Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-13125

Undocumented change: Core dumps can now be enabled dynamically

Details

    • Bug
    • Status: Closed (View Workflow)
    • Minor
    • Resolution: Not a Bug
    • None
    • N/A
    • Server
    • None

    Description

      Hi,

      In MariaDB 10.1.23 and 10.2.6, core dumps can be enabled by changing the process limit:

      prlimit --core=unlimited: --pid `pidof mysqld`
      

      In MariaDB 10.1.19 this required a restart.

      Way to test:
      1. Install the repository
      2. Install 10.1.23
      3. change the process limit
      4. kill -SIGSEGV `pidof mysqld`
      5. Observer there is a core dump

      Do the same with MariaDB 10.1.19 and observe there is no core dump.

      This change was undocumented but it is very useful desired behaviour. Please add this change to the release notes and/or find out what changed it.

      Attachments

        Issue Links

          Activity

            I think this change may be caused by the same change that caused this bug.

            michaeldg Michaël de groot added a comment - I think this change may be caused by the same change that caused this bug.

            Are you sure it was a change in the server? May be something in your environment?
            If the server — in what version exactly did it occur?

            serg Sergei Golubchik added a comment - Are you sure it was a change in the server? May be something in your environment? If the server — in what version exactly did it occur?
            elenst Elena Stepanova added a comment - - edited

            The change in behavior happened in 10.1.20 with this revision:

            commit 76546a099c24a06ad89238f61435051a4663b112
            Author: Sergei Golubchik <serg@mariadb.org>
            Date:   Mon Dec 5 15:51:24 2016 +0100
             
                MDEV-10382 Using systemd, mariadb doesn't restart on crashes
                
                when crashing on a signal, don't exit(), but re-signal it, so that
                the caller could check WIFSIGNALED()
            

            Assigning to serg to confirm it was intentional and will stick, as it does appear to be convenient.

            elenst Elena Stepanova added a comment - - edited The change in behavior happened in 10.1.20 with this revision: commit 76546a099c24a06ad89238f61435051a4663b112 Author: Sergei Golubchik <serg@mariadb.org> Date: Mon Dec 5 15:51:24 2016 +0100   MDEV-10382 Using systemd, mariadb doesn't restart on crashes when crashing on a signal, don't exit(), but re-signal it, so that the caller could check WIFSIGNALED() Assigning to serg to confirm it was intentional and will stick, as it does appear to be convenient.

            Very convenient! Please make this change documented, serg

            michaeldg Michaël de groot added a comment - Very convenient! Please make this change documented, serg

            This was a side effect. The goal of the commit was to properly pass the killing signal information to the parent process. Apparently this fixed coredumps too.

            serg Sergei Golubchik added a comment - This was a side effect. The goal of the commit was to properly pass the killing signal information to the parent process. Apparently this fixed coredumps too.

            People

              serg Sergei Golubchik
              michaeldg Michaël de groot
              Votes:
              0 Vote for this issue
              Watchers:
              3 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.