[Triolith-users] Important: Triolith two week service stop starting August 26th

[Mats Kronberg]

Dear Triolith users,

Summary:

Starting August 26th, Triolith will be unavailable due to it being
expanded by 400 nodes and moved to NSC's new computer room. The outage
will last approximately two weeks, but parts of Triolith will be
available again from August 27nd.


Details:

As you may or may not know, Triolith will be expanded with 400
additional nodes. NSC has also built a new computer room / data center
building, and Triolith will be moved to it.

Starting August 26th at 09:00 CEST, Triolith will be shut down, and
the process of disassembling and moving it will start.

No jobs will be started unless they can finish before 09:00 on the
26th. You might need to adjust the wall time limit of your jobs if you
need them to start during the period August 19th - 26th.

Due to the complexity of the move, it is hard to predict exactly how
long the entire process will take. If we run into any unexpected
technical problems, things can be delayed. It's also entirely possible
that the move will go faster than expected.

Our current best schedule estimate is:

Monday week 35 (2013-08-26): Service stop starts at 09:00. The login
nodes and a small number (24-48) of compute nodes should be available
again in the evening. The compute nodes will probably be restricted to
short jobs (like the current "devel" nodes).

Tuesday or Wednesday week 35: Additional compute nodes (bringing the
total to ~288) becomes available.

The remaining 912 compute nodes will be brought back online as soon as
they can be moved, connected to the system and tested. The vendor
estimates that all 1200 original nodes will be ready by the end of
week 36.

Once the original 1200 nodes have been moved, the additional 400 nodes
will be installed and tested. The acceptance test period for these
nodes will last approximately a month. Sometime during this period
there will be one or more short (up to a day) service stops of the
entire Triolith system in order to run full system tests.


If you have any questions regarding this, please contact support at nsc.liu.se.

-- 
Mats Kronberg, NSC Support <support at nsc.liu.se>