Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-24036 Stateful operators in continuous processing
  3. SPARK-24236

continuous replacement for ShuffleExchangeExec

    XMLWordPrintableJSON

Details

    • Sub-task
    • Status: Resolved
    • Major
    • Resolution: Incomplete
    • 2.4.0
    • None
    • Structured Streaming

    Description

      Two major differences from ShuffleExchangeExec:

      • The post-shuffle RDD should contain all the tasks, rather than reporting scheduler-level dependencies.
      • The operator will have to manually register the continuous shuffle dependencies, since they're not reported to the scheduler.

      Attachments

        Activity

          People

            Unassigned Unassigned
            joseph.torres Jose Torres
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: