Init script slowness

[Henrique de Moraes Holschuh]

On Tue, 04 Oct 2011, Steven Kurylo wrote:
> On Tue, Oct 4, 2011 at 10:28 AM, Henrique de Moraes Holschuh
> <hmh at debian.org> wrote:
> > On Tue, 04 Oct 2011, Ondřej Surý wrote:
> >> I agree on fixing SIGQUIT behaviour.
> >
> > Does anyone have an easy way to test whether sigquit is doing the right
> > thing on open but idle connections of the various protocols?  After all, if
> > does not have an open transaction _now_, it should get immediately closed by
> > a sigquit...
> 
> Checking in proc, my user has 5 imapd processes right now.
> 
> I attached strace to them one at a time, then issued a sigquit.  Each
> time the process exited immediately.  The logs show something like:
> 
> Oct  4 10:36:50 thetis cyrus/imaps[12021]: fetching user_deny.db entry
> for 'stevenkurylo'
> Oct  4 10:37:55 thetis cyrus/imaps[12021]: auditlog: traffic
> sessionid=<cyrus-12021-1317741678-1> bytes_in=<16484>
> bytes_out=<80756>
> Oct  4 10:37:55 thetis cyrus/master[1325]: process 12021 exited, status 75
> Oct  4 10:37:55 thetis cyrus/master[1325]: service imaps pid 12021 in
> BUSY state: terminated abnormally
> 
> I haven't had an issue with shut down taking the full 30 seconds.
> Seeing a strace and the log files from a pid on a machine exhibiting
> this behavoir would be interesting.

And also the protocol trace (cyrus calls it "telemetry").  That,
together with the strace/ltrace, would allow us to nail the problem down
rather fast and submit a bug report upstream.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh