LSF/Slurm equivalent command

Question

There are few bjobs commande that I can't reproduce with slurm.
I'm looking for the equivalence of :

• alloc_slot
• nalloc_slot
• slot
• memory usage per node

Especially, I'm wondering whether it's possible to the retrieve the requested and used number of slot per node.
Like, it happens that user requests a full node, so lets say 122 cores, but actually use only 4 cores.

Have someone already faced this situation ?
Thanks.

Answer 1

Especially, I'm wondering whether it's possible to the retrieve the requested and used number of slot per node.

You can have a look at scontrol show nodes. You will get information such as CPUAlloc=0, the number of allocated slots on the node and CPULoad=0.01 the actual load of the node, which should match the number CPUAlloc.

Like, it happens that user requests a full node, so lets say 122 cores, but actually use only 4 cores.

In such a situation you would see CPUAlloc=122 and CPULoad=4.00.

memory usage per node

The same command will output AllocMem=0 which is the sum of memory requests for the node and FreeMem=xxx which is what is reported by the free command and is the amount of memory not used by any program. You will need to compare those numbers with the total available memory on the node. In case of perfect memory usage by the jobs and no memory usage by the OS, AllocMem + FreeMem should be equal to RealMemory.