Thanks for the suggestions, Robert. I'll give that a try on the next pass.
At 6 hours a shot, I don't get very many tries during a day! :( Currently,
I've turned off the JIT (I've got a couple more hours to see if that
worked). For what I want to do now, the loss in performance probably won't
be too much of a problem, and when we are able to port to Linux in a few
months the problem should go away.
Thanks again!
Jim
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]On Behalf Of Robert
> Schulz
> Sent: Wednesday, May 02, 2001 7:37 PM
> To: '[EMAIL PROTECTED]'
> Subject: RE: [JBoss-user] What happens at 6 hours and 35 minutes?
>
>
>
> Not sure whether this will help, but here are a couple of
> suggestions ... Write a thread which dumps memory usage,
> calls System.gc() and dumps the time it takes into a file
> every 10 seconds or so. This might tell you whether the
> gc makes the JVM die. Next step is to implement a "state logging"
> singelton into which you call whenever you enter/leave "suspect
> parts" of the code and make the thread dump the state as well
> continously ... this might help to narrow it down.
>
> Does the crash depend on load, is it always after thye _exact_
> same amount of time?
>
> Overall yours is a pretty nasty problem, as you'll have to
> wait 6 hours to see whether it makes a difference |-(
>
> Good luck.
>
> R.
> > -----Original Message-----
> > From: Jim Brownfield [mailto:[EMAIL PROTECTED]]
> > Sent: Thursday, May 03, 2001 8:52
> > To: [EMAIL PROTECTED]
> > Subject: RE: [JBoss-user] What happens at 6 hours and 35 minutes?
> >
> >
> > Thanks, Toby.
> >
> > There's no backtrace on the threads, just the typical "Memory
> > fault: core
> > dumped" message. I will try the kill -SIGQUIT, and see if
> > that gives any
> > insight. The java profiling didn't provide anything useful.
> >
> > Thanks again!
> > Jim
> >
> > > -----Original Message-----
> > > From: [EMAIL PROTECTED]
> > > [mailto:[EMAIL PROTECTED]]On Behalf Of
> > Toby Allsopp
> > > Sent: Wednesday, May 02, 2001 3:01 PM
> > > To: [EMAIL PROTECTED]
> > > Subject: Re: [JBoss-user] What happens at 6 hours and 35 minutes?
> > >
> > >
> > > On Wed, May 02, 2001 at 02:42:43PM -0700, Jim Brownfield wrote:
> > > > Thanks for the suggestion, but it's not 6:35A.M, it's 6 hours
> > > and 35 minutes
> > > > running time, and it doesn't matter when I start JBoss, but you
> > > can set your
> > > > clock by when the JVM will fail after you've started it. I'm
> > > sure there's a
> > > > JVM problem with SCO's implementation, but unfortunately,
> > > there's nothing I
> > > > can do about that, <sarcasm>SCO being such an awesome
> > > > implementation</sarcasm>. I was hoping that someone
> > could say something
> > > > like, "oh yeah, JBoss does <whatever> at about that time", and
> > > I'd have a
> > > > place to go to try to make a temporary workaround in the
> > JBoss code. It
> > > > could be that the JVM is doing something itself, but if
> > so, it's still
> > > > related somehow to JBoss (or Tomcat, I suppose), since we have
> > > other Java
> > > > programs that run longer than 395 minutes.
> > >
> > > Ok, I'll make a more helpful effort than "fix JVM".
> > >
> > > When the JVM crashes, does it dump stacktraces for the running
> > > threads? I've
> > > seen many JVMs do this, and it might give you a clue as to
> > what's going on
> > > at the time.
> > >
> > > You could also try taking thread dumps (kill -3) up until
> > the time when it
> > > crashes.
> > >
> > > Toby.
> > >
> > > _______________________________________________
> > > JBoss-user mailing list
> > > [EMAIL PROTECTED]
> > > http://lists.sourceforge.net/lists/listinfo/jboss-user
> >
> >
> > _______________________________________________
> > JBoss-user mailing list
> > [EMAIL PROTECTED]
> > http://lists.sourceforge.net/lists/listinfo/jboss-user
> >
>
> _______________________________________________
> JBoss-user mailing list
> [EMAIL PROTECTED]
> http://lists.sourceforge.net/lists/listinfo/jboss-user
_______________________________________________
JBoss-user mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/jboss-user
