Bug#796625: [PATCH] clvm: ship native systemd units in debian package

[Andreas Henriksson]

Hello Bastian Blank.

Thanks for finally giving some feedback on this bug report.

On Tue, Jun 14, 2016 at 07:47:56PM +0200, Bastian Blank wrote:
> On Mon, Jun 13, 2016 at 01:31:49PM +0200, Andreas Henriksson wrote:
> > Hope this helps. Would be much appreciated if we could resolve this
> > bug report (very) soon. I'd like to avoid NMUing this package but
> > if progress is blocked on this bug report I'll go ahead if there's
> > no feedback by then.
> 
> Did you at least test it once?  My gut say: no.

Then your gut is misleading you. I've tested this as best as I could.
I don't have access to a cluster so could not fully test it
but I've done my best to verify it should run thing similarly
to how the init script starts things. I was hoping that was
not completely broken but maybe I made a mistake.

> 
> > --- lvm2-2.02.153/debian/patches/clvm-systemd-unit-debian-adaptions.patch	1970-01-01 01:00:00.000000000 +0100
> > +++ lvm2-2.02.153/debian/patches/clvm-systemd-unit-debian-adaptions.patch	2016-06-13 13:05:16.000000000 +0200
> 
> There is already a patch for systemd modifications, use it.

Could you please give some details? I've obviously not found it
otherwise I wouldn't have wasted so much of my time on this.

Why not use the upstream units? Why has this patch not been
referenced in this bug report? How are people going to
find out about it if we don't use the bug tracking system
to record information on what's going on?

Please feel free to post regular (weekly?) status updates
to this bug report about progress you're making on this issue
including all the details that could help avoid duplicate work
and wasted time.

> 
> > +--- a/scripts/lvm2_clvmd_systemd_red_hat.service.in
> > ++++ b/scripts/lvm2_clvmd_systemd_red_hat.service.in
> > +@@ -1,9 +1,10 @@
> > + [Unit]
> > + Description=Clustered LVM daemon
> > + Documentation=man:clvmd(8)
> > +-After=dlm.service corosync.service
> > ++After=cman.service corosync.service
> 
> Where did you find a cman.service?  The package does not even build cman
> support.

I found cman in the (Debian) init script. It's referenced under
Should-Start. This LSB header maps to the After= and Wants= systemd
unit directives. This is a loose "best effort" relationship, contrary
to the Requires-Start LSB header which would map to the stricter
"Requires=" (and After=) systemd unit directives which actually requires
the unit to successfully start up.
As the aim is to mask the init script with a matching service it should
really describe basically the same things.

If you have specific pointers about what's wrong I'd like to hear them,
otherwise I can't really do much else than trust whats written in the
init script. I have no better description of what the service really
wants and how to write its dependencies/relationships than that.

As was mentioned in the initial bug report there are people willing
to help you out if you reach out and describe what it is that your
service actually needs/wants/depends-on/provides, but lacking that
information the best we can do to help is to try to interpret your
init script and assume that's correct (but there are several signs
of it not being, including lintian even detecting outstanding issues
automatically, but without any other information it's not easy to
try to help you out. At the current state I don't see much I can
do at all to improve the situation further so rely on your cooperation.
An alternative approach would be to simply drop building the clvm
binary from src:lvm2 for now given it seems, which you already
acknowledged, need a complete overhaul to get in shape and restore
it once there's someone interested in giving it the maintenance
work it needs/lacked).

Fwiw, This bug in clvm is actually breaking users systems and gets
"random" services removed from the startup breaking their systems. I'm
inclined to raise the severity to grave because of this.
from upstream which did not have the issue at all).

Regards,
Andreas Henriksson