🔒 Paid
Graceful Shutdown Drops In-Flight Requests
Every rolling deploy drops a steady trickle of checkout requests. Replica count looks fine, CPU looks fine — but pods are being killed before their in-flight requests finish, because the shutdown hook never waits for them to drain. Diagnose the lifecycle bug and choose a fix: a faster drain timeout, or a slower, safer deploy strategy.
- Difficulty
- Senior
- Expected Time
- ~15 min
- Category
- Infrastructure & Capacity
Skills
- Distinguishing a shutdown-lifecycle bug from a scaling/capacity problem
- Reading per-pod termination behavior instead of aggregate replica health
- Weighing drain-timeout vs. deploy-strategy tradeoffs (deploy speed vs. request loss)
- Recognizing that scaling out does not fix a per-instance lifecycle defect
System Architecture
- Deploy Controller
- Orders API
- Load Balancer
- Order Events Queue
This is a Pro scenario. Upgrade to Pro to unlock it.
Upgrade to ProBack to Incidents