"standby clone": perform clone operations

dbutils.c

@@ -1142,6 +1142,7 @@ get_node_record_by_name(PGconn *conn, const char *node_name, t_node_info *node_i
 	int 		result;
 
 	initPQExpBuffer(&query);
+
 	appendPQExpBuffer(&query,
 		"SELECT node_id, type, upstream_node_id, node_name, conninfo, slot_name, priority, active"
 		"  FROM repmgr.nodes "
@@ -1152,6 +1153,8 @@ get_node_record_by_name(PGconn *conn, const char *node_name, t_node_info *node_i
 
 	result = _get_node_record(conn, query.data, node_info);
 
+	termPQExpBuffer(&query);
+
 	if (result == 0)
 	{
 		log_verbose(LOG_DEBUG, "get_node_record(): no record found for node %s",
@@ -1670,3 +1673,105 @@ get_slot_record(PGconn *conn, char *slot_name, t_replication_slot *record)
 
 	return 1;
 }
+
+
+/* ================ */
+/* backup functions */
+/* ================ */
+
+// XXX is first_wal_segment actually used anywhere?
+bool
+start_backup(PGconn *conn, char *first_wal_segment, bool fast_checkpoint, int server_version_num)
+{
+	PQExpBufferData	  query;
+	PGresult   *res;
+
+	initPQExpBuffer(&query);
+
+	if (server_version_num >= 100000)
+	{
+		appendPQExpBuffer(&query,
+						  "SELECT pg_catalog.pg_walfile_name(pg_catalog.pg_start_backup('repmgr_standby_clone_%ld', %s))",
+						  time(NULL),
+						  fast_checkpoint ? "TRUE" : "FALSE");
+	}
+	else
+	{
+		appendPQExpBuffer(&query,
+						  "SELECT pg_catalog.pg_xlogfile_name(pg_catalog.pg_start_backup('repmgr_standby_clone_%ld', %s))",
+						  time(NULL),
+						  fast_checkpoint ? "TRUE" : "FALSE");
+	}
+
+	log_verbose(LOG_DEBUG, "start_backup():\n  %s", query.data);
+
+	res = PQexec(conn, query.data);
+
+	termPQExpBuffer(&query);
+
+	if (PQresultStatus(res) != PGRES_TUPLES_OK)
+	{
+		log_error(_("unable to start backup:\n  %s"), PQerrorMessage(conn));
+		PQclear(res);
+		return false;
+	}
+
+	if (first_wal_segment != NULL)
+	{
+		char	   *first_wal_seg_pq = PQgetvalue(res, 0, 0);
+		size_t		buf_sz = strlen(first_wal_seg_pq);
+
+		first_wal_segment = pg_malloc0(buf_sz + 1);
+		snprintf(first_wal_segment, buf_sz + 1, "%s", first_wal_seg_pq);
+	}
+
+	PQclear(res);
+
+	return true;
+}
+
+
+bool
+stop_backup(PGconn *conn, char *last_wal_segment, int server_version_num)
+{
+	PQExpBufferData	  query;
+	PGresult   *res;
+
+	initPQExpBuffer(&query);
+
+	if (server_version_num >= 100000)
+	{
+		appendPQExpBuffer(&query,
+						  "SELECT pg_catalog.pg_walfile_name(pg_catalog.pg_stop_backup())");
+	}
+	else
+	{
+		appendPQExpBuffer(&query,
+						  "SELECT pg_catalog.pg_xlogfile_name(pg_catalog.pg_stop_backup())");
+	}
+
+	res = PQexec(conn, query.data);
+	termPQExpBuffer(&query);
+
+	if (PQresultStatus(res) != PGRES_TUPLES_OK)
+	{
+		log_error(_("unable to stop backup:\n  %s"), PQerrorMessage(conn));
+		PQclear(res);
+		return false;
+	}
+
+	if (last_wal_segment != NULL)
+	{
+		char	   *last_wal_seg_pq = PQgetvalue(res, 0, 0);
+		size_t		buf_sz = strlen(last_wal_seg_pq);
+
+		last_wal_segment = pg_malloc0(buf_sz + 1);
+		snprintf(last_wal_segment, buf_sz + 1, "%s", last_wal_seg_pq);
+	}
+
+	PQclear(res);
+
+	return true;
+}
+
+

dbutils.h

@@ -184,5 +184,9 @@ bool		create_replication_slot(PGconn *conn, char *slot_name, int server_version_
 
 int			get_slot_record(PGconn *conn, char *slot_name, t_replication_slot *record);
 
+/* backup functions */
+bool		start_backup(PGconn *conn, char *first_wal_segment, bool fast_checkpoint, int server_version_num);
+bool		stop_backup(PGconn *conn, char *last_wal_segment, int server_version_num);
+
 #endif
 

errcode.h

@@ -26,6 +26,7 @@
 #define ERR_SWITCHOVER_FAIL 18
 #define ERR_BARMAN 19
 #define ERR_REGISTRATION_SYNC 20
+#define ERR_OUT_OF_MEMORY 21
 
 #endif	 /* _ERRCODE_H_ */
 

repmgr-client-global.h

@@ -109,6 +109,11 @@ extern bool local_command(const char *command, PQExpBufferData *outputbuf);
 extern bool check_upstream_config(PGconn *conn, int server_version_num, bool exit_on_error);
 extern standy_clone_mode get_standby_clone_mode(void);
 
+extern int  copy_remote_files(char *host, char *remote_user, char *remote_path,
+							  char *local_path, bool is_directory, int server_version_num);
+
 extern void print_error_list(ItemList *error_list, int log_level);
 
+extern char * make_pg_path(char *file);
+
 #endif

repmgr-client.c

@@ -17,6 +17,7 @@
 
 #include <unistd.h>
 
+
 #include "repmgr.h"
 #include "repmgr-client.h"
 #include "repmgr-client-global.h"
@@ -24,6 +25,8 @@
 #include "repmgr-action-standby.h"
 #include "repmgr-action-cluster.h"
 
+#include <storage/fd.h>         /* for PG_TEMP_FILE_PREFIX */
+
 
 /* globally available variables *
  * ============================ */
@@ -40,6 +43,8 @@ char	 pg_bindir[MAXLEN] = "";
 char	 repmgr_slot_name[MAXLEN] = "";
 char	*repmgr_slot_name_ptr = NULL;
 
+char  path_buf[MAXLEN] = "";
+
 /*
  * if --node-id/--node-name provided, place that node's record here
  * for later use
@@ -1580,3 +1585,123 @@ get_standby_clone_mode(void)
 
 	return mode;
 }
+
+
+char *
+make_pg_path(char *file)
+{
+	maxlen_snprintf(path_buf, "%s%s", pg_bindir, file);
+
+	return path_buf;
+}
+
+
+int
+copy_remote_files(char *host, char *remote_user, char *remote_path,
+				  char *local_path, bool is_directory, int server_version_num)
+{
+	PQExpBufferData 	rsync_flags;
+	char		script[MAXLEN];
+	char		host_string[MAXLEN];
+	int			r;
+
+	initPQExpBuffer(&rsync_flags);
+
+	if (*config_file_options.rsync_options == '\0')
+	{
+		appendPQExpBuffer(&rsync_flags, "%s",
+						  "--archive --checksum --compress --progress --rsh=ssh");
+	}
+	else
+	{
+		appendPQExpBuffer(&rsync_flags, "%s",
+						  config_file_options.rsync_options);
+	}
+
+	if (runtime_options.force)
+	{
+		appendPQExpBuffer(&rsync_flags, "%s",
+						  " --delete --checksum");
+	}
+
+	if (!remote_user[0])
+	{
+		maxlen_snprintf(host_string, "%s", host);
+	}
+	else
+	{
+		maxlen_snprintf(host_string, "%s@%s", remote_user, host);
+	}
+
+	/*
+	 * When copying the main PGDATA directory, certain files and contents
+	 * of certain directories need to be excluded.
+	 *
+	 * See function 'sendDir()' in 'src/backend/replication/basebackup.c' -
+	 * we're basically simulating what pg_basebackup does, but with rsync rather
+	 * than the BASEBACKUP replication protocol command.
+	 *
+	 * *However* currently we'll always copy the contents of the 'pg_replslot'
+	 * directory and delete later if appropriate.
+	 */
+	if (is_directory)
+	{
+		/* Files which we don't want */
+		appendPQExpBuffer(&rsync_flags, "%s",
+						  " --exclude=postmaster.pid --exclude=postmaster.opts --exclude=global/pg_control");
+
+		appendPQExpBuffer(&rsync_flags, "%s",
+						  " --exclude=recovery.conf --exclude=recovery.done");
+
+		if (server_version_num >= 90400)
+		{
+			/*
+			 * Ideally we'd use PG_AUTOCONF_FILENAME from utils/guc.h, but
+			 * that has too many dependencies for a mere client program.
+			 */
+			appendPQExpBuffer(&rsync_flags, "%s",
+							  " --exclude=postgresql.auto.conf.tmp");
+		}
+
+		/* Temporary files which we don't want, if they exist */
+		appendPQExpBuffer(&rsync_flags, " --exclude=%s*",
+						  PG_TEMP_FILE_PREFIX);
+
+		/* Directories which we don't want */
+
+		if (server_version_num >= 100000)
+		{
+			appendPQExpBuffer(&rsync_flags, "%s",
+							  " --exclude=pg_wal/*");
+		}
+		else
+		{
+			appendPQExpBuffer(&rsync_flags, "%s",
+							  " --exclude=pg_xlog/*");
+		}
+
+		appendPQExpBuffer(&rsync_flags, "%s",
+						  " --exclude=pg_log/* --exclude=pg_stat_tmp/*");
+
+		maxlen_snprintf(script, "rsync %s %s:%s/* %s",
+						rsync_flags.data, host_string, remote_path, local_path);
+	}
+	else
+	{
+		maxlen_snprintf(script, "rsync %s %s:%s %s",
+						rsync_flags.data, host_string, remote_path, local_path);
+	}
+
+	log_info(_("rsync command line: '%s'"), script);
+
+	r = system(script);
+
+	log_debug("copy_remote_files(): r = %i; WIFEXITED: %i; WEXITSTATUS: %i", r, WIFEXITED(r), WEXITSTATUS(r));
+
+	/* exit code 24 indicates vanished files, which isn't a problem for us */
+	if (WIFEXITED(r) && WEXITSTATUS(r) && WEXITSTATUS(r) != 24)
+		log_verbose(LOG_WARNING, "copy_remote_files(): rsync returned unexpected exit status %i", WEXITSTATUS(r));
+
+	return r;
+}
+