/* * set backend connection close timer */ void pool_connection_pool_timer(POOL_CONNECTION_POOL *backend) { POOL_CONNECTION_POOL *p = pool_connection_pool; int i; pool_debug("pool_connection_pool_timer: set close time %ld", time(NULL)); MASTER_CONNECTION(backend)->closetime = time(NULL); /* set connection close time */ if (pool_config->connection_life_time == 0) return; /* look for any other timeout */ for (i=0;i<pool_config->max_pool;i++, p++) { if (!MASTER_CONNECTION(p)) continue; if (!MASTER_CONNECTION(p)->sp) continue; if (MASTER_CONNECTION(p)->sp->user == NULL) continue; if (p != backend && MASTER_CONNECTION(p)->closetime) return; } /* no other timer found. set my timer */ pool_debug("pool_connection_pool_timer: set alarm after %d seconds", pool_config->connection_life_time); pool_signal(SIGALRM, pool_backend_timer_handler); alarm(pool_config->connection_life_time); }
void pool_backend_timer(void) { #define TMINTMAX 0x7fffffff POOL_CONNECTION_POOL *p = pool_connection_pool; int i, j; time_t now; time_t nearest = TMINTMAX; ConnectionInfo *info; POOL_SETMASK(&BlockSig); now = time(NULL); pool_debug("pool_backend_timer_handler called at %ld", now); for (i=0;i<pool_config->max_pool;i++, p++) { if (!MASTER_CONNECTION(p)) continue; if (!MASTER_CONNECTION(p)->sp) continue; if (MASTER_CONNECTION(p)->sp->user == NULL) continue; /* timer expire? */ if (MASTER_CONNECTION(p)->closetime) { int freed = 0; pool_debug("pool_backend_timer_handler: expire time: %ld", MASTER_CONNECTION(p)->closetime+pool_config->connection_life_time); if (now >= (MASTER_CONNECTION(p)->closetime+pool_config->connection_life_time)) { /* discard expired connection */ pool_debug("pool_backend_timer_handler: expires user %s database %s", MASTER_CONNECTION(p)->sp->user, MASTER_CONNECTION(p)->sp->database); pool_send_frontend_exits(p); for (j=0;j<NUM_BACKENDS;j++) { if (!VALID_BACKEND(j)) continue; if (!freed) { pool_free_startup_packet(CONNECTION_SLOT(p, j)->sp); freed = 1; } pool_close(CONNECTION(p, j)); free(CONNECTION_SLOT(p, j)); } info = p->info; memset(p, 0, sizeof(POOL_CONNECTION_POOL)); p->info = info; memset(p->info, 0, sizeof(ConnectionInfo) * MAX_NUM_BACKENDS); /* prepare to shutdown connections to system db */ if(pool_config->system_db_dynamic_connection && (pool_config->parallel_mode || pool_config->enable_query_cache)) { if (system_db_info->pgconn) pool_close_libpq_connection(); if (pool_system_db_connection() && pool_system_db_connection()->con) { pool_send_frontend_exit(pool_system_db_connection()->con); pool_close(pool_system_db_connection()->con); } if( system_db_info->connection ) { free( system_db_info->connection ); memset(system_db_info->connection, 0, sizeof(POOL_CONNECTION_POOL_SLOT)); system_db_info->connection = NULL; } } } else { /* look for nearest timer */ if (MASTER_CONNECTION(p)->closetime < nearest) nearest = MASTER_CONNECTION(p)->closetime; } } } /* any remaining timer */ if (nearest != TMINTMAX) { nearest = pool_config->connection_life_time - (now - nearest); if (nearest <= 0) nearest = 1; pool_signal(SIGALRM, pool_backend_timer_handler); alarm(nearest); } POOL_SETMASK(&UnBlockSig); }
/* * Read startup packet * * Read the startup packet and parse the contents. */ static StartupPacket *read_startup_packet(POOL_CONNECTION *cp) { StartupPacket *sp; StartupPacket_v2 *sp2; int protov; int len; char *p; sp = (StartupPacket *)calloc(sizeof(*sp), 1); if (!sp) { pool_error("read_startup_packet: out of memory"); return NULL; } if (pool_config->authentication_timeout > 0) { pool_signal(SIGALRM, authentication_timeout); alarm(pool_config->authentication_timeout); } /* read startup packet length */ if (pool_read(cp, &len, sizeof(len))) { pool_error("read_startup_packet: incorrect packet length (%d)", len); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } len = ntohl(len); len -= sizeof(len); if (len <= 0 || len >= MAX_STARTUP_PACKET_LENGTH) { pool_error("read_startup_packet: incorrect packet length (%d)", len); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } sp->startup_packet = calloc(len, 1); if (!sp->startup_packet) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } /* read startup packet */ if (pool_read(cp, sp->startup_packet, len)) { pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } sp->len = len; memcpy(&protov, sp->startup_packet, sizeof(protov)); sp->major = ntohl(protov)>>16; sp->minor = ntohl(protov) & 0x0000ffff; p = sp->startup_packet; switch(sp->major) { case PROTO_MAJOR_V2: /* V2 */ sp2 = (StartupPacket_v2 *)(sp->startup_packet); sp->database = calloc(SM_DATABASE+1, 1); if (!sp->database) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } strncpy(sp->database, sp2->database, SM_DATABASE); sp->user = calloc(SM_USER+1, 1); if (!sp->user) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } strncpy(sp->user, sp2->user, SM_USER); break; case PROTO_MAJOR_V3: /* V3 */ p += sizeof(int); /* skip protocol version info */ while(*p) { if (!strcmp("user", p)) { p += (strlen(p) + 1); sp->user = strdup(p); if (!sp->user) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } } else if (!strcmp("database", p)) { p += (strlen(p) + 1); sp->database = strdup(p); if (!sp->database) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } } /* * From 9.0, the start up packet may include * application name. After receiving such that packet, * backend sends parameter status of application_name. * Upon reusing connection to backend, we need to * emulate this behavior of backend. So we remember * this and send parameter status packet to frontend * instead of backend in * connect_using_existing_connection(). */ else if (!strcmp("application_name", p)) { p += (strlen(p) + 1); sp->application_name = p; pool_debug("read_startup_packet: application_name: %s", p); } p += (strlen(p) + 1); } break; case 1234: /* cancel or SSL request */ /* set dummy database, user info */ sp->database = calloc(1, 1); if (!sp->database) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } sp->user = calloc(1, 1); if (!sp->user) { pool_error("read_startup_packet: out of memory"); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } break; default: pool_error("read_startup_packet: invalid major no: %d", sp->major); pool_free_startup_packet(sp); alarm(0); pool_signal(SIGALRM, SIG_IGN); return NULL; } pool_debug("Protocol Major: %d Minor: %d database: %s user: %s", sp->major, sp->minor, sp->database, sp->user); alarm(0); pool_signal(SIGALRM, SIG_IGN); return sp; }
/* * fork escalation process */ pid_t fork_escalation_process(void) { pid_t pid; pid = fork(); if (pid != 0) { if (pid == -1) ereport(NOTICE, (errmsg("failed to fork a escalation process"))); return pid; } on_exit_reset(); processType = PT_WATCHDOG_UTILITY; POOL_SETMASK(&UnBlockSig); init_ps_display("", "", "", ""); pool_signal(SIGTERM, wd_exit); pool_signal(SIGINT, wd_exit); pool_signal(SIGQUIT, wd_exit); pool_signal(SIGCHLD, SIG_DFL); pool_signal(SIGHUP, SIG_IGN); pool_signal(SIGPIPE, SIG_IGN); MemoryContextSwitchTo(TopMemoryContext); set_ps_display("watchdog escalation", false); ereport(LOG, (errmsg("watchdog: escalation started"))); /* * STEP 1 clear shared memory cache */ if (pool_config->memory_cache_enabled && pool_is_shmem_cache() && pool_config->clear_memqcache_on_escalation) { ereport(LOG, (errmsg("watchdog escalation"), errdetail("clearing all the query cache on shared memory"))); pool_clear_memory_cache(); } /* * STEP 2 execute escalation command provided by user in pgpool conf file */ if (strlen(pool_config->wd_escalation_command)) { int r = system(pool_config->wd_escalation_command); if (WIFEXITED(r)) { if (WEXITSTATUS(r) == EXIT_SUCCESS) ereport(LOG, (errmsg("watchdog escalation successful"))); else { ereport(WARNING, (errmsg("watchdog escalation command failed with exit status: %d", WEXITSTATUS(r)))); } } else { ereport(WARNING, (errmsg("watchdog escalation command exit abnormally"))); } } /* * STEP 3 bring up the delegate IP */ if (strlen(pool_config->delegate_IP) != 0) { if (wd_IP_up() != WD_OK) ereport(WARNING, (errmsg("watchdog escalation failed to acquire delegate IP"))); } exit(0); }
/* * fork de-escalation process */ pid_t fork_plunging_process(void) { pid_t pid; pid = fork(); if (pid != 0) { if (pid == -1) ereport(NOTICE, (errmsg("failed to fork a de-escalation process"))); return pid; } on_exit_reset(); processType = PT_WATCHDOG_UTILITY; POOL_SETMASK(&UnBlockSig); init_ps_display("", "", "", ""); pool_signal(SIGTERM, wd_exit); pool_signal(SIGINT, wd_exit); pool_signal(SIGQUIT, wd_exit); pool_signal(SIGCHLD, SIG_DFL); pool_signal(SIGHUP, SIG_IGN); pool_signal(SIGPIPE, SIG_IGN); MemoryContextSwitchTo(TopMemoryContext); set_ps_display("watchdog de-escalation", false); ereport(LOG, (errmsg("watchdog: de-escalation started"))); /* * STEP 1 execute de-escalation command provided by user in pgpool conf * file */ if (strlen(pool_config->wd_de_escalation_command)) { int r = system(pool_config->wd_de_escalation_command); if (WIFEXITED(r)) { if (WEXITSTATUS(r) == EXIT_SUCCESS) ereport(LOG, (errmsg("watchdog de-escalation successful"))); else { ereport(WARNING, (errmsg("watchdog de-escalation command failed with exit status: %d", WEXITSTATUS(r)))); } } else { ereport(WARNING, (errmsg("watchdog de-escalation command exit abnormally"))); } } /* * STEP 2 bring down the delegate IP */ if (strlen(pool_config->delegate_IP) != 0) { if (wd_IP_down() != WD_OK) ereport(WARNING, (errmsg("watchdog de-escalation failed to bring down delegate IP"))); } exit(0); }
void pool_backend_timer(void) { #define TMINTMAX 0x7fffffff POOL_CONNECTION_POOL *p = pool_connection_pool; int i; time_t now; time_t nearest = TMINTMAX; POOL_SETMASK(&BlockSig); now = time(NULL); pool_debug("pool_backend_timer_handler called at %d", now); for (i=0;i<pool_config.max_pool;i++, p++) { if (!MASTER_CONNECTION(p)) continue; if (MASTER_CONNECTION(p)->sp->user == NULL) continue; /* timer expire? */ if (MASTER_CONNECTION(p)->closetime) { pool_debug("pool_backend_timer_handler: expire time: %d", MASTER_CONNECTION(p)->closetime+pool_config.connection_life_time); if (now >= (MASTER_CONNECTION(p)->closetime+pool_config.connection_life_time)) { /* discard expired connection */ pool_debug("pool_backend_timer_handler: expires user %s database %s", MASTER_CONNECTION(p)->sp->user, MASTER_CONNECTION(p)->sp->database); pool_send_frontend_exits(p); pool_free_startup_packet(MASTER_CONNECTION(p)->sp); pool_close(MASTER_CONNECTION(p)->con); free(MASTER_CONNECTION(p)); if (DUAL_MODE) { pool_close(SECONDARY_CONNECTION(p)->con); free(SECONDARY_CONNECTION(p)); } memset(p, 0, sizeof(POOL_CONNECTION_POOL)); } else { /* look for nearest timer */ if (MASTER_CONNECTION(p)->closetime < nearest) nearest = MASTER_CONNECTION(p)->closetime; } } } /* any remaining timer */ if (nearest != TMINTMAX) { nearest = pool_config.connection_life_time - (now - nearest); if (nearest <= 0) nearest = 1; pool_signal(SIGALRM, pool_backend_timer_handler); alarm(nearest); } POOL_SETMASK(&UnBlockSig); }
/* * pgpool main program */ int main(int argc, char **argv) { int opt; int i; int pid; int size; int retrycnt; int sys_retrycnt; myargc = argc; myargv = argv; snprintf(conf_file, sizeof(conf_file), "%s/%s", DEFAULT_CONFIGDIR, POOL_CONF_FILE_NAME); snprintf(pcp_conf_file, sizeof(pcp_conf_file), "%s/%s", DEFAULT_CONFIGDIR, PCP_PASSWD_FILE_NAME); snprintf(hba_file, sizeof(hba_file), "%s/%s", DEFAULT_CONFIGDIR, HBA_CONF_FILE_NAME); while ((opt = getopt(argc, argv, "a:cdf:F:hm:nv")) != -1) { switch (opt) { case 'a': /* specify hba configuration file */ if (!optarg) { usage(); exit(1); } strncpy(hba_file, optarg, sizeof(hba_file)); break; case 'c': /* clear cache option */ clear_cache = 1; break; case 'd': /* debug option */ debug = 1; break; case 'f': /* specify configuration file */ if (!optarg) { usage(); exit(1); } strncpy(conf_file, optarg, sizeof(conf_file)); break; case 'F': /* specify PCP password file */ if (!optarg) { usage(); exit(1); } strncpy(pcp_conf_file, optarg, sizeof(pcp_conf_file)); break; case 'h': usage(); exit(0); break; case 'm': /* stop mode */ if (!optarg) { usage(); exit(1); } if (*optarg == 's' || !strcmp("smart", optarg)) stop_sig = SIGTERM; /* smart shutdown */ else if (*optarg == 'f' || !strcmp("fast", optarg)) stop_sig = SIGINT; /* fast shutdown */ else if (*optarg == 'i' || !strcmp("immediate", optarg)) stop_sig = SIGQUIT; /* immediate shutdown */ else { usage(); exit(1); } break; case 'n': /* no detaching control ttys */ not_detach = 1; break; case 'v': show_version(); exit(0); default: usage(); exit(1); } } mypid = getpid(); if (pool_init_config()) exit(1); if (pool_get_config(conf_file, INIT_CONFIG)) { pool_error("Unable to get configuration. Exiting..."); exit(1); } if (pool_config->enable_pool_hba) load_hba(hba_file); /* * If a non-switch argument remains, then it should be either "reload" or "stop". */ if (optind == (argc - 1)) { if (!strcmp(argv[optind], "reload")) { pid_t pid; pid = read_pid_file(); if (pid < 0) { pool_error("could not read pid file"); pool_shmem_exit(1); exit(1); } if (kill(pid, SIGHUP) == -1) { pool_error("could not reload configuration file pid: %d. reason: %s", pid, strerror(errno)); pool_shmem_exit(1); exit(1); } pool_shmem_exit(0); exit(0); } if (!strcmp(argv[optind], "stop")) { stop_me(); pool_shmem_exit(0); exit(0); } else { usage(); pool_shmem_exit(1); exit(1); } } /* * else if no non-switch argument remains, then it should be a start request */ else if (optind == argc) { pid = read_pid_file(); if (pid > 0) { if (kill(pid, 0) == 0) { fprintf(stderr, "pid file found. is another pgpool(%d) is running?\n", pid); exit(1); } else fprintf(stderr, "pid file found but it seems bogus. Trying to start pgpool anyway...\n"); } } /* * otherwise an error... */ else { usage(); exit(1); } /* set signal masks */ poolinitmask(); if (not_detach) write_pid_file(); else daemonize(); if (pool_semaphore_create(MAX_NUM_SEMAPHORES)) { pool_error("Unable to create semaphores. Exiting..."); pool_shmem_exit(1); exit(1); } /* * Restore previous backend status if possible */ read_status_file(); /* clear cache */ if (clear_cache && pool_config->enable_query_cache && SYSDB_STATUS == CON_UP) { Interval interval[1]; interval[0].quantity = 0; interval[0].unit = second; pool_clear_cache_by_time(interval, 1); } /* set unix domain socket path */ snprintf(un_addr.sun_path, sizeof(un_addr.sun_path), "%s/.s.PGSQL.%d", pool_config->socket_dir, pool_config->port); /* set up signal handlers */ pool_signal(SIGPIPE, SIG_IGN); /* create unix domain socket */ unix_fd = create_unix_domain_socket(un_addr); /* create inet domain socket if any */ if (pool_config->listen_addresses[0]) { inet_fd = create_inet_domain_socket(pool_config->listen_addresses, pool_config->port); } size = pool_config->num_init_children * pool_config->max_pool * sizeof(ConnectionInfo); con_info = pool_shared_memory_create(size); if (con_info == NULL) { pool_error("failed to allocate connection informations"); myexit(1); } memset(con_info, 0, size); size = pool_config->num_init_children * (sizeof(ProcessInfo)); pids = pool_shared_memory_create(size); if (pids == NULL) { pool_error("failed to allocate pids"); myexit(1); } memset(pids, 0, size); for (i = 0; i < pool_config->num_init_children; i++) { pids[i].connection_info = &con_info[i * pool_config->max_pool]; } /* create fail over/switch over event area */ Req_info = pool_shared_memory_create(sizeof(POOL_REQUEST_INFO)); if (Req_info == NULL) { pool_error("failed to allocate Req_info"); myexit(1); } /* initialize Req_info */ Req_info->kind = NODE_UP_REQUEST; memset(Req_info->node_id, -1, sizeof(int) * MAX_NUM_BACKENDS); Req_info->master_node_id = get_next_master_node(); Req_info->conn_counter = 0; InRecovery = pool_shared_memory_create(sizeof(int)); if (InRecovery == NULL) { pool_error("failed to allocate InRecovery"); myexit(1); } *InRecovery = 0; /* * We need to block signal here. Otherwise child might send some * signals, for example SIGUSR1(fail over). Children will inherit * signal blocking but they do unblock signals at the very beginning * of process. So this is harmless. */ POOL_SETMASK(&BlockSig); /* fork the children */ for (i=0;i<pool_config->num_init_children;i++) { pids[i].pid = fork_a_child(unix_fd, inet_fd, i); pids[i].start_time = time(NULL); } /* set up signal handlers */ pool_signal(SIGTERM, exit_handler); pool_signal(SIGINT, exit_handler); pool_signal(SIGQUIT, exit_handler); pool_signal(SIGCHLD, reap_handler); pool_signal(SIGUSR1, failover_handler); pool_signal(SIGUSR2, wakeup_handler); pool_signal(SIGHUP, reload_config_handler); /* create pipe for delivering event */ if (pipe(pipe_fds) < 0) { pool_error("failed to create pipe"); myexit(1); } pool_log("pgpool successfully started"); /* fork a child for PCP handling */ snprintf(pcp_un_addr.sun_path, sizeof(pcp_un_addr.sun_path), "%s/.s.PGSQL.%d", pool_config->pcp_socket_dir, pool_config->pcp_port); pcp_unix_fd = create_unix_domain_socket(pcp_un_addr); /* maybe change "*" to pool_config->pcp_listen_addresses */ pcp_inet_fd = create_inet_domain_socket("*", pool_config->pcp_port); pcp_pid = pcp_fork_a_child(pcp_unix_fd, pcp_inet_fd, pcp_conf_file); retrycnt = 0; /* reset health check retry counter */ sys_retrycnt = 0; /* reset SystemDB health check retry counter */ /* * This is the main loop */ for (;;) { CHECK_REQUEST; /* do we need health checking for PostgreSQL? */ if (pool_config->health_check_period > 0) { int sts; int sys_sts = 0; unsigned int sleep_time; if (retrycnt == 0) { pool_debug("starting health checking"); } else { pool_debug("retrying %d th health checking", retrycnt); } if (pool_config->health_check_timeout > 0) { /* * set health checker timeout. we want to detect * communication path failure much earlier before * TCP/IP stack detects it. */ pool_signal(SIGALRM, health_check_timer_handler); alarm(pool_config->health_check_timeout); } /* * do actual health check. trying to connect to the backend */ errno = 0; health_check_timer_expired = 0; POOL_SETMASK(&UnBlockSig); sts = health_check(); POOL_SETMASK(&BlockSig); if (pool_config->parallel_mode || pool_config->enable_query_cache) sys_sts = system_db_health_check(); if ((sts > 0 || sys_sts < 0) && (errno != EINTR || (errno == EINTR && health_check_timer_expired))) { if (sts > 0) { sts--; if (!pool_config->parallel_mode) { pool_log("set %d th backend down status", sts); Req_info->kind = NODE_DOWN_REQUEST; Req_info->node_id[0] = sts; failover(); /* need to distribute this info to children */ } else { retrycnt++; pool_signal(SIGALRM, SIG_IGN); /* Cancel timer */ if (retrycnt > NUM_BACKENDS) { /* retry count over */ pool_log("set %d th backend down status", sts); Req_info->kind = NODE_DOWN_REQUEST; Req_info->node_id[0] = sts; failover(); retrycnt = 0; } else { /* continue to retry */ sleep_time = pool_config->health_check_period/NUM_BACKENDS; pool_debug("retry sleep time: %d seconds", sleep_time); pool_sleep(sleep_time); continue; } } } if (sys_sts < 0) { sys_retrycnt++; pool_signal(SIGALRM, SIG_IGN); if (sys_retrycnt > NUM_BACKENDS) { pool_log("set SystemDB down status"); SYSDB_STATUS = CON_DOWN; sys_retrycnt = 0; } else if (sts == 0) /* goes to sleep only when SystemDB alone was down */ { sleep_time = pool_config->health_check_period/NUM_BACKENDS; pool_debug("retry sleep time: %d seconds", sleep_time); pool_sleep(sleep_time); continue; } } } if (pool_config->health_check_timeout > 0) { /* seems ok. cancel health check timer */ pool_signal(SIGALRM, SIG_IGN); } sleep_time = pool_config->health_check_period; pool_sleep(sleep_time); } else { for (;;) { int r; struct timeval t = {3, 0}; POOL_SETMASK(&UnBlockSig); r = pool_pause(&t); POOL_SETMASK(&BlockSig); if (r > 0) break; } } } pool_shmem_exit(0); }