Uploaded image for project: 'Hadoop Common'
  1. Hadoop Common
  2. HADOOP-3503

Race condition when client and namenode start block recovery simultaneously

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Closed
    • Major
    • Resolution: Fixed
    • 0.18.0
    • 0.18.0
    • None
    • None
    • Reviewed

    Description

      When a client detects a error while writing to a block, it starts the generation-stamp-protocol to remove stale replicas. At the same time, if the namenode experiences a lease expiry event for that file, the namenode starts the generation-stamp-protocol for the same block. Now, the client and thr namenode ping-pongs trying to stamp the block replicas. This ping-pong can continue for a long time.

      Attachments

        1. pingPong.patch
          19 kB
          Dhruba Borthakur
        2. pingPong2.patch
          22 kB
          Tsz-wo Sze

        Issue Links

          Activity

            People

              dhruba Dhruba Borthakur
              dhruba Dhruba Borthakur
              Votes:
              0 Vote for this issue
              Watchers:
              0 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: