Key takeaways
- Watch resources, connections, feeds and scheduled tasks.
- Alert on symptoms clients would notice first.
- Avoid alert fatigue by tuning thresholds.
What to monitor
- CPU, memory, disk and network on each server.
- Client connections per access server.
- Gateway and liquidity provider connections.
- Quote flow per symbol.
- Backups and scheduled reports.
- Errors in server and plugin logs.
Alerting
Alert on things that affect trading: quotes stopping, gateways disconnecting, access servers full or unreachable. Send lower-priority signals to daily reports instead of phones.
Response
Every alert should have an owner and a short runbook. An outsourced administrator with team backup covers nights and weekends. See Hire an Administrator.