This evening, starting at around 11PM, I hope to get the host machines up on Linux 3.3.8 and using a pre-emptive kernel.

     This will require reboots, maybe more than one if the first attempt doesn’t work, and each reboot involves hibernating all the guests, rebooting, then restoring them.

     Each server will freeze for approximately 1/2 hour during the process.  If you are on one of these and you are using x2go or nx clients, you will be able to re-attach to your session after the server comes back to life and pick up where you left off.

     I now have all the EL6 derived guest machines (CentOS 6.x and Scientific Linux 6x) on version 3.3.8 with pre-emptive kernels and the non-EL6 guest machines on 3.19.3 with pre-emptive kernels.

     This makes it possible for example to go to a server with x2go, fire up Firefox, go to youtube and, if you have an adequately fast internet connection, play a video reasonably smoothly.  It improves the general responsiveness of the servers.  With the old kernels, even if you had the bandwidth you’d only see around 10 frames per second on the video.

     I have resolved the issues between the old NFSv4 implementation and the new so older pre 3.4.0 kernels now communicate properly with post 3.4.0 kernels and id mapping works properly across the board with NFSv4.