Uploaded image for project: 'HPCC'
  1. HPCC
  2. HPCC-26961

Upgrade to 7.12, hash join stopped working - suspected memory corruption in LZ4



    • Bug
    • Status: New
    • Not specified
    • Resolution: Unresolved
    • 7.12.82
    • None
    • Thor
    • None
    • This is affecting our cert thor and dev thors on version 7.12.86


      We are testing a thor upgrade to 7.12, and started with 7.12.66.  Our QA team ran a build for regression and got an unexpected process termination on 3 consecutive tries.  At this point we recreated the issue on our dev thor, also on 7.12.66.  We upgraded to 7.12.86 and are still seeing the issue.

      In dev on 7.12.86 I was able to get the join to work by removing the hash keyword, but the join obviously took forever.  I also tried distributing the left and right manually and changing to local join, but that also failed with Process Termination.

      I ran the same WU on another of our thors that is still on 7.8.80 and the hash join worked and took 14 minutes. 

      Attached is a list of WUs, thor IPs, and screenshots showing the results described above.  I have zap reports too and can email them if needed.


        Issue Links



              mckellyln Mark Kelly
              lundpe01 Perek Lundquist
              0 Vote for this issue
              5 Start watching this issue