gauravkm commented on PR #3854: URL: https://github.com/apache/celeborn/pull/3854#issuecomment-5747922449
I found two P1 issues in the default probe configuration: **[P1] The master startup probe deadlocks a fresh default installation** (`charts/celeborn/values.yaml:137–143`). With three replicas and `OrderedReady`, master-0 waits for all master DNS records before starting the HTTP server. A startup probe also prevents the container becoming Ready until it succeeds, even without a readiness probe ([kubelet behavior](https://github.com/kubernetes/kubernetes/blob/v1.34.0/pkg/kubelet/prober/prober_manager.go#L270-L316)). Master-1/2 therefore never get created, and master-0 keeps restarting. `publishNotReadyAddresses` cannot publish records for pods that do not exist. Please reconcile the bootstrap/probe defaults while preserving StatefulSet upgrade compatibility, and add a fresh-install test with the default three replicas. Increasing the timeout does not break this cycle. **[P1] The enabled probes require an endpoint missing from the default image** (`charts/celeborn/values.yaml:137–140, 326–329`). The rendered manifests still select `docker.io/apache/celeborn:0.7.0` through `Chart.yaml`'s `appVersion`, but the [0.7.0 HTTP resources](https://github.com/apache/celeborn/tree/v0.7.0/service/src/main/scala/org/apache/celeborn/server/common/http/api) do not implement `/healthz`. Even after resolving the ordering issue, these probes cannot succeed against that release: masters fail startup probes and workers remain NotReady. Please align the image/probe defaults or keep probes opt-in for unsupported releases, and validate against the actual default image. I ran Helm lint and all 92 chart tests successfully and inspected the rendered manifests. These findings are based on source/configuration checks; I did not run a live Kubernetes deployment. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
