Selector for scratch free space in jobs header
Today we have problems running some REMOTE_SETUP jobs. According to support there was a job in one specific host using 100% of scratch/temp disk space. It seems that all of our attempts of running REMOTE_SETUP arrived in that same node and crashing in the compilation process.
Operations recommended using this specification chain in the job header to avoid this kind of situations:
#BSUB -R "select[(scratch<90)]"
This would ensure scratch is at 90% or less of usage, so there would be a 10% free.