Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-27609

point-select test-case show regression with 4 numa

    XMLWordPrintable

Details

    • Bug
    • Status: Open (View Workflow)
    • Major
    • Resolution: Unresolved
    • 10.8.0
    • None
    • Server
    • None

    Description

      1. Start server on 4 numa nodes.
      2. TPS as reported by 10.6.6: 942187.16
      3. TPS as reported by 10.8: 685072.85
      4. Closer look suggested increased contention for MDL rwlock

      10.8:

      wait/synch/rwlock/sql/MDL_lock::rwlock 2269649.2516 483785033

      So per instance wait: 4.69

      10.6:

      wait/synch/rwlock/sql/MDL_lock::rwlock 883513.5605 658128191

      So per instance wait: 1.34

      10.8 rw-lock wait is 3.5 times more. Not sure what is causing the increased contention for MDL_lock. Don't see any other mutex contention on InnoDB side or other mutexes.

      10.8: 5595ed9d9f20aeb463ac26e075c9cba41927e85e
      10.6.6: 21778b8aa88280458f2196aab93044149cf0a18c

      Attachments

        Activity

          People

            Unassigned Unassigned
            krunalbauskar Krunal Bauskar
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:

              Git Integration

                Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.