Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-30602 SPIP: Support push-based shuffle to improve shuffle efficiency
  3. SPARK-32923

Add support to properly handle different type of stage retries

    XMLWordPrintableJSON

    Details

    • Type: Sub-task
    • Status: Open
    • Priority: Major
    • Resolution: Unresolved
    • Affects Version/s: 3.1.0
    • Fix Version/s: None
    • Component/s: Shuffle, Spark Core
    • Labels:
      None

      Description

      In SPARK-23243 and SPARK-25341, the concept of an INDETERMINATE stage was introduced, which would be handled differently if retried.

      Since these was added to address a data correctness issue, we should also add support for these in push-based shuffle, so that we would be able to rollback the merged shuffle partitions of a shuffle map stage if it's an INDETERMINATE stage.

        Attachments

          Activity

            People

            • Assignee:
              Unassigned
              Reporter:
              mshen Min Shen
            • Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

              • Created:
                Updated: