X-Git-Url: https://code.grnet.gr/git/ganeti-local/blobdiff_plain/fb460cf7e9bda225e4f1c070cd6b4fac1b3f6696..3234695b0e09d78b79d2e3df076e4a9c646ef6d0:/lib/bootstrap.py diff --git a/lib/bootstrap.py b/lib/bootstrap.py index 5893cb6..69ba1ef 100644 --- a/lib/bootstrap.py +++ b/lib/bootstrap.py @@ -1,7 +1,7 @@ # # -# Copyright (C) 2006, 2007, 2008, 2010, 2011 Google Inc. +# Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 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 @@ -28,6 +28,7 @@ import os.path import re import logging import time +import tempfile from ganeti import rpc from ganeti import ssh @@ -41,9 +42,9 @@ from ganeti import serializer from ganeti import hypervisor from ganeti import bdev from ganeti import netutils -from ganeti import backend from ganeti import luxi from ganeti import jstore +from ganeti import pathutils # ec_id for InitConfig's temporary reservation manager @@ -60,7 +61,7 @@ def _InitSSHSetup(): permitted hosts and adds the hostkey to its own known hosts. """ - priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) + priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER) for name in priv_key, pub_key: if os.path.exists(name): @@ -92,12 +93,12 @@ 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): + nodecert_file=pathutils.NODED_CERT_FILE, + rapicert_file=pathutils.RAPI_CERT_FILE, + spicecert_file=pathutils.SPICE_CERT_FILE, + spicecacert_file=pathutils.SPICE_CACERT_FILE, + hmackey_file=pathutils.CONFD_HMAC_KEY, + cds_file=pathutils.CLUSTER_DOMAIN_SECRET_FILE): """Updates the cluster certificates, keys and secrets. @type new_cluster_cert: bool @@ -208,7 +209,7 @@ def _InitGanetiServerSetup(master_name): # Generate cluster secrets GenerateClusterCrypto(True, False, False, False, False) - result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED]) + result = utils.RunCmd([pathutils.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" % @@ -222,7 +223,9 @@ def _WaitForNodeDaemon(node_name): """ def _CheckNodeDaemon(): - result = rpc.RpcRunner.call_version([node_name])[node_name] + # Pylint bug + # pylint: disable=E1101 + result = rpc.BootstrapRunner().call_version([node_name])[node_name] if result.fail_msg: raise utils.RetryAgain() @@ -253,6 +256,61 @@ def _WaitForMasterDaemon(): " %s seconds" % _DAEMON_READY_TIMEOUT) +def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose, + use_cluster_key, ask_key, strict_host_check, data): + """Runs a command to configure something on a remote machine. + + @type cluster_name: string + @param cluster_name: Cluster name + @type node: string + @param node: Node name + @type basecmd: string + @param basecmd: Base command (path on the remote machine) + @type debug: bool + @param debug: Enable debug output + @type verbose: bool + @param verbose: Enable verbose output + @type use_cluster_key: bool + @param use_cluster_key: See L{ssh.SshRunner.BuildCmd} + @type ask_key: bool + @param ask_key: See L{ssh.SshRunner.BuildCmd} + @type strict_host_check: bool + @param strict_host_check: See L{ssh.SshRunner.BuildCmd} + @param data: JSON-serializable input data for script (passed to stdin) + + """ + cmd = [basecmd] + + # Pass --debug/--verbose to the external script if set on our invocation + if debug: + cmd.append("--debug") + + if verbose: + cmd.append("--verbose") + + family = ssconf.SimpleStore().GetPrimaryIPFamily() + srun = ssh.SshRunner(cluster_name, + ipv6=(family == netutils.IP6Address.family)) + scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER, + utils.ShellQuoteArgs(cmd), + batch=False, ask_key=ask_key, quiet=False, + strict_host_check=strict_host_check, + use_cluster_key=use_cluster_key) + + tempfh = tempfile.TemporaryFile() + try: + tempfh.write(serializer.DumpJson(data)) + tempfh.seek(0) + + result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh) + finally: + tempfh.close() + + if result.failed: + raise errors.OpExecError("Command '%s' failed: %s" % + (result.cmd, result.fail_reason)) + + def _InitFileStorage(file_storage_dir): """Initialize if needed the file storage. @@ -282,14 +340,16 @@ def _InitFileStorage(file_storage_dir): return file_storage_dir -def InitCluster(cluster_name, mac_prefix, # pylint: disable=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): +def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914 + 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, ipolicy=None, + prealloc_wipe_disks=False, use_external_mip_script=False, + hv_state=None, disk_state=None): """Initialise the cluster. @type candidate_pool_size: int @@ -310,24 +370,22 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 " 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) + primary_ip_version, errors.ECODE_INVAL) 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)) + " address." % (hostname.ip, primary_ip_version), + errors.ECODE_INVAL) 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), + (hostname.ip, pathutils.ETC_HOSTS), errors.ECODE_ENVIRON) if not ipcls.Own(hostname.ip): @@ -359,6 +417,14 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 " 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), + errors.ECODE_INVAL) + else: + master_netmask = ipcls.iplen + if vg_name is not None: # Check if volume group is valid vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, @@ -382,6 +448,13 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 curr_helper), errors.ECODE_INVAL) + logging.debug("Stopping daemons (if any are running)") + result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"]) + if result.failed: + raise errors.OpExecError("Could not stop daemons, command %s" + " had exitcode %s and error '%s'" % + (result.cmd, result.exit_code, result.output)) + if constants.ENABLE_FILE_STORAGE: file_storage_dir = _InitFileStorage(file_storage_dir) else: @@ -402,26 +475,68 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 (master_netdev, result.output.strip()), errors.ECODE_INVAL) - dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)] + dirs = [(pathutils.RUN_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) + full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy) + if ndparams is not None: utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES) else: ndparams = dict(constants.NDC_DEFAULTS) + # This is ugly, as we modify the dict itself + # FIXME: Make utils.ForceDictType pure functional or write a wrapper + # around it + if hv_state: + for hvname, hvs_data in hv_state.items(): + utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES) + hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data) + else: + hv_state = dict((hvname, constants.HVST_DEFAULTS) + for hvname in enabled_hypervisors) + + # FIXME: disk_state has no default values yet + if disk_state: + for storage, ds_data in disk_state.items(): + if storage not in constants.DS_VALID_TYPES: + raise errors.OpPrereqError("Invalid storage type in disk state: %s" % + storage, errors.ECODE_INVAL) + for ds_name, state in ds_data.items(): + utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES) + ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state) + # 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)), + errors.ECODE_INVAL) + utils.ForceDictType(dt_params, constants.DISK_DT_TYPES) + try: + utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS) + except errors.OpPrereqError, err: + raise errors.OpPrereqError("While verify diskparam options: %s" % err, + errors.ECODE_INVAL) + # set up ssh config and /etc/hosts - sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB) + sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB) sshkey = sshline.split(" ")[1] if modify_etc_hosts: @@ -457,6 +572,7 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 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, @@ -466,6 +582,7 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 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, @@ -477,6 +594,10 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 default_iallocator=default_iallocator, primary_ip_family=ipcls.family, prealloc_wipe_disks=prealloc_wipe_disks, + use_external_mip_script=use_external_mip_script, + ipolicy=full_ipolicy, + hv_state_static=hv_state, + disk_state_static=disk_state, ) master_node_config = objects.Node(name=hostname.name, primary_ip=hostname.ip, @@ -488,15 +609,15 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 ) InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) cfg = config.ConfigWriter(offline=True) - ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE) + ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE) cfg.Update(cfg.GetClusterInfo(), logging.error) - backend.WriteSsconfFiles(cfg.GetSsconfValues()) + ssconf.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"]) + result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"]) if result.failed: raise errors.OpExecError("Could not start daemons, command %s" " had exitcode %s and error %s" % @@ -506,7 +627,7 @@ def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913 def InitConfig(version, cluster_config, master_node_config, - cfg_file=constants.CLUSTER_CONF_FILE): + cfg_file=pathutils.CLUSTER_CONF_FILE): """Create the initial cluster configuration. It will contain the current node, which will also be the master @@ -534,6 +655,7 @@ def InitConfig(version, cluster_config, master_node_config, uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID), name=constants.INITIAL_NODE_GROUP_NAME, members=[master_node_config.name], + diskparams={}, ) nodegroups = { default_nodegroup.uuid: default_nodegroup, @@ -544,6 +666,7 @@ def InitConfig(version, cluster_config, master_node_config, nodegroups=nodegroups, nodes=nodes, instances={}, + networks={}, serial_no=1, ctime=now, mtime=now) utils.WriteFile(cfg_file, @@ -560,18 +683,31 @@ def FinalizeClusterDestroy(master): """ cfg = config.ConfigWriter() modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup - result = rpc.RpcRunner.call_node_stop_master(master) + 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" " the node: %s", msg) -def SetupNodeDaemon(cluster_name, node, ssh_key_check): +def SetupNodeDaemon(opts, cluster_name, node): """Add a node to the cluster. This function must be called before the actual opcode, and will ssh @@ -580,38 +716,19 @@ def SetupNodeDaemon(cluster_name, node, ssh_key_check): @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 """ - 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 - 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, - utils.ShellQuote(bind_address))) - - result = sshrunner.Run(node, 'root', mycommand, batch=False, - ask_key=ssh_key_check, - 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)) + data = { + constants.NDS_CLUSTER_NAME: cluster_name, + constants.NDS_NODE_DAEMON_CERTIFICATE: + utils.ReadFile(pathutils.NODED_CERT_FILE), + constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(), + constants.NDS_START_NODE_DAEMON: True, + } + + RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP, + opts.debug, opts.verbose, + True, opts.ssh_key_check, opts.ssh_key_check, data) _WaitForNodeDaemon(node) @@ -646,7 +763,7 @@ def MasterFailover(no_voting=False): " as master candidates. Only these nodes" " can become masters. Current list of" " master candidates is:\n" - "%s" % ('\n'.join(mc_no_master)), + "%s" % ("\n".join(mc_no_master)), errors.ECODE_STATE) if not no_voting: @@ -692,11 +809,22 @@ 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) + 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" - " %s, please disable manually: %s", old_master, msg) + " %s, please disable manually: %s", old_master, msg) logging.info("Checking master IP non-reachability...") @@ -721,7 +849,8 @@ def MasterFailover(no_voting=False): logging.info("Starting the master daemons on the new master") - result = rpc.RpcRunner.call_node_start_master_daemons(new_master, 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" @@ -777,7 +906,7 @@ 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") @@ -791,7 +920,8 @@ 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