General status

We're at the commissioning phase of the SL6 T3-Higgs. For that, Computing is providing a mirrored infra-structure, with all elements in SL6. You will have an interactive node called "t3-higgs-sl6.ultralight.org" which is connected to SL6 worker-nodes.

All features that you had before should work transparently but to make sure there is no risk, we're going through commissioning.

Assumptions :

• Jobs submitted from the usual SL5 interactive node (t3-higgs.ultralight.org) will land on SL5 worker-nodes. ~300 cores.
• Jobs submitted from the new SL6 interactive node (t3-higgs-sl6.ultralight.org) , will land on SL6 worker-nodes. 32 cores for now.

• Workflows jobs run properly
• Home directories are fine and behave as expected (no visible performance issue)
• Compiling on home directories issue - it's a good idea to try it again and see if it's gone. We're using newer versions of NFS clients.

Attention

The login node, is a 2-core, 8 GB RAM machine. So be gentle with it. Once we're happy we will invert -- SL5 interactive will be "slow" and SL6 will be "fast". But until then please be aware that this will be a not-so-fast shared server between the users.

How to use it

Not much different from before, just make sure you login to $USER@t3-higgs-sl6.ultralight.org and work as usual.

Known problems

None at the moment. But please note that the whole infra-structure was not deployed with automated configuration-management tools (those are coming in few days).

That means that I did check that this works : ComputingInteractiveNodeRequirements but something could have slipped. Feel free to submit a ticket in the usual place. We will handle it as soon as possible.

-- Main.samir - 2015-02-26