[Gluster-users] bonnie hangs with glusterFS 2.0.4

[Julien Cornuwel]

Le vendredi 07 août 2009 à 19:24 +0530, Vijay Bellur a écrit :
> Julien Cornuwel wrote:
> > Well, no luck. Adding the readahead translator made bonnie hang at the
> > same moment :
> >
> > # bonnie -d /test -u user
> > Using uid:1000, gid:1000.
> > Writing with putc()...done
> > Writing intelligently...done
> > Rewriting...
> >
> > And nothing more : load average dropped at 0.00, 0.00, 0.00, so I guess
> > bonnie is waiting for something that will never come... 
> >   
> 
> Can you please try with 2.0.6rc4 available at  
> http://ftp.gluster.com/pub/gluster/glusterfs/qa-releases/glusterfs-2.0.6rc4.tar.gz 
> ? We have fixed some issues which could potentially cause GlusterFS to hang.

So, I ran the test again with 2.0.6rc4. This time, bonnie didn't hang.
But the load average was fluctuating a lot. More precisely, it repeated
the following pattern :
- Reasonably high (around 0.45)
- Constant slowing down (till 0.00)
- Stale a few seconds at 0.00
- Rose around 0.45 again, and so on...

The test took ages, literally. I stopped it after 24 hours and the
rewrite part wasn't finished yet.
Still nothing in logs except usual connection messages.
Note : Even when bonnie hang, the volume was still accessible via other
programs (ie shell).

These tests are taking too long, so I think I will drop bonnie and do
simpler tests (kernel compilations). I'll keep you posted.