University of Cambridge Computer Laboratory - Notice history

The GPU VM cluster which hosts dev-gpu-* and dev-cpu-* virtual machines, and the associated storage server, requires some urgent software updates and hardware maintenance in order to rectify a couple of known problems. We propose to do this on Sunday; however this could be rescheduled if this would be particularly disruptive (contact mas90 ASAP if so).

All dev-gpu-* and dev-cpu-* VMs plus the shared servers dev-gpu-1, dev-gpu-2, dev-cpu-1, dev-gpu-acs and dev-cpu-acs must be shut down during this maintenance, as the storage server that holds VM disks and home directories will be unavailable for a short time. Capacity to host VMs will gradually be restored during the maintenance as each VM host is updated and brought back online.

Following on from the earlier unscheduled VM storage outage, we need to replace a failed memory module in the storage server that backs one of our departmental VM pools in order to restore performance and reliablity.

This requires us to shut down all VMs on xene-pool1, which will affect the following departmental services:

• cl-student-ssh - Undergraduate SSH server

• MSA (partial outage, one of two servers affected)

• Request Tracker

• VPN2 (partial outage, one of two servers affected and new connections are already steered towards the other server)

• Departmental database server (SQL Server / svr-win-db / db-*)

• Windows Remote Desktop service

• dbwebserver

• WSUS (Windows Updates)

And it will affect the following user VMs:

• cl-teaching-ecad

• dev-compilers0

• egress

• knot

• lmserv-mentor

• svr-papers

• svr-www-ecad

• svr-yg386-web

These will be shut down soon after 5pm and will remain off for approximately an hour. The at-risk window is given as 2.5 hours due to uncertainty with the exact timing.

We will take the opportunity to do some routine maintenance (software and firmware updates) of the storage system at the same time, in order to avoid a future need to do more scheduled maintenance.