Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 30 Nov 2006 16:42:01 -0500
From:      Robert Noland <rnoland@2hip.net>
To:        freebsd-ports@freebsd.org
Subject:   Something is really not right with azureus 2.5 / diablo-jdk 1.5
Message-ID:  <1164922921.61502.13.camel@rnoland-ibm.acs.internap.com>
In-Reply-To: 571063.87057.qm@web50304.mail.yahoo.com

next in thread | raw e-mail | index | archive | help

> On Thu, Nov 30, 2006 at 11:36:10AM -0800, Tim Clewlow wrote:
> > 
> > --- Robert Noland <rnoland at 2hip.net> wrote:
> > 
> > > From: Robert Noland <rnoland at 2hip.net>
> > > To: freebsd-ports at freebsd.org
> > > Date: Thu, 30 Nov 2006 13:06:40 -0500
> > > Subject: Something is really not right with azureus 2.5 / diablo-jdk 1.5
> > > 
> > > 
> > > > Hello,
> > > > 
> > > > I have azureus-2.5.0.0 which requires diablo-jdk-1.5.0.07.01_1 to build
> > > (among
> > > > other things). It builds and runs fine, however, azureus limps along for
> > > about
> > > > 10 minutes with very low download and upload speeds, then it seems to grind
> > > to
> > > > a halt, the d/l and u/l basically stop. There are no errors to report
> > > because
> > > > the application keeps going, but the problem is it stops actually d/l or
> > > u/l
> > > > anything (sometimes you might get short bursts of extremely slow speeds,
> > > but it
> > > > is mostly 0 B/s).
> > > > 
> > > > I suspect that diablo-jdk that is at fault, but in truth this is just a
> > > hunch.
> > > 
> > > This is a problem with azureus, not the jdk.  The same issue exists with
> > > sun jdk15.  I am attempting to look at the code, but someone with more
> > > java foo than I would probably be much more productive.  It appears to
> > > be an issue with socket handling...
> > > 
> > > robert.
> > > 
> > 
> > FYI - I discovered the internet connection was constantly at 100% u/l due to a
> > number of people on the LAN doing p2p, so I shaped the external link to only
> > allow 80% of the real uplink capacity (which creates a 20% buffer to attempt to
> > mininize collisions on the uplink) and azureus suddenly came back to life. Note
> > that azureus on the windows box was happy before I shaped the link, but the BSD
> > version was not. This makes me think the BSD version of azureus gets into
> > trouble when outgoing packets get lost due to packet collision.
> > 
> > Tim.
> 
> Does this thing use nio ?

Yes, it does.

robert.





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1164922921.61502.13.camel>