Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-23974

main.flush_block_commit_notembedded failed on powerpc after 10.5.6 upload to Debian

Details

    Description

      After uploading MariaDB 10.5.6 I noticed at https://buildd.debian.org/status/package.php?p=mariadb-10.5 the following failure for arch powerpc:

      main.flush_block_commit_notembedded 'innodb' w8 [ fail ]  Found warnings/errors in server log file!
              Test ended at 2020-10-16 01:18:07
      line
      2020-10-16  1:17:59 6 [Warning] InnoDB: Trying to delete tablespace 'test/t1' but there are 1 flushes and 0 pending i/o's on it.
      ^ Found warnings in /<<PKGBUILDDIR>>/builddir/mysql-test/var/8/log/mysqld.1.err
      ok
       
       - saving '/<<PKGBUILDDIR>>/builddir/mysql-test/var/8/log/main.flush_block_commit_notembedded-innodb/' to '/<<PKGBUILDDIR>>/builddir/mysql-test/var/log/main.flush_block_commit_notembedded-innodb/'
      

      Full log at https://buildd.debian.org/status/fetch.php?pkg=mariadb-10.5&arch=powerpc&ver=1%3A10.5.6-1&stamp=1602811263&raw=0

      Unfortunately there is no more info available in the log.

      Attachments

        Issue Links

          Activity

            otto Otto Kekäläinen added a comment - Filed downstream as https://bugs.debian.org/972350

            This and a few other tests are executing heavy buffer pool flushing, and the warning is possible on slower systems. I added some suppressions.

            marko Marko Mäkelä added a comment - This and a few other tests are executing heavy buffer pool flushing, and the warning is possible on slower systems. I added some suppressions .

            This is potentially related to MDEV-27955

            otto Otto Kekäläinen added a comment - This is potentially related to MDEV-27955

            otto, the only relation that I can identify with MDEV-27955 is that a regression test was badly written, or it assumed that more resources are available for running the test. Did you mean that, or did you actually mean some other ticket?

            A follow-up fix was needed. I was unaware that call mtr.add_suppression() causes something to be written to the binlog.

            marko Marko Mäkelä added a comment - otto , the only relation that I can identify with MDEV-27955 is that a regression test was badly written, or it assumed that more resources are available for running the test. Did you mean that, or did you actually mean some other ticket? A follow-up fix was needed. I was unaware that call mtr.add_suppression() causes something to be written to the binlog.

            Did you mean that, or did you actually mean some other ticket?

            Yes I meant it, Thanks for the answer, all good.

            otto Otto Kekäläinen added a comment - Did you mean that, or did you actually mean some other ticket? Yes I meant it, Thanks for the answer, all good.

            This had been fixed by suppressing the warning on MemorySanitizer builds and Valgrind instrumented test runs. But, as noted in MDEV-33599, occasionally (once in several months) this would fail on normal debug instrumented builds as well:

            10.5 9b18275623026ccfbffb348d56cb629b6255f574

            innodb.innodb-64k-crash 'innodb'         w2 [ fail ]  Found warnings/errors in server log file!
                    Test ended at 2024-04-16 14:04:02
            line
            2024-04-16 14:03:37 3 [Warning] InnoDB: Trying to delete tablespace 'test/t1' but there are 16 pending operations on it.
            ^ Found warnings in /home/buildbot/amd64-ubuntu-2004-fulltest/build/mysql-test/var/2/log/mysqld.1.err
            ok
             - saving '/home/buildbot/amd64-ubuntu-2004-fulltest/build/mysql-test/var/2/log/innodb.innodb-64k-crash-innodb/' to '/home/buildbot/amd64-ubuntu-2004-fulltest/build/mysql-test/var/log/innodb.innodb-64k-crash-innodb/'
            

            I would not suppress this warning on tests that are run against release builds, because it could be a sign of serious performance regression (some new bottleneck in the I/O subsystem).

            marko Marko Mäkelä added a comment - This had been fixed by suppressing the warning on MemorySanitizer builds and Valgrind instrumented test runs. But, as noted in MDEV-33599 , occasionally (once in several months) this would fail on normal debug instrumented builds as well: 10.5 9b18275623026ccfbffb348d56cb629b6255f574 innodb.innodb-64k-crash 'innodb' w2 [ fail ] Found warnings/errors in server log file! Test ended at 2024-04-16 14:04:02 line 2024-04-16 14:03:37 3 [Warning] InnoDB: Trying to delete tablespace 'test/t1' but there are 16 pending operations on it. ^ Found warnings in /home/buildbot/amd64-ubuntu-2004-fulltest/build/mysql-test/var/2/log/mysqld.1.err ok - saving '/home/buildbot/amd64-ubuntu-2004-fulltest/build/mysql-test/var/2/log/innodb.innodb-64k-crash-innodb/' to '/home/buildbot/amd64-ubuntu-2004-fulltest/build/mysql-test/var/log/innodb.innodb-64k-crash-innodb/' I would not suppress this warning on tests that are run against release builds, because it could be a sign of serious performance regression (some new bottleneck in the I/O subsystem).

            People

              marko Marko Mäkelä
              otto Otto Kekäläinen
              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.