def beeline(self, **kargs): """Connect to thriftserver command line """ logger.debug('thriftserver_command_beeline') _check_spark() cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) ths_props_path = path_of_fb['thrift_properties'] cmd = 'source {}; echo ${}'.format(ths_props_path, 'HIVE_HOST') host = sp.check_output(cmd, shell=True).decode('utf-8').strip() cmd = 'source {}; echo ${}'.format(ths_props_path, 'HIVE_PORT') port = sp.check_output(cmd, shell=True).decode('utf-8').strip() spark_env = _get_env() base_cmd = '{}/beeline'.format(spark_env['spark_bin']) options = { 'u': 'jdbc:hive2://{}:{}'.format(host, port), 'n': os.environ['USER'] } for key, value in kargs.items(): options[key] = value for key, value in options.items(): base_cmd += ' -{} {}'.format(key, value) logger.debug(base_cmd) msg = message.get('try_connection') logger.info(msg) os.system(base_cmd)
def delete(self, cluster_id): """Delete cluster It is automatically backed up with timestamps as tags :param cluster_id: target cluster id """ if not cluster_util.validate_id(cluster_id): raise ClusterIdError(cluster_id) center = Center() center.update_ip_port() success = center.check_hosts_connection() if success: center.stop_redis(force=True) path_of_fb = config.get_path_of_fb(cluster_id) props_path = path_of_fb['redis_properties'] hosts = config.get_props(props_path, 'sr2_redis_master_hosts', []) if not center.check_include_localhost(hosts): hosts += [config.get_local_ip()] tag = time.strftime("%Y%m%d%H%M%S", time.gmtime()) cluster_backup_dir = 'cluster_{}_bak_{}'.format(cluster_id, tag) for host in hosts: center.cluster_backup(host, cluster_id, cluster_backup_dir) msg = message.get('cluster_delete_complete') msg = msg.format(cluster_id=cluster_id) logger.info(msg)
def failover_with_dir(self, server, dir): """Find masters that use the specified directory path and do failover with its slave :param server: IP or hostname :param dir: directory path """ center = Center() center.update_ip_port() logger.debug('failover_with_dir') master_nodes = center.get_master_obj_list() cluster_id = config.get_cur_cluster_id() lib_path = config.get_ld_library_path(cluster_id) path_of_fb = config.get_path_of_fb(cluster_id) sr2_redis_bin = path_of_fb['sr2_redis_bin'] env_cmd = [ 'GLOBIGNORE=*;', 'export LD_LIBRARY_PATH={};'.format(lib_path['ld_library_path']), 'export DYLD_LIBRARY_PATH={};'.format( lib_path['dyld_library_path']), ] redis_cli_cmd = os.path.join(sr2_redis_bin, 'redis-cli') # Find masters with dir ret = RedisCliUtil.command_all_async('config get dir', slave=True) outs = '' meta = [] m_endpoint = [] for node in master_nodes: m_endpoint.append(node['addr']) for _, host, port, res, stdout in ret: if res == 'OK': flat_stdout = '\n'.join([outs, stdout]) line = flat_stdout.splitlines() if self.compare_ip(host, server) and dir in line[2]: endpoint = '{}:{}'.format(socket.gethostbyname(host), port) if endpoint in m_endpoint: meta.append(endpoint) else: logger.warning("FAIL {}:{} {}".format(host, port, stdout)) for endpoint in meta: for master_node in master_nodes: if endpoint == master_node['addr']: for slave_node in master_node['slaves']: addr = slave_node['addr'] (s_host, s_port) = addr.split(':') sub_cmd = 'cluster failover takeover' command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, s_host, s_port, sub_cmd, ) self._print( message.get('try_failover_takeover').format( slave=addr)) stdout = subprocess.check_output(command, shell=True) self._print(stdout)
def command_all_async(sub_cmd, slave=True): def _async_target_func(m_s, pre_cmd, host, port, sub_cmd, ret): try: command = '{} -h {} -p {} {}'.format(pre_cmd, host, port, sub_cmd) logger.debug(command) stdout = subprocess.check_output(command, shell=True) stdout = stdout.decode('utf-8').strip() ret.append((m_s, host, port, 'OK', stdout)) except Exception as ex: stderr = str(ex) logger.debug(stderr) ret.append((m_s, host, port, 'FAIL', stderr)) cluster_id = config.get_cur_cluster_id() master_host_list = config.get_master_host_list(cluster_id) master_port_list = config.get_master_port_list(cluster_id) if slave: slave_host_list = config.get_slave_host_list(cluster_id) slave_port_list = config.get_slave_port_list(cluster_id) path_of_fb = config.get_path_of_fb(cluster_id) sr2_redis_bin = path_of_fb['sr2_redis_bin'] logger.debug('command_all_async') cluster_id = config.get_cur_cluster_id() lib_path = config.get_ld_library_path(cluster_id) env_cmd = [ 'export LD_LIBRARY_PATH={};'.format(lib_path['ld_library_path']), 'export DYLD_LIBRARY_PATH={};'.format( lib_path['dyld_library_path']), ] env = ' '.join(env_cmd) threads = [] ret = [] # (m/s, host, port, result, message) pre_cmd = '{} {}/redis-cli -c'.format(env, sr2_redis_bin) for host in master_host_list: for port in master_port_list: t = Thread( target=_async_target_func, args=('Master', pre_cmd, host, port, sub_cmd, ret), ) threads.append(t) if slave: for host in slave_host_list: for port in slave_port_list: t = Thread( target=_async_target_func, args=('Slave', pre_cmd, host, port, sub_cmd, ret), ) threads.append(t) for th in threads: th.start() time.sleep(0.02) for th in threads: th.join() logger.debug(ret) return ret
def get_state(self, cluster_id, host='127.0.0.1'): path_of_fb = config.get_path_of_fb(cluster_id) cluster_path = path_of_fb['cluster_path'] state_file = os.path.join(cluster_path, '.deploy.state') client = net.get_ssh(host) if net.is_exist(client, state_file): return PENDING if net.is_dir(client, cluster_path): return DEPLOYED return CLEAN
def stop(self): """Stop thriftserver """ logger.debug('thriftserver_command_stop') _check_spark() cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) ths_props_path = path_of_fb['thrift_properties'] source_cmd = 'source {}'.format(ths_props_path) base_cmd = '$SPARK_SBIN/stop-thriftserver.sh' cmd = '{}; {}'.format(source_cmd, base_cmd) logger.debug(cmd) os.system(cmd)
def is_pending(self, cluster_id, nodes=['127.0.0.1']): if isinstance(nodes, str): nodes = [nodes] for node in nodes: path_of_fb = config.get_path_of_fb(cluster_id) cluster_path = path_of_fb['cluster_path'] deploy_state = os.path.join(cluster_path, '.deploy.state') client = net.get_ssh(node) if net.is_exist(client, deploy_state): client.close() return True client.close() return False
def save_redis_template_config(key, value): """Save redis template config to file :param key: key :param value: value """ key = key.strip() cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) master_template = path_of_fb['master_template'] msg = message.get('save_config_to_template') logger.info(msg) RedisCliUtil._save_config(master_template, key, value)
def _get_env(): if 'SPARK_HOME' not in os.environ: raise EnvError('you should set env SPARK_HOME') ret = { 'spark_home': os.environ['SPARK_HOME'], } env_list = ['SPARK_BIN', 'SPARK_SBIN', 'SPARK_LOG', 'SPARK_CONF'] cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) ths_props_path = path_of_fb['thrift_properties'] for env in env_list: cmd = 'source {}; echo ${}'.format(ths_props_path, env.upper()) stdout = sp.check_output(cmd, shell=True).decode('utf-8').strip() ret[env.lower()] = stdout return ret
def thriftserver(self): """Edit 'thriftserver.properties' """ cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) target_path = path_of_fb['thrift_properties'] self._edit_conf(target_path, syntax='sh') center = Center() center.update_ip_port() success = center.check_hosts_connection() if not success: return success = center.sync_file(target_path) if success: msg = message.get('complete_conf_edit') logger.info(msg)
def slave(self): """Edit 'redis-slave.conf.template' """ cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) target_path = path_of_fb['slave_template'] self._edit_conf(target_path, syntax='sh') center = Center() center.update_ip_port() success = center.check_hosts_connection() if not success: return success = center.sync_file(target_path) if success: msg = message.get('complete_conf_edit') logger.info(msg)
def install(self, host, cluster_id, name): logger.debug('Deploy cluster {} at {}...'.format(cluster_id, host)) path_of_fb = config.get_path_of_fb(cluster_id) release_path = path_of_fb['release_path'] cluster_path = path_of_fb['cluster_path'] if '/' in name: name = os.path.basename(name) installer_path = os.path.join(release_path, name) command = '''chmod 755 {0}; \ PATH=${{PATH}}:/usr/sbin; \ {0} --full {1}'''.format(installer_path, cluster_path) client = net.get_ssh(host) if not net.is_exist(client, installer_path): raise FileNotExistError(installer_path, host=host) net.ssh_execute(client=client, command=command) client.close() logger.debug('OK')
def force_failover(self, server): """ Find all masters on the server and convert them to slaves. Finally, in the server, only slaves will be remained. :param server: IP or hostname """ logger.debug('force_failover') center = Center() center.update_ip_port() master_nodes = center.get_master_obj_list() cluster_id = config.get_cur_cluster_id() lib_path = config.get_ld_library_path(cluster_id) path_of_fb = config.get_path_of_fb(cluster_id) sr2_redis_bin = path_of_fb['sr2_redis_bin'] env_cmd = [ 'GLOBIGNORE=*;', 'export LD_LIBRARY_PATH={};'.format(lib_path['ld_library_path']), 'export DYLD_LIBRARY_PATH={};'.format( lib_path['dyld_library_path']), ] redis_cli_cmd = os.path.join(sr2_redis_bin, 'redis-cli') outs = '' meta = [] m_endpoint = [] for node in master_nodes: addr = node['addr'] (host, port) = addr.split(':') # if host == server: if self.compare_ip(host, server): for slave_node in node['slaves']: addr = slave_node['addr'] (s_host, s_port) = addr.split(':') sub_cmd = 'cluster failover takeover' command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, s_host, s_port, sub_cmd, ) self._print( message.get('try_failover_takeover').format( slave=addr)) stdout = subprocess.check_output(command, shell=True) self._print(stdout)
def delete(self, cluster_id): """Delete cluster It is automatically backed up with timestamps as tags :param cluster_id: target cluster id """ if not cluster_util.validate_id(cluster_id): raise ClusterIdError(cluster_id) path_of_fb = config.get_path_of_fb(cluster_id) props_path = path_of_fb['redis_properties'] hosts = config.get_props(props_path, 'sr2_redis_master_hosts', []) tag = time.strftime("%Y%m%d%H%M%S", time.gmtime()) cluster_backup_dir = 'cluster_{}_bak_{}'.format(cluster_id, tag) for host in hosts: Center().cluster_backup(host, cluster_id, cluster_backup_dir) msg = message.get('cluster_delete_complete') msg = msg.format(cluster_id=cluster_id) logger.info(msg)
def start(self): """Start thriftserver """ logger.debug('thriftserver_command_start') _check_spark() cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) ths_props_path = path_of_fb['thrift_properties'] source_cmd = 'source {}'.format(ths_props_path) hive_opts = _get_hive_opts_str() base_cmd = '$SPARK_SBIN/start-thriftserver.sh {}'.format(hive_opts) cmd = '{}; {}'.format(source_cmd, base_cmd) logger.debug(cmd) os.system(cmd) spark_log = os.path.join(os.environ['SPARK_HOME'], 'logs') if _find_files_with_regex(spark_log, ROLLING_LOGFILE_REGEX): for file in _find_files_with_regex(spark_log, NOHUP_LOGFILE_REGEX): os.remove(os.path.join(spark_log, file))
def reset_distribution(self): """ Reset the distribution of masters and slaves with original setting """ center = Center() center.update_ip_port() logger.debug('reset_distribution') cluster_id = config.get_cur_cluster_id() lib_path = config.get_ld_library_path(cluster_id) path_of_fb = config.get_path_of_fb(cluster_id) sr2_redis_bin = path_of_fb['sr2_redis_bin'] env_cmd = [ 'GLOBIGNORE=*;', 'export LD_LIBRARY_PATH={};'.format(lib_path['ld_library_path']), 'export DYLD_LIBRARY_PATH={};'.format( lib_path['dyld_library_path']), ] redis_cli_cmd = os.path.join(sr2_redis_bin, 'redis-cli') slave_nodes = center.get_slave_nodes() master_ports = center.master_port_list for slave_node in slave_nodes: (host, port) = slave_node.split(':') try: value = int(port) if value in master_ports: # failover takeover msg = message.get('try_failover_takeover').format( slave=slave_node) self._print(msg) sub_cmd = 'cluster failover takeover' command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, host, port, sub_cmd, ) stdout = subprocess.check_output(command, shell=True) outs = '' outs = '\n'.join([outs, stdout]) self._print(outs) except ValueError: pass
def run_monitor(n=10, t=2): """Monitoring logs of redis. :param n: number of lines to print log :param t: renewal cycle(sec) """ if not isinstance(n, int): msg = message.get('error_option_type_not_number').format(option='n') logger.error(msg) return if not isinstance(t, int) and not isinstance(t, float): msg = message.get('error_option_type_not_float').format(option='t') logger.error(msg) return try: sp.check_output('which tail', shell=True) except Exception: msg = message.get('error_not_found_command_tail') logger.error(msg) return cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) sr2_redis_log = path_of_fb['sr2_redis_log'] log_files = '{}/servers*'.format(sr2_redis_log) host_list = config.get_master_host_list() target_host = ask_util.host_for_monitor(host_list) try: sp.check_output('which watch', shell=True) command = "ssh -t {} watch -n {} 'tail -n {} {}'".format( target_host, t, n, log_files) sp.call(command, shell=True) except Exception: msg = message.get('error_not_found_command_watch') logger.warning(msg) logger.info(message.get('message_for_exit')) command = "tail -F -s {} {}".format(t, log_files) client = net.get_ssh(target_host) net.ssh_execute_async(client, command)
def transfer_installer(self, host, cluster_id, installer_path): installer_path = os.path.expanduser(installer_path) path_of_fb = config.get_path_of_fb(cluster_id) name = os.path.basename(installer_path) dst = os.path.join(path_of_fb['release_path'], name) client = net.get_ssh(host) sftp = net.get_sftp(client) if not net.is_dir(client, path_of_fb['release_path']): logger.debug("Not exist releases directory at '{}'".format(host)) sftp.mkdir(path_of_fb['release_path']) logger.debug("Create releases directory at '{}'".format(host)) logger.debug('Check {}...'.format(name)) if not net.is_exist(client, dst): logger.debug('FAIL') logger.debug("Transfer '{}' to '{}'...".format(name, host)) sftp.put(installer_path, '{}.download'.format(dst)) command = 'mv {0}.download {0}'.format(dst) net.ssh_execute(client=client, command=command) sftp.close() client.close() logger.debug('OK')
def monitor(self): """Monitoring log of thriftserver" """ logger.debug('thriftserver_command_monitor') _check_spark() cluster_id = config.get_cur_cluster_id() path_of_fb = config.get_path_of_fb(cluster_id) ths_props_path = path_of_fb['thrift_properties'] source_cmd = 'source {}'.format(ths_props_path) spark_log = _get_env()['spark_log'] # log_file_path = ''.join([ # spark_log, # '/spark-{}-org.apache'.format(os.environ['USER']), # '.spark.sql.hive.thriftserver.HiveThriftServer2-1-*.out', # ]) log_file_path = os.path.join(spark_log, NOHUP_LOGFILE) if _find_files_with_regex(spark_log, ROLLING_LOGFILE_REGEX): log_file_path = os.path.join(spark_log, ROLLING_LOGFILE) base_cmd = 'tail -F {}'.format(log_file_path) cmd = '{}; {}'.format(source_cmd, base_cmd) logger.debug(cmd) msg = message.get('message_for_exit') logger.info(msg) os.system(cmd)
def _deploy(cluster_id, history_save, clean): deploy_state = DeployUtil().get_state(cluster_id) if deploy_state == DEPLOYED: msg = message.get('ask_deploy_again') msg = msg.format(cluster_id=cluster_id) msg = color.yellow(msg) yes = ask_util.askBool(msg, default='n') if not yes: logger.info(message.get('cancel')) return restore_yes = None no_localhost = False current_time = time.strftime("%Y%m%d%H%M%S", time.gmtime()) cluster_backup_dir = 'cluster_{}_bak_{}'.format(cluster_id, current_time) conf_backup_dir = 'cluster_{}_conf_bak_{}'.format(cluster_id, current_time) tmp_backup_dir = 'cluster_{}_conf_bak_{}'.format(cluster_id, 'tmp') meta = [['NAME', 'VALUE']] path_of_fb = config.get_path_of_fb(cluster_id) conf_path = path_of_fb['conf_path'] props_path = path_of_fb['redis_properties'] cluster_path = path_of_fb['cluster_path'] path_of_cli = config.get_path_of_cli(cluster_id) conf_backup_path = path_of_cli['conf_backup_path'] tmp_backup_path = os.path.join(conf_backup_path, tmp_backup_dir) local_ip = config.get_local_ip() # ask installer installer_path = ask_util.installer() installer_name = os.path.basename(installer_path) meta.append(['installer', installer_name]) # ask restore conf if deploy_state == DEPLOYED: restore_yes = ask_util.askBool(message.get('ask_restore_conf')) meta.append(['restore', restore_yes]) # input props hosts = [] if deploy_state == DEPLOYED: if restore_yes: meta += DeployUtil().get_meta_from_props(props_path) hosts = config.get_props(props_path, 'sr2_redis_master_hosts') else: if not os.path.isdir(conf_backup_path): os.mkdir(conf_backup_path) if os.path.exists(tmp_backup_path): msg = message.get('ask_load_history_of_previous_modification') yes = ask_util.askBool(msg) if not yes: shutil.rmtree(tmp_backup_path) if not os.path.exists(tmp_backup_path): os.mkdir(tmp_backup_path) shutil.copy(os.path.join(conf_path, 'redis.properties'), os.path.join(tmp_backup_path, 'redis.properties')) tmp_props_path = os.path.join(tmp_backup_path, 'redis.properties') editor.edit(tmp_props_path, syntax='sh') meta += DeployUtil().get_meta_from_props(tmp_props_path) hosts = config.get_props(tmp_props_path, 'sr2_redis_master_hosts') else: # new deploy props_dict = ask_util.props(cluster_id, save=history_save) hosts = props_dict['hosts'] meta += DeployUtil().get_meta_from_dict(props_dict) utils.print_table(meta) msg = message.get('confirm_deploy_information') yes = ask_util.askBool(msg) if not yes: logger.info(message.get('cancel')) return # check node status success = Center().check_hosts_connection(hosts, True) if not success: msg = message.get('error_exist_unavailable_host') logger.error(msg) return logger.debug('Connection of all hosts ok.') success = Center().check_include_localhost(hosts) if not success: no_localhost = True # get port info if deploy_state == DEPLOYED: if restore_yes: key = 'sr2_redis_master_ports' m_ports = config.get_props(props_path, key, []) key = 'sr2_redis_slave_ports' s_ports = config.get_props(props_path, key, []) replicas = len(s_ports) // len(m_ports) else: key = 'sr2_redis_master_ports' m_ports = config.get_props(tmp_props_path, key, []) key = 'sr2_redis_slave_ports' s_ports = config.get_props(tmp_props_path, key, []) replicas = len(s_ports) // len(m_ports) else: m_ports = props_dict['master_ports'] s_ports = props_dict['slave_ports'] replicas = props_dict['replicas'] while True: msg = message.get('check_port') logger.info(msg) host_ports_list = [] for host in hosts: host_ports_list.append((host, m_ports + s_ports)) conflict = Center().check_port_is_enable(host_ports_list) if not conflict: logger.info("OK") break utils.print_table([["HOST", "PORT"]] + conflict) msg = message.get('ask_port_collision') msg = color.yellow(msg) yes = ask_util.askBool(msg) if yes: logger.info("OK") break m_ports = ask_util.master_ports(False, cluster_id) replicas = ask_util.replicas(False) s_ports = ask_util.slave_ports(cluster_id, len(m_ports), replicas) if deploy_state == DEPLOYED: if restore_yes: key = 'sr2_redis_master_ports' value = cluster_util.convert_list_2_seq(m_ports) config.set_props(props_path, key, value) key = 'sr2_redis_slave_ports' value = cluster_util.convert_list_2_seq(s_ports) config.set_props(props_path, key, value) else: key = 'sr2_redis_master_ports' value = cluster_util.convert_list_2_seq(m_ports) config.set_props(tmp_props_path, key, value) key = 'sr2_redis_slave_ports' value = cluster_util.convert_list_2_seq(s_ports) config.set_props(tmp_props_path, key, value) else: props_dict['master_ports'] = m_ports props_dict['slave_ports'] = s_ports props_dict['replicas'] = replicas # if pending, delete legacy on each hosts if no_localhost: if DeployUtil().get_state(cluster_id, local_ip) == PENDING: client = net.get_ssh(local_ip) command = 'rm -rf {}'.format(cluster_path) net.ssh_execute(client=client, command=command) client.close() for host in hosts: if DeployUtil().get_state(cluster_id, host) == PENDING: client = net.get_ssh(host) command = 'rm -rf {}'.format(cluster_path) net.ssh_execute(client=client, command=command) client.close() # added_hosts = post_hosts - pre_hosts msg = message.get('check_cluster_exist') logger.info(msg) added_hosts = set(hosts) meta = [] if deploy_state == DEPLOYED: pre_hosts = config.get_props(props_path, 'sr2_redis_master_hosts') added_hosts -= set(pre_hosts) can_deploy = True if no_localhost: added_hosts |= set([local_ip]) for host in added_hosts: client = net.get_ssh(host) is_localhost = Center().is_localhost(host) if is_localhost: if no_localhost: continue if os.path.exists(cluster_path + '/remote'): meta.append([host, color.green('CLEAN')]) continue if net.is_exist(client, cluster_path): meta.append([host, color.red('CLUSTER EXIST')]) can_deploy = False continue meta.append([host, color.green('CLEAN')]) if meta: utils.print_table([['HOST', 'STATUS']] + meta) if not can_deploy: msg = message.get('error_cluster_collision') logger.error(msg) return # if not force: # logger.error("If you trying to force, use option '--force'") # return logger.info('OK') # cluster stop and clean if deploy_state == DEPLOYED and clean: center = Center() cur_cluster_id = config.get_cur_cluster_id(allow_empty_id=True) run_cluster_use(cluster_id) center.update_ip_port() center.stop_redis() center.remove_all_of_redis_log_force() center.cluster_clean() run_cluster_use(cur_cluster_id) # backup conf if deploy_state == DEPLOYED: Center().conf_backup(local_ip, cluster_id, conf_backup_dir) # backup cluster backup_hosts = [] if deploy_state == DEPLOYED: backup_hosts += set(pre_hosts) # if force: # backup_hosts += added_hosts for host in backup_hosts: cluster_path = path_of_fb['cluster_path'] client = net.get_ssh(host) Center().cluster_backup(host, cluster_id, cluster_backup_dir) client.close() # transfer & install msg = message.get('transfer_and_execute_installer') logger.info(msg) target_hosts = hosts + [local_ip] if no_localhost else hosts for host in target_hosts: if not (no_localhost and Center().is_localhost(host)): logger.info(' - {}'.format(host)) client = net.get_ssh(host) cmd = 'mkdir -p {0} && touch {0}/.deploy.state'.format(cluster_path) net.ssh_execute(client=client, command=cmd) client.close() DeployUtil().transfer_installer(host, cluster_id, installer_path) try: DeployUtil().install(host, cluster_id, installer_name) except SSHCommandError as ex: msg = message.get('error_execute_installer') msg = msg.format(installer=installer_path) logger.error(msg) logger.exception(ex) return # setup props if deploy_state == DEPLOYED: if restore_yes: tag = conf_backup_dir else: tag = tmp_backup_dir Center().conf_restore(local_ip, cluster_id, tag) else: key = 'sr2_redis_master_hosts' config.make_key_enable(props_path, key) config.set_props(props_path, key, props_dict['hosts']) key = 'sr2_redis_master_ports' config.make_key_enable(props_path, key) value = cluster_util.convert_list_2_seq(props_dict['master_ports']) config.set_props(props_path, key, value) key = 'sr2_redis_slave_hosts' config.make_key_enable(props_path, key) config.set_props(props_path, key, props_dict['hosts']) config.make_key_disable(props_path, key) if props_dict['replicas'] > 0: key = 'sr2_redis_slave_hosts' config.make_key_enable(props_path, key) key = 'sr2_redis_slave_ports' config.make_key_enable(props_path, key) value = cluster_util.convert_list_2_seq(props_dict['slave_ports']) config.set_props(props_path, key, value) key = 'ssd_count' config.make_key_enable(props_path, key) config.set_props(props_path, key, props_dict['ssd_count']) key = 'sr2_redis_data' config.make_key_enable(props_path, key, v1_flg=True) config.make_key_enable(props_path, key, v1_flg=True) config.make_key_disable(props_path, key) config.set_props(props_path, key, props_dict['prefix_of_db_path']) key = 'sr2_redis_db_path' config.make_key_enable(props_path, key, v1_flg=True) config.make_key_enable(props_path, key, v1_flg=True) config.make_key_disable(props_path, key) config.set_props(props_path, key, props_dict['prefix_of_db_path']) key = 'sr2_flash_db_path' config.make_key_enable(props_path, key, v1_flg=True) config.make_key_enable(props_path, key, v1_flg=True) config.make_key_disable(props_path, key) config.set_props(props_path, key, props_dict['prefix_of_db_path']) # synk props msg = message.get('sync_conf') logger.info(msg) for node in hosts: if socket.gethostbyname(node) in config.get_local_ip_list(): continue client = net.get_ssh(node) if not client: msg = message.get('error_ssh_connection').format(host=node) logger.error(msg) return net.copy_dir_to_remote(client, conf_path, conf_path) client.close() # set deploy state complete if os.path.exists(tmp_backup_path): shutil.rmtree(tmp_backup_path) for node in target_hosts: path_of_fb = config.get_path_of_fb(cluster_id) cluster_path = path_of_fb['cluster_path'] client = net.get_ssh(node) cmd = 'rm -rf {}'.format(os.path.join(cluster_path, '.deploy.state')) net.ssh_execute(client=client, command=cmd) client.close() if no_localhost: os.system('touch {}/remote'.format(cluster_path)) msg = message.get('complete_deploy').format(cluster_id=cluster_id) logger.info(msg) Cluster().use(cluster_id) msg = message.get('suggest_after_deploy') logger.info(msg)
def _deploy_zero_downtime(cluster_id): logger.debug("zero downtime update cluster {}".format(cluster_id)) center = Center() center.update_ip_port() m_hosts = center.master_host_list m_ports = center.master_port_list s_hosts = center.slave_host_list s_ports = center.slave_port_list path_of_fb = config.get_path_of_fb(cluster_id) cluster_path = path_of_fb['cluster_path'] # check master alive m_count = len(m_hosts) * len(m_ports) alive_m_count = center.get_alive_master_redis_count() if alive_m_count < m_count: logger.error(message.get('error_exist_disconnected_master')) return if not config.is_slave_enabled: logger.error(message.get('error_need_to_slave')) return # select installer installer_path = ask_util.installer() installer_name = os.path.basename(installer_path) # backup info current_time = time.strftime("%Y%m%d%H%M%S", time.gmtime()) conf_backup_dir = 'cluster_{}_conf_bak_{}'.format(cluster_id, current_time) cluster_backup_dir = 'cluster_{}_bak_{}'.format(cluster_id, current_time) local_ip = config.get_local_ip() # backup conf center.conf_backup(local_ip, cluster_id, conf_backup_dir) # backup cluster for host in s_hosts: client = net.get_ssh(host) center.cluster_backup(host, cluster_id, cluster_backup_dir) client.close() # transfer & install logger.info(message.get('transfer_and_execute_installer')) for host in m_hosts: logger.info(' - {}'.format(host)) client = net.get_ssh(host) cmd = 'mkdir -p {0} && touch {0}/.deploy.state'.format(cluster_path) net.ssh_execute(client=client, command=cmd) client.close() DeployUtil().transfer_installer(host, cluster_id, installer_path) try: DeployUtil().install(host, cluster_id, installer_name) except SSHCommandError as ex: msg = message.get('error_execute_installer') msg = msg.format(installer=installer_path) logger.error(msg) logger.exception(ex) return # restore conf center.conf_restore(local_ip, cluster_id, conf_backup_dir) # set deploy state complete for node in m_hosts: path_of_fb = config.get_path_of_fb(cluster_id) cluster_path = path_of_fb['cluster_path'] client = net.get_ssh(node) cmd = 'rm -rf {}'.format(os.path.join(cluster_path, '.deploy.state')) net.ssh_execute(client=client, command=cmd) client.close() # restart slave center.stop_current_nodes(master=False, slave=True) center.configure_redis() center.sync_conf() center.start_current_nodes(master=False, slave=True) center.wait_until_all_redis_process_up() slaves_for_failover = center.get_slave_nodes() key = 'cluster-node-timeout' origin_m_value = center.cli_config_get(key, m_hosts[0], m_ports[0]) origin_s_value = center.cli_config_get(key, s_hosts[0], s_ports[0]) logger.debug('config set: cluster-node-timeout 2000') RedisCliConfig().set(key, '2000', all=True) # cluster failover (with no option) logger.info(message.get('failover_on_deploy')) logger.debug(slaves_for_failover) try_count = 0 while try_count < 10: try_count += 1 success = True for slave_addr in slaves_for_failover: host, port = slave_addr.split(':') stdout = center.run_failover("{}:{}".format(host, port)) logger.debug("failover {}:{} {}".format(host, port, stdout)) if stdout != "ERR You should send CLUSTER FAILOVER to a slave": # In some cases, the cluster failover is not complete # even if stdout is OK # If redis changed to master completely, # return 'ERR You should send CLUSTER FAILOVER to a slave' success = False if success: break msg = message.get('retry').format(try_count=try_count) logger.info(msg) time.sleep(5) logger.debug('restore config: cluster-node-timeout') center.cli_config_set_all(key, origin_m_value, m_hosts, m_ports) center.cli_config_set_all(key, origin_s_value, s_hosts, s_ports) if not success: logger.error(message.get('error_redis_failover')) return # restart master (current slave) center.stop_current_nodes(master=False, slave=True) center.configure_redis(slave=False) center.sync_conf() center.start_current_nodes(master=False, slave=True) center.wait_until_all_redis_process_up()
def restore(self, cluster_id, tag=None): """Restore cluster :param cluster_id: target cluster id :param tag: Tag of backup, if omitted, restore the most recent backup file """ logger.debug('cluster restore: cluster_id={}, tag={}'.format( cluster_id, tag )) if not cluster_util.validate_id(cluster_id): raise ClusterIdError(cluster_id) # find restore folder with tag (local) path_of_fb = config.get_path_of_fb(cluster_id) cluster_backup_path = path_of_fb['cluster_backup_path'] if tag is None: backup_list = os.listdir(cluster_backup_path) pattern = 'cluster_{}_bak_'.format(cluster_id) filtered = filter(lambda x: x.startswith(pattern), backup_list) sorted_list = sorted(list(filtered)) if not sorted_list: msg = message.get('error_not_found_any_backup') logger.error('BackupNotExistError: ' + msg) return tag = sorted_list[-1] logger.debug("tag option is empty, auto select: {}".format(tag)) cluster_restore_dir = tag backup_path = os.path.join(cluster_backup_path, cluster_restore_dir) if not os.path.isdir(backup_path): msg = message.get('error_not_found_backup').format(tag=tag) logger.error('BackupNotExistError: ' + msg) return # get hosts from cluster props props_path = os.path.join( backup_path, 'tsr2-assembly-1.0.0-SNAPSHOT', 'conf', 'redis.properties' ) hosts = config.get_props(props_path, 'sr2_redis_master_hosts', []) # check status of hosts success = Center().check_hosts_connection(hosts, True) if not success: msg = message.get('error_exist_unavailable_host') logger.error(msg) return logger.debug('Connection of all hosts ok.') success = Center().check_include_localhost(hosts) if not success: msg = message.get('error_not_include_localhost') logger.error(msg) return # check all host tag folder: OK / NOT FOUND msg = message.get('check_backup_info') logger.info(msg) buf = [] for host in hosts: client = net.get_ssh(host) if not net.is_dir(client, backup_path): logger.debug('cannot find backup dir: {}-{}'.format( host, cluster_restore_dir )) buf.append([host, color.red('NOT FOUND')]) client.close() if buf: utils.print_table([['HOST', 'RESULT'] + buf]) return logger.info('OK') # backup cluster new_tag = time.strftime("%Y%m%d%H%M%S", time.gmtime()) cluster_backup_dir = 'cluster_{}_bak_{}'.format(cluster_id, new_tag) for host in hosts: Center().cluster_backup(host, cluster_id, cluster_backup_dir) # restore cluster command = "cp -r {} {}/cluster_{}".format( backup_path, path_of_fb['base_directory'], cluster_id ) for host in hosts: msg = message.get('restore_cluster') msg = msg.format(tag=cluster_backup_dir, host=host) logger.info(msg) client = net.get_ssh(host) net.ssh_execute(client, command) client.close() logger.info("OK")
def do_replicate(self, slave, master): """ Replicate a slave node to a master node. Use like 'cluster replicate {slave's ip}:{slave's port} {master's ip}:{master's port} :param slave: {slave's ip or hostname}:{slave's port} :param master: {master's ip or hostname}:{master's port} """ logger.debug('do_replicate') # Get master's uuid s_hostname, s_port = slave.split(':') m_hostname, m_port = master.split(':') s_host = socket.gethostbyname(s_hostname) m_host = socket.gethostbyname(m_hostname) cluster_id = config.get_cur_cluster_id() lib_path = config.get_ld_library_path(cluster_id) path_of_fb = config.get_path_of_fb(cluster_id) sr2_redis_bin = path_of_fb['sr2_redis_bin'] env_cmd = [ 'GLOBIGNORE=*;', 'export LD_LIBRARY_PATH={};'.format(lib_path['ld_library_path']), 'export DYLD_LIBRARY_PATH={};'.format( lib_path['dyld_library_path']), ] redis_cli_cmd = os.path.join(sr2_redis_bin, 'redis-cli') sub_cmd = 'cluster nodes' command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, s_host, s_port, sub_cmd, ) stdout = subprocess.check_output(command, shell=True) outs = '' outs = '\n'.join([outs, stdout]) lines = outs.splitlines() m_ip_port = m_host + str(':') + m_port filtered_lines = (filter(lambda x: m_ip_port in x, lines)) if len(filtered_lines) == 0: msg = message.get('error_need_cluster_meet') self._print(msg) msg = message.get('cluster_meet') self._print(msg) # Cluster meet sub_cmd = 'cluster meet {ip} {port}'.format(ip=m_host, port=m_port) command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, s_host, s_port, sub_cmd, ) stdout = subprocess.check_output(command, shell=True) self._print(stdout) # Get master's uuid sub_cmd = 'cluster nodes' command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, s_host, s_port, sub_cmd, ) stdout = subprocess.check_output(command, shell=True) outs = '' outs = '\n'.join([outs, stdout]) lines = outs.splitlines() filtered_lines = (filter(lambda x: m_ip_port in x, lines)) m_uuid = filtered_lines[0].split()[0] else: m_uuid = filtered_lines[0].split()[0] if len(m_uuid) == 0: msg = message.get('error_no_uuid') raise ClusterRedisError(msg) # Replicate msg = message.get('start_replicate') self._print(msg) sub_cmd = 'cluster replicate {uuid}'.format(uuid=m_uuid) command = '{} {} -h {} -p {} {}'.format( ' '.join(env_cmd), redis_cli_cmd, s_host, s_port, sub_cmd, ) stdout = subprocess.check_output(command, shell=True) outs = '' outs = '\n'.join([outs, stdout]) self._print(outs)