Skip to main contentOperational visibility relies on metrics, logs, and alerts tuned to your workload.
Observability stack
- Prometheus for metrics collection
- Grafana for dashboards and visualizations
- Loki (or ELK) for centralized log aggregation
Key service metrics
- Kafka consumer lag
- WebSocket active connection count
- Redis memory utilization and cache hit ratio
- TiDB region health and TiKV store availability
Alerting recommendations
- Sustained CPU utilization above 80%
- Database query latency exceeding 100 ms
- Kafka consumer lag breaching thresholds
- WebSocket connection drops or failure rate spikes
- Tune thresholds to traffic patterns and workload characteristics
