github/repmgr
1
0
Fork 0
mirror of https://github.com/EnterpriseDB/repmgr.git synced 2026-03-27 08:56:29 +00:00
Code Issues Packages Projects Releases 1 Wiki Activity

Better detect completion of demotion candidate shutdown

Browse Source 
If a connection attempt fails, keep pinging the server until it
finally away, or the timeout kicks in.

Addresses issue reported in GitHub #188 and previously noted in
repmgr.c
This commit is contained in:
Ian Barwick
2016-06-30 21:27:52 +09:00
parent c30447ac90
commit 746c9793ed
1 changed files with 19 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
repmgr.c
View File

@@ -2769,8 +2769,8 @@ do_standby_switchover(void)
char repmgr_db_cli_params[MAXLEN] = ""; char repmgr_db_cli_params[MAXLEN] = "";
int query_result; int query_result;
t_node_info remote_node_record; t_node_info remote_node_record;
bool connection_success; bool connection_success,
shutdown_success;
/* /*
* If --remote_pg_bindir supplied, use that to build the path on the * If --remote_pg_bindir supplied, use that to build the path on the
@@ -3187,37 +3187,42 @@ do_standby_switchover(void)
termPQExpBuffer(&command_output); termPQExpBuffer(&command_output);
connection_success = false; shutdown_success = false;
/* loop for timeout waiting for current primary to stop */ /* loop for timeout waiting for current primary to stop */
for(i = 0; i < options.reconnect_attempts; i++) for (i = 0; i < options.reconnect_attempts; i++)
{ {
/* Check whether primary is available */ /* Check whether primary is available */
remote_conn = test_db_connection(remote_conninfo, false); /* don't fail on error */ remote_conn = test_db_connection(remote_conninfo, false); /* don't fail on error */
/* XXX failure to connect doesn't mean the server is necessarily /*
* completely stopped - we need to better detect the reason for * If we're unable to connect, keep PQping-ing the server until it
* connection failure ("server not listening" vs "shutting down") * finally goes away
*
* -> check is_pgup()
*/ */
if (PQstatus(remote_conn) != CONNECTION_OK) if (PQstatus(remote_conn) != CONNECTION_OK)
{ {
connection_success = true; PGPing ping_res = PQping(remote_conninfo);
log_notice(_("current master has been stopped\n")); /* database server could not be contacted */
break; if (ping_res == PQPING_NO_RESPONSE)
{
/* XXX we should double-check access to the physical server here */
shutdown_success = true;
log_notice(_("current master has been stopped\n"));
break;
}
} }
PQfinish(remote_conn); PQfinish(remote_conn);
// configurable? /* XXX make configurable? */
sleep(options.reconnect_interval); sleep(options.reconnect_interval);
i++; i++;
} }
if (connection_success == false) if (shutdown_success == false)
{ {
log_err(_("master server did not shut down\n")); log_err(_("master server did not shut down\n"));
log_hint(_("check the master server status before performing any further actions")); log_hint(_("check the master server status before performing any further actions"));