[dm-devel] [PATCH v2 11/11] multipathd: remove unhelpful startup / shutdown messages

[Martin Wilck]

On Mon, 2022-03-21 at 19:43 -0500, Benjamin Marzinski wrote:
> On Fri, Mar 18, 2022 at 5:33 PM <mwilck at suse.com> wrote:
> > 
> > From: Martin Wilck <mwilck at suse.com>
> > 
> > These messages are noisy in the system log without being actually
> > helpful.
> 
> I've actually found the "start up" and "shut down" messages useful a
> number of times, for tracking when multipathd starts up and shuts
> down. 

Makes sense ;-)

Currently we see the following messages for multipathd startup and
shutdown:

Mar 11 09:30:00 bremer systemd[1]: Starting Device-Mapper Multipath Device Controller.
Mar 11 09:30:01 bremer multipathd[363]: --------start up--------
Mar 11 09:30:01 bremer systemd[1]: Started Device-Mapper Multipath Device Controller.
Mar 11 09:30:01 bremer multipathd[363]: read /etc/multipath.conf
Mar 11 09:30:01 bremer multipathd[363]: path checkers start up
...
Mar 11 09:30:52 bremer systemd[1]: Stopping Device-Mapper Multipath Device Controller...
Mar 11 09:30:52 bremer multipathd[363]: exit (signal)
Mar 11 09:30:52 bremer multipathd[363]: --------shut down-------
Mar 11 09:30:52 bremer systemd[1]: Stopped Device-Mapper Multipath Device Controller.

To my taste, this is too much. Of course, not everyone is using
systemd. Without systemd and with the part of my patch you acked, we'd
be down from 9 to 3 messages. IMO either the "exit" message or the
"shut down" message could be hidden at -v2. I suppose we could decrease
the verbosity level of handle_signals() to -v3 instead. Would you agree
with that?

> Since people generally run multipathd constantly, they rarely
> appear more than a couple of times per boot. I would prefer if they
> could stay.  I'm fine with removing the others.

Ok, fine with me. Do you insist on the "--------", too? It's mainly
that which bothers me. If you look at the typical boot messages of
contemporary Linux servers, no other daemon uses this strong emphasis
for an informational message. The informational value would be higher
if we printed a detailed version number including HEAD commit ID, like
other daemons do.

Martin