SplitNCigarReads uses more than 1 core with no options to set number of cores used

whheaton

If running on a cluster system that kills jobs if they use more than their allocation, you need to be able to either know how many cores are required or be able to limit the number of cores used. Are there options I am missing? Otherwise I would like to know the recommended number of cores to request for this tool. Is the number of cores used based on what the machine has? If so that needs to be limited for a cluster or cloud system. Right now I am requesting 4 cores per job but the jobs spend most of their time at < 200% CPU usage.

shlee

Hi @whheaton,

For the publically accessible data listed in https://github.com/gatk-workflows/gatk3-4-rnaseq-germline-snps-indels/blob/master/rna-germline-variant-calling.inputs.json, SplitNCigarReads task lists the following runtime attributes:

runtime {
        disks: "local-disk " + sub(((size(input_bam,"GB")+1)*5 + size(ref_fasta,"GB")),"\\..*","") + " HDD"
        docker: "broadinstitute/genomes-in-the-cloud:2.3.1-1504795437"
        memory: "4 GB"
        preemptible: preemptible_count
    }

These can be used as starting points for figuring out roughly what the tool requires for a certain sized BAM. Let me know if this helps or if you need more guidance.