Load balancers sit at a strategic point for observability. Every request passes through.
Log these fields for each request:
- Timestamp and duration
- Client IP and requested URL
- Backend server selected
- Response status code
- Bytes transferred
Monitor these metrics:
- Requests per second (throughput)
- Error rates (, responses)
- Latency percentiles (p50, p95, p99)
- Backend health status
- Active connections per server
Alerts on error rates and latency catch problems before users complain.