Uploaded image for project: 'Hadoop YARN'
  1. Hadoop YARN
  2. YARN-9235

If linux container executor is not set for a GPU cluster GpuResourceHandlerImpl is not initialized and NPE is thrown

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • 3.0.0, 3.1.0
    • 3.3.0, 3.2.1, 3.1.3
    • yarn
    • None
    • Reviewed

    Description

      If GPU plugin is enabled for the NodeManager, it is possible to run jobs with GPU.

      However, if LinuxContainerExecutor is not configured, an NPE is thrown when calling 

      GpuResourcePlugin.getNMResourceInfo

      Also, there are no warns in the log if GPU is misconfigured like this. 

      Attachments

        1. YARN-9235.001.patch
          3 kB
          Antal Bálint Steinbach
        2. YARN-9235.002.patch
          6 kB
          Antal Bálint Steinbach
        3. YARN-9235.003.patch
          6 kB
          Antal Bálint Steinbach
        4. YARN-9235.004.patch
          6 kB
          Szilard Nemeth
        5. YARN-9235.004.patch
          6 kB
          Antal Bálint Steinbach
        6. YARN-9235.branch-3.1.001.patch
          6 kB
          Szilard Nemeth
        7. YARN-9235.branch-3.2.001.patch
          6 kB
          Szilard Nemeth

        Issue Links

          depends upon

          Improvement - An improvement or enhancement to an existing feature or task. YARN-9121 Replace GpuDiscoverer.getInstance() to a readable object for easy access control

          • Major - Major loss of function.
          • Resolved

          Activity

            People

              adam.antal Adam Antal
              bsteinbach Antal Bálint Steinbach
              Votes:
              0 Vote for this issue
              Watchers:
              8 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: