dongjoon-hyun opened a new pull request, #893: URL: https://github.com/apache/spark-kubernetes-operator/pull/893
### What changes were proposed in this pull request? This PR aims to handle an evicted or deactivated Kueue `Workload` of a running `SparkCluster`, like SPARK-59800 does before the master is requested. - An evicted cluster enters `Suspended` and releases its master and workers like `spec.suspend`. Once the pods are gone and Kueue's requeue backoff elapses, its `Workload` is deleted and the cluster is queued again with a new `Workload`. - A deactivated `Workload` is kept, so the cluster stays `Suspended` until the `Workload` is reactivated. - A `PodsReadyTimeout` eviction is only reported by a new `KueueEvictionIgnored` warning event, since the operator does not report `PodsReady`. ### Why are the changes needed? Currently, a running `SparkCluster` ignores a Kueue eviction. Kueue keeps the quota of an evicted `Workload` until its integration releases it, so a preemption never completes. In contrast, a deactivated `Workload` is no longer counted, so the cluster runs outside of the quota. ### Does this PR introduce _any_ user-facing change? No for the released versions, because the Kueue integration is not released yet in v1.0.0. ### How was this patch tested? Pass the CIs with the newly added test cases. ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Opus 5.5 -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
