Have do_standby_follow_internal() not abort on error

Pass the error code back to the caller instead, mainly so
"repmgr node rejoin" can better report errors.

HISTORY

@@ -1,3 +1,7 @@
+4.0.3   2018-02-
+        repmgr: improve switchover handling when "pg_ctl" used to control the
+          server and logging output is not explicitly redirected (Ian)
+
 4.0.2   2018-01-
         repmgr: add missing -W option to getopt_long() invocation; GitHub #350 (Ian)
         repmgr: automatically create slot name if missing; GitHub #343 (Ian)

errcode.h

@@ -43,5 +43,6 @@
 #define ERR_BARMAN 19
 #define ERR_REGISTRATION_SYNC 20
 #define ERR_OUT_OF_MEMORY 21
+#define ERR_REJOIN_FAIL 22
 
 #endif							/* _ERRCODE_H_ */

repmgr-action-node.c

@@ -1590,6 +1590,7 @@ do_node_rejoin(void)
 
 	bool		success = true;
 	int			server_version_num = UNKNOWN_SERVER_VERSION_NUM;
+	int			follow_error_code = SUCCESS;
 
 	/* check node is not actually running */
 
@@ -1859,7 +1860,31 @@ do_node_rejoin(void)
 
 	success = do_standby_follow_internal(upstream_conn,
 										 &primary_node_record,
-										 &follow_output);
+										 &follow_output,
+										 &follow_error_code);
+
+	if (success == false)
+	{
+		log_notice(_("NODE REJOIN failed"));
+		log_detail("%s", follow_output.data);
+
+		create_event_notification(upstream_conn,
+								  &config_file_options,
+								  config_file_options.node_id,
+								  "node_rejoin",
+								  success,
+								  follow_output.data);
+
+		PQfinish(upstream_conn);
+
+		termPQExpBuffer(&follow_output);
+		exit(follow_error_code);
+	}
+
+	/*
+	 * XXX add checks that node actually started and connected to primary,
+	 * if not exit with ERR_REJOIN_FAIL
+	 */
 
 	create_event_notification(upstream_conn,
 							  &config_file_options,
@@ -1870,19 +1895,12 @@ do_node_rejoin(void)
 
 	PQfinish(upstream_conn);
 
-	if (success == false)
-	{
-		log_notice(_("NODE REJOIN failed"));
-		log_detail("%s", follow_output.data);
-
-		termPQExpBuffer(&follow_output);
-		exit(ERR_DB_QUERY);
-	}
-
 	log_notice(_("NODE REJOIN successful"));
 	log_detail("%s", follow_output.data);
 
 	termPQExpBuffer(&follow_output);
+
+	return;
 }
 
 

repmgr-action-standby.c

@@ -1590,6 +1590,7 @@ do_standby_follow(void)
 
 	PQExpBufferData follow_output;
 	bool		success = false;
+	int			follow_error_code = SUCCESS;
 
 	uint64		local_system_identifier = UNKNOWN_SYSTEM_IDENTIFIER;
 	t_conninfo_param_list repl_conninfo;
@@ -1703,7 +1704,8 @@ do_standby_follow(void)
 
 	success = do_standby_follow_internal(primary_conn,
 										 &primary_node_record,
-										 &follow_output);
+										 &follow_output,
+										 &follow_error_code);
 
 	create_event_notification(primary_conn,
 							  &config_file_options,
@@ -1720,7 +1722,7 @@ do_standby_follow(void)
 		log_detail("%s", follow_output.data);
 
 		termPQExpBuffer(&follow_output);
-		exit(ERR_DB_QUERY);
+		exit(follow_error_code);
 	}
 
 	log_notice(_("STANDBY FOLLOW successful"));
@@ -1740,7 +1742,7 @@ do_standby_follow(void)
  * this function.
  */
 bool
-do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_record, PQExpBufferData *output)
+do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_record, PQExpBufferData *output, int *error_code)
 {
 	t_node_info local_node_record = T_NODE_INFO_INITIALIZER;
 	int			original_upstream_node_id = UNKNOWN_NODE_ID;
@@ -1763,8 +1765,9 @@ do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_recor
 	{
 		log_error(_("unable to retrieve record for node %i"),
 				  config_file_options.node_id);
-		PQfinish(primary_conn);
+
-		exit(ERR_BAD_CONFIG);
+		*error_code = ERR_BAD_CONFIG;
+		return false;
 	}
 
 	/*
@@ -1893,8 +1896,9 @@ do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_recor
 
 	if (!create_recovery_file(&local_node_record, &recovery_conninfo, config_file_options.data_directory))
 	{
-		PQfinish(primary_conn);
+		/* XXX ERR_RECOVERY_FILE ??? */
-		exit(ERR_BAD_CONFIG);
+		*error_code = ERR_BAD_CONFIG;
+		return false;
 	}
 
 	/*
@@ -1931,8 +1935,9 @@ do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_recor
 		if (success == false)
 		{
 			log_error(_("unable to %s server"), action);
-			PQfinish(primary_conn);
+
-			exit(ERR_NO_RESTART);
+			*error_code = ERR_NO_RESTART;
+			return false;
 		}
 	}
 
@@ -2059,6 +2064,7 @@ do_standby_switchover(void)
 				  config_file_options.node_id);
 
 		PQfinish(local_conn);
+
 		exit(ERR_DB_QUERY);
 	}
 

repmgr-action-standby.h

@@ -28,7 +28,7 @@ extern void do_standby_switchover(void);
 
 extern void do_standby_help(void);
 
-extern bool do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_record, PQExpBufferData *output);
+extern bool do_standby_follow_internal(PGconn *primary_conn, t_node_info *primary_node_record, PQExpBufferData *output, int *error_code);