[Triolith-users] Triolith service stop next Thursday

[Mats Kronberg]

Dear Triolith users,

On Thursday next week (June 27th) from 14:00 CEST most compute nodes
in Triolith will be unavailable.

This is due to a required upgrade of the firmware in the core
Infiniband network switch. This requires that no jobs are running.

As soon as the upgrade and testing is complete, we will allow queued
jobs to start again. It's hard to say exactly how long this will take,
but we estimate somewhere between 30 minutes and 4 hours. We will also
reboot the login nodes during this service stop. This will take 5-15
minutes, but except during that time, the login nodes will be
available, so you can still access your data and submit jobs.

A number of development nodes will also be available during the
service stop (use with e.g "sbatch --reservation=devel", max walltime
for these is 1h).

No running or queued jobs will be affected. Instead we have configured
the system to not allow any jobs to start if they can not finish
before Thursday 14:00. E.g right now, the longest job that is allowed
to start is slightly less than 7 days. If you need to run jobs between
now and Thursday, make sure that you request a short enough wall time
that the job can finish before Thursday 14:00.

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

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