version push

fix: store the master connection user name on standby follow

README.rst

@@ -625,18 +625,18 @@ Now restore to the original configuration by stopping
 primary server, then bringing up "node2" as a standby with a valid
 ``recovery.conf`` file.
 
-Stop the "node2" server::
+Stop the "node2" server and type the following on "node1" server::
 
   repmgr -f /var/lib/pgsql/repmgr/repmgr.conf standby promote
 
-Now the original primary, "node1" is acting again as primary.
+Now the original primary, "node1", is acting again as primary.
 
-Start the "node2" server and type this on "node1"::
+Start the "node2" server and type this on "node2"::
 
   repmgr standby clone --force -h node2 -p 5432 -U postgres -R postgres --verbose
 
-Verify the roles have reversed by attempting to insert a record on "node"
+Verify the roles have reversed by attempting to insert a record on "node1"
-and on "node1".
+and on "node2".
 
 The servers are now again acting as primary on "node1" and standby on "node2".
 

config.c

@@ -191,7 +191,9 @@ trim (char *s)
 		++s1;
 
 	/* Copy finished string */
-	strcpy (s, s1);
+	memmove (s, s1, s2 - s1);
+	s[s2 - s1 + 1] = '\0';
+
 	return s;
 }
 

dbutils.c

@@ -18,6 +18,8 @@
  */
 
 #include <unistd.h>
+#include <time.h>
+#include <sys/time.h>
 
 #include "repmgr.h"
 #include "strutil.h"
@@ -426,34 +428,65 @@ getMasterConnection(PGconn *standby_conn, char *schema, char *cluster,
  * return 1 if Ok; 0 if any error ocurred; -1 if timeout reached
  */
 int
-wait_connection_availability(PGconn *conn, int timeout)
+wait_connection_availability(PGconn *conn, long long timeout)
 {
 	PGresult   *res;
+	fd_set      read_set;
+	int         sock = PQsocket(conn);
+	struct timeval tmout, before, after;
+	struct timezone tz;
 
-	while(timeout-- >= 0)
+	/* recalc to microseconds */
+	timeout *= 1000000;
+
+	while (timeout > 0)
 	{
 		if (PQconsumeInput(conn) == 0)
 		{
 			log_warning(_("wait_connection_availability: could not receive data from connection. %s\n"),
-			            PQerrorMessage(conn));
+						PQerrorMessage(conn));
 			return 0;
 		}
 
 		if (PQisBusy(conn) == 0)
 		{
-			res = PQgetResult(conn);
+			do {
-			if (res == NULL)
+				res = PQgetResult(conn);
-				break;
+				PQclear(res);
-			PQclear(res);
+			} while(res != NULL);
+
+			break;
 		}
-		sleep(1);
+
+
+		tmout.tv_sec = 0;
+		tmout.tv_usec = 250000;
+
+		FD_ZERO(&read_set);
+		FD_SET(sock, &read_set);
+
+		gettimeofday(&before, &tz);
+		if (select(sock, &read_set, NULL, NULL, &tmout) == -1)
+		{
+			log_warning(
+				_("wait_connection_availability: select() returned with error: %s"),
+				strerror(errno));
+			return -1;
+		}
+		gettimeofday(&after, &tz);
+
+		timeout -= (after.tv_sec * 1000000 + after.tv_usec) -
+			(before.tv_sec * 1000000 + before.tv_usec);
 	}
+
+
 	if (timeout >= 0)
+	{
 		return 1;
-	else {
-		log_warning(_("wait_connection_availability: timeout reached"));
-		return -1;
 	}
+
+	log_warning(_("wait_connection_availability: timeout reached"));
+	return -1;
 }
 
 

dbutils.h

@@ -39,6 +39,6 @@ const char	 *get_cluster_size(PGconn *conn);
 PGconn *getMasterConnection(PGconn *standby_conn, char *schema, char *cluster,
                             int *master_id, char *master_conninfo_out);
 
-int wait_connection_availability(PGconn *conn, int timeout);
+int wait_connection_availability(PGconn *conn, long long timeout);
 bool CancelQuery(PGconn *conn, int timeout);
 #endif

debian/repmgr.repmgrd.default

@@ -1,14 +1,18 @@
-#!/bin/sh
 # default settings for repmgrd. This file is source by /bin/sh from
 # /etc/init.d/repmgrd
 
-# Options for repmgrd
+# disable repmgrd by default so it won't get started upon installation
-REPMGRD_OPTS=""
+# valid values: yes/no
+REPMGRD_ENABLED=no
+
+# Options for repmgrd (required)
+#REPMGRD_OPTS="--config_file /path/to/repmgr.conf"
+
+# User to run repmgrd as
+#REPMGRD_USER=postgres
 
 # repmgrd binary
-REPMGR_BIN="/usr/bin/repmgr"
+#REPMGR_BIN=/usr/bin/repmgr
 
 # pid file
-REPMGR_PIDFILE="/var/run/repmgrd.pid"
+#REPMGR_PIDFILE=/var/run/repmgrd.pid
-
-

debian/repmgr.repmgrd.init

@@ -1,48 +1,101 @@
 #!/bin/sh
 ### BEGIN INIT INFO
-# Provides:          repmgrd
+# Provides:	     repmgrd
-# Required-Start:    $local_fs $remote_fs $network $syslog $postgresql
+# Required-Start:    $local_fs $remote_fs $network $syslog postgresql
-# Required-Stop:     $local_fs $remote_fs $network $syslog $postgresql
+# Required-Stop:     $local_fs $remote_fs $network $syslog postgresql
-# Should-Start:      $syslog $postgresql
+# Should-Start:	     $syslog postgresql
-# Should-Start:      $syslog $postgresql
 # Default-Start:     2 3 4 5
-# Default-Stop:      0 1 6
+# Default-Stop:	     0 1 6
 # Short-Description: Start/stop repmgrd
+# Description:       Enable repmgrd replication management and monitoring daemon for PostgreSQL
 ### END INIT INFO
 
 set -e
 
-if test -f /etc/default/repmgrd; then
+DESC="PostgreSQL replication management and monitoring daemon"
-    . /etc/default/repmgrd
+NAME=repmgrd
-fi
 
-if [ -z "$REPMGRD_BIN" ]; then
+REPMGRD_ENABLED=no
-    REPMGRD_BIN="/usr/bin/repmgrd"
+REPMGRD_OPTS=
-fi
+REPMGRD_USER=postgres
+REPMGRD_BIN=/usr/bin/repmgrd
+REPMGRD_PIDFILE=/var/run/repmgrd.pid
 
-if [ -z "$REPMGRD_PIDFILE" ]; then
+# Read configuration variable file if it is present
-    REPMGRD_PIDFILE="/var/run/repmgrd.pid"
+[ -r /etc/default/$NAME ] && . /etc/default/$NAME
-fi
 
 test -x $REPMGRD_BIN || exit 0
 
-case "$1" in
+case "$REPMGRD_ENABLED" in
-    start)
+    [Yy]*)
-        start-stop-daemon --start --quiet --make-pidfile --pidfile $REPMGRD_PIDFILE --exec $REPMGRD_BIN $REPMGRD_OPTS
+	break
-        ;;
+	;;
-
-    stop)
-        start-stop-daemon --stop --oknodo --quiet --pidfile $REPMGRD_PIDFILE
-        ;;
-
-    restart)
-        $0 stop && $0 start || exit 1
-        ;;
-
     *)
-        echo "Usage: $0 {start|stop|restart}"
+	exit 0
-        exit 1
+	;;
-        ;;
+esac
+
+# Define LSB log_* functions.
+. /lib/lsb/init-functions
+
+if [ -z "$REPMGRD_OPTS" ]
+then
+    log_warning_msg "Not starting $NAME, REPMGRD_OPTS not set in /etc/default/$NAME"
+    exit 0
+fi
+
+do_start()
+{
+	# Return
+	#   0 if daemon has been started
+	#   1 if daemon was already running
+	#   other if daemon could not be started or a failure occured
+	start-stop-daemon --start --quiet --chuid $REPMGRD_USER --make-pidfile --pidfile $REPMGRD_PIDFILE --exec $REPMGRD_BIN -- $REPMGRD_OPTS
+}
+
+do_stop()
+{
+	# Return
+	#   0 if daemon has been stopped
+	#   1 if daemon was already stopped
+	#   other if daemon could not be stopped or a failure occurred
+	start-stop-daemon --stop --quiet --retry=TERM/30/KILL/5 --pidfile $REPMGRD_PIDFILE --exec $REPMGRD_BIN
+}
+
+case "$1" in
+  start)
+	log_daemon_msg "Starting $DESC" "$NAME"
+	do_start
+	case "$?" in
+		0) log_end_msg 0 ;;
+		1) log_progress_msg "already started"
+		   log_end_msg 0 ;;
+		*) log_end_msg 1 ;;
+	esac
+
+	;;
+  stop)
+	log_daemon_msg "Stopping $DESC" "$NAME"
+	do_stop
+	case "$?" in
+		0) log_end_msg 0 ;;
+		1) log_progress_msg "already stopped"
+		   log_end_msg 0 ;;
+		*) log_end_msg 1 ;;
+	esac
+
+	;;
+  restart|force-reload)
+	$0 stop
+	$0 start
+	;;
+  status)
+	status_of_proc -p $PIDFILE $DAEMON $NAME && exit 0 || exit $?
+	;;
+  *)
+	echo "Usage: $SCRIPTNAME {start|stop|restart|force-reload|status}" >&2
+	exit 3
+	;;
 esac
 
 exit 0

log.c

@@ -58,6 +58,8 @@ void stderr_log_with_level(const char *level_name, int level, const char *fmt, .
 		vfprintf(stderr, fmt1, ap);
 
 		va_end(ap);
+
+		fflush(stderr);
 	}
 }
 
@@ -142,7 +144,14 @@ bool logger_init(t_configuration_options *opts, const char* ident, const char* l
 
 	if (*opts->logfile)
 	{
-		freopen(opts->logfile, "a", stderr);
+		FILE *fd;
+		fd = freopen(opts->logfile, "a", stderr);
+
+		if (fd == NULL)
+		{
+			fprintf(stderr, "error reopening stderr to '%s': %s",
+					opts->logfile, strerror(errno));
+		}
 	}
 
 	return true;

repmgr.c

@@ -575,13 +575,15 @@ do_master_register(void)
 							  repmgr_schema, options.node);
 			log_debug(_("master register: %s\n"), sqlquery);
 
-			if (!PQexec(conn, sqlquery))
+			res = PQexec(conn, sqlquery);
+			if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 			{
 				log_warning(_("Cannot delete node details, %s\n"),
 							PQerrorMessage(conn));
 				PQfinish(conn);
 				exit(ERR_BAD_CONFIG);
 			}
+			PQclear(res);
 		}
 
 		/* Ensure there isn't any other master already registered */
@@ -603,13 +605,15 @@ do_master_register(void)
 	                  options.conninfo, options.priority);
 	log_debug(_("master register: %s\n"), sqlquery);
 
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		log_warning(_("Cannot insert node details, %s\n"),
 		            PQerrorMessage(conn));
 		PQfinish(conn);
 		exit(ERR_BAD_CONFIG);
 	}
+	PQclear(res);
 
 	PQfinish(conn);
 	log_notice(_("Master node correctly registered for cluster %s with id %d (conninfo: %s)\n"),
@@ -735,7 +739,8 @@ do_standby_register(void)
 
 		log_debug(_("standby register: %s\n"), sqlquery);
 
-		if (!PQexec(master_conn, sqlquery))
+		res = PQexec(master_conn, sqlquery);
+		if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 		{
 			log_err(_("Cannot delete node details, %s\n"),
 			        PQerrorMessage(master_conn));
@@ -743,6 +748,7 @@ do_standby_register(void)
 			PQfinish(conn);
 			exit(ERR_BAD_CONFIG);
 		}
+		PQclear(res);
 	}
 
 	sqlquery_snprintf(sqlquery, "INSERT INTO %s.repl_nodes(id, cluster, name, conninfo, priority) "
@@ -1488,6 +1494,7 @@ do_standby_follow(void)
 	 */
 	strncpy(runtime_options.host, PQhost(master_conn), MAXLEN);
 	strncpy(runtime_options.masterport, PQport(master_conn), MAXLEN);
+	strncpy(runtime_options.username, PQuser(master_conn), MAXLEN);
 	PQfinish(master_conn);
 
 	log_info(_("%s Changing standby's master\n"),progname);
@@ -1695,7 +1702,8 @@ do_witness_create(void)
 	                  repmgr_schema, options.node, options.cluster_name, options.node_name, options.conninfo, options.priority);
 
 	log_debug(_("witness create: %s"), sqlquery);
-	if (!PQexec(masterconn, sqlquery))
+	res = PQexec(masterconn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		log_err(_("Cannot insert node details, %s\n"), PQerrorMessage(masterconn));
 		PQfinish(masterconn);
@@ -2068,16 +2076,49 @@ static bool
 create_schema(PGconn *conn)
 {
 	char 		sqlquery[QUERY_STR_LEN];
+	PGresult    *res;
 
 	sqlquery_snprintf(sqlquery, "CREATE SCHEMA %s", repmgr_schema);
 	log_debug(_("master register: %s\n"), sqlquery);
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		log_err(_("Cannot create the schema %s: %s\n"),
 		        repmgr_schema, PQerrorMessage(conn));
 		PQfinish(conn);
 		exit(ERR_BAD_CONFIG);
 	}
+	PQclear(res);
+
+	/* to avoid confusion of the time_lag field and provide a consistent UI we
+	 * use these functions for providing the latest update timestamp */
+	sqlquery_snprintf(sqlquery,
+					  "CREATE FUNCTION %s.repmgr_update_last_updated() RETURNS TIMESTAMP WITH TIME ZONE "
+					  "AS '$libdir/repmgr_funcs', 'repmgr_update_last_updated' "
+					  " LANGUAGE C STRICT", repmgr_schema);
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
+	{
+		fprintf(stderr, "Cannot create the function repmgr_update_last_updated: %s\n",
+		        PQerrorMessage(conn));
+		return false;
+	}
+	PQclear(res);
+
+
+	sqlquery_snprintf(sqlquery,
+					  "CREATE FUNCTION %s.repmgr_get_last_updated() RETURNS TIMESTAMP WITH TIME ZONE "
+					  "AS '$libdir/repmgr_funcs', 'repmgr_get_last_updated' "
+					  "LANGUAGE C STRICT", repmgr_schema);
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
+	{
+		fprintf(stderr, "Cannot create the function repmgr_get_last_updated: %s\n",
+		        PQerrorMessage(conn));
+		return false;
+	}
+	PQclear(res);
+
 
 	/* ... the tables */
 	sqlquery_snprintf(sqlquery, "CREATE TABLE %s.repl_nodes (        "
@@ -2088,86 +2129,101 @@ create_schema(PGconn *conn)
 	                  "  priority  integer not null,    "
 	                  "  witness   boolean not null default false)", repmgr_schema);
 	log_debug(_("master register: %s\n"), sqlquery);
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		log_err(_("Cannot create the table %s.repl_nodes: %s\n"),
 		        repmgr_schema, PQerrorMessage(conn));
 		PQfinish(conn);
 		exit(ERR_BAD_CONFIG);
 	}
+	PQclear(res);
 
 	sqlquery_snprintf(sqlquery, "CREATE TABLE %s.repl_monitor ( "
 	                  "  primary_node                   INTEGER NOT NULL, "
 	                  "  standby_node                   INTEGER NOT NULL, "
 	                  "  last_monitor_time              TIMESTAMP WITH TIME ZONE NOT NULL, "
+	                  "  last_apply_time                TIMESTAMP WITH TIME ZONE, "
 	                  "  last_wal_primary_location      TEXT NOT NULL,   "
 	                  "  last_wal_standby_location      TEXT,  "
 	                  "  replication_lag                BIGINT NOT NULL, "
 	                  "  apply_lag                      BIGINT NOT NULL) ", repmgr_schema);
 	log_debug(_("master register: %s\n"), sqlquery);
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		log_err(_("Cannot create the table %s.repl_monitor: %s\n"),
 		        repmgr_schema, PQerrorMessage(conn));
 		PQfinish(conn);
 		exit(ERR_BAD_CONFIG);
 	}
+	PQclear(res);
 
 	/* a view */
 	sqlquery_snprintf(sqlquery, "CREATE VIEW %s.repl_status AS "
 	                  " SELECT primary_node, standby_node, name AS standby_name, last_monitor_time, "
 	                  "        last_wal_primary_location, last_wal_standby_location, "
 	                  "        pg_size_pretty(replication_lag) replication_lag, "
+	                  "        age(now(), last_apply_time) AS replication_time_lag, "
 	                  "        pg_size_pretty(apply_lag) apply_lag, "
-	                  "        age(now(), last_monitor_time) AS time_lag "
+	                  "        age(now(), CASE WHEN pg_is_in_recovery() THEN %s.repmgr_get_last_updated() ELSE last_monitor_time END) AS communication_time_lag "
 	                  "   FROM %s.repl_monitor JOIN %s.repl_nodes ON standby_node = id "
 	                  "  WHERE (standby_node, last_monitor_time) IN (SELECT standby_node, MAX(last_monitor_time) "
 	                  "                                                FROM %s.repl_monitor GROUP BY 1)",
-	                  repmgr_schema, repmgr_schema, repmgr_schema, repmgr_schema);
+	                  repmgr_schema, repmgr_schema, repmgr_schema, repmgr_schema, repmgr_schema);
 	log_debug(_("master register: %s\n"), sqlquery);
-	if (!PQexec(conn, sqlquery))
+
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		log_err(_("Cannot create the view %s.repl_status: %s\n"),
 		        repmgr_schema, PQerrorMessage(conn));
 		PQfinish(conn);
 		exit(ERR_BAD_CONFIG);
 	}
+	PQclear(res);
 
 	/* an index to improve performance of the view */
 	sqlquery_snprintf(sqlquery, "CREATE INDEX idx_repl_status_sort "
 	                  "    ON %s.repl_monitor (last_monitor_time, standby_node) ",
 	                  repmgr_schema);
 	log_debug(_("master register: %s\n"), sqlquery);
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
-		log_err(_("Cannot indexing table %s.repl_monitor: %s\n"),
+		log_err(_("Can't index table %s.repl_monitor: %s\n"),
 		        repmgr_schema, PQerrorMessage(conn));
 		PQfinish(conn);
 		exit(ERR_BAD_CONFIG);
 	}
+	PQclear(res);
 
 	/* XXX Here we MUST try to load the repmgr_function.sql not hardcode it here */
 	sqlquery_snprintf(sqlquery,
 	                  "CREATE OR REPLACE FUNCTION %s.repmgr_update_standby_location(text) RETURNS boolean "
 	                  "AS '$libdir/repmgr_funcs', 'repmgr_update_standby_location' "
 	                  "LANGUAGE C STRICT ", repmgr_schema);
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		fprintf(stderr, "Cannot create the function repmgr_update_standby_location: %s\n",
 		        PQerrorMessage(conn));
 		return false;
 	}
+	PQclear(res);
 
 	sqlquery_snprintf(sqlquery,
 	                  "CREATE OR REPLACE FUNCTION %s.repmgr_get_last_standby_location() RETURNS text "
 	                  "AS '$libdir/repmgr_funcs', 'repmgr_get_last_standby_location' "
 	                  "LANGUAGE C STRICT ", repmgr_schema);
-	if (!PQexec(conn, sqlquery))
+	res = PQexec(conn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		fprintf(stderr, "Cannot create the function repmgr_get_last_standby_location: %s\n",
 		        PQerrorMessage(conn));
 		return false;
 	}
+	PQclear(res);
 
 	return true;
 }
@@ -2182,7 +2238,8 @@ copy_configuration(PGconn *masterconn, PGconn *witnessconn)
 
 	sqlquery_snprintf(sqlquery, "TRUNCATE TABLE %s.repl_nodes", repmgr_schema);
 	log_debug("copy_configuration: %s\n", sqlquery);
-	if (!PQexec(witnessconn, sqlquery))
+	res = PQexec(witnessconn, sqlquery);
+	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
 		fprintf(stderr, "Cannot clean node details in the witness, %s\n",
 		        PQerrorMessage(witnessconn));
@@ -2208,7 +2265,8 @@ copy_configuration(PGconn *masterconn, PGconn *witnessconn)
 		                  atoi(PQgetvalue(res, i, 3)),
 		                  PQgetvalue(res, i, 4));
 
-		if (!PQexec(witnessconn, sqlquery))
+		res = PQexec(witnessconn, sqlquery);
+		if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 		{
 			fprintf(stderr, "Cannot copy configuration to witness, %s\n",
 			        PQerrorMessage(witnessconn));

repmgr.conf.sample

@@ -22,7 +22,7 @@ reconnect_attempts=6
 reconnect_interval=10
 
 # Autofailover options
-failover=automatic
+failover=manual
 priority=-1
 promote_command='repmgr standby promote -f /path/to/repmgr.conf'
 follow_command='repmgr standby follow -f /path/to/repmgr.conf -W'

repmgrd.c

@@ -144,17 +144,24 @@ static void terminate(int retval);
 static void setup_event_handlers(void);
 #endif
 
-static void do_daemonize();
+static void do_daemonize(void);
 static void check_and_create_pid_file(const char *pid_file);
 
-#define CloseConnections()	\
+static void
-	if (PQisBusy(primaryConn) == 1) \
+CloseConnections() {
-		(void) CancelQuery(primaryConn, local_options.master_response_timeout); \
+	if (primaryConn != NULL && PQisBusy(primaryConn) == 1)
-	if (myLocalConn != NULL) \
+		CancelQuery(primaryConn, local_options.master_response_timeout);
-		PQfinish(myLocalConn);	\
+
-	if (primaryConn != NULL && primaryConn != myLocalConn) \
+	if (myLocalConn != NULL)
+		PQfinish(myLocalConn);
+
+	if (primaryConn != NULL && primaryConn != myLocalConn)
 		PQfinish(primaryConn);
 
+	primaryConn = NULL;
+	myLocalConn = NULL;
+}
+
 
 int
 main(int argc, char **argv)
@@ -172,6 +179,7 @@ main(int argc, char **argv)
 	int			optindex;
 	int			c, ret;
 	bool		daemonize = false;
+	FILE       *fd;
 
 	char standby_version[MAXVERSIONSTR], *ret_ver;
 
@@ -241,8 +249,19 @@ main(int argc, char **argv)
 		terminate(ERR_BAD_CONFIG);
 	}
 
-	freopen("/dev/null", "r", stdin);
+	fd = freopen("/dev/null", "r", stdin);
-	freopen("/dev/null", "w", stdout);
+	if (fd == NULL)
+	{
+		fprintf(stderr, "error reopening stdin to '/dev/null': %s",
+				strerror(errno));
+	}
+
+	fd = freopen("/dev/null", "w", stdout);
+	if (fd == NULL)
+	{
+		fprintf(stderr, "error reopening stdout to '/dev/null': %s",
+				strerror(errno));
+	}
 
 	logger_init(&local_options, progname, local_options.loglevel, local_options.logfacility);
 	if (verbose)
@@ -250,7 +269,13 @@ main(int argc, char **argv)
 
 	if (log_type == REPMGR_SYSLOG)
 	{
-		freopen("/dev/null", "w", stderr);
+		fd = freopen("/dev/null", "w", stderr);
+
+		if (fd == NULL)
+		{
+			fprintf(stderr, "error reopening stderr to '/dev/null': %s",
+					strerror(errno));
+		}
 	}
 
 	snprintf(repmgr_schema, MAXLEN, "%s%s", DEFAULT_REPMGR_SCHEMA_PREFIX, local_options.cluster_name);
@@ -358,7 +383,14 @@ main(int argc, char **argv)
 
 							if (*local_options.logfile)
 							{
-								freopen(local_options.logfile, "a", stderr);
+								FILE *fd;
+								fd = freopen(local_options.logfile, "a", stderr);
+								if (fd == NULL)
+								{
+									fprintf(stderr, "error reopening stderr to '%s': %s",
+											local_options.logfile, strerror(errno));
+								}
+
 							}
 
 							update_registration();
@@ -431,10 +463,6 @@ main(int argc, char **argv)
 
 	} while (true);
 
-	/* Prevent a double-free */
-	if (primaryConn == myLocalConn)
-		myLocalConn = NULL;
-
 	/* close the connection to the database and cleanup */
 	CloseConnections();
 
@@ -530,6 +558,7 @@ StandbyMonitor(void)
 	char last_wal_primary_location[MAXLEN];
 	char last_wal_standby_received[MAXLEN];
 	char last_wal_standby_applied[MAXLEN];
+	char last_wal_standby_applied_timestamp[MAXLEN];
 
 	unsigned long long int lsn_primary;
 	unsigned long long int lsn_standby_received;
@@ -546,11 +575,15 @@ StandbyMonitor(void)
 
 	if (!CheckConnection(myLocalConn, "standby"))
 	{
+		log_err("Failed to connect to local node, exiting!\n");
 		terminate(1);
 	}
 
 	if (PQstatus(primaryConn) != CONNECTION_OK)
 	{
+		PQfinish(primaryConn);
+		primaryConn = NULL;
+
 		if (local_options.failover == MANUAL_FAILOVER)
 		{
 			log_err(_("We couldn't reconnect to master. Now checking if another node has been promoted.\n"));
@@ -637,7 +670,7 @@ StandbyMonitor(void)
 	sqlquery_snprintf(
 	    sqlquery,
 	    "SELECT CURRENT_TIMESTAMP, pg_last_xlog_receive_location(), "
-	    "pg_last_xlog_replay_location()");
+	    "pg_last_xlog_replay_location(), pg_last_xact_replay_timestamp()");
 
 	res = PQexec(myLocalConn, sqlquery);
 	if (PQresultStatus(res) != PGRES_TUPLES_OK)
@@ -651,6 +684,7 @@ StandbyMonitor(void)
 	strncpy(monitor_standby_timestamp, PQgetvalue(res, 0, 0), MAXLEN);
 	strncpy(last_wal_standby_received , PQgetvalue(res, 0, 1), MAXLEN);
 	strncpy(last_wal_standby_applied , PQgetvalue(res, 0, 2), MAXLEN);
+	strncpy(last_wal_standby_applied_timestamp, PQgetvalue(res, 0, 3), MAXLEN);
 	PQclear(res);
 
 	/* Get primary xlog info */
@@ -678,9 +712,10 @@ StandbyMonitor(void)
 	sqlquery_snprintf(sqlquery,
 	                  "INSERT INTO %s.repl_monitor "
 	                  "VALUES(%d, %d, '%s'::timestamp with time zone, "
-	                  " '%s', '%s', "
+	                  " '%s'::timestamp with time zone, '%s', '%s', "
 	                  " %lld, %lld)", repmgr_schema,
 	                  primary_options.node, local_options.node, monitor_standby_timestamp,
+	                  last_wal_standby_applied_timestamp,
 	                  last_wal_primary_location,
 	                  last_wal_standby_received,
 	                  (lsn_primary - lsn_standby_received),
@@ -770,7 +805,12 @@ do_failover(void)
 
 		/* if we can't see the node just skip it */
 		if (PQstatus(nodeConn) != CONNECTION_OK)
+		{
+			if (nodeConn != NULL)
+				PQfinish(nodeConn);
+
 			continue;
+		}
 
 		visible_nodes++;
 		nodes[i].is_visible = true;
@@ -953,6 +993,7 @@ do_failover(void)
 
 	/* Close the connection to this server */
 	PQfinish(myLocalConn);
+	myLocalConn = NULL;
 
 	/*
 	 * determine which one is the best candidate to promote to primary
@@ -1269,6 +1310,8 @@ terminate(int retval)
 		unlink(pid_file);
 	}
 
+	log_info("Terminating...\n");
+
 	exit(retval);
 }
 
@@ -1323,6 +1366,7 @@ do_daemonize()
 {
 	char *ptr, path[MAXLEN];
 	pid_t pid = fork();
+	int ret;
 
 	switch (pid)
 	{
@@ -1371,7 +1415,12 @@ do_daemonize()
 			*path = '/';
 		}
 
-		chdir(path);
+		ret = chdir(path);
+		if (ret != 0)
+		{
+			log_err("Error changing directory to '%s': %s", path,
+					strerror(errno));
+		}
 
 		break;
 
@@ -1387,6 +1436,7 @@ check_and_create_pid_file(const char *pid_file)
 	FILE *fd;
 	char buff[MAXLEN];
 	pid_t pid;
+	size_t nread;
 
 	if (stat(pid_file, &st) != -1)
 	{
@@ -1400,7 +1450,14 @@ check_and_create_pid_file(const char *pid_file)
 			exit(ERR_BAD_CONFIG);
 		}
 
-		fread(buff, MAXLEN - 1, 1, fd);
+		nread = fread(buff, MAXLEN - 1, 1, fd);
+
+		if (nread == 0 && ferror(fd))
+		{
+			log_err("Error reading PID file '%s', giving up...\n", pid_file);
+			exit(ERR_BAD_CONFIG);
+		}
+
 		fclose(fd);
 
 		pid = atoi(buff);

sql/repmgr_funcs.c

@@ -15,6 +15,7 @@
 #include "storage/shmem.h"
 #include "storage/spin.h"
 #include "utils/builtins.h"
+#include "utils/timestamp.h"
 
 /* same definition as the one in xlog_internal.h */
 #define MAXFNAMELEN 	64
@@ -28,6 +29,7 @@ typedef struct repmgrSharedState
 {
     LWLockId    lock;           		/* protects search/modification */
     char		location[MAXFNAMELEN];	/* last known xlog location */
+	TimestampTz last_updated;
 } repmgrSharedState;
 
 /* Links to shared memory state */
@@ -49,6 +51,12 @@ Datum repmgr_get_last_standby_location(PG_FUNCTION_ARGS);
 PG_FUNCTION_INFO_V1(repmgr_update_standby_location);
 PG_FUNCTION_INFO_V1(repmgr_get_last_standby_location);
 
+Datum repmgr_update_last_updated(PG_FUNCTION_ARGS);
+Datum repmgr_get_last_updated(PG_FUNCTION_ARGS);
+
+PG_FUNCTION_INFO_V1(repmgr_update_last_updated);
+PG_FUNCTION_INFO_V1(repmgr_get_last_updated);
+
 
 /*
  * Module load callback
@@ -187,3 +195,38 @@ repmgr_update_standby_location(PG_FUNCTION_ARGS)
 
 	PG_RETURN_BOOL(repmgr_set_standby_location(locationstr));
 }
+
+/* update and return last updated with current timestamp */
+Datum
+repmgr_update_last_updated(PG_FUNCTION_ARGS)
+{
+	TimestampTz last_updated = GetCurrentTimestamp();
+
+	/* Safety check... */
+	if (!shared_state)
+		PG_RETURN_NULL();
+
+	LWLockAcquire(shared_state->lock, LW_SHARED);
+	shared_state->last_updated = last_updated;
+	LWLockRelease(shared_state->lock);
+
+	PG_RETURN_TIMESTAMPTZ(last_updated);
+}
+
+
+/* get last updated timestamp */
+Datum
+repmgr_get_last_updated(PG_FUNCTION_ARGS)
+{
+	TimestampTz last_updated;
+
+    /* Safety check... */
+    if (!shared_state)
+        PG_RETURN_NULL();
+
+	LWLockAcquire(shared_state->lock, LW_EXCLUSIVE);
+	last_updated = shared_state->last_updated;
+	LWLockRelease(shared_state->lock);
+
+	PG_RETURN_TIMESTAMPTZ(last_updated);
+}

sql/repmgr_funcs.sql.in

@@ -1,6 +1,6 @@
 /*
  * repmgr_function.sql
- * Copyright (c) 2ndQuadrant, 2010
+ * Copyright (c) 2ndQuadrant, 2010-2014
  *
  */
 
@@ -13,3 +13,11 @@ LANGUAGE C STRICT;
 CREATE FUNCTION repmgr_get_last_standby_location() RETURNS text
 AS 'MODULE_PATHNAME', 'repmgr_get_last_standby_location'
 LANGUAGE C STRICT;
+
+CREATE FUNCTION repmgr_update_last_updated() RETURNS TIMESTAMP WITH TIME ZONE
+AS 'MODULE_PATHNAME', 'repmgr_update_last_updated'
+LANGUAGE C STRICT;
+
+CREATE FUNCTION repmgr_get_last_updated() RETURNS TIMESTAMP WITH TIME ZONE
+AS 'MODULE_PATHNAME', 'repmgr_get_last_updated'
+LANGUAGE C STRICT;

sql/uninstall_repmgr_funcs.sql

@@ -1,2 +1,11 @@
+/*
+ * uninstall_repmgr_funcs.sql
+ * Copyright (c) 2ndQuadrant, 2010-2014
+ *
+ */
+
 DROP FUNCTION repmgr_update_standby_location(text);
 DROP FUNCTION repmgr_get_last_standby_location();
+
+DROP FUNCTION repmgr_update_last_updated();
+DROP FUNCTION repmgr_get_last_updated();

version.h

@@ -1,5 +1,5 @@
 #ifndef _VERSION_H_
 #define _VERSION_H_
 
-#define REPMGR_VERSION "2.0RC1"
+#define REPMGR_VERSION "2.0"
 #endif