University of Cambridge Computer Laboratory - GN09 datacentre electrical upgrade – Maintenance details

Maintenance has been completed successfully. All affected servers and network switches are back up (most as of several hours ago).

The UPS electrical maintenance work has finished; we will now start re-energising circuits within the datacentre GN09. This will be a gradual process so expect the disruption to continue for the time being. An update will be posted when all rack power feeds are live.

Work to upgrade power resilience in GN09 will take place from 10th until 13th January 2023, requiring a shutdown of all circuits supplied by our primary UPS for most of 13th January.

Some servers used for research are supplied solely from those circuits, and will need to be turned off for the day unless alternative arrangements are made. We have some limited capacity to provide alternative power feeds for a subset of servers. Owners of affected machines have been contacted, and must get in touch as instructed if they require their machine to have a temporary power feed on 13th January. Otherwise, such machines will be powered off for the duration.

A larger set of servers will lose networking for the duration unless an alternative power feed is set up for their rack switch. Owners of such affected machines have also been contacted and must get in touch if the network outage would be too disruptive.

Servers' out-of-band management (Caelum Console) will be partially unavailable.

Besides the above, most other machines in GN09 (except the core network and filer which have a secondary UPS) will be without power resilience for the day. Servers with redundant PSUs will temporarily be vulnerable to a single PSU failure. Immediate widespread disruption is possible in the event of a problem with the mains power.

Work to upgrade power resilience in GN09 will take place from 10th until 13th January 2023, requiring a shutdown of all circuits supplied by our primary UPS for most of 13th January.

Some servers used for research are supplied solely from those circuits, and will need to be turned off for the day unless alternative arrangements are made. We have some limited capacity to provide alternative power feeds for a subset of servers. Owners of affected machines have been contacted, and must get in touch as instructed if they require their machine to have a temporary power feed on 13th January. Otherwise, such machines will be powered off for the duration.

A larger set of servers will lose networking for the duration unless an alternative power feed is set up for their rack switch. Owners of such affected machines have also been contacted and must get in touch if the network outage would be too disruptive.

Servers' out-of-band management (Caelum Console) will be partially unavailable.

Besides the above, most other machines in GN09 (except the core network and filer which have a secondary UPS) will be without power resilience for the day. Servers with redundant PSUs will temporarily be vulnerable to a single PSU failure. Immediate widespread disruption is possible in the event of a problem with the mains power.