These were accepted as a reasonable account of the last meeting.
According to Angus this shouldn't be possible, but there's at least one
edge case where things can go wrong. That's where (e.g.) a
file is sent directly using lpr. If the printer can't interpret what it's
sent and prints out pages of junk then the actual page count can end up
greater than the expected value, and hence the credit. Services Unit
are looking into ways to prevent this.
This was a US thing, and was perhaps once used for CSBE. RAT to be removed from the action.
Still to do. Upgrades have been restarted.
Graham doesn't believe there are any, but we should do a test to check. Defer to shortly before Easter.
After a bit of discussion there was a need identified to document the ability to do this more clearly to users, but it wasn't clear which users or where this should be done. Services Unit to think about this some more.
Actions added or revived:
... which is apparently 15th-19th February
Deferred to shortly before Easter.
Reinstate the recurrent deferred action (October) for this.
Added, with Graham as "owner".
HoS says that exams have priority, so we can pull the plug as necessary without referring upwards.
On the MPU list.
(There should be a recurrent deferred action for this...)
Report from Computing Executive Group
There was a general feeling of unease about the amount of information received. We wouldn't expect full details, but some pointers as to things we might all watch out for ourselves would have been good to have. Do they have contingency plans in place, covering the kinds of things that could be foreseen to arise when running a CA service?
ToDo lists are likely to be discussed at tomorrow's "security" meetings.
Alastair had received some comments, but would welcome more by the end of the week to meet the deadline. The 2017 section is unlikely to contain much detail.
Reports from Units
Alan B's response should be added to the RT ticket (#75278).
While we were clear that we didn't currently offer a service for totally-self-managed VMs, there was some surprise regarding the trust model (no end-user login or equivalent access at all) described in the MPU report. Services Unit have a number of web servers, where the base machine is managed by the Unit but the actual apache configuration is managed by the end users. RaT Unit also have VMs with user access (e.g. for MSc projects). Other types of use might give even more general access (cgi-bin, some US machines).
One suggestion was to have different classes of VM host for different classes of VMs. Different reboot regimes might have to be applied as a consequence. MPU will add it to their list for further consideration.
Stephen reiterated that non-CO <develop> should be unusual, and that in all cases there should be a clear comment added to the profile explaining who, why and for how long these non-<stable> machines would be expected to persist. ALL to check and tidy up any of theirs.
Should the clusters be on separate not-routed-outside-EdLAN subnets? Probably. RaT and Inf Units to discuss
Topics for discussion
The 2016 Computing Plan was discussed here, but minuted above in the CEG section.
The next meeting will be on 27th January 2016 at 10:00 in IF-4.31.
Please contact us with any comments or corrections.
Unless explicitly stated otherwise, all material is copyright The University of Edinburgh