repmgr: add "--node-id" option to "cluster cleanup"

Implements GitHub #493.

HISTORY

@@ -1,6 +1,7 @@
 4.2.0   2018-??-??
         repmgr: add parameter "shutdown_check_timeout" for use by "standby switchover";
           GitHub #504 (Ian)
+        repmgr: add "--node-id" option to "repmgr cluster cleanup"; GitHub #493 (Ian)
 
 4.1.1   2018-09-05
         logging: explicitly log the text of failed queries as ERRORs to

dbutils.c

@@ -4167,7 +4167,7 @@ add_monitoring_record(PGconn *primary_conn,
 
 
 int
-get_number_of_monitoring_records_to_delete(PGconn *primary_conn, int keep_history)
+get_number_of_monitoring_records_to_delete(PGconn *primary_conn, int keep_history, int node_id)
 {
 	PQExpBufferData query;
 	int				record_count = -1;
@@ -4181,8 +4181,15 @@ get_number_of_monitoring_records_to_delete(PGconn *primary_conn, int keep_histor
 					  " WHERE pg_catalog.age(pg_catalog.now(), last_monitor_time) >= '%d days'::interval",
 					  keep_history);
 
-	res = PQexec(primary_conn, query.data);
+	if (node_id != UNKNOWN_NODE_ID)
+	{
+		appendPQExpBuffer(&query,
+						  "  AND standby_node_id = %i", node_id);
+	}
 
+	log_verbose(LOG_DEBUG, "get_number_of_monitoring_records_to_delete():\n  %s", query.data);
+
+	res = PQexec(primary_conn, query.data);
 
 	if (PQresultStatus(res) != PGRES_TUPLES_OK)
 	{
@@ -4202,7 +4209,7 @@ get_number_of_monitoring_records_to_delete(PGconn *primary_conn, int keep_histor
 
 
 bool
-delete_monitoring_records(PGconn *primary_conn, int keep_history)
+delete_monitoring_records(PGconn *primary_conn, int keep_history, int node_id)
 {
 	PQExpBufferData query;
 	bool			success = true;
@@ -4210,12 +4217,18 @@ delete_monitoring_records(PGconn *primary_conn, int keep_history)
 
 	initPQExpBuffer(&query);
 
-	if (keep_history > 0)
+	if (keep_history > 0 || node_id != UNKNOWN_NODE_ID)
 	{
 		appendPQExpBuffer(&query,
 						  "DELETE FROM repmgr.monitoring_history "
 						  " WHERE pg_catalog.age(pg_catalog.now(), last_monitor_time) >= '%d days'::INTERVAL ",
 						  keep_history);
+
+		if (node_id != UNKNOWN_NODE_ID)
+		{
+			appendPQExpBuffer(&query,
+							  "  AND standby_node_id = %i", node_id);
+		}
 	}
 	else
 	{

dbutils.h

@@ -491,8 +491,8 @@ add_monitoring_record(PGconn *primary_conn,
 					  long long unsigned int apply_lag_bytes
 );
 
-int			get_number_of_monitoring_records_to_delete(PGconn *primary_conn, int keep_history);
+int			get_number_of_monitoring_records_to_delete(PGconn *primary_conn, int keep_history, int node_id);
-bool		delete_monitoring_records(PGconn *primary_conn, int keep_history);
+bool		delete_monitoring_records(PGconn *primary_conn, int keep_history, int node_id);
 
 
 

doc/appendix-release-notes.sgml

@@ -20,6 +20,7 @@
     <para><emphasis>???, 2018</emphasis></para>
     <para>
     </para>
+
     <sect2>
       <title>Configuration file changes</title>
       <para>
@@ -39,6 +40,23 @@
       </para>
 
     </sect2>
+
+    <sect2>
+      <title>repmgr enhancements</title>
+      <para>
+       <itemizedlist>
+
+          <listitem>
+            <para>
+              <command><link linkend="repmgr-cluster-cleanup">repmgr cluster cleanup</link></command>
+              now accepts the <option>--node-id</option> option to delete records for only one
+              node. (GitHub #493).
+            </para>
+          </listitem>
+
+        </itemizedlist>
+      </para>
+    </sect2>
   </sect1>
 
   <sect1 id="release-4.1.1">

doc/repmgr-cluster-cleanup.sgml

@@ -51,6 +51,20 @@
     </para>
   </refsect1>
 
+  <refsect1>
+    <title>Options</title>
+    <variablelist>
+
+      <varlistentry>
+        <term><option>--node-id</option></term>
+        <listitem>
+          <para>
+            Only delete monitoring records for the specified node.
+          </para>
+        </listitem>
+      </varlistentry>
+    </variablelist>
+  </refsect1>
 
   <refsect1>
     <title>See also</title>

repmgr-action-cluster.c

@@ -1372,7 +1372,9 @@ do_cluster_cleanup(void)
 
 	log_debug(_("number of days of monitoring history to retain: %i"), runtime_options.keep_history);
 
-	entries_to_delete = get_number_of_monitoring_records_to_delete(primary_conn, runtime_options.keep_history);
+	entries_to_delete = get_number_of_monitoring_records_to_delete(primary_conn,
+																   runtime_options.keep_history,
+																   runtime_options.node_id);
 
 	if (entries_to_delete < 0)
 	{
@@ -1392,7 +1394,7 @@ do_cluster_cleanup(void)
 
 	initPQExpBuffer(&event_details);
 
-	if (delete_monitoring_records(primary_conn, runtime_options.keep_history) == false)
+	if (delete_monitoring_records(primary_conn, runtime_options.keep_history, runtime_options.node_id) == false)
 	{
 		appendPQExpBuffer(&event_details,
 						  _("unable to delete monitoring records"));
@@ -1418,8 +1420,21 @@ do_cluster_cleanup(void)
 		log_detail("%s", PQerrorMessage(primary_conn));
 	}
 
-	appendPQExpBuffer(&event_details,
+	if (runtime_options.keep_history == 0)
-					  _("monitoring records deleted"));
+	{
+		appendPQExpBuffer(&event_details,
+						  _("all monitoring records deleted"));
+	}
+	else
+	{
+		appendPQExpBuffer(&event_details,
+						  _("monitoring records deleted"));
+	}
+
+	if (runtime_options.node_id != UNKNOWN_NODE_ID)
+		appendPQExpBuffer(&event_details,
+						  _(" for node %i"),
+						  runtime_options.node_id);
 
 	if (runtime_options.keep_history > 0)
 		appendPQExpBuffer(&event_details,
@@ -1433,18 +1448,11 @@ do_cluster_cleanup(void)
 							  true,
 							  event_details.data);
 
+	log_notice("%s", event_details.data);
+
 	termPQExpBuffer(&event_details);
 	PQfinish(primary_conn);
 
-	if (runtime_options.keep_history > 0)
-	{
-		log_notice(_("monitoring records older than %i day(s) deleted"),
-				   runtime_options.keep_history);
-	}
-	else
-	{
-		log_info(_("all monitoring records deleted"));
-	}
 
 	return;
 }

repmgr-client.c

@@ -1520,6 +1520,7 @@ check_cli_parameters(const int action)
 			case PRIMARY_UNREGISTER:
 			case STANDBY_UNREGISTER:
 			case WITNESS_UNREGISTER:
+			case CLUSTER_CLEANUP:
 			case CLUSTER_EVENT:
 			case CLUSTER_MATRIX:
 			case CLUSTER_CROSSCHECK: