35 @@ def _WaitForNodeDaemon(node_name): """ def _CheckNodeDaemon(): - result = rpc.RpcRunner.call_version([node_name])[node_name] + result = rpc.BootstrapRunner().call_version([node_name])[node_name] if result.fail_msg: 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 +261,11 @@ 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 +282,15 @@ 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, +def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 + master_netmask, 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, diskparams=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): + maintain_node_health=False, drbd_helper=None, uid_pool=None, + default_iallocator=None, primary_ip_version=None, + prealloc_wipe_disks=False, use_external_mip_script=False): """Initialise the cluster. @type candidate_pool_size: int @@ -248,13 +311,9 @@ def InitCluster(cluster_name, mac_prefix, " entries: %s" % invalid_hvs, errors.ECODE_INVAL) - - ipcls = None - if primary_ip_version == constants.IP4_VERSION: - ipcls = netutils.IP4Address - elif primary_ip_version == constants.IP6_VERSION: - ipcls = netutils.IP6Address - else: + try: + ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version) + except errors.ProgrammerError: raise errors.OpPrereqError("Invalid primary ip version: %d." % primary_ip_version) @@ -298,6 +357,13 @@ def InitCluster(cluster_name, mac_prefix, " but it does not belong to this host." % secondary_ip, errors.ECODE_ENVIRON) + if master_netmask is not None: + if not ipcls.ValidateNetmask(master_netmask): + raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " % + (master_netmask, primary_ip_version)) + else: + master_netmask = ipcls.iplen + if vg_name is not None: # Check if volume group is valid vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, @@ -321,7 +387,15 @@ 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, @@ -336,16 +410,33 @@ def InitCluster(cluster_name, mac_prefix, dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)] utils.EnsureDirs(dirs) + objects.UpgradeBeParams(beparams) utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES) 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) hv_class = hypervisor.GetHypervisor(hv_name) hv_class.CheckParameterSyntax(hv_params) + # diskparams is a mapping of disk-template->diskparams dict + for template, dt_params in diskparams.items(): + param_keys = set(dt_params.keys()) + default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys()) + if not (param_keys <= default_param_keys): + unknown_params = param_keys - default_param_keys + raise errors.OpPrereqError("Invalid parameters for disk template %s:" + " %s" % (template, + utils.CommaJoin(unknown_params))) + utils.ForceDictType(dt_params, constants.DISK_DT_TYPES) + # set up ssh config and /etc/hosts sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB) sshkey = sshline.split(" ")[1] @@ -364,6 +455,12 @@ 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() @@ -377,13 +474,17 @@ def InitCluster(cluster_name, mac_prefix, tcpudp_port_pool=set(), master_node=hostname.name, master_ip=clustername.ip, + master_netmask=master_netmask, 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, + diskparams=diskparams, candidate_pool_size=candidate_pool_size, modify_etc_hosts=modify_etc_hosts, modify_ssh_setup=modify_ssh_setup, @@ -395,6 +496,7 @@ def InitCluster(cluster_name, mac_prefix, default_iallocator=default_iallocator, primary_ip_family=ipcls.family, prealloc_wipe_disks=prealloc_wipe_disks, + use_external_mip_script=use_external_mip_script, ) master_node_config = objects.Node(name=hostname.name, primary_ip=hostname.ip, @@ -402,6 +504,7 @@ 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) @@ -412,10 +515,14 @@ 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, @@ -445,8 +552,9 @@ 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], + diskparams=cluster_config.diskparams, ) nodegroups = { default_nodegroup.uuid: default_nodegroup, @@ -473,11 +581,24 @@ def FinalizeClusterDestroy(master): """ cfg = config.ConfigWriter() modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup - result = rpc.RpcRunner.call_node_stop_master(master, True) + runner = rpc.BootstrapRunner() + + master_params = cfg.GetMasterNetworkParameters() + master_params.name = master + ems = cfg.GetUseExternalMipScript() + result = runner.call_node_deactivate_master_ip(master_params.name, + master_params, ems) + + msg = result.fail_msg + if msg: + logging.warning("Could not disable the master IP: %s", msg) + + result = runner.call_node_stop_master(master) msg = result.fail_msg if msg: logging.warning("Could not disable the master role: %s", msg) - result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup) + + result = runner.call_node_leave_cluster(master, modify_ssh_setup) msg = result.fail_msg if msg: logging.warning("Could not shutdown the node daemon and cleanup" @@ -498,27 +619,7 @@ 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: @@ -530,11 +631,12 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check): # either by being constants or by the checks above sshrunner.CopyFileToNode(node, constants.NODED_CERT_FILE) sshrunner.CopyFileToNode(node, constants.RAPI_CERT_FILE) + sshrunner.CopyFileToNode(node, constants.SPICE_CERT_FILE) + sshrunner.CopyFileToNode(node, constants.SPICE_CACERT_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, @@ -624,7 +726,18 @@ def MasterFailover(no_voting=False): logging.info("Stopping the master daemon on node %s", old_master) - result = rpc.RpcRunner.call_node_stop_master(old_master, True) + runner = rpc.BootstrapRunner() + master_params = cfg.GetMasterNetworkParameters() + master_params.name = old_master + ems = cfg.GetUseExternalMipScript() + result = runner.call_node_deactivate_master_ip(master_params.name, + master_params, ems) + + msg = result.fail_msg + if msg: + logging.warning("Could not disable the master IP: %s", msg) + + result = runner.call_node_stop_master(old_master) msg = result.fail_msg if msg: logging.error("Could not disable the master role on the old master" @@ -634,6 +747,7 @@ def MasterFailover(no_voting=False): master_ip = sstore.GetMasterIP() total_timeout = 30 + # Here we have a phase where no master should be running def _check_ip(): if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT): @@ -646,9 +760,14 @@ def MasterFailover(no_voting=False): " continuing but activating the master on the current" " node will probably fail", total_timeout) + if jstore.CheckDrainFlag(): + logging.info("Undraining job queue") + jstore.SetDrainFlag(False) + logging.info("Starting the master daemons on the new master") - result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting) + result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master, + no_voting) msg = result.fail_msg if msg: logging.error("Could not start the master role on the new master" @@ -704,7 +823,7 @@ def GatherMasterVotes(node_list): if not node_list: # no nodes left (eventually after removing myself) return [] - results = rpc.RpcRunner.call_master_info(node_list) + results = rpc.BootstrapRunner().call_master_info(node_list) if not isinstance(results, dict): # this should not happen (unless internal error in rpc) logging.critical("Can't complete rpc call, aborting master startup") @@ -718,7 +837,8 @@ def GatherMasterVotes(node_list): if msg: logging.warning("Error contacting node %s: %s", node, msg) fail = True - # for now we accept both length 3 and 4 (data[3] is primary ip version) + # for now we accept both length 3, 4 and 5 (data[3] is primary ip version + # and data[4] is the master netmask) elif not isinstance(data, (tuple, list)) or len(data) < 3: logging.warning("Invalid data received from node %s: %s", node, data) fail = True