[Gluster-users] huge performance problem with 3.0.*

Service Technique Viewsurf tech at viewsurf.com
Thu Apr 29 10:06:12 UTC 2010


We are using glusterfs 2.0.9 in OpenVZ Debian containers. Since we have 
few problems (so soft like nginx/apache2 won't restart on logrotate), 
we've tried to use 3.0.3 and 3.0.4 versions.

Using 2.0.9, despite the few issues, i works quite well. Using any 
version of 3.0.*, we see an immediate and huge load average (mostly 
wait-I/O), and all scheduled processes seemed to be in stalled state, so 
that there is an accumulation of such processes, until the container is 
completely unreachable.

The hardware node runs 2.6.32 kernel (patched for openvz) and we use 
debian packages for all versions of gluster.

does anybody have an idea about this problem? I haven't seen something 
similar in the list archives


Florian Iragne

More information about the Gluster-users mailing list