hcq/MaxScale
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

MXS-1490: Perform failover only after failcount monitor loops

Browse Source 
The same failcount variable is used for the detect_standalone_master-
feature.
This commit is contained in:
Esa Korhonen
2017-11-15 16:41:12 +02:00
parent 703230a930
commit 84d1ea0bff
1 changed files with 35 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files

69
server/modules/monitor/mysqlmon/mysql_mon.cc
View File

@ -2884,53 +2884,54 @@ bool mon_process_failover(MYSQL_MONITOR* monitor, uint32_t failover_timeout)
MXS_CONFIG* cnf = config_get_global_options(); MXS_CONFIG* cnf = config_get_global_options();
MXS_MONITORED_SERVER* failed_master = NULL; MXS_MONITORED_SERVER* failed_master = NULL;
for (MXS_MONITORED_SERVER *ptr = monitor->monitor->monitored_servers; ptr; ptr = ptr->next) if (!cnf->passive)
{ {
if (ptr->new_event && !cnf->passive && for (MXS_MONITORED_SERVER *ptr = monitor->monitor->monitored_servers; ptr; ptr = ptr->next)
ptr->server->last_event == MASTER_DOWN_EVENT)
{ {
if (failed_master) if (ptr->new_event && ptr->server->last_event == MASTER_DOWN_EVENT)
{ {
MXS_ALERT("Multiple failed master servers detected: " if (failed_master)
"'%s' is the first master to fail but server "
"'%s' has also triggered a master_down event.",
failed_master->server->unique_name,
ptr->server->unique_name);
return false;
}
if (ptr->server->active_event)
{
// MaxScale was active when the event took place
failed_master = ptr;
ptr->new_event = false;
}
else if (monitor->monitor->master_has_failed)
{
/**
* If a master_down event was triggered when this MaxScale was
* passive, we need to execute the failover script again if no new
* masters have appeared.
*/
int64_t timeout = SEC_TO_HB(failover_timeout);
int64_t t = hkheartbeat - ptr->server->triggered_at;
if (t > timeout)
{ {
MXS_WARNING("Failover of server '%s' did not take place within " MXS_ALERT("Multiple failed master servers detected: "
"%u seconds, failover needs to be re-triggered", "'%s' is the first master to fail but server "
ptr->server->unique_name, failover_timeout); "'%s' has also triggered a master_down event.",
failed_master->server->unique_name,
ptr->server->unique_name);
return false;
}
if (ptr->server->active_event)
{
// MaxScale was active when the event took place
failed_master = ptr; failed_master = ptr;
ptr->new_event = false; }
else if (monitor->monitor->master_has_failed)
{
/**
* If a master_down event was triggered when this MaxScale was
* passive, we need to execute the failover script again if no new
* masters have appeared.
*/
int64_t timeout = SEC_TO_HB(failover_timeout);
int64_t t = hkheartbeat - ptr->server->triggered_at;
if (t > timeout)
{
MXS_WARNING("Failover of server '%s' did not take place within "
"%u seconds, failover needs to be re-triggered",
ptr->server->unique_name, failover_timeout);
failed_master = ptr;
}
} }
} }
} }
} }
if (failed_master) if (failed_master && failed_master->mon_err_count >= monitor->failcount)
{ {
MXS_NOTICE("Performing automatic failover to replace failed master '%s'.", MXS_NOTICE("Performing automatic failover to replace failed master '%s'.",
failed_master->server->unique_name); failed_master->server->unique_name);
failed_master->new_event = false;
rval = do_failover(monitor); rval = do_failover(monitor);
} }