What should a Kubernetes health check actually check?
Design startup, readiness, and liveness endpoints around the action Kubernetes will take when each one fails.
Read postPractical notes on Kubernetes, cloud infrastructure, and the small details that keep systems working.
Browse the posts13 posts
No posts match your search.
Design startup, readiness, and liveness endpoints around the action Kubernetes will take when each one fails.
Read postRead scheduling failures, CPU throttling, and OOM kills as three different signals.
Read postWait for a rollout, test something a user would do, and know which revision is safe to restore.
Read postFollow the effective route table, the destination, and the return path one decision at a time.
Read postRecover a previous object to a scratch directory and verify its contents before touching the live key.
Read postUse a reviewable plan to catch replacements and understand what applying it might interrupt.
Read postPut dependency manifests ahead of frequently edited source, and check what reaches the final image.
Read postGive a CI build one path for reusable inputs and another for the exact output that gets deployed.
Read postBuild a p95 graph that combines replicas correctly and shows the traffic behind the number.
Read postFind the transaction holding up a query before reaching for a restart or a larger instance.
Read postAsk the authoritative server and your normal resolver to tell a stale answer from a bad record.
Read postMeasure a local wired transfer and a wireless transfer before blaming the internet connection.
Read postCapture a reversible change, a request that proves it works, and a clear trigger for going back.
Read post