RAxML within sgt_constructor uses all available resources

[anoukwillemsen]

Hello,

I am running sgt_constructor.py on a slurm compute cluster. I am limiting my run to 50 threads (see script below I use for the run), however, when RAxML is executed it starts using all available resources (In my case >1500 CPU cores, see attached image) on the specific node where its running. Of course this is not fair to other users on the cluster and makes sgt_constructor not suitable for running on a compute cluster. Could you please modify the sg_contructor.smk script (line 203) so that raxml-ng is limited to the resources specified?

Thanks and all the best,

Anouk

!/bin/bash

#

SBATCH --ntasks=5

SBATCH --ntasks-per-node=5

SBATCH --nodes=1

SBATCH --mem=150G

SBATCH --nice=1

SBATCH --partition=basic

SBATCH --time=3-20:00:00

SBATCH --output=log_phylofisher_GeneTree_4.out

SBATCH --error=log_phylofisher_GeneTree_4.err

export LC_ALL="C"; export LC_ALL="en_US.utf8";

load modules

module load conda; conda activate phylofisher-1.2.10;

module load ncbiblastplus/2.13.0; #################################

Gene Tree Construction

sgt_constructor.py -t 50 -i working_dataset_constructor_out_Jul.27.2023/ -o sgt_constructor_out_Aug.08.2023;

[robert-ervin-jones]

Hi Anouk,

I apologize for this. I did not realize raxml-ng used 8 threads by default. I will update the code so that raxml-ng only uses 1 thread.

Best, Robert

[anoukwillemsen]

Thanks for the update!