Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-20624 SPIP: Add better handling for node shutdown
  3. SPARK-39200

Stream is corrupted Exception while fetching the blocks from fallback storage system

    XMLWordPrintableJSON

Details

    • Sub-task
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • 3.2.0
    • 3.3.1, 3.2.3, 3.4.0
    • Shuffle
    • None

    Description

      When executor decommissioning and fallback storage is enabled - the shuffle reads are failing with `FetchFailedException: Stream is corrupted` 

      ref: https://issues.apache.org/jira/browse/SPARK-18105 (search for decommission)

       

      This is happening when the shuffle block is bigger than `inputstream.read` can read in one attempt. The code path is not reading the block fully (`readFully`) and the partial read is causing the exception.

      Attachments

        Activity

          People

            ukby1234 Frank Yin
            rajendrag Rajendra Gujja
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: