Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-24284

InnoDB: Record field 64 len 18446744...

Details

    Description

      We have some mariadb 10.3 instances where the server has log entries similar to below:

      [ERROR] InnoDB: Record field 64 len 18446744073709544479
      

      Maybe few times per day at most, I haven't found what the exact trigger is

      The value of "record field" 64 may vary, but the "len" is always this 18446744073709544479.

      We haven't noticed any operational issues on the servers so far but still interested to learn if this is something to look into deeper
      For example, running CHECK TABLE against all tables doesn't come up with his message, either.

      I don't really have much information on when/how these might have began to happen.
      But what seems to be somewhat related is that the servers where this happens are part of galera cluster, single instance servers do not have these entries.

      Attachments

        Issue Links

          Activity

            mreibert Mark Reibert added a comment -

            marko — thank you for the additional information and thoughts. For my large (6TB) Galera cluster I can wipe a node and let it recreate the DB from scratch via a SST, but it seems I should first upgrade to 10.4.18 (I am currently on 10.4.14) on account of MDEV-24449.

            mreibert Mark Reibert added a comment - marko — thank you for the additional information and thoughts. For my large (6TB) Galera cluster I can wipe a node and let it recreate the DB from scratch via a SST, but it seems I should first upgrade to 10.4.18 (I am currently on 10.4.14) on account of MDEV-24449 .

            Yes, you can recreate DB from scratch via a SST but do not use rsync or mariabackup methods, you need to use mysqldump method.

            jplindst Jan Lindström (Inactive) added a comment - Yes, you can recreate DB from scratch via a SST but do not use rsync or mariabackup methods, you need to use mysqldump method.
            mreibert Mark Reibert added a comment -

            Hello jplindst — Unfortunately with a 6TB database mariabackup is my only practical option given the 6 hours it takes for a SST and the other methods blocking the donor during that time.

            mreibert Mark Reibert added a comment - Hello jplindst — Unfortunately with a 6TB database mariabackup is my only practical option given the 6 hours it takes for a SST and the other methods blocking the donor during that time.

            10.3 will EOL soon.

            janlindstrom Jan Lindström added a comment - 10.3 will EOL soon.

            I think that this corruption should affect MariaDB 10.4 as well. The root cause could be a hard-to-reproduce InnoDB crash recovery failure. In 10.5, the recovery was rewritten as part of MDEV-12353 and many related changes.

            marko Marko Mäkelä added a comment - I think that this corruption should affect MariaDB 10.4 as well. The root cause could be a hard-to-reproduce InnoDB crash recovery failure. In 10.5, the recovery was rewritten as part of MDEV-12353 and many related changes.

            People

              jplindst Jan Lindström (Inactive)
              rlam Roope Pääkkönen (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              5 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.