What dashboards are must-have for a critical API?

Deep dive

Dashboards should answer "are we healthy" and "where is it broken":

• RED: request rate, error rate, duration (p50/p95/p99).
• Saturation: CPU, memory, thread pools, queue depth.
• Dependencies: DB latency, cache hit ratio, downstream status.
• SLO: burn-rate over 1h/6h windows.

Examples

A good top row:

RPS | Error % | p95/p99 latency | SLO burn-rate

Common pitfalls

• Too many charts without a narrative.
• Missing breakdowns by version or region.
• No links to traces/logs from the dashboard.

Interview follow-ups

• How do you choose time windows for p95/p99