Hi everyone, I’m currently testing

nf-core/eager

on our HPC cluster (IBEX) using Singularity and Slurm, and the pipeline stops at the MALT metagenomics profiling step:

NFCORE_EAGER:EAGER:METAGENOMICS:METAGENOMICS_PROFILING:MALT_RUN (doublestranded_0)

The error seems to be related to Java memory:

java.lang.OutOfMemoryError: Java heap space

From the Slurm report, the job used around ~273 GB of memory, but it looks like MALT may not be receiving enough Java heap memory inside the container. I was wondering where would be the best place to add extra Java memory options for this process in

nf-core/eager

? For example, should I set something like:

JAVA_TOOL_OPTIONS="-Xmx700G"
_JAVA_OPTIONS="-Xmx700G"
INSTALL4J_ADD_VM_PARAMS="-Xmx700G"

inside my Slurm submission script, in a custom Nextflow config using

withName: '.*MALT_RUN.*'

, or somewhere else? For context, I’m running the dev version/revision:

nf-core/eager 3.0.0dev
revision: f77d2464a7c6fe06b67ce1832f2fe72ba21c8fde

and the metagenomics settings include:

--run_metagenomics
--metagenomics_profiling_tool malt
--metagenomics_profiling_database /path/to/MALT_index
--metagenomics_malt_memorymode map

Any advice on the recommended way to increase Java heap memory specifically for MALT would be greatly appreciated. Thank you!

Hi Maru! How are you providing the additional memory to malt_run? looking at the implementation in eager 2, malt takes the option

-J-Xmx

to set the max java heap (instead of the usual

-Xmx

). For the dev version, that option should be provided to the

args

of the process in a

withName:

clause. Please let us know if that is unclear, or if that does/doesn’t work. I think the module might actually need updating to simplify this

Hi! Ahh that makes sense, thank you! I was adding the extra memory in my custom Nextflow config like this:

withName: /.*MALT_RUN.*/ {
    cpus = 1
    memory = '700 GB'
    time = '120h'
    maxForks = 1
    errorStrategy = 'retry'
    maxRetries = 1

    beforeScript = '''
        export JAVA_TOOL_OPTIONS="-Xmx600G"
        export _JAVA_OPTIONS="-Xmx600G"
        export INSTALL4J_ADD_VM_PARAMS="-Xmx600G"
    '''
}

So I was increasing the Slurm/Nextflow memory and trying to set the Java heap through environment variables, but I hadn’t passed the MALT-specific

-J-Xmx

option directly to the process args. I’ll try changing it to something like this:

withName: /.*MALT_RUN.*/ {
    cpus = 1
    memory = '700 GB'
    time = '120h'
    maxForks = 1
    errorStrategy = 'retry'
    maxRetries = 1

    args = '-J-Xmx600G'
}

Does this look like the right way to pass the Java heap option to

malt-run

in eager? Thanks again for the help!

yes, that looks more like the intended way to do it