Fil des news

(le retour de la revanche)

Acoustics measurements campaign is BACK!

• All queues and partitions will be blocked from 23-09-2022

• All remaining jobs will be killed on 27-09-2022 afternoon

“Sous réserve de conditions météorologiques favorables, la réalisation des mesures acoustiques MLE / SING est programmée dans la nuit du Mardi 27 septembre au Mercredi 28 septembre 2022.”

Under favorable weather conditions, acoustic measurements (yes, again-again) will occur during the night from Tuesday 27th of september to Wednesday 28th of septembre 2022.

We may need to shutdown all of compute nodes and parts of storage to meet the required power/cooling conditions of the measurements, if the date is confirmed.

Backup dates 4 to 5 of october 2022 (same day, one week later).

Cascade cluster will be re-started after manul re-cabling (see previous news)

Planning is fluid.

in order to :

• make room for “Cascade part 2” servers (~80 nodes)
• better placement of scratches (in rack)
• add scratch/Cral to Cascade
• enable 4 new visualization servers (probably in Lake)
• moving and filtering out E5 nodes (for deb11+slurm)

we need to :

• stop and move scratch/Cascade to another rack
• re-cable Cascade s92node[65-78]
• re-cable Cascade infiniband network
• power off and de-rack E5 nodes

Some of it will be done right after “Acoustics Measurements” (see next news).

deb9 → deb11 migration update

• production nodes for partition E5 (deb11+slurm) are coming up.

new E5 login node: c8220node1

(it replace e5-2670comp[1-2] in the new configuration)

• +24 nodes in Lake partition (from Friday 9th Sept)

deb9 → deb11 migration update

We are powering off old E5 cluster, by queues, to:

• make room for “Cascade part 2” servers (~80 nodes)
• moving and filtering E5 nodes we will keep active (in deb11+slurm)

test nodes for E5/deb11+slurm will be drain of jobs and powered off soon to make room for real E5 production nodes (upgraded to deb11+slurm).

A block of nodes (~24n, 768c) will be added to Lake partition, without scratch, next week.

There is an urgent and mandatory OS upgrade awaiting.

All debian11 nodes will be rebooted after jobs finish.

update: you're doing it wrong! also apply to PSMN Staff. Stay tuned…

update: reboot deb11 done, all jobs restarting…