[Triolith-users] Triolith service stop on Tuesday from 13:00

[Mats Kronberg]

Dear Triolith users,

Unfortunately, we need to stop all Triolith compute nodes on Tuesday
(2012-10-30) to repair the core Infiniband network switch. This fix
will restore full performance to the Infiniband (fast interconnect)
network. The degradation in performance caused by this fault has
probably been too small to be noticeable by most applications.

Until the stop, only jobs that can finish before Tuesday 13:00 will be
started. If you want jobs to run before that time, make sure that you
adjust the wall time limit accordingly.

No running jobs will be affected. Queued jobs will remain in the queue
until the stop is over. During the stop, you can still access your
files from one of the login nodes and submit new jobs to the queue.

During the service stop, we will also update and restart the login
nodes (but one at a time, so you will be able to access your files and
submit new jobs the whole time).

During the stop, we will also run an additional benchmark test of the
whole system. Depending on the initial results of that test, we will
either end the service stop on Tuesday afternoon, or extend it until
Wednesday morning.

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