multithreading and parallel processing files

Hiba_Shaban · December 14, 2018, 12:02am

I’ve initialized a Galaxy instance on aws and am an admin on there.
I am wondering how to parallelize the use of some tools on there. For example tophat2 has an option -p where on the command line it allows for parallel processing but that option is not one of the options that can be configured in the User Interface tool. Is there a way around this?

There is a notification on the homepage of Cloudman UI that says this version of Galaxy has configured some tools to run on 4 cores in the file “job_conf.xml.cloud”, but is it possible to edit this file to use more processors?

Thanks

jxtx · December 14, 2018, 8:50pm

Yes, you can edit the file. You will need to create a new destination with the number of cores you want set. Something like:

        <destination id="slurm_cluster_cpu16" runner="slurm">
            <param id="nativeSpecification">--nodes=1 --ntasks=16</param>
        </destination>

The default can be seen here:

github.com

galaxyproject/ansible-cloudman-galaxy-setup/blob/master/files/galaxy/job_conf.xml.cloud

<?xml version="1.0"?>
<job_conf>
    <plugins workers="4">
        <plugin id="slurm" type="runner" load="galaxy.jobs.runners.slurm:SlurmJobRunner">
            <param id="drmaa_library_path">/mnt/galaxy/slurm-drmaa/lib/libdrmaa.so</param>
        </plugin>
        <plugin id="pulsar_rest" type="runner" load="galaxy.jobs.runners.pulsar:PulsarRESTJobRunner" />
        <plugin id="local" type="runner" load="galaxy.jobs.runners.local:LocalJobRunner" workers="4" />
    </plugins>

    <handlers default="main">
        <handler id="main" tags="handlers">
            <plugin id="slurm" />
        </handler>
    </handlers>

    <destinations default="slurm_cluster">
        <destination id="slurm_cluster" runner="slurm"/>
        <destination id="slurm_cluster_cpu4" runner="slurm">
            <param id="nativeSpecification">--nodes=1 --ntasks=4</param>

This file has been truncated. show original

@enis, if this file is edited, how should it be persisted?

enis · December 17, 2018, 4:54pm

To persist job_conf.xml, copy the original template from /mnt/cm/cm/conftemplates/job_conf.xml to /opt/cloudman/config/conftemplates/job_conf.xml and modify as desired before restarting Galaxy via CloudMan. (Note that you’ll have to create the /opt/cloudman/config/conftemplates/ directory.)

Hiba_Shaban · December 17, 2018, 7:30pm

Thank you, I’m trying this now. hopefully I’ll be able to get it to work!

Hiba_Shaban · December 17, 2018, 7:54pm

which of the files do I modify after copying?

enis · December 17, 2018, 8:42pm

Should be just job_conf.xml to add a specific destination and tools you want to run with more cores.
These slides may help with some examples of how the given file is composed: https://galaxyproject.github.io/dagobah-training/2018-cape-town/16-compute-cluster/compute-cluster.html#7

Hiba_Shaban · December 17, 2018, 8:49pm

Sorry I should have been more clear, I meant do I edit the /mnt/cm/cm/conftemplates/job_conf.xml or /opt/cloudman/config/conftemplates/job_conf.xml that I copied from the first destination

enis · December 17, 2018, 9:06pm

You should copy /mnt/cm/cm/conftemplates/job_conf.xml to /opt/cloudman/config/conftemplates/job_conf.xml.

Hiba_Shaban · December 19, 2018, 7:38pm

Thank you @enis and @jxtx, it worked.

vebaev · July 22, 2019, 3:47pm

Do anyone have a list of tools that work in multi-core more so I can populate my job_conf file?