Uploaded image for project: 'Beam'
  1. Beam
  2. BEAM-6698

Portable Validates Runner Tests on Flink flaky after update to gradle5

Details

    • Bug
    • Status: Resolved
    • P3
    • Resolution: Fixed
    • None
    • Missing
    • test-failures
    • None

    Description

      After upgrade to gradle 5 [1], the two portable runner test projects on jenkins

      • beam_PostCommit_Java_PVR_Flink_Streaming [2]
      • beam_PostCommit_Java_PVR_Flink_Batch [3]

      became flaky.

      First investigation seems to point the tests to be failing on direct buffer memory (e.g. [4]) while staging files.

      Although I am unsure, whether this is really the root cause, or something that shows up after some other failure.

      INFO: Transport failed
      org.apache.beam.vendor.grpc.v1p13p1.io.netty.util.internal.OutOfDirectMemoryError: failed to allocate 16777216 byte(s) of direct memory (used: 1895825695, max: 1908932608)
      

      As far as I know, we do not set `-XX:MaxDirectMemorySize` anywhere in our setup, neither does gradle itself. At least on my machine both gradle 4 and gradle 5 stick to the same jvm default

      ###########################################
      sun.misc.VM.maxDirectMemory(): 1908932608 Bytes
      sun.misc.VM.maxDirectMemory(): 1820 MB
      ###########################################
      

       

      Unfortunately this does not reproduce on (my) local machine. We might try to workaround here by increasing ` -XX:MaxDirectMemorySize==3G` but this would probably only hide the problem? But might still be helpful to increase temporarily on branch just to be sure, that this is indeed the root cause?

      [1] https://issues.apache.org/jira/browse/BEAM-6630
      [2] https://builds.apache.org/job/beam_PostCommit_Java_PVR_Flink_Streaming/
      [3] https://builds.apache.org/job/beam_PostCommit_Java_PVR_Flink_Batch/
      [4] https://scans.gradle.com/s/tpo3yffjznfxa/tests/yobvrae4rwsg4-go44ti5iq45vq

       

      Attachments

        Issue Links

          Activity

            People

              Unassigned Unassigned
              michel Michael Luckey
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Time Tracking

                  Estimated:
                  Original Estimate - Not Specified
                  Not Specified
                  Remaining:
                  Remaining Estimate - 0h
                  0h
                  Logged:
                  Time Spent - 4h 10m
                  4h 10m

                  Slack

                    Issue deployment