X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/bb3776b4b5f9a8caf80ddb2a8eaf21242c336603..a3de2ae783ce4fac89ca914a88e35ed81a1cb722:/lib/bootstrap.py diff --git a/lib/bootstrap.py b/lib/bootstrap.py index 4ef199d..d18e68b 100644 --- a/lib/bootstrap.py +++ b/lib/bootstrap.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007, 2008 Google Inc. +# Copyright (C) 2006, 2007, 2008, 2010 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 @@ -39,6 +39,17 @@ from ganeti import objects from ganeti import ssconf from ganeti import serializer 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(): @@ -76,62 +87,85 @@ def GenerateHmacKey(file_name): backup=True) -def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_hmac_key, - rapi_cert_pem=None): +def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key, + new_cds, rapi_cert_pem=None, cds=None, + nodecert_file=constants.NODED_CERT_FILE, + rapicert_file=constants.RAPI_CERT_FILE, + hmackey_file=constants.CONFD_HMAC_KEY, + cds_file=constants.CLUSTER_DOMAIN_SECRET_FILE): """Updates the cluster certificates, keys and secrets. @type new_cluster_cert: bool @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_hmac_key: bool - @param new_hmac_key: Whether to generate a new HMAC key + @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 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 hmackey_file: string + @param hmackey_file: optional override of the hmac key file path """ - # SSL certificate - cluster_cert_exists = os.path.exists(constants.SSL_CERT_FILE) + # noded SSL certificate + cluster_cert_exists = os.path.exists(nodecert_file) if new_cluster_cert or not cluster_cert_exists: if cluster_cert_exists: - utils.CreateBackup(constants.SSL_CERT_FILE) + utils.CreateBackup(nodecert_file) - logging.debug("Generating new cluster certificate at %s", - constants.SSL_CERT_FILE) - utils.GenerateSelfSignedSslCert(constants.SSL_CERT_FILE) + logging.debug("Generating new cluster certificate at %s", nodecert_file) + utils.GenerateSelfSignedSslCert(nodecert_file) - # HMAC key - if new_hmac_key or not os.path.exists(constants.HMAC_CLUSTER_KEY): - logging.debug("Writing new HMAC key to %s", constants.HMAC_CLUSTER_KEY) - GenerateHmacKey(constants.HMAC_CLUSTER_KEY) + # confd HMAC key + if new_confd_hmac_key or not os.path.exists(hmackey_file): + logging.debug("Writing new confd HMAC key to %s", hmackey_file) + GenerateHmacKey(hmackey_file) # RAPI - rapi_cert_exists = os.path.exists(constants.RAPI_CERT_FILE) + rapi_cert_exists = os.path.exists(rapicert_file) if rapi_cert_pem: # Assume rapi_pem contains a valid PEM-formatted certificate and key - logging.debug("Writing RAPI certificate at %s", - constants.RAPI_CERT_FILE) - utils.WriteFile(constants.RAPI_CERT_FILE, data=rapi_cert_pem, backup=True) + logging.debug("Writing RAPI certificate at %s", rapicert_file) + utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True) elif new_rapi_cert or not rapi_cert_exists: if rapi_cert_exists: - utils.CreateBackup(constants.RAPI_CERT_FILE) + utils.CreateBackup(rapicert_file) - logging.debug("Generating new RAPI certificate at %s", - constants.RAPI_CERT_FILE) - utils.GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE) + logging.debug("Generating new RAPI certificate at %s", rapicert_file) + utils.GenerateSelfSignedSslCert(rapicert_file) + + # Cluster domain secret + if cds: + logging.debug("Writing cluster domain secret to %s", cds_file) + utils.WriteFile(cds_file, data=cds, backup=True) + + elif new_cds or not os.path.exists(cds_file): + logging.debug("Generating new cluster domain secret at %s", cds_file) + GenerateHmacKey(cds_file) def _InitGanetiServerSetup(master_name): """Setup the necessary configuration for the initial node daemon. This creates the nodepass file containing the shared password for - the cluster and also generates the SSL certificate. + the cluster, generates the SSL certificate and starts the node daemon. + + @type master_name: str + @param master_name: Name of the master node """ # Generate cluster secrets - GenerateClusterCrypto(True, False, False) + GenerateClusterCrypto(True, False, False, False) result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED]) if result.failed: @@ -152,17 +186,72 @@ 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) -def InitCluster(cluster_name, mac_prefix, + 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): + """Initialize if needed the file storage. + + @param file_storage_dir: the user-supplied value + @return: either empty string (if file storage was disabled at build + 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) + + if not os.path.exists(file_storage_dir): + try: + os.makedirs(file_storage_dir, 0750) + except OSError, err: + raise errors.OpPrereqError("Cannot create file storage directory" + " '%s': %s" % (file_storage_dir, err), + errors.ECODE_ENVIRON) + + if not os.path.isdir(file_storage_dir): + raise errors.OpPrereqError("The file storage directory '%s' is not" + " a directory." % file_storage_dir, + errors.ECODE_ENVIRON) + return file_storage_dir + + +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): + 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 @@ -183,39 +272,56 @@ def InitCluster(cluster_name, mac_prefix, " entries: %s" % invalid_hvs, errors.ECODE_INVAL) - hostname = utils.GetHostInfo() - if hostname.ip.startswith("127."): - raise errors.OpPrereqError("This host's IP resolves to the private" - " range (%s). Please fix DNS or %s." % + ipcls = None + if primary_ip_version == constants.IP4_VERSION: + ipcls = netutils.IP4Address + elif primary_ip_version == constants.IP6_VERSION: + ipcls = netutils.IP6Address + else: + raise errors.OpPrereqError("Invalid primary ip version: %d." % + primary_ip_version) + + hostname = netutils.GetHostname(family=ipcls.family) + if not ipcls.IsValid(hostname.ip): + raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d" + " address." % (hostname.ip, primary_ip_version)) + + if ipcls.IsLoopback(hostname.ip): + raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback" + " address. Please fix DNS or %s." % (hostname.ip, constants.ETC_HOSTS), errors.ECODE_ENVIRON) - if not utils.OwnIpAddress(hostname.ip): + if not ipcls.Own(hostname.ip): raise errors.OpPrereqError("Inconsistency: this host's name resolves" " to %s,\nbut this ip address does not" - " belong to this host. Aborting." % + " belong to this host" % hostname.ip, errors.ECODE_ENVIRON) - clustername = utils.GetHostInfo(utils.HostInfo.NormalizeName(cluster_name)) + clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family) - if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, - timeout=5): - raise errors.OpPrereqError("Cluster IP already active. Aborting.", + if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5): + raise errors.OpPrereqError("Cluster IP already active", errors.ECODE_NOTUNIQUE) - if secondary_ip: - if not utils.IsValidIP(secondary_ip): - raise errors.OpPrereqError("Invalid secondary ip given", + if not secondary_ip: + if primary_ip_version == constants.IP6_VERSION: + raise errors.OpPrereqError("When using a IPv6 primary address, a valid" + " IPv4 address must be given as secondary", errors.ECODE_INVAL) - if (secondary_ip != hostname.ip and - not utils.OwnIpAddress(secondary_ip)): - raise errors.OpPrereqError("You gave %s as secondary IP," - " but it does not belong to this host." % - secondary_ip, errors.ECODE_ENVIRON) - else: secondary_ip = hostname.ip + if not netutils.IP4Address.IsValid(secondary_ip): + raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid" + " IPv4 address." % secondary_ip, + errors.ECODE_INVAL) + + if not netutils.IP4Address.Own(secondary_ip): + raise errors.OpPrereqError("You gave %s as secondary IP," + " but it does not belong to this host." % + secondary_ip, errors.ECODE_ENVIRON) + if vg_name is not None: # Check if volume group is valid vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, @@ -225,24 +331,21 @@ def InitCluster(cluster_name, mac_prefix, " you are not using lvm" % vgstatus, errors.ECODE_INVAL) - 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) - - if not os.path.exists(file_storage_dir): + if drbd_helper is not None: try: - os.makedirs(file_storage_dir, 0750) - except OSError, err: - raise errors.OpPrereqError("Cannot create file storage directory" - " '%s': %s" % (file_storage_dir, err), + curr_helper = bdev.BaseDRBD.GetUsermodeHelper() + except errors.BlockDeviceError, err: + raise errors.OpPrereqError("Error while checking drbd helper" + " (specify --no-drbd-storage if you are not" + " using drbd): %s" % str(err), errors.ECODE_ENVIRON) + if drbd_helper != curr_helper: + raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s" + " is the current helper" % (drbd_helper, + curr_helper), + errors.ECODE_INVAL) - if not os.path.isdir(file_storage_dir): - raise errors.OpPrereqError("The file storage directory '%s' is not" - " a directory." % file_storage_dir, - errors.ECODE_ENVIRON) + file_storage_dir = _InitFileStorage(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, @@ -261,25 +364,36 @@ 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) hv_class = hypervisor.GetHypervisor(hv_name) hv_class.CheckParameterSyntax(hv_params) - # set up the inter-node password and certificate - _InitGanetiServerSetup(hostname.name) - # set up ssh config and /etc/hosts sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB) sshkey = sshline.split(" ")[1] if modify_etc_hosts: - utils.AddHostToEtcHosts(hostname.name) + utils.AddHostToEtcHosts(hostname.name, hostname.ip) if modify_ssh_setup: _InitSSHSetup() + if default_iallocator is not None: + alloc_script = utils.FindFile(default_iallocator, + constants.IALLOCATOR_SEARCH_PATH, + os.path.isfile) + if alloc_script is None: + raise errors.OpPrereqError("Invalid default iallocator script '%s'" + " specified" % default_iallocator, + errors.ECODE_INVAL) + now = time.time() # init of cluster config file @@ -298,13 +412,19 @@ 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, modify_ssh_setup=modify_ssh_setup, + 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, @@ -312,16 +432,25 @@ 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() + cfg = config.ConfigWriter(offline=True) ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE) cfg.Update(cfg.GetClusterInfo(), logging.error) + backend.WriteSsconfFiles(cfg.GetSsconfValues()) + + # set up the inter-node password and certificate + _InitGanetiServerSetup(hostname.name) + + 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)) - # 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) + _WaitForMasterDaemon() def InitConfig(version, cluster_config, master_node_config, @@ -341,13 +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, @@ -389,51 +531,28 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check): @param ssh_key_check: whether to do a strict key check """ - sshrunner = ssh.SshRunner(cluster_name) - - noded_cert = utils.ReadFile(constants.SSL_CERT_FILE) - rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE) - hmac_key = utils.ReadFile(constants.HMAC_CLUSTER_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, 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 hmac_key.endswith("\n"): - hmac_key += "\n" + family = ssconf.SimpleStore().GetPrimaryIPFamily() + sshrunner = ssh.SshRunner(cluster_name, + ipv6=(family == netutils.IP6Address.family)) + + bind_address = constants.IP4_ADDRESS_ANY + if family == netutils.IP6Address.family: + bind_address = constants.IP6_ADDRESS_ANY # set up inter-node password and certificate and restarts the node daemon # 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 - 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" % - (constants.SSL_CERT_FILE, noded_cert, - constants.RAPI_CERT_FILE, rapi_cert, - constants.HMAC_CLUSTER_KEY, hmac_key, - constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE, - constants.HMAC_CLUSTER_KEY, - constants.DAEMON_UTIL, constants.NODED)) + 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," @@ -497,23 +616,51 @@ 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 + def _check_ip(): + if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT): + raise utils.RetryAgain() - # instantiate a real config writer, as we now know we have the - # configuration data - cfg = config.ConfigWriter() + try: + utils.Retry(_check_ip, (1, 1.5, 5), total_timeout) + except utils.RetryTimeout: + logging.warning("The master IP is still reachable after %s seconds," + " continuing but activating the master on the current" + " node will probably fail", total_timeout) - 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 @@ -522,6 +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 @@ -562,7 +710,7 @@ def GatherMasterVotes(node_list): @return: list of (node, votes) """ - myself = utils.HostInfo().name + myself = netutils.Hostname.GetSysName() try: node_list.remove(myself) except ValueError: @@ -584,6 +732,7 @@ 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) elif not isinstance(data, (tuple, list)) or len(data) < 3: logging.warning("Invalid data received from node %s: %s", node, data) fail = True