On 9/28/06, Tom London <selinux(a)gmail.com&gt; wrote: > Running latest Rawhide. > > During shutdown, I see 'flashing' on the console several python-type > error messages. The best I have been able to copy down is that they > complain about > /usr/lib/python2.4/logging/handlers.py:677 and :617. These lines are: > > self._connect_unixsocket(self.address) > and > self.socket.connect(address) > > These messages do not appear anywhere in /var/log that I can find. > > Can anyone verify this? I'll try to do a better job grabbing the text > of the messages... > Tracing this out a bit, it appears to be generated just after the 'action $"Sending all processes the TERM signal..." /sbin/killall5 -15' and just before the 'action $"Sending all processes the KILL signal..." /sbin/killall5 -9' in /etc/rc.d/rc6.d/S01reboot. Sound familar to anyone?

On Thu, 2006-09-28 at 20:46 -0700, Tom London wrote: > On 9/28/06, John Dennis <jdennis(a)redhat.com&gt; 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(a)redhat.com&gt; > > > 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(a)redhat.com&gt;