For the runit-rpm project, I submitted a pull request.
The patch includes the following motivation:
>From systemd's viewpoint, the *service *is runsvdir, and not
runsvdir-start. The latter is simply a wrapper script for the former and
needed because of initttab's limitations. With systemd, a service file can
contain all the information in that script and more. Thus, the essence of
runsvdir-start is moved into the service file. Caveats:
* I'm not aware of a way to *completely empty *the enrivonment. Systemd
sets environment variables as described in systemd.exec(7)
* The distributed version of runsvdir-start contains a very strange and
IMO incorrect PATH. In general, sysadmins appreciate having PATH follow the
order:
sbin > bin
local > usr
usr > /
That is:
/usr/local/sbin:/usr/sbin:/sbin:/usr/local/bin:/usr/bin:/bin
An acceptable alternative is to regard sbin and bin as equals, thus:
/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
But I consider the previous one to be best-practice.
* For Linux specifically, OOMScoreAdjust is set to disable OOMKiller on
runsvdir. I think everyone sees the obvious necessity here.
* Finally, the meta-daemon should be restarted if it's ever stopped.
Quite possible this is not correct behavior, but in my view, if someone is
using systemd, they should use systemd to stop/start runsvdir, and not stop
it with signals.
--
Otheus
oth...@gmail.com
+43.699.1049.7813
