Uploaded image for project: 'Hadoop YARN'
  1. Hadoop YARN
  2. YARN-5483

Optimize RMAppAttempt#pullJustFinishedContainers

    XMLWordPrintableJSON

Details

    • Improvement
    • Status: Closed
    • Major
    • Resolution: Fixed
    • 2.6.0
    • 2.8.0, 2.6.5, 2.7.4, 3.0.0-alpha1
    • None
    • None
    • Reviewed

    Description

      about 1000 app running on cluster, jprofiler found pullJustFinishedContainers cost too much cpu.

      Attachments

        1. YARN-5483-branch-2.7.patch.02
          5 kB
          sandflee
        2. YARN-5483-branch-2.6.patch.02
          3 kB
          sandflee
        3. YARN-5483-branch-2.6.patch
          3 kB
          sandflee
        4. YARN-5483-branch-2.7.patch
          5 kB
          sandflee
        5. YARN-5483.04.patch
          6 kB
          sandflee
        6. YARN-5483.03.patch
          5 kB
          sandflee
        7. YARN-5483.02.patch
          5 kB
          sandflee
        8. YARN-5483.01.patch
          2 kB
          sandflee
        9. jprofiler-cpu.png
          40 kB
          sandflee

        Issue Links

          fixes

          Bug - A problem which impairs or prevents the functions of the product. YARN-4741 RM is flooded with RMNodeFinishedContainersPulledByAMEvents in the async dispatcher event queue

          • Critical - Crashes, loss of data, severe memory leak.
          • Resolved
          relates to

          Sub-task - The sub-task of the issue YARN-1372 Ensure all completed containers are reported to the AMs across RM restart

          • Major - Major loss of function.
          • Closed

          Activity

            People

              sandflee sandflee
              sandflee sandflee
              Votes:
              0 Vote for this issue
              Watchers:
              10 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: