Shutdown errors about 'logging/handlers.py' .... ?

[John Dennis]

On Thu, 2006-09-28 at 20:46 -0700, Tom London wrote:
> On 9/28/06, John Dennis <jdennis at redhat.com> wrote:
> > It's possible this might be coming from setroubleshootd which uses the
> > python logging library to write to syslog (the same could be said for
> > any daemon written in python). But there are two odd things, the code in
> > the python logging library is wrapped in a try/except so it shouldn't be
> > generating error messages and if setroubleshootd is the culprit then it
> > should have been shutdown by init (e.g. service stop) before this point.
> > The client portion of setroubleshoot (sealert) uses the same code but it
> > should have been shutdown when the desktop login session was terminated.
> > Hmm....
> > --
> > John Dennis <jdennis at redhat.com>
> >
> John,
> 
> Good call.  I inserted a 'ps agx >/tmp/before.15' in S01reboot. I
> attach. Looks like setroubleshootd is still running:
> 
>  2179 ?        Ss     0:00 irqbalance
>  2220 ?        Ssl    0:01 /usr/bin/python -E /usr/sbin/setroubleshootd
>  2442 ?        S<     0:00 [krfcommd]
> 
> I'll tried explictly killing it off before I shutdown....That seems to
> fix things.

Thanks Tom for tracking this down. I suspect what is happening is the
setroubleshootd process is not exiting because one of its threads is
still running. In python there is a flag to control if child threads
exit when the main thread does and some behavior I have to investigate
more with respect to threads and signals. We had put some code in to
address some of this but it sure looks like its not working as expected.
Could you tell me what version of setroubleshoot you are running so I
can see if that version had the code changes? I'll also add this issue
to me to-do list for today.
-- 
John Dennis <jdennis at redhat.com>