Comment 6 for bug 1820325

Reviewed: https://review.openstack.org/644250
Committed: https://git.openstack.org/cgit/openstack/kolla-ansible/commit/?id=b25c0ee4772fdf04fe6f263a8e81571204f706ff
Submitter: Zuul
Branch: master

commit b25c0ee4772fdf04fe6f263a8e81571204f706ff
Author: Mark Goddard <email address hidden>
Date: Sun Mar 17 17:48:54 2019 +0000

    Fix MariaDB 10.3 upgrade

    Upgrading MariaDB from Rocky to Stein currently fails, with the new
    container left continually restarting. The problem is that the Rocky
    container does not shutdown cleanly, leaving behind state that the new
    container cannot recover. The container does not shutdown cleanly
    because we run dumb-init with a --single-child argument, causing it to
    forward signals to only the process executed by dumb-init. In our case
    this is mysqld_safe, which ignores various signals, including SIGTERM.
    After a (default 10 second) timeout, Docker then kills the container.

    A Kolla change [1] removes the --single-child argument from dumb-init
    for the MariaDB container, however we still need to support upgrading
    from Rocky images that don't have this change. To do that, we add new
    handlers to execute 'mysqladmin shutdown' to cleanly shutdown the
    service.

    A second issue with the current upgrade approach is that we don't
    execute mysql_upgrade after starting the new service. This can leave the
    database state using the format of the previous release. This patch also
    adds handlers to execute mysql_upgrade.

    [1] https://review.openstack.org/644244

    Depends-On: https://review.openstack.org/644244
    Depends-On: https://review.openstack.org/645990
    Change-Id: I08a655a359ff9cfa79043f2166dca59199c7d67f
    Closes-Bug: #1820325