Uploaded image for project: 'HPCC'
  1. HPCC
  2. HPCC-16751

Potential join crash on stop()


    • Type: Bug
    • Status: Resolved
    • Priority: Major
    • Resolution: Fixed
    • Affects Version/s: 6.0.4
    • Fix Version/s: 6.2.0
    • Component/s: Dali, Thor
    • Labels:
    • Environment:
      Linux - Centos 7 hosted within VMWare


      If an upstream activity called Join::stop() twice (which it shouldn't), it caused a crash.
      The particular case where this was spotted, was where a lightweight selfjoin called it's input stop() twice.
      It had to be a lightweight variety, where the code called through to the input stop() directly and then again in the slave activity base class.

      Description at time issue was reported:

      Here a short description :

      In our project, we are using SALT for internal linking and it generates several ECL script (For hygiene, specificities and iterate and so on … ) based on a definition file. We have faced on this issue during the step of the iteration processing (i.e BWR_ITERATE file execution). This ecl file “BWR_ITERATE” calls Proc_Iterate function which is a function generated by SALT. This function executes matches function to find the link between entities. The problem is that the function “matches” calls JOIN, HASH and distribute

      For example: A piece of code of the function “matches” :

      mj0 := JOIN( dn0_deduped, dn0_deduped, LEFT.entity_id > RIGHT.entity_id
      AND LEFT.name = RIGHT.name
      AND ( ~left.name_isnull AND ~right.name_isnull )
      ATMOST(LEFT.name = RIGHT.name,10000),HASH);

      I would like to know why you have jenkins during the iteration execution.

      0000018F 2016-12-09 15:57:26.102 12558 12558 "ERROR: 4: /var/lib/jenkins2/workspace/LN-Candidate-with-Plugins-6.0.8-1/LN/centos-7.0-x86_64/HPCC-Platform/thorlcr/graph/thgraphmaster.cpp(1418) : broadcastSendAsync : MP link closed ("

      Please find ecl files attached ,
      Please let me know if you require further information




            • Assignee:
              jakesmith Jake Smith
              chris.langridge@rbi.co.uk Chris Langridge
            • Votes:
              0 Vote for this issue
              4 Start watching this issue


              • Created: