Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-36042

[Dynamic allocation] Executor grace period (ExecutorIdleTimeout) ignored due to nulll startTime for pods in pending state

    XMLWordPrintableJSON

    Details

    • Type: Bug
    • Status: Open
    • Priority: Major
    • Resolution: Unresolved
    • Affects Version/s: 3.1.1
    • Fix Version/s: None
    • Component/s: Kubernetes
    • Labels:
      None
    • Environment:

      AWS EKS with dynamic allocation 

      Description

      Pending executor are always timeouted due to null startTime and funtion returning true in case of exception in parsing startTime.

       

      In class ExecutorPodsAllocator:

      private def isExecutorIdleTimedOut(state: ExecutorPodState, currentTime: Long): Boolean = {
      try {
       val startTime = Instant.parse(state.pod.getStatus.getStartTime).toEpochMilli()
       currentTime - startTime > executorIdleTimeout
      catch {
        case _: Exception =>
        logDebug(s"Cannot get startTime of pod ${state.pod}")}}
        true
      }}

        Attachments

          Activity

            People

            • Assignee:
              Unassigned
              Reporter:
              apclement Alexandre CLEMENT
            • Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

              Dates

              • Created:
                Updated: