github/repmgr
1
0
Fork 0
mirror of https://github.com/EnterpriseDB/repmgr.git synced 2026-03-22 22:56:29 +00:00
Code Issues Packages Projects Releases 1 Wiki Activity

EXPERIMENTAL: synchronise try_primary_reconnect()'s reconnection loop

Browse Source 
Per proposal in GitHub #662, this patch attempts to synchronise each
repmgrd's primary reconnection attempts to prevent potential race
conditions. This relies on each node's clock being correcly
synchronised.

Currently this change is experimental and is not enabled by default.
It can be enabled by setting the repmgr.conf parameter
"reconnect_loop_sync".
This commit is contained in:
Ian Barwick
2020-10-06 13:22:38 +09:00
parent 467d19bcd4
commit e10d9fd393
3 changed files with 53 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
configdata.c
View File

@@ -881,6 +881,20 @@ struct ConfigFileSetting config_file_settings[] =
{ .strmaxlen = sizeof(config_file_options.ssh_options) },
{}
},
/* ==================================
* undocumented experimental settings
* ==================================
*/
/* reconnect_loop_sync */
{
"reconnect_loop_sync",
CONFIG_BOOL,
{ .boolptr = &config_file_options.reconnect_loop_sync },
{ .booldefault = false },
{},
{},
{}
},
/* ==========================
* undocumented test settings
* ==========================

12
configfile.h
View File

@@ -238,7 +238,17 @@ typedef struct
char rsync_options[MAXLEN];
char ssh_options[MAXLEN];
/* undocumented test settings */
/*
* undocumented settings
*
* These settings are for testing or experimential features
* and may be changed without notice.
*/
/* experimental settings */
bool reconnect_loop_sync;
/* test settings */
int promote_delay;
int failover_delay;
char connection_check_query[MAXLEN];

32
repmgrd-physical.c
View File

@@ -5377,6 +5377,11 @@ try_primary_reconnect(PGconn **conn, PGconn *local_conn, t_node_info *node_info)
for (i = 0; i < max_attempts; i++)
{
time_t started_at = time(NULL);
int up_to;
bool sleep_now = false;
bool max_sleep_seconds;
log_info(_("checking state of node \"%s\" (ID: %i), %i of %i attempts"),
node_info->node_name,
node_info->node_id,
@@ -5442,12 +5447,31 @@ try_primary_reconnect(PGconn **conn, PGconn *local_conn, t_node_info *node_info)
node_info->node_id);
}
if (i + 1 < max_attempts)
/*
* Experimental behaviour, see GitHub #662.
*/
if (config_file_options.reconnect_loop_sync == true)
{
up_to = (time(NULL) - started_at);
max_sleep_seconds = (up_to == 0)
? config_file_options.reconnect_interval
: (up_to % config_file_options.reconnect_interval);
if (i + 1 <= max_attempts)
sleep_now = true;
}
else
{
max_sleep_seconds = config_file_options.reconnect_interval;
if (i + 1 < max_attempts)
sleep_now = true;
}
if (sleep_now == true)
{
int j;
log_info(_("sleeping %i seconds until next reconnection attempt"),
config_file_options.reconnect_interval);
for (j = 0; j < config_file_options.reconnect_interval; j++)
log_info(_("sleeping up to %i seconds until next reconnection attempt"),
max_sleep_seconds);
for (j = 0; j < max_sleep_seconds; j++)
{
int new_primary_node_id;
if (get_new_primary(local_conn, &new_primary_node_id) == true && new_primary_node_id != UNKNOWN_NODE_ID)