dongjoon-hyun commented on PR #895: URL: https://github.com/apache/spark-kubernetes-operator/pull/895#issuecomment-5864904805
Thank you so much for the thorough review, @MaxGekk. I addressed the comments in ea8b9997. - **Catch-all for a spec failure**: The catch is narrowed to the spec build of the API lookup, which runs only when the informer cache has no driver, and only for a suspended application. A spec failure for a cached driver propagates again like before this PR and `ClusterInitStep`, and a non-suspended application still gets `SchedulingFailure`. New tests pin both cases. - **Repeated lookup**: `driverRequested` is computed once in the suspend branch and passed to `holdForKueueAdmission`, so the driver found there is not looked up again. - **Duplicated driver check**: `SparkAppContext.isLiveDriverPod` is now shared by the cache and API server checks of `getCurrentAttemptDriverPod()` and by the lookup of `isDriverRequested`. The stale `Workload` cache and the side effects of building the driver spec also apply to the existing lookups of SPARK-59800 and `getCurrentAttemptDriverPod()`. I'll handle them together in a follow-up JIRA by looking up the driver by a name derived without building the spec. -- 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]
