Docker and CI/CD · Senior · Question 23 of 75
Why do Kubernetes pods sometimes drop requests during a rolling update despite SIGTERM handling?
- ✓Correct answer: Endpoint removal lags SIGTERM; a short
preStopsleep covers it - 2Kubernetes sends SIGKILL first and only then sends SIGTERM
- 3Services keep routing to old pods until the deployment ends
Explanation
SIGTERM and removal from load balancers happen concurrently. If the app stops accepting connections immediately, in-flight routing still sends it traffic for a moment.