Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-33260

Crash at startup when unclean shutdown

Details

    Description

      When unclean shutdown happens and recovery kicks in during launch, the server crashes.

      This happens on 10.11.6 version, version 10.11.5 is unaffected.

      Unfortunately I'm unable to produce backtrace, because quay.io/mariadb-foundation/mariadb-debug:10.11 image does not seem to contain correct debug symbols. Tried to post this to Zulip, so far without response. I have core dump but no debug symbols.

      I fixed the crash by downgrading to 10.11.5, doing clean shutdown and upgrading to 10.11.6 again. Then 10.11.6 correctly starts.

      Attachments

        Issue Links

          relates to

          Bug - A problem which impairs or prevents the functions of the product. MDBF-637 missing debuginfo packages due to reprepro

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          Bug - A problem which impairs or prevents the functions of the product. MDEV-26263 Investigate possible race on vc_templ recreation

          • Major - Major loss of function.
          • Open

          Activity

            Ascending order - Click to sort in descending order
            View 17 older comments
            cuchac Cuchac added a comment - - edited

            Hello,

            is it possible that this issue is also caused by MDEV-33275 ?

            I was able to reproduce this crash when removed `innodb_adaptive_flushing=OFF` from configuration file. When `innodb_adaptive_flushing=OFF` is set, this error does not happen.

            cuchac Cuchac added a comment - - edited Hello, is it possible that this issue is also caused by MDEV-33275 ? I was able to reproduce this crash when removed `innodb_adaptive_flushing=OFF` from configuration file. When `innodb_adaptive_flushing=OFF` is set, this error does not happen.
            cuchac Cuchac added a comment - - edited

            Hello,
            this is bad. I upgraded to 10.11.7 where MDEV-33275 should be fixed and the crash still happens. Also I was wrong and `innodb_adaptive_flushing=OFF` this time did not fix the issue.

            I have uploaded full stacktrace from 10.11.7

            cuchac Cuchac added a comment - - edited Hello, this is bad. I upgraded to 10.11.7 where MDEV-33275 should be fixed and the crash still happens. Also I was wrong and `innodb_adaptive_flushing=OFF` this time did not fix the issue. I have uploaded full stacktrace from 10.11.7
            nikitamalyavin Nikita Malyavin added a comment -

            cuchac Hello! Are you fine with changing the indexed fields in your table schema from GENERATED VIRTUAL to GENERATED STORED?
            Only those fields that you have indexes on.

            This can help, unless you are using long unique indexes, i.e. UNIQUE on TEXT fields, or on long enough VARCHARs.

            nikitamalyavin Nikita Malyavin added a comment - cuchac Hello! Are you fine with changing the indexed fields in your table schema from GENERATED VIRTUAL to GENERATED STORED? Only those fields that you have indexes on. This can help, unless you are using long unique indexes, i.e. UNIQUE on TEXT fields, or on long enough VARCHARs.
            cuchac Cuchac added a comment - - edited

            Hello,
            I switched to 10.11.5 to finish crash recovery and then back to 10.11.7 and it is working again.

            If I understand it correctly GENERATED VIRTUAL fields with index are known to be causing some errors and switching them to GENERATED STORED can help? Currently I have "downgrade" workaround so if there is a plan to fix this issue in the near future, I would like to avoid changing DB structure.

            cuchac Cuchac added a comment - - edited Hello, I switched to 10.11.5 to finish crash recovery and then back to 10.11.7 and it is working again. If I understand it correctly GENERATED VIRTUAL fields with index are known to be causing some errors and switching them to GENERATED STORED can help? Currently I have "downgrade" workaround so if there is a plan to fix this issue in the near future, I would like to avoid changing DB structure.
            ShiroDN Tomas Leypold added a comment -

            Hi,
            Is there anything else I can do to help debug this issue?

            ShiroDN Tomas Leypold added a comment - Hi, Is there anything else I can do to help debug this issue?

            People

              nikitamalyavin Nikita Malyavin
              cuchac Cuchac
              Votes:
              0 Vote for this issue
              Watchers:
              6 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.