From 668b2c9b59e626c2efce0c031c0a0c866221ea94 Mon Sep 17 00:00:00 2001
From: Ian Barwick <ian@2ndquadrant.com>
Date: Fri, 22 Feb 2019 14:04:37 +0900
Subject: [PATCH] repmgrd: use PQping() as a first test of whether an upstream
 node is available

It's possible the upstream node may be temporarily not accepting connections
but is still running, so we only confirm that connections are not possible once
PQping() reports a negative result.

This feature has been adapted from repmgr4.
---
 dbutils.c | 16 ++++++++++++++++
 dbutils.h |  1 +
 repmgrd.c |  9 +++++----
 3 files changed, 22 insertions(+), 4 deletions(-)

diff --git a/dbutils.c b/dbutils.c
index 3f126eb6..fe9ff684 100644
--- a/dbutils.c
+++ b/dbutils.c
@@ -2137,3 +2137,19 @@ get_last_wal_receive_location(PGconn *conn)
 
 	return ptr;
 }
+
+
+bool
+is_server_available(const char *conninfo)
+{
+	PGPing		status = PQping(conninfo);
+
+	log_verbose(LOG_DEBUG, "is_server_available(): ping status for \"%s\" is %i\n", conninfo, (int)status);
+
+	if (status == PQPING_OK)
+		return true;
+
+	log_warning("is_server_available(): ping status for \"%s\" is %i\n", conninfo, (int)status);
+
+	return false;
+}
diff --git a/dbutils.h b/dbutils.h
index 9b16877c..8b226e31 100644
--- a/dbutils.h
+++ b/dbutils.h
@@ -146,5 +146,6 @@ int			get_data_checksum_version(const char *data_directory);
 /* backported from repmgr 4.x */
 XLogRecPtr	parse_lsn(const char *str);
 XLogRecPtr	get_last_wal_receive_location(PGconn *conn);
+bool		is_server_available(const char *conninfo);
 
 #endif
diff --git a/repmgrd.c b/repmgrd.c
index 676554c9..f9cdc496 100644
--- a/repmgrd.c
+++ b/repmgrd.c
@@ -879,9 +879,7 @@ standby_monitor(void)
 	 * local_options.reconnect_interval seconds
 	 */
 
-	check_connection(&upstream_conn, upstream_node_type, upstream_conninfo);
-
-	if (PQstatus(upstream_conn) != CONNECTION_OK)
+	if (!check_connection(&upstream_conn, upstream_node_type, upstream_conninfo))
 	{
 		int previous_master_node_id = master_options.node;
 
@@ -2137,8 +2135,11 @@ check_connection(PGconn **conn, const char *type, const char *conninfo)
 {
 	int			connection_retries;
 
+	if (conninfo != NULL && is_server_available(conninfo))
+		return true;
+
 	/*
-	 * Check if the node is still available if after
+	 * Check if the node is still available; if after
 	 * local_options.reconnect_attempts * local_options.reconnect_interval
 	 * seconds of retries we cannot reconnect return false
 	 */