GATK - applybqsr - not a valid memory

[complexgenome]

Hi,
Suddenly with minimum edits to the code. I'm heading into an error in the nextflow log as:

Caused by: nextflow.exception.AbortOperationException: Not a valid 'memory' value in process definition: 60 GB60 GB

I've process of applyBSQR as:

process applybqsr{

        maxForks 3
        debug true
        errorStrategy 'retry'
    maxRetries 10
publishDir path: "${params.outdir}/applybqsr/normal", mode: 'copy', pattern: '*_N*'
    publishDir path: "${params.outdir}/applybqsr/tumor", mode: 'copy', pattern: '*_T*'
        input:
        tuple val(sample_id_tumor), path(markbam_tumor , stageAs: 'tumor_applybqsr_bam/*')
        tuple val(sample_id_tumor), path(recaltable_tumor,  stageAs: 'tumor_applybqsr_recal/*')
        tuple val(sample_id_normal), path(markbam_normal , stageAs: 'normal_applybqsr_bam/*')
        tuple val(sample_id_normal), path(recaltable_normal  ,stageAs: 'normal_applybqsr_recal/*')

        output:
        tuple val(sample_id_tumor), path("${sample_id_tumor}.sorted.markdup.recal.bam"), emit: recal_bam_tumor
        tuple val(sample_id_tumor), path("${sample_id_tumor}.sorted.markdup.recal.bai"),  emit: recal_bai_tumor
        tuple val(sample_id_normal), path("${sample_id_normal}.sorted.markdup.recal.bam"), emit: recal_bam_normal
        tuple val(sample_id_normal), path("${sample_id_normal}.sorted.markdup.recal.bai"),  emit: recal_bai_normal
        
        script:
        """
        ml gatk/4.1.3.0 samtools/1.9
        samtools index ${markbam_tumor}
        samtools index ${markbam_normal}

        gatk ApplyBQSR \\
        -R $params.hg38genome \\
        -I ${markbam_tumor} \\
        --bqsr-recal-file ${recaltable_tumor} \\
        -O ${sample_id_tumor}.sorted.markdup.recal.bam \\
        -L $params.target \\
        --create-output-bam-index true \\
        --java-options "-Xmx${task.memory.toGiga()}g"
        gatk ApplyBQSR \\
        -R $params.hg38genome \\
        -I ${markbam_normal} \\
        --bqsr-recal-file ${recaltable_normal} \\
        -O ${sample_id_normal}.sorted.markdup.recal.bam \\
        -L $params.target \\
        --create-output-bam-index true \\
        --java-options "-Xmx${task.memory.toGiga()}g"


        """

}
workflow.onComplete { 
        log.info ( workflow.success ? "Done applyBQSR!" : "Oops .. something went wrong in applyBQSR" )
}

The config file is as:

profiles {

        minerva {
                process.executor = 'LSF'
                process.queue = 'premium'
        process.memory = '80GB'
                process.cpus = 4
                process.time = '50 h'
                process.clusterOptions =  ' -P acc_pareks02a  -R affinity[core(12)] '
        }
}
process {
        withName: applybqsr {
                memory={'60 GB'* task.attempt }
                 cpus = { 2 * task.attempt }
        }
}

Please see attached log file.
log_applybqsr.txt

If I look at the .command.sh it looks fine:

#!/bin/bash -ue
ml gatk/4.1.3.0 samtools/1.9
samtools index tumor_applybsr_bam/MM-3463_T.dedup.sorted.bam
samtools index normal_applybsr_bam/MM-3463_N.dedup.sorted.bam

gatk ApplyBQSR \
       -R /sc/arion/projects/pareks02a/Daphni2_Pipeline_resources/gencode38/bwa/GRCh38.primary_assembly.genome.fa \
       -I tumor_applybsr_bam/MM-3463_T.dedup.sorted.bam \
       --bqsr-recal-file tumor_applybsr_recal/MM-0574_T_table.recal \
       -O MM-0574_T.sorted.markdup.recal.bam \
       -L /sc/arion/projects/pareks02a/Daphni2_Pipeline_resources/Twist_Exome_Target_hg38.padded.interval_list.bed/Twist_Exome_Target_hg38.padded.interval_list.bed \
       --create-output-bam-index true \
       --java-options -Xmx60g

What is the issue here? Similar line of java-options work fine with gatk markduplicates, recalibrate, setupnmdtags
It was working fine until 2-3 days ago with similar memory. Only things changes are publishDir.

Kindly let me know if any more details re required.

[bentsherman]

Fun fact about Groovy, multiplying a string by a number repeats the string...

memory = { '60 GB' * task.attempt }

Python does this too I believe 😅

Try this instead:

memory = { 60.GB * task.attempt }

[complexgenome]

@bentsherman Thanks.
I don't know. It worked fine in another set of samples that were launched separately. Also, gatk mark duplicates, setupnmdtag, recalibrate had same string multiplication all were through without this duplication of string.

I've edited the config file and waiting for it to run applybqsr.

Why is it random?

[SPPearce]

Why is it random?

(just a passerby who came across this question)
If task.attempt is one, then it'll work fine, you'll only have issues where the first time the process was attempted failed - generally large samples where you want to try and run again with more memory.