You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@flink.apache.org by "Gyula Fora (Jira)" <ji...@apache.org> on 2024/01/30 13:44:00 UTC

[jira] [Created] (FLINK-34311) Do not change min resource requirements when rescaling for adaptive scheduler

Gyula Fora created FLINK-34311:
----------------------------------

             Summary: Do not change min resource requirements when rescaling for adaptive scheduler
                 Key: FLINK-34311
                 URL: https://issues.apache.org/jira/browse/FLINK-34311
             Project: Flink
          Issue Type: Improvement
          Components: Kubernetes Operator
            Reporter: Gyula Fora
            Assignee: Gyula Fora
             Fix For: kubernetes-operator-1.8.0


when applying the rescale api to change parallelism we should not change the min parallelism.
The problem currently is that if we cannot aquire the new resources within {{jobmanager.adaptive-scheduler.resource-wait-timeout}} the job will completely fail

The {{jobmanager.adaptive-scheduler.resource-stabilization-timeout}} still allows us to wait for quite long if necessary to get the target parallelism but failing completely because of the wait timeout seems very unfortunate

It's best to keep the min resources unchanged and let the adaptive scheduler take care of the parallelism changes together with the timeout settings.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)