Purushotham Pushpavanthar created SPARK-61268:
-------------------------------------------------

             Summary: Track Unschedulable executor pods as a distinct pod state 
and log the scheduler reason
                 Key: SPARK-61268
                 URL: https://issues.apache.org/jira/browse/SPARK-61268
             Project: Spark
          Issue Type: Improvement
          Components: Kubernetes
    Affects Versions: 5.0.0
            Reporter: Purushotham Pushpavanthar


ExecutorPodsSnapshot maps every Pending pod to PodPending and ignores the 
PodScheduled=False, reason=Unschedulable condition, so a pod that cannot be 
placed looks identical to one about to start and the reason is only visible via 
kubectl describe. Add a PodUnschedulable state (a sub-state of pending for all 
existing counting) carrying the scheduler's message, and have 
ExecutorPodsAllocator log it once per pod. Prerequisite for SPARK-61258 but 
useful on its own.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to