On Tue, Jun 22, 2010 at 4:00 PM, RW <gluster at tauceti.net> wrote: > What distribution, kernel and GlusterFS versions are you using? Oops, sorry :) I forgot to say. Clients: glusterfs-common-3.0.4-1 glusterfs-client-3.0.4-1 Servers: glusterfs-server-3.0.4-1 glusterfs-client-3.0.4-1 glusterfs-common-3.0.4-1 OS for all 7 machines: Distribution - RHEL 5.5 Kernel - stock 2.6.18-194.3.1.el5 > I've had problems with some kernels between 2.6.31 and 2.6.32-r10 > (-r10 is a Gentoo release number) with GlusterFS 3.0.4 inside a KVM > and KVM versions < 0.12.4. Kernel 2.6.30 or latest 2.6.32-r11 > and KVM 0.12.4 seem to work fine now. We also have had some > NFSv4 problems with the kernels mentioned above. 2.6.30 and > 2.6.32-r11 doesn't have this problem. > > The problem was a race condition in KVM's virtio-net driver. > Maybe you have some problems with your network(driver) too. Hummmm, will see if it relates to the network driver, but since I didn't see any kernel messages or packet errors, I doubt it's that :( > Some weeks ago we have moved a directory which was used by a > Drupal cache module from GlusterFS to NFSv4 since we've had a > lot of problems with it. There was also heavy writing and reading > into this directory. Currently I'm not sure anymore if this was > a kernel or GlusterFS problem. I've never tested it again with > GlusterFS. Some other directories with mainly reading are working > fine. Seems like our scenario... We'll move to the shared memory based solution, but meanwhile ... :) A friend is testing some other features like quick-read to see how it behaves too. -- []'s Hugo www.devin.com.br
