Yikun edited a comment on pull request #35640: URL: https://github.com/apache/spark/pull/35640#issuecomment-1061267316
@dongjoon-hyun I'm not sure there were some misunderstanding, something below might help you understand if something confused. We configure pod (driver and executors) one by one but we only return the PreAdditionalK8SResource once before driver creation. So there are only one podgroup is created for one job in current implementation（the pre kubernetes only calls in driver side). If we are in the same page then: > It seems that you decided not to use volcano API for some reasons No more hidden reason, just make sure it can be more generic, then it can be created by [kubernetesClient.resourceList(preKubernetesResources: _*).createOrReplace()]( https://github.com/apache/spark/blob/master/resource-managers/kubernetes/core/src/main/scala/org/apache/spark/deploy/k8s/submit/KubernetesClientApplication.scala#L140-L140). So we only return volcano model in here. > Can we use new DefaultVolcanoClient().podGroups().load()? if you meaned use this to build volcano-model, client side yaml load could be a alternative way to complete this, but I think we still need to left placeholder in yaml and do format in feature step. This perhaps not flexiable, and a little bit hard to maintain. and also introduce a volcano client module deps. :), feel free to left any question you had, I will try my best to make work better. -- 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]
