[ClusterLabs] Antw: Re: Antw: Re: Antw: Re: Coming in Pacemaker 2.0.0: /var/log/pacemaker/pacemaker.log

[Ken Gaillot]

On Mon, 2018-01-15 at 12:40 +0000, Adam Spiers wrote:
> Ulrich Windl <Ulrich.Windl at rz.uni-regensburg.de> wrote:
> > > > > Vladislav Bogdanov <bubble at hoster-ok.com> schrieb:
> > > 
> > > 15.01.2018 11:23, Ulrich Windl wrote:
> > > > > > > Vladislav Bogdanov <bubble at hoster-ok.com> schrieb:
> > > > > 
> > > > > 11.01.2018 18:39, Ken Gaillot wrote:
> > > > > 
> > > > > [...]
> > > > > > > > I thought one option aired at the summit to address
> > > > > > > > this was
> > > > > > > > /var/log/clusterlabs, but it's entirely possible my
> > > > > > > > memory's
> > > > > > > > playing
> > > > > > > > tricks on me again.
> > > > > > 
> > > > > > I don't remember that, but it sounds like a good choice.
> > > > > > However we'd
> > > > > > still have the same issue of needing a single package to
> > > > > > own it.
> > > > > 
> > > > > In rpm world several packages may own a directory if it is
> > > > > consistently
> > > > > marked as '%dir' in a filelist.
> > > > 
> > > > Sure? I mean a package using a directory should include it as
> > > > %dir, but what
> > > 
> > > if multiple packages use the same dir with different owners,
> > > maybe?
> > > 
> > > Then they will conflict.
> > > 
> > > I just rechecked, creating two dummy packages owning one
> > > directory. If
> > > owner/mode matches, then packages are correctly installed and
> > > directory
> > > is reported to be owned by both.
> > > If there is mismatch, then rpm refuses to install.
> > 
> > But for a general solution, do you think it's more clean to have
> > the
> > same directory with identical properties in multiple packages, or
> > to
> > have one package that owns that directory?
> 
> This question is somewhat redundant in the context of this upstream
> mailing list, since each distribution will make that decision and
> choose a packaging strategy consistent with their own policies.  The
> important point was the previous one, which is that rpm would support
> and handle these approaches correctly, so it's not of sufficient
> concern per se to avoid /var/log/clusterlabs.  (I'm assuming that
> .deb
> does too, although that should be checked.)

Distributions can be stricter than what the rpm command accepts.

E.g. Fedora package review guidelines (and thus RHEL and derivatives):
"Packages must not own files or directories already owned by other
packages." (with rare exceptions)

Debian Policy (and thus derivatives): "Log files should usually be
named /var/log/package.log. If you have many log files, or need a
separate directory for permission reasons (/var/log is writable only by
root), you should usually create a directory named /var/log/package and
place your log files there." (again with rare exceptions)

It makes my life easier to stick with /var/log/pacemaker :-) because I
don't have to coordinate with all the other projects on the
/var/log/cluster vs /var/log/clusterlabs question, and it will work on
all distros without modification. But if there's a strong sentiment for
a common log directory, we still have time to do it.
-- 
Ken Gaillot <kgaillot at redhat.com>