Re: Deciphering error message

[Gerrit Pape <pape@smarden.org>]

On Wed, Aug 10, 2005 at 08:44:33AM +1200, Paul Dorman wrote:
> I have a test machine (running inside a Xen VM) and have installed the
> runit package for sarge to see how everything might look. I have
> an /etc/runit directory, where a getty-5 example directory was created
> with the following structure:
> 
> sysadmin:/etc/runit# ls -l getty-5/
> total 8
> -rwxr-xr-x  1 root root 31 2005-03-26 08:47 finish
> -rwxr-xr-x  1 root root 44 2005-03-26 08:47 run
> lrwxrwxrwx  1 root root 16 2005-08-09 15:21 supervise
> -> /var/run/getty-5

The symlink in /var/run/ is just for convenience in Debian, to have
/etc/ possibly mount read-only.  runsv creates the ./supervise/
directory automatically, also if it's a dangling symlink.

> I created a symlink to the getty-5 directory in /var/service, and I can
> see the following in /var/run/getty-5/:

Yes, but it shouldn't interest you that much, this directory is meant to
be used by the runit programs.

> I copied the structure of the /etc/runit/getty-5 and modified it to
> start apache2, and noted that there is an empty /var/run/apache2
> directory which I have linked to /etc/runit/apache2/supervise. 
> 
> Apache is running, and I note that it continues to run when I kill -9
> the Apache processes, so it looks like Apache is running under runit's
> supervision. Great!
> 
> The issue I now have is that I see this unhelpful line in my process
> list (all on one line):
> 
> root     11332  0.0  0.0    88   28 ?        Ss   08:29   0:00
> runsvdir /var/service log:  lock supervise/lock: temporary failure?runsv
> apache2: fatal: unable to lock supervise/lock: temporary failure?runsv
> getty-5: fatal: unable to lock supervise/lock: temporary failure?runsv
> apache2: fatal: unable to lock supervise/lock: temporary failure?runsv
> getty-5: fatal: unable to lock supervise/lock: temporary failure?runsv
> apache2: fatal: unable to lock supervise/lock: temporary failure?

> Could someone please tell me what this message means? Is there something
> I need to do to make it go away? If it's not a real error, it sure looks
> like one!

It looks like there're already runsv processes running in
/var/service/getty-5/ and /var/service/apache2/, and runsvdir tries to
start up new ones; tell the currently running runsv processes to exit so
that runsvdir can start up and monitor the supervisors, this may fix the
problem the log shows.
 # runsvctrl exit /service/getty-5/ /service/apache2/

What you see in the process listing is the 'last resort' readproctitle
log of runsvdir, it shows messages written to stderr by the ./run and
./finish scripts or service daemons.  runsvdir will write a dot at the
and of this log every 15 minutes.

HTH, Gerrit.