linux.cs.ncl.ac.uk has been updated to point to
borg, named after Anita Borg. This machine runs the same Liux distribution as is available in our cluster in The Mill.
The previous version of this service will continue to run on
ritchie for a few weeks. It was named after Dennis Ritchie. If you find that you need to back off and start using
ritchie for your day to day work, please inform us. We plan to withdraw
ritchie after the start of semester 2.
On request of one of our research groups, CS support have set up an experimental Internet Relay Chat (IRC) service. If you are familiar with IRC, point your clients to
irc.ncl.ac.uk and say hello! The service is currently available to connect to on campus only. We have installed some IRC tools onto ritchie, our time share server: the client
irssi, the bouncer tool
bip and the
tmux terminal multiplexers. If you are familiar with UNIX/Linux console operation but unfamiliar with IRC, I’d recommend the
irssi client. For Windows users, we are yet to perform a survey of modern IRC clients but MIRC may be a good place to start.
I was very happy to action this request, personally, as I’ve been hankering for a Newcastle IRC service for years, as have a small number of my former colleagues in ISS.
#cs exists for general Computing Science discussion and
#social is available for anything-goes chat. Feel free to create your own channels as you see fit.
If you have any questions, comments or suggestions for this service, please get in touch. Comments on this post are one good way to share public suggestions. The next steps for this service are some introductory guides for IRC for the uninitiated and some preparations to make the server publically accessible, to aid extra-institution collaboration.
I’m pleased to report that we have re-opened the Mill Laboratory, having completed some hardware alterations and updated the operating system from Fedora GNU/Linux version 15 to 19.
As always, if you have any problems, questions or comments regarding this facility please don’t hesitate to get in touch.
With apologies for the very short notice:
There have been a number of problems reported recently regarding the availability of machines in the Mill computer cluster. We are aware of approximately 12 machines that are not reachable on the network at present and a further 9 which are reachable but cannot be logged into. We think there may be further problems impacting people’s profiles across all machines.
CS Support are closing the Mill tomorrow, Tuesday 5ᵗʰ April in order to perform a number of fixes and improvements in order to resolve these problems. On the hardware side, this includes a significant amount of rewiring work to correct machines that have been tampered with. We also expect to identify some machines with power supply issues.
We’re also taking the opportunity to perform some tests on the network switch configuration for the lab to try and improve the networking performance.
On the software side, we aim to resolve the issue that prevents logins to work on the 9 machines, networking in general on any of the dozen which do not turn out to have a hardware problem, and begin diagnosis of the profile issues.
CS Support will update you tomorrow on the situation at the close of the day.
Thanks for your patience!
There’s a bug in the version of gnome-shell which we run on the Mill cluster. Occasionally on logout the process does not exit properly and instead starts “spinning” in a busy loop. This does not appear to stop others from using the machine locally, but does impact the performance of compute jobs scheduled to run on the cluster during Evenings.
To address this in the short term, we’ve developed a policy that automatically kills any user process that has been running for more than 24 hours. It’s quite possible that this will impact people who are using the Mill to run longer-term jobs. If so please get in touch with us. We can look at “whitelisting” particular users, or alternative ways of running your jobs.
This policy is now live on mills
mill010 through to
mill019. I hope to apply it to the entire cluster by the end of the week.
Longer term we hope that this bug will be resolved properly when we upgrade the Mill cluster to a newer version of the Fedora operating system.
Please let me know if you experience any problems on the Mill cluster, in particular on any numbered between 010 and 019 inclusive.