mono-beagled -- major memory leak

Peter Robinson pbrobinson at gmail.com
Wed Mar 8 13:56:17 UTC 2006


> > > > >Is anyone else seeing this?  I'm assuming that it is not supposed to
> > > > >do that.  Any ideas about how to debug this?
> > > > >
> > Yes I've experienced the same behaviour a couple times. The most recent
> > this afternoon. It took 5 minutes to have the system respond enough so I
> > could kill it. It swallowed 300MB of my 512MB RAM and was seriously
> > thrashing my storage.
> >
> > > Well, I didn't mention high CPU usage as I didn't really notice that.
> > > I also am not really sure exactly what the above means.  I did notice
> > > that something locked up my disk really well accessing it for a short
> > > while.  The I/O wait cycles tied up the entire CPU, so I wasn't able
> > > to do much to find out what program it was, but it seems likely to
> > > have been beagled.
> >
> > I had load numbers nearing 9 when I finally got the beagle program
> > killed.
> >
> >
> I too have just tracked down some nasty problems to beagled malicing my
> system.  It was when I was gaming every 30 seconds or so my entire IO
> system would stop responding for a second or so.  I also found it was
> locking up my system when I was running eclipse and working on the
> repository in my home directory.
>
> The most frustrating part was my inability to turn it off.  If I removed
> it from gnome-session it always found its way back on login, and then
> having a cron job to start it up as well is just wrong.
>
> I have come to terms with accepting mono apps on fedora ( I love tomboy
> and f-spot )  but beagle seems to still need some work.

This sounds like it might be the reason for my desktop suddenly and
for no apparent reason locking solid. It wouldn't respond to anything
but then if I had an active unlocked Windows rdesktop session active i
could still work on that so it was weird. The usual but painful
solution I've been using is the lovely Ctrl+Alt+Backspace combo to
kill X.

Pete




More information about the fedora-test-list mailing list