X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/ec17d09c01f897bc6951e4125db152162caa0aa2..435e4bd6ce5fdbe5a8623dd3c8fb215427cb34ab:/lib/bootstrap.py diff --git a/lib/bootstrap.py b/lib/bootstrap.py index d28135a..ea2a829 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 @@ -25,9 +25,9 @@ import os import os.path -import sha import re import logging +import time from ganeti import rpc from ganeti import ssh @@ -37,17 +37,22 @@ from ganeti import config from ganeti import constants 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 + +# ec_id for InitConfig's temporary reservation manager +_INITCONF_ECID = "initconfig-ecid" -def _InitSSHSetup(node): - """Setup the SSH configuration for the cluster. +def _InitSSHSetup(): + """Setup the SSH configuration for the cluster. This generates a dsa keypair for root, adds the pub key to the permitted hosts and adds the hostkey to its own known hosts. - Args: - node: the name of this host as a fqdn - """ priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) @@ -63,133 +68,304 @@ def _InitSSHSetup(node): raise errors.OpExecError("Could not generate ssh keypair, error %s" % result.output) - f = open(pub_key, 'r') - try: - utils.AddAuthorizedKey(auth_keys, f.read(8192)) - finally: - f.close() + utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) -def _InitGanetiServerSetup(): - """Setup the necessary configuration for the initial node daemon. +def GenerateHmacKey(file_name): + """Writes a new HMAC key. - This creates the nodepass file containing the shared password for - the cluster and also generates the SSL certificate. + @type file_name: str + @param file_name: Path to output file """ - result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024", - "-days", str(365*5), "-nodes", "-x509", - "-keyout", constants.SSL_CERT_FILE, - "-out", constants.SSL_CERT_FILE, "-batch"]) - if result.failed: - raise errors.OpExecError("could not generate server ssl cert, command" - " %s had exitcode %s and error message %s" % - (result.cmd, result.exit_code, result.output)) + utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400, + backup=True) + + +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_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 + + """ + # 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(nodecert_file) + + logging.debug("Generating new cluster certificate at %s", nodecert_file) + utils.GenerateSelfSignedSslCert(nodecert_file) + + # 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(rapicert_file) + + if rapi_cert_pem: + # Assume rapi_pem contains a valid PEM-formatted certificate and key + 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(rapicert_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, generates the SSL certificate and starts the node daemon. - os.chmod(constants.SSL_CERT_FILE, 0400) + @type master_name: str + @param master_name: Name of the master node - result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"]) + """ + # Generate cluster secrets + GenerateClusterCrypto(True, False, False, False) + result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED]) if result.failed: raise errors.OpExecError("Could not start the node daemon, command %s" " had exitcode %s and error %s" % (result.cmd, result.exit_code, result.output)) + _WaitForNodeDaemon(master_name) -def InitCluster(cluster_name, mac_prefix, def_bridge, - master_netdev, file_storage_dir, - secondary_ip=None, - vg_name=None, beparams=None, hvparams=None, - enabled_hypervisors=None, default_hypervisor=None): - """Initialise the cluster. + +def _WaitForNodeDaemon(node_name): + """Wait for node daemon to become responsive. """ - if config.ConfigWriter.IsCluster(): - raise errors.OpPrereqError("Cluster is already initialised") + def _CheckNodeDaemon(): + result = rpc.RpcRunner.call_version([node_name])[node_name] + if result.fail_msg: + raise utils.RetryAgain() - hostname = utils.HostInfo() + try: + utils.Retry(_CheckNodeDaemon, 1.0, 10.0) + except utils.RetryTimeout: + raise errors.OpExecError("Node daemon on %s didn't answer queries within" + " 10 seconds" % node_name) - if hostname.ip.startswith("127."): - raise errors.OpPrereqError("This host's IP resolves to the private" - " range (%s). Please fix DNS or %s." % - (hostname.ip, constants.ETC_HOSTS)) - if not utils.OwnIpAddress(hostname.ip): - raise errors.OpPrereqError("Inconsistency: this host's name resolves" - " to %s,\nbut this ip address does not" - " belong to this host." - " Aborting." % hostname.ip) - - clustername = utils.HostInfo(cluster_name) - - if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, - timeout=5): - raise errors.OpPrereqError("Cluster IP already active. Aborting.") - - if secondary_ip: - if not utils.IsValidIP(secondary_ip): - raise errors.OpPrereqError("Invalid secondary ip given") - 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) - else: - secondary_ip = hostname.ip +def _InitFileStorage(file_storage_dir): + """Initialize if needed the file storage. - if vg_name is not None: - # Check if volume group is valid - vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, - constants.MIN_VG_SIZE) - if vgstatus: - raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" - " you are not using lvm" % vgstatus) + @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.") + " 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)) + " '%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) + " a directory." % file_storage_dir, + errors.ECODE_ENVIRON) + 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, prealloc_wipe_disks=False): + """Initialise the cluster. + + @type candidate_pool_size: int + @param candidate_pool_size: master candidate pool size + + """ + # TODO: complete the docstring + if config.ConfigWriter.IsCluster(): + raise errors.OpPrereqError("Cluster is already initialised", + errors.ECODE_STATE) + + if not enabled_hypervisors: + raise errors.OpPrereqError("Enabled hypervisors list must contain at" + " least one member", errors.ECODE_INVAL) + invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES + if invalid_hvs: + raise errors.OpPrereqError("Enabled hypervisors contains invalid" + " 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: + 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 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" % + hostname.ip, errors.ECODE_ENVIRON) + + clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family) + + if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5): + raise errors.OpPrereqError("Cluster IP already active", + errors.ECODE_NOTUNIQUE) + + 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) + 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, + constants.MIN_VG_SIZE) + if vgstatus: + raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" + " you are not using lvm" % vgstatus, + errors.ECODE_INVAL) + + if drbd_helper is not None: + try: + 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) + + 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) + raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix, + errors.ECODE_INVAL) result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) if result.failed: raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % (master_netdev, - result.output.strip())) + result.output.strip()), errors.ECODE_INVAL) - 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) + dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)] + utils.EnsureDirs(dirs) - # set up the inter-node password and certificate - _InitGanetiServerSetup() + utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES) + utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES) + objects.NIC.CheckParameterSyntax(nicparams) + + # 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 ssh config and /etc/hosts - f = open(constants.SSH_HOST_RSA_PUB, 'r') - try: - sshline = f.read() - finally: - f.close() + sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB) sshkey = sshline.split(" ")[1] - utils.AddHostToEtcHosts(hostname.name) - _InitSSHSetup(hostname.name) + if modify_etc_hosts: + 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 cluster_config = objects.Cluster( @@ -198,7 +374,6 @@ def InitCluster(cluster_name, mac_prefix, def_bridge, highest_used_port=(constants.FIRST_DRBD_PORT - 1), mac_prefix=mac_prefix, volume_group_name=vg_name, - default_bridge=def_bridge, tcpudp_port_pool=set(), master_node=hostname.name, master_ip=clustername.ip, @@ -206,21 +381,42 @@ def InitCluster(cluster_name, mac_prefix, def_bridge, cluster_name=clustername.name, file_storage_dir=file_storage_dir, enabled_hypervisors=enabled_hypervisors, - default_hypervisor=default_hypervisor, - beparams={constants.BEGR_DEFAULT: beparams}, + beparams={constants.PP_DEFAULT: beparams}, + nicparams={constants.PP_DEFAULT: nicparams}, 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, + 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, - secondary_ip=secondary_ip) - - cfg = InitConfig(constants.CONFIG_VERSION, - cluster_config, master_node_config) + secondary_ip=secondary_ip, + 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) 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) # start the master ip # TODO: Review rpc call from bootstrap - rpc.RpcRunner.call_node_start_master(hostname.name, True) + # TODO: Warn on failed start master + rpc.RpcRunner.call_node_start_master(hostname.name, True, False) def InitConfig(version, cluster_config, master_node_config, @@ -231,31 +427,42 @@ def InitConfig(version, cluster_config, master_node_config, node, and no instances. @type version: int - @param version: Configuration version - @type cluster_config: objects.Cluster - @param cluster_config: Cluster configuration - @type master_node_config: objects.Node - @param master_node_config: Master node configuration - @type file_name: string - @param file_name: Configuration file path - - @rtype: ssconf.SimpleConfigWriter - @returns: Initialized config instance + @param version: configuration version + @type cluster_config: L{objects.Cluster} + @param cluster_config: cluster configuration + @type master_node_config: L{objects.Node} + @param master_node_config: master node configuration + @type cfg_file: string + @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="default", + 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) - cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file) - cfg.Save() - - return cfg + serial_no=1, + ctime=now, mtime=now) + utils.WriteFile(cfg_file, + data=serializer.Dump(config_data.ToDict()), + mode=0600) def FinalizeClusterDestroy(master): @@ -265,118 +472,210 @@ def FinalizeClusterDestroy(master): begun in cmdlib.LUDestroyOpcode. """ - if not rpc.RpcRunner.call_node_stop_master(master, True): - logging.warning("Could not disable the master role") - if not rpc.RpcRunner.call_node_leave_cluster(master): - logging.warning("Could not shutdown the node daemon and cleanup the node") - - -def SetupNodeDaemon(node, ssh_key_check): + cfg = config.ConfigWriter() + modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup + result = rpc.RpcRunner.call_node_stop_master(master, True) + 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) + msg = result.fail_msg + if msg: + logging.warning("Could not shutdown the node daemon and cleanup" + " the node: %s", msg) + + +def SetupNodeDaemon(cluster_name, node, ssh_key_check): """Add a node to the cluster. This function must be called before the actual opcode, and will ssh to the remote node, copy the needed files, and start ganeti-noded, allowing the master to do the rest via normal rpc calls. - Args: - node: fully qualified domain name for the new node + @param cluster_name: the cluster name + @param node: the name of the new node + @param ssh_key_check: whether to do a strict key check """ - cfg = ssconf.SimpleConfigReader() - sshrunner = ssh.SshRunner(cfg.GetClusterName()) - gntpem = utils.ReadFile(constants.SSL_CERT_FILE) + 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 - if re.search('^!EOF\.', gntpem, re.MULTILINE): - raise errors.OpExecError("invalid PEM encoding in the SSL certificate") - if not gntpem.endswith("\n"): - raise errors.OpExecError("PEM must end with newline") + # 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" + + 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%s restart" % - (constants.SSL_CERT_FILE, gntpem, - constants.NODE_INITD_SCRIPT)) + 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)) 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," " output: %s" % (node, result.fail_reason, result.output)) - return 0 + _WaitForNodeDaemon(node) -def MasterFailover(): +def MasterFailover(no_voting=False): """Failover the master node. This checks that we are not already the master, and will cause the current master to cease being master, and the non-master to become new master. + @type no_voting: boolean + @param no_voting: force the operation without remote nodes agreement + (dangerous) + """ - cfg = ssconf.SimpleConfigWriter() + sstore = ssconf.SimpleStore() - new_master = utils.HostInfo().name - old_master = cfg.GetMasterNode() - node_list = cfg.GetNodeList() + old_master, new_master = ssconf.GetMasterAndMyself(sstore) + node_list = sstore.GetNodeList() + mc_list = sstore.GetMasterCandidates() if old_master == new_master: raise errors.OpPrereqError("This commands must be run on the node" " where you want the new master to be." " %s is already the master" % - old_master) - - vote_list = GatherMasterVotes(node_list) - - if vote_list: - voted_master = vote_list[0][0] - if voted_master is None: - raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not" - " respond.") - elif voted_master != old_master: - raise errors.OpPrereqError("I have wrong configuration, I believe the" - " master is %s but the other nodes voted for" - " %s. Please resync the configuration of" - " this node." % (old_master, voted_master)) + old_master, errors.ECODE_INVAL) + + if new_master not in mc_list: + mc_no_master = [name for name in mc_list if name != old_master] + raise errors.OpPrereqError("This node is not among the nodes marked" + " as master candidates. Only these nodes" + " can become masters. Current list of" + " master candidates is:\n" + "%s" % ('\n'.join(mc_no_master)), + errors.ECODE_STATE) + + if not no_voting: + vote_list = GatherMasterVotes(node_list) + + if vote_list: + voted_master = vote_list[0][0] + if voted_master is None: + raise errors.OpPrereqError("Cluster is inconsistent, most nodes did" + " not respond.", errors.ECODE_ENVIRON) + elif voted_master != old_master: + raise errors.OpPrereqError("I have a wrong configuration, I believe" + " the master is %s but the other nodes" + " voted %s. Please resync the configuration" + " of this node." % + (old_master, voted_master), + errors.ECODE_STATE) # end checks rcode = 0 logging.info("Setting master to %s, old master: %s", new_master, old_master) - if not rpc.RpcRunner.call_node_stop_master(old_master, True): + 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", old_master) + " %s, please disable manually: %s", old_master, msg) - cfg.SetMasterNode(new_master) - cfg.Save() + 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() - if not rpc.RpcRunner.call_upload_file(cfg.GetNodeList(), - constants.CLUSTER_CONF_FILE): - logging.error("Could not distribute the new configuration" - " to the other nodes, please check.") + 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) + logging.info("Starting the master daemons on the new master") - if not rpc.RpcRunner.call_node_start_master(new_master, True): + result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting) + msg = result.fail_msg + if msg: logging.error("Could not start the master role on the new master" - " %s, please check", new_master) + " %s, please check: %s", new_master, msg) rcode = 1 + logging.info("Master failed over from %s to %s", old_master, new_master) return rcode +def GetMaster(): + """Returns the current master node. + + This is a separate function in bootstrap since it's needed by + gnt-cluster, and instead of importing directly ssconf, it's better + to abstract it in bootstrap, where we do use ssconf in other + functions too. + + """ + sstore = ssconf.SimpleStore() + + old_master, _ = ssconf.GetMasterAndMyself(sstore) + + return old_master + + def GatherMasterVotes(node_list): """Check the agreement on who is the master. @@ -393,12 +692,12 @@ def GatherMasterVotes(node_list): @type node_list: list @param node_list: the list of nodes to query for master info; the current - node wil be removed if it is in the list + node will be removed if it is in the list @rtype: list @return: list of (node, votes) """ - myself = utils.HostInfo().name + myself = netutils.Hostname.GetSysName() try: node_list.remove(myself) except ValueError: @@ -411,17 +710,25 @@ def GatherMasterVotes(node_list): # this should not happen (unless internal error in rpc) logging.critical("Can't complete rpc call, aborting master startup") return [(None, len(node_list))] - positive = negative = 0 - other_masters = {} votes = {} for node in results: - if not isinstance(results[node], (tuple, list)) or len(results[node]) < 3: - # here the rpc layer should have already logged errors + nres = results[node] + data = nres.payload + msg = nres.fail_msg + fail = False + 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 + if fail: if None not in votes: votes[None] = 0 votes[None] += 1 continue - master_node = results[node][2] + master_node = data[2] if master_node not in votes: votes[master_node] = 0 votes[master_node] += 1