Operational Meeting Infrastructure Unit Report -- 13th June 2018

The new date for this is likely to be Saturday 14th and Sunday 15th July. Watch this space. Note:

• We have not yet made any changes to the non-UPS power-bar turn-on settings, as discussed previously.
• The rack-0 UPS is still very heavily loaded, and is not likely to offer a lot of protection for the machines connected to it. Specifically, you might well not get the full five minute low-battery warning. The rack-3 UPS is also heavily loaded, though not quite as much, so machines connected to it should get their full five-minute low-battery warning.

Here's Paul Hutton's email regarding server room access.

February 1st 2019 is going to be "DNS flag day". inf.ed.ac.uk passes the test. (The blog is worth a read too.)

dtisite.inf has moved back to linnaeus in order to simplify the access controls for netmonBayes on hazlewood. Once the building is complete we'll freeze the site and transfer it to group-space, as was previously done for the Forum's newbuildpics.

The testloghost is now rattle. hickox (a R200) has been decomissioned.

To accommodate additional machines we're going to put another "block of three" racks into use in the B.02 server room. Ideally they'll be made up from the two Dell racks by the door, along with the Dell rack which currently houses the infrastructure servers, that rack being replaced by the new one. The idea would be to ease the attachment of the third rack to the rest of the run, to provide a smooth transition for the rack-top cable baskets, and to use our standard existing PDU-mounting brackets (as well as to try to preserve the aesthetic appeal of the room...).

We would of course have to move kit out of the rack. Probably the least painful way to do this would be to take knussen, dutoit and hazlewood down, one at a time, park them on the floor of the adjacent comms rack, bring them up, do the rack-shuffle, and then move them back one at a time. We could live without rattle and linnaeus for the duration.

What about charmoz, the Forum console server?

This does not mean that you can leave old machines in place. We are still going to be tight for space, and you will still need to remove things which are no longer needed. And we may still have to ask you to move "small" machines to defragment the space to allow bigger ones in.

The Forum and AT core switches are overdue for replacement (RT#89204, RT#89032). We can do these one at a time. The two things which won't just fail over transparently are:

1. self-managed routing, which will have to be moved to one of the other switches a few days before core0 and atc0 are done to allow leases to renew; and
2. The AT EdLAN link, which will be moved to the hot-spare port on atc1, which should cause just a couple of seconds break while the fibre is moved.

We'll then use one of the ex-AT 5406 switches to replace core3, which is an even older 5304 which has had a fan-fault for some time now. This also gives us a reasonable upgrade path to 10GbaseLR for our EdLAN-OC link. We also have an old 5304 in the 2A closet, normally turned off, which can take over as a backstop router should the server room go off. It would make sense to use the other 5406 to replace this. If we do, we'll probably then leave it running but with higher costs than the main core.