We are “hot-“modifying the InfiniBand network of Cascade cluster. You may experience some temporarly access problems on Cascade scratches.
/scratch/Cascade
is 99% full, hence performing as badly as possible (it's even not performing at all, for now).
You should have cleanup while you can, because now, I'm in charge…: Erasing all files & directory older than 180 days (that's six months old) on /scratch/Cascade
.
Do NOT complain, documentation is crystal-clear (http://www.ens-lyon.fr/PSMN/Documentation/filesystems/scratch.html)
EDIT 17:00: due to a misconfiguration we cannot find, half of scratch nodes (Cascade and Cral) are not connected anymore.