Add "repmgr node status"

Outputs an overview of a node's status, and emits warnings if any
issues detected.

Makefile.in

@@ -19,6 +19,8 @@ SCRIPTS_built = repmgr4 repmgrd4
 PG_CPPFLAGS = -std=gnu89 -I$(libpq_srcdir) -Wall -Wmissing-prototypes -Wmissing-declarations $(EXTRA_CFLAGS)
 SHLIB_LINK = $(libpq)
 
+HEADERS = $(wildcard *.h)
+
 OBJS = \
 	repmgr.o
 
@@ -27,7 +29,7 @@ include Makefile.global
 $(info Building against PostgreSQL $(MAJORVERSION))
 
 REPMGR_CLIENT_OBJS = repmgr-client.o \
-  repmgr-action-primary.o repmgr-action-standby.o repmgr-action-bdr.o repmgr-action-cluster.o \
+  repmgr-action-primary.o repmgr-action-standby.o repmgr-action-bdr.o repmgr-action-cluster.o repmgr-action-node.o \
   configfile.o log.o strutil.o dbutils.o dirutil.o compat.o controldata.o
 REPMGRD_OBJS = repmgrd.o repmgrd-physical.o repmgrd-bdr.o configfile.o log.o dbutils.o strutil.o
 
@@ -39,6 +41,8 @@ repmgr4: $(REPMGR_CLIENT_OBJS)
 repmgrd4: $(REPMGRD_OBJS)
 	$(CC) $(CFLAGS) $(REPMGRD_OBJS) $(libpq_pgport) $(LDFLAGS) $(LDFLAGS_EX) $(LIBS) -o $@$(X)
 
+$(REPMGR_CLIENT_OBJS): $(HEADERS)
+$(REPMGRD_OBJS): $(HEADERS)
 
 # Ensure Makefiles are up-to-date (should we move this to Makefile.global?)
 Makefile: Makefile.in config.status configure
@@ -56,6 +60,7 @@ additional-clean:
 	rm -f repmgr-action-primary.o
 	rm -f repmgr-action-standby.o
 	rm -f repmgr-action-bdr.o
+	rm -f repmgr-action-node.o
 	rm -f repmgr-action-cluster.o
 	rm -f repmgrd.o
 	rm -f repmgrd-physical.o

dbutils.c

@@ -2005,6 +2005,7 @@ delete_node_record(PGconn *conn, int node)
 	log_verbose(LOG_DEBUG, "delete_node_record():\n  %s", query.data);
 
 	res = PQexec(conn, query.data);
+	termPQExpBuffer(&query);
 
 	if (!res || PQresultStatus(res) != PGRES_COMMAND_OK)
 	{
@@ -2019,6 +2020,43 @@ delete_node_record(PGconn *conn, int node)
 	return true;
 }
 
+void
+get_node_replication_stats(PGconn *conn, t_node_info *node_info)
+{
+	PQExpBufferData	  query;
+	PGresult   *res;
+
+	initPQExpBuffer(&query);
+
+	appendPQExpBuffer(
+		&query,
+		" SELECT current_setting('max_wal_senders')::INT AS max_wal_senders, "
+		"        (SELECT COUNT(*) FROM pg_catalog.pg_stat_replication) AS attached_wal_receivers, "
+		"        current_setting('max_replication_slots')::INT AS max_replication_slots, "
+		"        (SELECT COUNT(*) FROM pg_catalog.pg_replication_slots WHERE active = TRUE)  AS active_replication_slots, "
+		"        (SELECT COUNT(*) FROM pg_catalog.pg_replication_slots WHERE active = FALSE) AS inactive_replication_slots ");
+
+	res = PQexec(conn, query.data);
+	termPQExpBuffer(&query);
+
+	if (PQresultStatus(res) != PGRES_TUPLES_OK)
+	{
+		log_warning(_("unable to retrieve node replication statistics:\n  %s"),
+					PQerrorMessage(conn));
+		PQclear(res);
+		return;
+	}
+
+	node_info->max_wal_senders = atoi(PQgetvalue(res, 0, 0));
+	node_info->attached_wal_receivers = atoi(PQgetvalue(res, 0, 1));
+	node_info->max_replication_slots = atoi(PQgetvalue(res, 0, 2));
+	node_info->active_replication_slots = atoi(PQgetvalue(res, 0, 3));
+	node_info->inactive_replication_slots = atoi(PQgetvalue(res, 0, 4));
+
+	return;
+}
+
+
 
 void
 clear_node_info_list(NodeInfoList *nodes)

dbutils.h

@@ -84,6 +84,12 @@ typedef struct s_node_info
 	PGconn		 *conn;
 	/* for ad-hoc use e.g. when working with a list of nodes */
     char		  details[MAXLEN];
+	/* various statistics */
+	int			  max_wal_senders;
+	int			  attached_wal_receivers;
+	int			  max_replication_slots;
+	int			  active_replication_slots;
+	int			  inactive_replication_slots;
 }	t_node_info;
 
 
@@ -104,7 +110,8 @@ typedef struct s_node_info
 	RECTYPE_UNKNOWN,  \
 	MS_NORMAL, \
 	NULL, \
-	"" \
+	"", \
+	-1, -1, -1, -1, -1 \
 }
 
 
@@ -306,6 +313,8 @@ bool		update_node_record_conn_priority(PGconn *conn, t_configuration_options *op
 
 void		clear_node_info_list(NodeInfoList *nodes);
 
+void		get_node_replication_stats(PGconn *conn, t_node_info *node_info);
+
 /* event functions */
 bool		create_event_record(PGconn *conn, t_configuration_options *options, int node_id, char *event, bool successful, char *details);
 bool		create_event_notification(PGconn *conn, t_configuration_options *options, int node_id, char *event, bool successful, char *details);

repmgr-action-node.c

@@ -0,0 +1,184 @@
+/*
+ * repmgr-action-node.c
+ *
+ * Implements actions available for any kind of node
+ *
+ * Copyright (c) 2ndQuadrant, 2010-2017
+ */
+
+
+#include "repmgr.h"
+
+#include "repmgr-client-global.h"
+#include "repmgr-action-node.h"
+
+void
+do_node_status(void)
+{
+	PGconn	   	   *conn;
+
+	int 			target_node_id = UNKNOWN_NODE_ID;
+	t_node_info 	node_info = T_NODE_INFO_INITIALIZER;
+	char			server_version[MAXLEN];
+	char			cluster_size[MAXLEN];
+	PQExpBufferData	output;
+
+	ItemList 		warnings = { NULL, NULL };
+	RecoveryType	recovery_type;
+
+	if (strlen(config_file_options.conninfo))
+		conn = establish_db_connection(config_file_options.conninfo, true);
+	else
+		conn = establish_db_connection_by_params(&source_conninfo, true);
+
+	if (runtime_options.node_id != UNKNOWN_NODE_ID)
+		target_node_id = runtime_options.node_id;
+	else
+		target_node_id = config_file_options.node_id;
+
+	/* Check node exists and is really a standby */
+
+	if (get_node_record(conn, target_node_id, &node_info) != RECORD_FOUND)
+	{
+		log_error(_("no record found for node %i"), target_node_id);
+		PQfinish(conn);
+		exit(ERR_BAD_CONFIG);
+	}
+	(void) get_server_version(conn, server_version);
+
+	if (get_cluster_size(conn, cluster_size) == false)
+		strncpy(cluster_size, _("unknown"), MAXLEN);
+
+	recovery_type = get_recovery_type(conn);
+
+	get_node_replication_stats(conn, &node_info);
+
+	initPQExpBuffer(&output);
+
+	appendPQExpBuffer(
+		&output,
+		"Node \"%s\":\n",
+		node_info.node_name);
+
+	appendPQExpBuffer(
+		&output,
+		"%cPostgreSQL version: %s\n",
+		'\t', server_version);
+
+	appendPQExpBuffer(
+		&output,
+		"%cTotal data size: %s\n",
+		'\t', cluster_size);
+
+	appendPQExpBuffer(
+		&output,
+		"%cConninfo: \"%s\"\n",
+		'\t', node_info.conninfo);
+
+	appendPQExpBuffer(
+		&output,
+		"%cRole: \"%s\"\n",
+		'\t', get_node_type_string(node_info.type));
+
+	switch (node_info.type)
+	{
+		case PRIMARY:
+			if (recovery_type == RECTYPE_STANDBY)
+			{
+				item_list_append(
+					&warnings,
+					_("- node is registered as primary but running as standby"));
+			}
+			break;
+		case STANDBY:
+			if (recovery_type == RECTYPE_PRIMARY)
+			{
+				item_list_append(
+					&warnings,
+					_("- node is registered as standby but running as primary"));
+			}
+			break;
+		case BDR:
+		default:
+			break;
+	}
+
+
+	if (node_info.max_wal_senders >= 0)
+	{
+		appendPQExpBuffer(
+			&output,
+			"%cReplication connections: %i (of maximal %i)\n",
+			'\t',
+			node_info.attached_wal_receivers,
+			node_info.max_wal_senders);
+
+	}
+	else if (node_info.max_wal_senders == 0)
+	{
+		appendPQExpBuffer(
+			&output,
+			"%cReplication connections: disabled\n",
+			'\t');
+	}
+
+	if (node_info.max_replication_slots > 0)
+	{
+		appendPQExpBuffer(
+			&output,
+			"%cReplication slots: %i (of maximal %i)",
+			'\t',
+			node_info.active_replication_slots + node_info.inactive_replication_slots,
+			node_info.max_replication_slots);
+
+		if (node_info.inactive_replication_slots > 0)
+		{
+			appendPQExpBuffer(
+				&output,
+				"; %i inactive",
+				node_info.inactive_replication_slots);
+
+			item_list_append_format(
+				&warnings,
+				_("- node has %i inactive replication slots"),
+				node_info.inactive_replication_slots);
+		}
+	}
+	else if (node_info.max_replication_slots == 0)
+	{
+		appendPQExpBuffer(
+			&output,
+			"%cReplication slots: disabled\n",
+			'\t');
+	}
+
+/*
+	appendPQExpBuffer(
+		&output,
+		"%c: \"%s\"\n",
+		'\t', );
+	appendPQExpBuffer(
+		&output,
+		"%c: \"%s\"\n",
+		'\t', node_info.);
+*/
+
+
+	puts(output.data);
+
+	termPQExpBuffer(&output);
+
+	if (warnings.head != NULL && runtime_options.terse == false)
+	{
+		log_warning(_("following issue(s) were detected:"));
+		print_item_list(&warnings);
+		log_hint(_("execute \"repmgr node check\" for more details"));
+	}
+
+}
+
+
+void
+do_node_check(void)
+{
+}

repmgr-action-node.h

@@ -0,0 +1,12 @@
+/*
+ * repmgr-action-node.h
+ * Copyright (c) 2ndQuadrant, 2010-2017
+ */
+
+#ifndef _REPMGR_ACTION_NODE_H_
+#define _REPMGR_ACTION_NODE_H_
+
+extern void do_node_status(void);
+extern void do_node_check(void);
+
+#endif /* _REPMGR_ACTION_NODE_H_ */

repmgr-action-standby.c

@@ -1389,8 +1389,8 @@ do_standby_follow(void)
 	 * replication slots are in use and we want to delete the old slot.
 	 */
 	record_status = get_node_record(primary_conn,
-									  config_file_options.node_id,
+									config_file_options.node_id,
-									  &local_node_record);
+									&local_node_record);
 
 	if (record_status != RECORD_FOUND)
 	{

repmgr-client.c

@@ -16,6 +16,9 @@
  * STANDBY UNREGISTER
  * STANDBY PROMOTE
  *
+ * BDR REGISTER
+ * BDR UNREGISTER
+ *
  * CLUSTER SHOW
  * CLUSTER EVENT
  */
@@ -30,6 +33,7 @@
 #include "repmgr-action-primary.h"
 #include "repmgr-action-standby.h"
 #include "repmgr-action-bdr.h"
+#include "repmgr-action-node.h"
 
 #include "repmgr-action-cluster.h"
 
@@ -588,6 +592,10 @@ main(int argc, char **argv)
 				action = PRIMARY_REGISTER;
 			else if (strcasecmp(repmgr_action, "UNREGISTER") == 0)
 				action = PRIMARY_UNREGISTER;
+			else if (strcasecmp(repmgr_action, "CHECK") == 0)
+				action = NODE_CHECK;
+			else if (strcasecmp(repmgr_action, "STATUS") == 0)
+				action = NODE_STATUS;
 		}
 
 		else if (strcasecmp(repmgr_node_type, "STANDBY") == 0)
@@ -608,6 +616,10 @@ main(int argc, char **argv)
 				action = STANDBY_ARCHIVE_CONFIG;
 			else if (strcasecmp(repmgr_action, "RESTORE-CONFIG") == 0)
 				action = STANDBY_RESTORE_CONFIG;
+			else if (strcasecmp(repmgr_action, "CHECK") == 0)
+				action = NODE_CHECK;
+			else if (strcasecmp(repmgr_action, "STATUS") == 0)
+				action = NODE_STATUS;
 		}
 		else if (strcasecmp(repmgr_node_type, "BDR") == 0)
 #else
@@ -618,6 +630,18 @@ main(int argc, char **argv)
 				action = BDR_REGISTER;
 			else if (strcasecmp(repmgr_action, "UNREGISTER") == 0)
 				action = BDR_UNREGISTER;
+			else if (strcasecmp(repmgr_action, "CHECK") == 0)
+				action = NODE_CHECK;
+			else if (strcasecmp(repmgr_action, "STATUS") == 0)
+				action = NODE_STATUS;
+		}
+
+		else if (strcasecmp(repmgr_node_type, "NODE") == 0)
+		{
+		    if (strcasecmp(repmgr_action, "CHECK") == 0)
+				action = NODE_CHECK;
+			else if (strcasecmp(repmgr_action, "STATUS") == 0)
+				action = NODE_STATUS;
 		}
 
 		else if (strcasecmp(repmgr_node_type, "CLUSTER") == 0)
@@ -929,6 +953,11 @@ main(int argc, char **argv)
 			do_bdr_unregister();
 			break;
 
+		/* NODE */
+		case NODE_STATUS:
+			do_node_status();
+			break;
+
 		/* CLUSTER */
 		case CLUSTER_SHOW:
 			do_cluster_show();

repmgr-client.h

@@ -22,13 +22,15 @@
 #define STANDBY_SWITCHOVER	   8
 #define STANDBY_ARCHIVE_CONFIG 9
 #define STANDBY_RESTORE_CONFIG 10
-#define CLUSTER_SHOW		   11
+#define BDR_REGISTER		   11
-#define CLUSTER_CLEANUP		   12
+#define BDR_UNREGISTER		   12
-#define CLUSTER_MATRIX		   13
+#define NODE_STATUS			   13
-#define CLUSTER_CROSSCHECK	   14
+#define NODE_CHECK			   14
-#define CLUSTER_EVENT		   15
+#define CLUSTER_SHOW		   15
-#define BDR_REGISTER		   16
+#define CLUSTER_CLEANUP		   16
-#define BDR_UNREGISTER		   17
+#define CLUSTER_MATRIX		   17
+#define CLUSTER_CROSSCHECK	   18
+#define CLUSTER_EVENT		   19
 
 /* command line options without short versions */
 #define OPT_HELP						   1