Fil des news

• new scratch/Cascade is (finally) available
• Lake-flix and Cascade-flix are open to everybody, for short duration (no longer than 2 days is best) small parallel and sequential jobs, with requeue in case of high priority jobs (see documentation)

We are restarting slowly:

• E5, E5-GPU, Epyc are “ALL GREEN”
• Lake cluster
  • with a new partition Lake-flix, with preemption mode, doc has been updated
  • c6420nodes are still in upgrade progress (this is taking waaaay too long)
• Cascade cluster
  • new scratch/Cascade is unavailable (cable/card IB problem)
  • with a new partition Cascade-flix, with preemption mode, doc has been updated

Any problem(s): please open a ticket.

As a reminder: the use of our web forms is not optional, it is our way of creating and tracking our intervention tickets :

Formulaires du PSMN

Thank you.

Apart for a few expected hiccups, this power outage went smoothly. We are doing our last upgrades on fileservers (which still need a few reboots).

Expected restart this evening, at best tomorrow, we hope.

EDIT:

• gateways are up, access to files should be OK.
• slurmctl is still down/drain, nodes are (still) performing upgrades

En prévision de la coupure, qui pour nous commence demain vendredi, de samedi qui vient, nous avons bloqué les noeuds de calcul (mode DRAIN) afin de purger un maximum de jobs avant la-dite coupure électrique.

In anticipation of the outage, which for us begins tomorrow Friday, this coming Saturday, we have blocked submission to compute nodes (DRAIN mode) in order to purge as many jobs as possible before the said power outage.

An electrical power outage for maintenance purpose will be held Saturday 21th of October 2023 (12 hours long maintenance).

• All PSMN infrastructure will be powered off from friday 20th (starting 12h00) to tuesday 24th
  • login nodes, compute nodes & visualization nodes will be POWERED OFF at noon (12h00) friday
• /scratch/Cascade will be replaced with new hardware, hence
  • all data on scratch/Cascade will be ERASED upon migration
• you have less than a WEEK to retrieve remaining data from /scratch/Cascade

See our online documentation:

• https://www.ens-lyon.fr/PSMN/Documentation/