Apache Drill
  1. Apache Drill
  2. DRILL-938

Simplifed SF100 TPCH query returns incorrect results

    Details

      Description

      This simplified query is dropping records. Should return 150M, returns 146,485,463.

      select
       count(*) 
      from
        customer c,
        orders o
      where
        c.c_custkey = o.o_custkey
      

        Activity

        Hide
        Aman Sinha added a comment -

        I was able to reproduce the wrong result on SF1 scale factor , max width set to 10. The issue was with resizing of the hash table. During resizing (which includes rehashing) the entries in the links vector were not set correctly if the hash chain spanned multiple batch holders. The attached patch has the fix for this. Note that I have tested it on SF1 scale factor only, not SF100.

        Show
        Aman Sinha added a comment - I was able to reproduce the wrong result on SF1 scale factor , max width set to 10. The issue was with resizing of the hash table. During resizing (which includes rehashing) the entries in the links vector were not set correctly if the hash chain spanned multiple batch holders. The attached patch has the fix for this. Note that I have tested it on SF1 scale factor only, not SF100.

          People

          • Assignee:
            DrillCommitter
            Reporter:
            Jacques Nadeau
          • Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development