"daemon start": initial implementation

configfile.c

@@ -371,17 +371,24 @@ _parse_config(t_configuration_options *options, ItemList *error_list, ItemList *
 	options->bdr_local_monitoring_only = false;
 	options->bdr_recovery_timeout = DEFAULT_BDR_RECOVERY_TIMEOUT;
 
-	/*-----------------
-	 * service settings
-	 *-----------------
+	/*-------------------------
+	 * service command settings
+	 *-------------------------
 	 */
 	memset(options->pg_ctl_options, 0, sizeof(options->pg_ctl_options));
-	memset(options->service_stop_command, 0, sizeof(options->service_stop_command));
 	memset(options->service_start_command, 0, sizeof(options->service_start_command));
+	memset(options->service_stop_command, 0, sizeof(options->service_stop_command));
 	memset(options->service_restart_command, 0, sizeof(options->service_restart_command));
 	memset(options->service_reload_command, 0, sizeof(options->service_reload_command));
 	memset(options->service_promote_command, 0, sizeof(options->service_promote_command));
 
+	/*---------------------------------
+	 * repmgrd service command settings
+	 *---------------------------------
+	 */
+	memset(options->repmgrd_service_start_command, 0, sizeof(options->repmgrd_service_start_command));
+	memset(options->repmgrd_service_stop_command, 0, sizeof(options->repmgrd_service_stop_command));
+
 	/*----------------------------
 	 * event notification settings
 	 *----------------------------
@@ -585,11 +592,11 @@ _parse_config(t_configuration_options *options, ItemList *error_list, ItemList *
 		else if (strcmp(name, "priority") == 0)
 			options->priority = repmgr_atoi(value, name, error_list, 0);
 		else if (strcmp(name, "location") == 0)
-			strncpy(options->location, value, MAXLEN);
+			strncpy(options->location, value, sizeof(options->location));
 		else if (strcmp(name, "promote_command") == 0)
-			strncpy(options->promote_command, value, MAXLEN);
+			strncpy(options->promote_command, value, sizeof(options->promote_command));
 		else if (strcmp(name, "follow_command") == 0)
-			strncpy(options->follow_command, value, MAXLEN);
+			strncpy(options->follow_command, value, sizeof(options->follow_command));
 		else if (strcmp(name, "reconnect_attempts") == 0)
 			options->reconnect_attempts = repmgr_atoi(value, name, error_list, 0);
 		else if (strcmp(name, "reconnect_interval") == 0)
@@ -621,41 +628,48 @@ _parse_config(t_configuration_options *options, ItemList *error_list, ItemList *
 
 		/* service settings */
 		else if (strcmp(name, "pg_ctl_options") == 0)
-			strncpy(options->pg_ctl_options, value, MAXLEN);
-		else if (strcmp(name, "service_stop_command") == 0)
-			strncpy(options->service_stop_command, value, MAXLEN);
+			strncpy(options->pg_ctl_options, value, sizeof(options->pg_ctl_options));
 		else if (strcmp(name, "service_start_command") == 0)
-			strncpy(options->service_start_command, value, MAXLEN);
+			strncpy(options->service_start_command, value, sizeof(options->service_start_command));
+		else if (strcmp(name, "service_stop_command") == 0)
+			strncpy(options->service_stop_command, value, sizeof(options->service_stop_command));
 		else if (strcmp(name, "service_restart_command") == 0)
-			strncpy(options->service_restart_command, value, MAXLEN);
+			strncpy(options->service_restart_command, value, sizeof(options->service_restart_command));
 		else if (strcmp(name, "service_reload_command") == 0)
-			strncpy(options->service_reload_command, value, MAXLEN);
+			strncpy(options->service_reload_command, value, sizeof(options->service_reload_command));
 		else if (strcmp(name, "service_promote_command") == 0)
-			strncpy(options->service_promote_command, value, MAXLEN);
+			strncpy(options->service_promote_command, value, sizeof(options->service_promote_command));
+
+		/* repmgrd service settings */
+		else if (strcmp(name, "repmgrd_service_start_command") == 0)
+			strncpy(options->repmgrd_service_start_command, value, sizeof(options->repmgrd_service_start_command));
+		else if (strcmp(name, "repmgrd_service_stop_command") == 0)
+			strncpy(options->repmgrd_service_stop_command, value, sizeof(options->repmgrd_service_stop_command));
+
 
 		/* event notification settings */
 		else if (strcmp(name, "event_notification_command") == 0)
-			strncpy(options->event_notification_command, value, MAXLEN);
+			strncpy(options->event_notification_command, value, sizeof(options->event_notification_command));
 		else if (strcmp(name, "event_notifications") == 0)
 		{
 			/* store unparsed value for comparison when reloading config */
-			strncpy(options->event_notifications_orig, value, MAXLEN);
+			strncpy(options->event_notifications_orig, value, sizeof(options->event_notifications_orig));
 			parse_event_notifications_list(options, value);
 		}
 
 		/* barman settings */
 		else if (strcmp(name, "barman_host") == 0)
-			strncpy(options->barman_host, value, MAXLEN);
+			strncpy(options->barman_host, value, sizeof(options->barman_host));
 		else if (strcmp(name, "barman_server") == 0)
-			strncpy(options->barman_server, value, MAXLEN);
+			strncpy(options->barman_server, value, sizeof(options->barman_server));
 		else if (strcmp(name, "barman_config") == 0)
-			strncpy(options->barman_config, value, MAXLEN);
+			strncpy(options->barman_config, value, sizeof(options->barman_config));
 
 		/* rsync/ssh settings */
 		else if (strcmp(name, "rsync_options") == 0)
-			strncpy(options->rsync_options, value, MAXLEN);
+			strncpy(options->rsync_options, value, sizeof(options->rsync_options));
 		else if (strcmp(name, "ssh_options") == 0)
-			strncpy(options->ssh_options, value, MAXLEN);
+			strncpy(options->ssh_options, value, sizeof(options->ssh_options));
 
 		/* undocumented settings for testing */
 		else if (strcmp(name, "promote_delay") == 0)

configfile.h

@@ -141,14 +141,18 @@ typedef struct
 
 	/* service settings */
 	char		pg_ctl_options[MAXLEN];
-	char		service_stop_command[MAXLEN];
-	char		service_start_command[MAXLEN];
-	char		service_restart_command[MAXLEN];
-	char		service_reload_command[MAXLEN];
-	char		service_promote_command[MAXLEN];
+	char		service_start_command[MAXPGPATH];
+	char		service_stop_command[MAXPGPATH];
+	char		service_restart_command[MAXPGPATH];
+	char		service_reload_command[MAXPGPATH];
+	char		service_promote_command[MAXPGPATH];
+
+	/* repmgrd service settings */
+	char		repmgrd_service_start_command[MAXPGPATH];
+	char		repmgrd_service_stop_command[MAXPGPATH];
 
 	/* event notification settings */
-	char		event_notification_command[MAXLEN];
+	char		event_notification_command[MAXPGPATH];
 	char		event_notifications_orig[MAXLEN];
 	EventNotificationList event_notifications;
 
@@ -205,6 +209,8 @@ typedef struct
 		false, DEFAULT_BDR_RECOVERY_TIMEOUT, \
 		/* service settings */ \
 		"", "", "", "", "", "", \
+		/* repmgrd service settings */ \
+		"", "",  \
 		/* event notification settings */ \
 		"", "", { NULL, NULL }, \
 		/* barman settings */ \

dbutils.c

@@ -1746,6 +1746,31 @@ repmgrd_get_local_node_id(PGconn *conn)
 }
 
 
+bool
+repmgrd_check_local_node_id(PGconn *conn)
+{
+	PGresult   *res = NULL;
+	bool		node_id_settable = true;
+	const char *sqlquery = "SELECT repmgr.get_local_node_id()";
+
+	res = PQexec(conn, sqlquery);
+
+	if (PQresultStatus(res) != PGRES_TUPLES_OK)
+	{
+		log_db_error(conn, sqlquery, _("repmgrd_get_local_node_id(): unable to execute query"));
+	}
+
+	if (PQgetisnull(res, 0, 0))
+	{
+		node_id_settable = false;
+	}
+
+	PQclear(res);
+
+	return node_id_settable;
+}
+
+
 /*
  * Function that checks if the primary is in exclusive backup mode.
  * We'll use this when executing an action can conflict with an exclusive

dbutils.h

@@ -431,6 +431,7 @@ TimeLineHistoryEntry *get_timeline_history(PGconn *repl_conn, TimeLineID tli);
 /* repmgrd shared memory functions */
 bool		repmgrd_set_local_node_id(PGconn *conn, int local_node_id);
 int			repmgrd_get_local_node_id(PGconn *conn);
+bool		repmgrd_check_local_node_id(PGconn *conn);
 BackupState	server_in_exclusive_backup_mode(PGconn *conn);
 void		repmgrd_set_pid(PGconn *conn, pid_t repmgrd_pid, const char *pidfile);
 pid_t		repmgrd_get_pid(PGconn *conn);

errcode.h

@@ -48,5 +48,6 @@
 #define ERR_REJOIN_FAIL 24
 #define ERR_NODE_STATUS 25
 #define ERR_REPMGRD_PAUSE 26
+#define ERR_REPMGRD_SERVICE 27
 
 #endif							/* _ERRCODE_H_ */

repmgr-action-daemon.c

@@ -384,6 +384,84 @@ fetch_node_records(PGconn *conn, NodeInfoList *node_list)
 }
 
 
+void
+do_daemon_start(void)
+{
+	PGconn	   *conn = NULL;
+	PQExpBufferData repmgrd_command;
+	PQExpBufferData output_buf;
+	bool success;
+
+	/*
+	 * if local connection available, check if repmgr.so is installed, and
+	 * whether repmgrd is running
+	 */
+	log_verbose(LOG_INFO, _("connecting to local node"));
+
+	if (strlen(config_file_options.conninfo))
+		conn = establish_db_connection(config_file_options.conninfo, false);
+	else
+		conn = establish_db_connection_by_params(&source_conninfo, false);
+
+	if (PQstatus(conn) != CONNECTION_OK)
+	{
+		log_warning(_("unable to connect to local node"));
+	}
+	else
+	{
+		check_shared_library(conn);
+
+		if (is_repmgrd_running(conn) == true)
+		{
+			log_error(_("repmgrd appears to be running already"));
+			PQfinish(conn);
+			exit(ERR_REPMGRD_SERVICE);
+		}
+	}
+
+	initPQExpBuffer(&repmgrd_command);
+
+	if (config_file_options.repmgrd_service_start_command[0] != '\0')
+	{
+		appendPQExpBufferStr(&repmgrd_command,
+							 config_file_options.repmgrd_service_start_command);
+	}
+	else
+	{
+		make_repmgrd_path(&repmgrd_command);
+	}
+
+	if (runtime_options.dry_run == true)
+	{
+		log_info(_("prerequisites for starting repmgrd met"));
+		log_detail("%s", repmgrd_command.data);
+		exit(SUCCESS);
+	}
+
+	log_debug("repmgrd start command: '%s'", repmgrd_command.data);
+
+	initPQExpBuffer(&output_buf);
+
+	success = local_command(repmgrd_command.data, &output_buf);
+	termPQExpBuffer(&repmgrd_command);
+
+	if (success == false)
+	{
+		log_error(_("unable to start repmgrd"));
+		if (output_buf.data[0] != '\0')
+			log_detail("%s", output_buf.data);
+		termPQExpBuffer(&output_buf);
+		exit(ERR_REPMGRD_SERVICE);
+	}
+
+	termPQExpBuffer(&output_buf);
+}
+
+
+void do_daemon_stop(void)
+{
+}
+
 void do_daemon_help(void)
 {
 	print_help_header();
@@ -392,6 +470,8 @@ void do_daemon_help(void)
 	printf(_("    %s [OPTIONS] daemon status\n"),  progname());
 	printf(_("    %s [OPTIONS] daemon pause\n"),   progname());
 	printf(_("    %s [OPTIONS] daemon unpause\n"), progname());
+	printf(_("    %s [OPTIONS] daemon start\n"),   progname());
+	printf(_("    %s [OPTIONS] daemon stop\n"),    progname());
 	puts("");
 
 	printf(_("DAEMON STATUS\n"));
@@ -416,6 +496,13 @@ void do_daemon_help(void)
 	printf(_("    --dry-run               check if nodes are reachable but don't unpause repmgrd\n"));
 	puts("");
 
+	printf(_("DAEMON START\n"));
+	puts("");
+	puts("XXX");
+
+	printf(_("DAEMON STOP\n"));
+	puts("");
+	puts("XXX");
 
 	puts("");
 }

repmgr-action-daemon.h

@@ -23,6 +23,8 @@
 extern void do_daemon_status(void);
 extern void do_daemon_pause(void);
 extern void do_daemon_unpause(void);
+extern void do_daemon_start(void);
+extern void do_daemon_stop(void);
 
 extern void do_daemon_help(void);
 #endif

repmgr-client-global.h

@@ -239,6 +239,8 @@ extern void get_superuser_connection(PGconn **conn, PGconn **superuser_conn, PGc
 extern bool remote_command(const char *host, const char *user, const char *command, PQExpBufferData *outputbuf);
 
 extern void make_remote_repmgr_path(PQExpBufferData *outputbuf, t_node_info *remote_node_record);
+extern void make_repmgrd_path(PQExpBufferData *output_buf);
+
 
 /* display functions */
 extern void print_help_header(void);
@@ -254,5 +256,7 @@ extern bool can_use_pg_rewind(PGconn *conn, const char *data_directory, PQExpBuf
 extern void drop_replication_slot_if_exists(PGconn *conn, int node_id, char *slot_name);
 
 extern bool check_node_can_attach(TimeLineID local_tli, XLogRecPtr local_xlogpos, PGconn *follow_target_conn, t_node_info *follow_target_node_record, bool is_rejoin);
+extern void check_shared_library(PGconn *conn);
+extern bool is_repmgrd_running(PGconn *conn);
 
 #endif							/* _REPMGR_CLIENT_GLOBAL_H_ */

repmgr-client.c

@@ -35,7 +35,8 @@
  * DAEMON STATUS
  * DAEMON PAUSE
  * DAEMON UNPAUSE
- *
+ * DAEMON START
+ * DAEMON STOP
  *
  * 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
@@ -935,6 +936,10 @@ main(int argc, char **argv)
 				action = DAEMON_PAUSE;
 			else if (strcasecmp(repmgr_action, "UNPAUSE") == 0)
 				action = DAEMON_UNPAUSE;
+			else if (strcasecmp(repmgr_action, "START") == 0)
+				action = DAEMON_START;
+			else if (strcasecmp(repmgr_action, "STOP") == 0)
+				action = DAEMON_STOP;
 		}
 		else
 		{
@@ -1344,6 +1349,11 @@ main(int argc, char **argv)
 		case DAEMON_UNPAUSE:
 			do_daemon_unpause();
 			break;
+		case DAEMON_START:
+			do_daemon_start();
+		case DAEMON_STOP:
+			do_daemon_stop();
+			break;
 
 		default:
 			/* An action will have been determined by this point  */
@@ -1834,6 +1844,8 @@ check_cli_parameters(const int action)
 			case NODE_SERVICE:
 			case DAEMON_PAUSE:
 			case DAEMON_UNPAUSE:
+			case DAEMON_START:
+			case DAEMON_STOP:
 				break;
 			default:
 				item_list_append_format(&cli_warnings,
@@ -1920,7 +1932,10 @@ action_name(const int action)
 			return "DAEMON PAUSE";
 		case DAEMON_UNPAUSE:
 			return "DAEMON UNPAUSE";
-
+		case DAEMON_START:
+			return "DAEMON START";
+		case DAEMON_STOP:
+			return "DAEMON STOP";
 	}
 
 	return "UNKNOWN ACTION";
@@ -2731,6 +2746,34 @@ make_remote_repmgr_path(PQExpBufferData *output_buf, t_node_info *remote_node_re
 }
 
 
+void
+make_repmgrd_path(PQExpBufferData *output_buf)
+{
+	if (config_file_options.repmgr_bindir[0] != '\0')
+	{
+		int			len = strlen(config_file_options.repmgr_bindir);
+
+		appendPQExpBufferStr(output_buf,
+							 config_file_options.repmgr_bindir);
+
+		/* Add trailing slash */
+		if (config_file_options.repmgr_bindir[len - 1] != '/')
+		{
+			appendPQExpBufferChar(output_buf, '/');
+		}
+	}
+	else if (pg_bindir[0] != '\0')
+	{
+		appendPQExpBufferStr(output_buf,
+							 pg_bindir);
+	}
+
+	appendPQExpBuffer(output_buf,
+					  "repmgrd -f %s ",
+					  config_file_path);
+}
+
+
 /* ======================== */
 /* server control functions */
 /* ======================== */
@@ -3358,3 +3401,47 @@ check_node_can_attach(TimeLineID local_tli, XLogRecPtr local_xlogpos, PGconn *fo
 	PQfinish(follow_target_repl_conn);
 	return success;
 }
+
+
+/*
+ * Simple check to see if "shared_preload_libraries" includes "repmgr".
+ * Parsing "shared_preload_libraries" is non-trivial, as it's potentially
+ * a comma-separated list, and worse may not be readable by the repmgr
+ * user.
+ *
+ * Instead, we check if a function which should return a value returns
+ * NULL; this indicates the shared library is not installed.
+ */
+void
+check_shared_library(PGconn *conn)
+{
+	bool ok = repmgrd_check_local_node_id(conn);
+
+	if (ok == true)
+		return;
+
+	log_error(_("repmgrd not configured for this node"));
+	log_hint(_("ensure \"shared_preload_libraries\" includes \"repmgr\" and restart PostgreSQL"));
+	PQfinish(conn);
+	exit(ERR_BAD_CONFIG);
+}
+
+
+bool
+is_repmgrd_running(PGconn *conn)
+{
+	pid_t		pid;
+	bool		is_running = false;
+
+	pid = repmgrd_get_pid(conn);
+
+	if (pid != UNKNOWN_PID)
+	{
+		if (kill(pid, 0) != -1)
+		{
+			is_running = true;
+		}
+	}
+
+	return is_running;
+}

repmgr-client.h

@@ -48,6 +48,8 @@
 #define DAEMON_STATUS		   22
 #define DAEMON_PAUSE		   23
 #define DAEMON_UNPAUSE		   24
+#define DAEMON_START 		   25
+#define DAEMON_STOP 		   26
 
 /* command line options without short versions */
 #define OPT_HELP						   1001

repmgr.conf.sample

@@ -322,6 +322,12 @@ ssh_options='-q -o ConnectTimeout=10'	# Options to append to "ssh"
 # repmgr provides options to override the default pg_ctl commands
 # used to stop, start, restart, reload and promote the PostgreSQL cluster
 #
+# These options are useful when PostgreSQL has been installed from a package
+# which provides OS-level service commands. In environments using an init system
+# such as systemd, which keeps track of the state of various services, it is
+# essential that the service commands are correctly configured and pg_ctl is
+# not executed directly.
+#
 # NOTE: These commands must be runnable on remote nodes as well for switchover
 # to function correctly.
 #
@@ -343,7 +349,7 @@ ssh_options='-q -o ConnectTimeout=10'	# Options to append to "ssh"
 #
 # Debian/Ubuntu users: use "sudo pg_ctlcluster" to execute service control commands.
 #
-# For more details, see: https://repmgr.org/docs/4.1/configuration-service-commands.html
+# For more details, see: https://repmgr.org/docs/current/configuration-service-commands.html
 
 #service_start_command = ''
 #service_stop_command = ''
@@ -355,6 +361,11 @@ ssh_options='-q -o ConnectTimeout=10'	# Options to append to "ssh"
 					# for "promote_command"; do not use "repmgr standby promote"
 					# (or a script which executes "repmgr standby promote") here.
 
+# Used by "repmgr daemon (start|stop)" to control repmgrd
+#
+#repmgrd_service_start_command = ''
+#repmgrd_service_stop_command = ''
+
 #------------------------------------------------------------------------------
 # Status check thresholds
 #------------------------------------------------------------------------------