repmgrd: on a cascaded standby, don't fail over if "failover=manual"

Addresses GitHub #531.

HISTORY

@@ -8,6 +8,8 @@
         repmgr: "standby switchover": improve handling of connection URIs when
           executing "node rejoin" on the demotion candidate; GitHub #525 (Ian)
         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)
 
 4.2.1   2018-??-??
         repmgr: add sanity check for correct extension version (Ian)

doc/appendix-release-notes.sgml

@@ -83,6 +83,13 @@
             </para>
           </listitem>
 
+          <listitem>
+            <para>
+              <application>repmgrd</application>:  on a cascaded standby, don't fail over if
+              <literal>failover=manual</literal>. GitHub #531.
+            </para>
+          </listitem>
+
         </itemizedlist>
       </para>
     </sect2>

doc/repmgrd-cascading-replication.sgml

@@ -17,6 +17,8 @@
   is promoted, a standby connected to another standby will not be affected
   and continue working as normal (even if the upstream standby it's connected
   to becomes the primary node). If however the node's direct upstream fails,
-  the "cascaded standby" will attempt to reconnect to that node's parent.
+  the "cascaded standby" will attempt to reconnect to that node's parent
+  (unless <varname>failover</varname> is set to <literal>manual</literal> in
+  <filename>repmgr.conf</filename>).
  </para>
 </chapter>

repmgrd-physical.c

@@ -950,7 +950,14 @@ monitor_streaming_standby(void)
 						}
 						else if (upstream_node_info.type == STANDBY)
 						{
+
 							failover_done = do_upstream_standby_failover();
+
+							if (failover_done == false)
+							{
+								monitoring_state = MS_DEGRADED;
+								INSTR_TIME_SET_CURRENT(degraded_monitoring_start);
+							}
 						}
 
 						/*
@@ -1225,7 +1232,15 @@ loop:
 						log_detail(_("waiting for upstream or another primary to reappear"));
 					}
 				}
-				else if (config_file_options.monitoring_history == true)
+
+				/*
+				 * Add update about monitoring updates.
+				 *
+				 * Note: with cascaded replication, it's possible we're still able to write
+				 * monitoring history to the primary even if the upstream is still reachable.
+				 */
+
+				if (PQstatus(primary_conn) == CONNECTION_OK && config_file_options.monitoring_history == true)
 				{
 					if (INSTR_TIME_IS_ZERO(last_monitoring_update))
 					{
@@ -2194,15 +2209,18 @@ update_monitoring_history(void)
 /*
  * do_upstream_standby_failover()
  *
- * Attach cascaded standby to primary
+ * Attach cascaded standby to another node, currently the primary.
  *
- * Currently we will try to attach to the cluster primary, as "repmgr
+ * Note that in contrast to a primary failover, where one of the downstrean
- * standby follow" doesn't support attaching to another node.
+ * standby nodes will become a primary, a cascaded standby failover (where the
+ * upstream standby has gone away) is "just" a case of attaching the standby to
+ * another node.
  *
- * If this becomes supported, it might be worth providing a selection
+ * Currently we will try to attach the node to the cluster primary.
- * of reconnection strategies as different behaviour might be desirable
+ *
- * in different situations;
+ * TODO: As of repmgr 4.3, "repmgr standby follow" supports attaching a standby to another
- * or maybe the option not to reconnect might be required?
+ * standby node. We need to provide a selection of reconnection strategies as different
+ * behaviour might be desirable in different situations.
  */
 
 static bool
@@ -2217,6 +2235,15 @@ do_upstream_standby_failover(void)
 
 	close_connection(&upstream_conn);
 
+	/*
+	 *
+	 */
+	if (config_file_options.failover == FAILOVER_MANUAL)
+	{
+		log_notice(_("this node is not configured for automatic failover"));
+		return false;
+	}
+
 	if (get_primary_node_record(local_conn, &primary_node_info) == false)
 	{
 		log_error(_("unable to retrieve primary node record"));