Advance warning of scheduled systems work and downtime

August 18-19th: Huxley building network upgrades

Over the weekend of August 18-19th, the network switches in the CWCs (comms wiring closets) on each floor of the Huxley building are being replaced with new units. There are two CWCs on each floor, one at the north side adjoining the Blackett building, near the stairwell, and the other opposite the southern lift shaft, and they will be upgraded in turn which means there will be a network outage of up to 2 hours for all sockets fed from each CWC.

Users will not need to take any action before leaving college on Friday but some systems that are left powered on all the time may not re-establish a network connection after a switch has been replaced, or they may not renew a DHCP lease (so they will not get an IP address). Power cycling the system is the easiest solution but if this cannot be done for some reason, such as ongoing computations running on the system, etc, restarting just the network services is possible on all modern operating systems but this obviously needs to be done when you near the computer in most cases.

Services provided by the Maths server room in Huxley 616 will obviously be affected and we have agreed with ICT for the work in CWC 100, which feeds this room as well as the network connections in rooms on the south side of level 6, to commence at about 7am on Saturday, 18th August and should be completed by 10am.

A network outage to Huxley 616 will particularly affect a small number of Maths compute cluster users whose cluster home directories are on the server in this server room - these home directories will become disconnected from the cluster in the City & Guilds datacentre, preventing data being read or written to these. These users have already been identified and notified to make them aware of this, and the opportunity has been taken to encourage them to have their cluster home directories moved from clustor to the new clustor2, which is in the same rack as the cluster itself. clustor2 offers much higher read/write performance as well as twice the storage capacity of clustor.

Other services that will be affected by upgrade works in CWC 100 will be various websites (including the LMS), access to clusters and servers hosted within Huxley 616 and the Maths SSH gateways.

August 22nd: disk upgrade for modal server

The modal compute server in the Statistics section will be shut down at around 11:00 on Wednesday morning, August 22nd for an additional user data disk to be added. This timing is to allow an existing user's compute jobs to complete and the downtime will be no more than about 5 minutes.

modal may continue to be used in the meantime as long as any jobs complete before 11am on Wednesday. Any jobs still running after that time will be terminated!. This shutdown has already had to be rescheduled twice owing to inaccurate estimates from users of when jobs will finish but it will not be rescheduled again.

There is no other scheduled work planned for the foreseeable future.

Andy Thomas

Research Computing Manacer,
Department of Mathematics

last updated: 16.8.2018