From f1667a7e98ece9301e21f25cffad648c7b6d3364 Mon Sep 17 00:00:00 2001
From: Ian Barwick <barwick@gmail.com>
Date: Thu, 7 Feb 2019 16:53:48 +0900
Subject: [PATCH] repmgrd: don't consider nodes where repmgrd is not running

If, for whatever reason, repmgrd is not running on a node, but that
node qualifies as promotion candidate, failover will not take place
as that node will never promote itself.

We therefore discount nodes where repmgrd is running as promotion
candidates, which will ensure one node is always promoted.

There is a slight risk here that the node(s) where repmgrd is not running
are further ahead, leading to a timeline fork. It might be possible
to mitigate that by having the "election" leader perform the promote
(or follow) operation.
---
 HISTORY                         |  2 ++
 doc/appendix-release-notes.sgml | 21 +++++++++++++++++++++
 repmgrd-physical.c              | 17 +++++++++++++++++
 3 files changed, 40 insertions(+)

diff --git a/HISTORY b/HISTORY
index f74f1b7e..6fbe8b15 100644
--- a/HISTORY
+++ b/HISTORY
@@ -20,6 +20,8 @@
         repmgrd: check binary and extension major versions match; GitHub #515 (Ian)
         repmgrd: on a cascaded standby, don't fail over if "failover=manual";
           GitHub #531 (Ian)
+		repmgrd: don't consider nodes where repmgrd is not running as promotion
+		  candidates (Ian)
 
 4.2.1   2018-??-??
         repmgr: add sanity check for correct extension version (Ian)
diff --git a/doc/appendix-release-notes.sgml b/doc/appendix-release-notes.sgml
index 9e054392..91672880 100644
--- a/doc/appendix-release-notes.sgml
+++ b/doc/appendix-release-notes.sgml
@@ -124,6 +124,27 @@ REPMGRD_OPTS="--daemonize=false"</programlisting>
       </para>
     </sect2>
 
+    <sect2>
+      <title>repmgrd enhancements</title>
+      <para>
+        <itemizedlist>
+
+          <listitem>
+            <para>
+              <application>repmgrd</application> will no longer consider nodes where <application>repmgrd</application>
+			  is not running as promotion candidates.
+            </para>
+			<para>
+			  Previously, if <application>repmgrd</application> was not running on a node, but
+			  that node qualified as the promotion candidate, it would never be promoted due to
+			  the absence of a running <application>repmgrd</application>.
+			</para>
+          </listitem>
+
+		</itemizedlist>
+	  </para>
+	</sect2>
+
     <sect2>
       <title>Bug fixes</title>
       <para>
diff --git a/repmgrd-physical.c b/repmgrd-physical.c
index ebe0c839..e827c042 100644
--- a/repmgrd-physical.c
+++ b/repmgrd-physical.c
@@ -3257,6 +3257,7 @@ do_election(void)
 			log_debug("node %i is witness, not querying state", cell->node_info->node_id);
 			continue;
 		}
+
 		/* don't check 0-priority nodes */
 		if (cell->node_info->priority == 0)
 		{
@@ -3265,6 +3266,22 @@ do_election(void)
 			continue;
 		}
 
+
+		/*
+		 * check if repmgrd running - skip if not
+		 *
+		 * TODO: include pid query in replication info query?
+		 *
+		 * NOTE: from Pg12 we could execute "pg_promote()" from a running repmgrd;
+		 * here we'll need to find a way of ensuring only one repmgrd does this
+		 */
+		if (repmgrd_get_pid(cell->node_info->conn) == UNKNOWN_PID)
+		{
+			log_warning(_("repmgrd not running on node %i, skipping"),
+						cell->node_info->node_id);
+			continue;
+		}
+
 		if (get_replication_info(cell->node_info->conn, &sibling_replication_info) == false)
 		{
 			log_warning(_("unable to retrieve replication information for node %i, skipping"),