Uploaded image for project: 'IronJacamar'
  1. IronJacamar
  2. JBJCA-1482

Performance breakdown after switching from Wildfly 26 to Wildfly 29.

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Major Major
    • None
    • 1.5.10.Final
    • Performance
    • None

      After switching from Wildfly 26 to Wildfly 29 we suffered a breakdown of the scalability of batches in our application running against DB2 and Oracle. PostgreSQL was not affected.

      Running a batch with a parallelity of 32 the performance deteriorated by 10%, when running with a parallelity of 128 the batches were hardly faster than with a parallelity of 32 and took between 3.5 to 4 times longer than usual.

      Eventmonitors from DB2 showed that the DB2 (shown in DB-Mon-26.png and DB-Mon-29.png) had to describe and prepare many more SQL-Statements than before.

      The Statement-Cache view of the Wildfly-console indicates that the reason is the different behaviour of the statement-cache which deletes a lot of SQL-Statements in Wildfly 29 while not deleting any in Wildfly 26. (See StmtCache_26.png and StmtCache_29.png).

      We detected Issue JBJCA-1461, downloaded the commit from GitHub and managed to compile the code with reverted changes. After patching the ironjacamar-jdbc-3.0.3.Final.jar module with the newly compiled class the behaviour of the Statement-Cache was back to what we are used to and the batch-times are also "normal" again.

        1. StmtCache_26.png
          StmtCache_26.png
          41 kB
        2. DB-Mon-29.png
          DB-Mon-29.png
          83 kB
        3. DB-Mon-26.png
          DB-Mon-26.png
          85 kB
        4. StmtCache_29.png
          StmtCache_29.png
          66 kB

              tadamski@redhat.com Tomasz Adamski
              moc00498 Andreas Penninger (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated: