In slurm-speak, 'partitions' are the different queues. The primary use of these on farm seems to be switching between small and big mem.
You specify partitions with -p
, so e.g.
#SBATCH -p bml
at the top of your sbatch script will specify the bml partition (big memory, low priority).
Without specifying a partition, I've been able to allocate 30 GB but not 130 GB.
We have access to:
- bml - low priority big mem node (up to a TB)
- bmm - medium priority big mem node
- bmh - high priority big mem node
as well as
- low2 - low priority compute nodes
- med2 - medium priority compute nodes
- high2 - high priority compute nodes
Per Bill Broadley, here is how things work:
bmh = get what you paid for within 1 minute. bmm = get more than you paid for, if there's any free resources (which is usually), but your job might be suspended if another user asks for nodes in bmh. bml = get more than you paid for (just like bmm) but your job might be killed and rescheduled.
Our group allocation has limits.
As of June 15, 2020, they are as follows.
On bmh/bmm/bml, we have one bigmem node, so:
- 1TB of memory max
- 96 CPU
On high2/med2/low2, we have two parallel nodes, so:
- 512 GB RAM total, 256 GB max per job
- 32 cores on one machine, 64 cores on the other machine