Steps to reproduce:
- Install PgSQL 11 + timescaledb
- Configure with 9 proxies, with vps ~ 2200
- Restart zabbix-server process
On restart, the zabbix server sends a large UPDATE query that eventually exhausts available memory on the pgsql Server. Server is configured with 64Gb RAM, 40Gb Swap, with large work_mem to satisfy requirements of zabbix select queries.
When this occurs, the history syncer processes do not update history and the database/server falls behind, generating multiple triggers and alerts.
See log file...
See memory dump...
History syncer does not generate queries that exhaust database memory.
Note: this appears to be the same issue reported in