Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-14804

innodb.update_time failed in buildbot with wrong result

Details

    Description

      http://buildbot.askmonty.org/buildbot/builders/bld-starfs-release/builds/577/steps/test/logs/stdio

      innodb.update_time 'innodb'              w2 [ fail ]
      		 
              Test ended at 2017-12-16 10:01:56
      		 
       
      		 
      CURRENT_TEST: innodb.update_time
      		 
      --- /opt/buildbot-slave/mariadb/starfsmenn2/build/mysql-test/suite/innodb/r/update_time.result	2017-12-14 18:23:31.087061690 -0500
      		 
      +++ /opt/buildbot-slave/mariadb/starfsmenn2/build/mysql-test/suite/innodb/r/update_time.reject	2017-12-16 10:01:56.799233773 -0500
      		 
      @@ -29,7 +29,7 @@
      		 
       SELECT COUNT(*) FROM information_schema.innodb_buffer_page
      		 
       WHERE table_name = '`test`.`t`';
      		 
       COUNT(*)
      		 
      -0
      		 
      +1
      		 
       SELECT COUNT(*) FROM information_schema.tables WHERE table_name = 't'
      		 
       AND update_time IS NOT NULL;
      		 
       COUNT(*)
      		 
       
      		 
      mysqltest: Result content mismatch

      Attachments

        Activity

          Ascending order - Click to sort in descending order
          marko Marko Mäkelä added a comment -

          The test contains an inherent race condition. If page flushing proceeds quicker than expected, then not all pages of the table might be evicted. This was rather easily repeatable in a 10.6-based branch that included the page flushing performance fix MDEV-26827.

          I tried to make the failure less probable by halving the buffer pool to the minimal size (innodb_buffer_pool_size=5M).

          marko Marko Mäkelä added a comment - The test contains an inherent race condition. If page flushing proceeds quicker than expected, then not all pages of the table might be evicted. This was rather easily repeatable in a 10.6-based branch that included the page flushing performance fix MDEV-26827 . I tried to make the failure less probable by halving the buffer pool to the minimal size ( innodb_buffer_pool_size=5M ).
          marko Marko Mäkelä added a comment -

          According to monty, this test still fails.

          On a closer look, I do not see the point of this test. The update_time is stored in the table metadata (dict_table_t); it has nothing to do with buffer pool page eviction or replacement. There is another test innodb.update_time_wl6658 that seems much more useful than this one.

          marko Marko Mäkelä added a comment - According to monty , this test still fails. On a closer look, I do not see the point of this test. The update_time is stored in the table metadata ( dict_table_t ); it has nothing to do with buffer pool page eviction or replacement. There is another test innodb.update_time_wl6658 that seems much more useful than this one.
          marko Marko Mäkelä added a comment -

          I removed the buffer pool related test. I retained a test that covers XA COMMIT after server restart.

          marko Marko Mäkelä added a comment - I removed the buffer pool related test . I retained a test that covers XA COMMIT after server restart.

          People

            marko Marko Mäkelä
            elenst Elena Stepanova
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Git Integration

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