Distributed Lock Failure

A fleet of workers coordinates exclusive access to a shared resource through a distributed lock service. A downstream dependency has slowed down, and individual jobs are now routinely taking longer than the configured lock lease. The shared resource must never apply two conflicting writes for the same unit of work — a duplicate or lost write here is a critical correctness incident, not a performance blip, and simply making the symptom less frequent is not the same as making the system safe. Diagnose why the same work is being processed more than once, and bring the system back to a state you can confidently hand back.

Difficulty

Staff

Expected Time

~15 min

Category

Distributed Systems