Engineering Lab
Incident Desk
A browser incident desk: logs, metrics, traces, database, cache, API, deploys, and changes. Name the root cause. Then read the postmortem.
Al Beltran Engineering Lab
Loading desk…
How the challenge works
- Open one channel at a time. You do not need every pane to name the cause.
- Hot channels usually explain the symptom. Noise channels are there so you practice skipping them.
- After you name a cause, the desk shows evidence, the fix, the trade-off, and what to monitor. This is a demo incident — not live telemetry.
Key concepts
Acquire time vs query time
If traces spend a second getting a connection and 10ms running SQL, you do not have a slow query. You have a pool problem.
Config is a deploy
A TTL change is a release. Treat it like one: diff, canary, and an abort.
Common mistakes
- Scaling API boxes because latency is up. Check the pool and the cache first.
- Reading every channel before forming a hypothesis.
Interview tips
- Say the symptom in one sentence, then pick the cheapest signal that would falsify it.
- Name the blast radius and the rollback before you name the elegant fix.
Related challenges
Built by Al Beltran