Ground truth. Not guesswork.
Production broke.
Find out why.
Train your incident instincts on simulated production failures. Follow the evidence, uncover the root cause, and become the engineer who sees past the symptoms.
A practice incident. No signup. No production risk.
Our ground rules
The detective’s toolkit
Less finger-pointing.
More fingerprints.
A red dashboard tells you something is wrong. A good investigation tells you what happened, why it happened, and what would prevent it next time.
Put your reasoning to the test- 01
Read the scene
Start with the incident timeline, request symptoms, and system clues. Separate what you know from what you assume.
- 02
Build a defensible theory
Connect the evidence across service boundaries. Find the failure mechanism, not just the last error in the chain.
- 03
Check your conclusion
Compare your explanation with the case debrief. Challenge your assumptions and sharpen your next investigation.
Your first investigation
The database looks guilty.
Is it?
Try this short, simulated case. Read the clues, form your hypothesis, then open the debrief. The aim isn’t speed. It’s a conclusion that holds up.
Case file 001
The connection that never came back
14:02A checkout release adds a new validation branch.
14:07Request latency climbs. Database CPU stays flat; queries remain fast.
14:10All 20 application database connections are checked out. New requests wait for a connection.
14:12Invalid checkouts return early after acquiring a connection, without releasing it.
What is the root cause, and what would you fix?
Reveal the case debrief
The application is leaking connections. The new early-return path skips cleanup. The pool eventually runs out, so requests time out while waiting—not while executing slow queries.
Release connections in a guaranteed cleanup path, such as a finally block, and test the invalid-checkout branch. Rolling back can stop the leak; restarting may clear the pool temporarily but does not fix the code.
Detective’s takeaway: the resource under pressure isn’t always the source of the failure.
Practice here. Take the mindset to production.
Every failed request
has a story.
RAT - Request Autopsy Tool starts with one habit: investigate before you guess. Explore the free case and see how evidence-led thinking changes your diagnosis.