Improve replication connection check

Previously the check verifying that a node has connected to its upstream
merely assumed the presence of a record in pg_stat_replication indicates
a successful replication connection. However the record may contain a
state other than "streaming", typically "startup" (which will occur when
a node has diverged from its upstream and will therefore never
transition to "streaming"), which needs to be taken into account when
considering the state of the replication connection to avoid false
positives.
This commit is contained in:
Ian Barwick
2020-08-27 16:05:12 +09:00
parent 53c9eacbc4
commit 0ad6aceceb
5 changed files with 90 additions and 35 deletions

View File

@@ -1797,7 +1797,7 @@ do_standby_register(void)
else
{
/* check our standby is connected */
if (is_downstream_node_attached(upstream_conn, config_file_options.node_name) == NODE_ATTACHED)
if (is_downstream_node_attached(upstream_conn, config_file_options.node_name, NULL) == NODE_ATTACHED)
{
log_verbose(LOG_INFO, _("local node is attached to specified upstream node %i"), runtime_options.upstream_node_id);
}
@@ -1856,7 +1856,7 @@ do_standby_register(void)
primary_node_id);
/* check our standby is connected */
if (is_downstream_node_attached(primary_conn, config_file_options.node_name) == NODE_ATTACHED)
if (is_downstream_node_attached(primary_conn, config_file_options.node_name, NULL) == NODE_ATTACHED)
{
log_verbose(LOG_INFO, _("local node is attached to primary"));
}
@@ -3073,7 +3073,9 @@ do_standby_follow(void)
for (timer = 0; timer < config_file_options.standby_follow_timeout; timer++)
{
NodeAttached node_attached = is_downstream_node_attached(follow_target_conn, config_file_options.node_name);
NodeAttached node_attached = is_downstream_node_attached(follow_target_conn,
config_file_options.node_name,
NULL);
if (node_attached == NODE_ATTACHED)
{
@@ -3711,7 +3713,7 @@ do_standby_switchover(void)
exit(ERR_BAD_CONFIG);
}
if (is_downstream_node_attached(remote_conn, local_node_record.node_name) != NODE_ATTACHED)
if (is_downstream_node_attached(remote_conn, local_node_record.node_name, NULL) != NODE_ATTACHED)
{
log_error(_("local node \"%s\" (ID: %i) is not attached to demotion candidate \"%s\" (ID: %i)"),
local_node_record.node_name,
@@ -5195,7 +5197,8 @@ do_standby_switchover(void)
*/
node_attached = is_downstream_node_attached(local_conn,
remote_node_record.node_name);
remote_node_record.node_name,
NULL);
if (node_attached == NODE_ATTACHED)
{
switchover_success = true;