Comment 2 for bug 1643439

No, I think this change to the /usr/libexec/mysqld is wrong for several reasons:

- There is no guarantee that mysqld lives in /usr/libexec on all systems
- There is no guarantee that the startup command is the same on all systems
- It is not OK to bypass systemd or any other init system, since they do a lot more than simply execute the start command (e.g. handling of dependencies and pid files)

Having said that, whilst the proposed fix is wrong, it is probably still correct to report that "systemctl start mariadb --wsrep-new-cluster" fails. So we should not simply close this bug. But it would have been more helpful if the bug report gave the exact error. I expect the problem is that systemctl tries to handle the --wsrep-new-cluster option which should not be there, but I don't have a system to test on right now, so I can't be sure.

Another related question is whether all Linux distributions commonly used for OpenStack now use systemd. I think they do, but I'm not sure. This would determine whether it's OK to use systemctl in OpenStack documentation or whether something more generic like service(8) should be used.