These were accepted as a reasonable account of the last meeting.
The underlying problem is fixed. The script is still pending.
Neil has produced some first notes on how this might be done. For further study.
Actions added or revived:
Toby's initial thoughts are here.
IS are looking at this, but it's likely to be after they roll out the Windows 10 desktop that anything will appear. We would have to pay for the necessary licences for the machines in order to cover non-Informatics users. Defer for 6 months.
It doesn't. However, the TiBS 2 client is fine with the TiBS 3 server, which will make upgrading easier. There's a project to cover the upgrade.
It looks as though Microsoft might licence it, but only in quantities that would be too large for us to be useful. There are so many NDA issues around finding out any more. We've documented what we can (see next item).
The computing.help page is here.
The document is here.
Blog articles suggested:
Report from Computing Executive Group
Reports from Units
The current rack UPS setup is only intended to give (just about) enough on-battery time for machines to do a clean shutdown. If anyone thinks this isn't enough they should make a case to CEG for more money.
There are workarounds for the way the UPSes and power-bars and
nut interact, but there are pros and cons for each.
Services-unit have 4 old fileservers and some disc arrays to decommision. At present there are only 3 free ports left on one of the two core switches.
Post-meeting: it seems that Graham's problem is due to the way
/etc/hosts is set up. Whether that's correct is for
It was agreed that
core0 should be rebooted at a lunch-time,
after a warning to sys-announce.
atc0 is likely to be done
"almost off peak", again after a warning.
If anyone has any suggestions for measurements, please let George know. There are probes like ours dotted around the world, and they can do things like ping, DNS and http responsiveness tests.
George will write a blog article
However, please don't just dispose of things without checking with potential owners first!
This has bugs, most notably in
sssd which seems to have
jumped a version. It's available for (careful) testing, but won't be rolled
out until after teaching and the bugs in it are fixed.
Iain speculated as to the purpose of some of the new servers' BIOS settings.
In an attempt to get to the bottom of this, Neil will start a wiki page
to which all should contribute. This might show up some pattern.
Inf-unit had previously
had a similar-sounding problem with
norrington, a SL6 R320.
For now, he has a script sending <return>s to his stuck machine's
console to try to flush all of the pending output!
Topics for discussion
Domain squatters (George).
The next meeting will be on Wednesday 22nd February at 10:00, location to be confirmed, chaired by George.
Please contact us with any comments or corrections.
Unless explicitly stated otherwise, all material is copyright The University of Edinburgh