Steps to reproduce:

1. RDS MySQL Database Minor Version Upgrades via Blue-Green Deployment
2. When the upgrade is complete and there is an instance switchover, the old instance becomes a read-only instance
3. Zabbix servers still connects to old instance due to DNS caching
4. Zabbix servers configured with ha mode stopped the zabbix service at the same time due to connected read-only instances

Result:
See screenshot...

**

Expected:
The zabbix server should stop the service when it detects that the database cannot be inserted and continues to fail to execute for several minutes before stopping the service.

Otherwise, this Zabbix high availability means nothing. Because this problem connecting to a read-only instance could probably be recovered after 1 minute, and it's so weird that the zabbix service actually just stops when it detects the anomaly.