Ceci est une ancienne révision du document !


S'abonner au fil des news (flux RSS)

Fil des news




20250630 / Cooling and scratch

  • Cooling system is still under heavy load (and awaits repairs)
    • Lake partitions are reduced, and frequencies set at minimum
    • Cascade partitons have frequencies set at minimum
  • Scratch Cascade is almost full
    • please cleanup, before I do…
    • find /scratch/Cascade/$USER/ --dry-run -type f -iname '*' -ctime +180 -exec rm -f {} \;
    • find /scratch/Cascade/$USER/ --dry-run -empty -type d -delete

remove –dry-run for thrilling action.

2025/06/30 06:55 · ltaulell

20250627 / Cooling in datacenter

“Depuis 11h, un problème de groupe froid (1/4 en rade) affecte le Data Center (salle SING).

Nous devons impérativement limiter la chauffe de nos installations en les arrêtant ou en limitant leur fréquence.”

There's a problem with datacenter cooling system. Frequencies will be reduced and unused nodes will be powered off during all week-end.

edit: Lake & Lake-premium are in drained mode for the WE: running jobs will finish, queued jobs will stay queued.

If it is not enough, Cascade & Cascade-premium will endure the same fate, or worst (poweroff…).

2025/06/27 17:38 · ltaulell

20250611 / scratch/Cral

Probably because a high energy particule went by, scratch/Cral went awol around 11:00.

(for real, reason unknown, but 2 servers of the pack suddendly decided that “no.”)

Plan B applied, with conviction, things should be back to normal.

2025/06/11 10:12 · ltaulell

20250605 / problem with data10

data10, home of chimie logins has been down for a few hours. Staff is looking into it.

EDIT 05-06-2025,14:30 good news, data are OK. bad news, network is HS.

EDIT 05-06-2025,16:00 GOOD news, Emmanuel and Cerasela resolved the problem.

2025/06/05 12:31 · ltaulell

20250429 / End of E5 (last reminder)

As of tomorrow (end of month), E5 partitions (E5-short, E5 & E5-long) will be stopped.

/scratch/E5N will be available on e5-2667v4comp[1-2] login nodes for another month (for users to retrieve data), then shutdown as well.

e5-2667v4comp[1-2] login nodes will stay available as long as possible, like x5570comp[1-2] login nodes.

2025/04/29 09:21 · ltaulell
news/blog.1392304970.txt.gz · Dernière modification : 2020/08/25 15:58 (modification externe)