Running Snakemake with SLURM (using profile)

Running Snakemake with SLURM (using profile)

This guide explains how to set up and run Snakemake (tested version 8.14.0) on an SLURM-based cluster using the snakemake-executor-plugin-slurm.

This guide is based on my GitHub comment on the Snakemake repository.

Installation

Install Snakemake and the SLURM executor plugin via Conda:

conda install -y -c bioconda -c conda-forge snakemake=8.14.0 snakemake-executor-plugin-slurm

Test Directory Structure

Before running, your directory should look like this:

test_run/
├── one.txt
├── profiles/
│   └── work_profile/
│       └── config.v8+.yaml
└── Snakefile

where the contents of files are:
profiles/work_profile/config.v8+.yaml

executor: slurm
default-resources:
  slurm_account: "account_name"
  slurm_partition: "queue_name"
  runtime: 2h
  mem_mb: 4096
  slurm_extra: "'--mail-type=END --mail-user=first.last@mail.com -o logs/smk.%N.%j.out -e logs/smk.%N.%j.err'"
printshellcmds: True
jobs: 100
restart-times: 3
latency-wait: 60
rerun-incomplete: True
use-conda: false
set-threads:
  rule1: 1
  rule2: 2
  test: 4
set-resources:
  test:
    slurm_partition: "queue_name"
    runtime: 4h
    mem_mb: 4096

See Snakemake profiles documentation for more info.

Snakefile

rule all:
    input:
        "done.txt"

rule test:
    input:
        "one.txt"
    output:
        "done.txt"
    log:
        "test.log"
    shell:
        """
        echo "Processing data with {threads} threads" > {log}
        cp {input} {output}
        /usr/bin/time -v sleep 10
        echo "Processing complete" >> {log}
        """

one.txt
Just a placeholder file:

empty file

Running the Workflow

Now we can execute snakemake on SLURM using the profiles/work_profile.
From within the test_run/ directory:

$ cd test_run
$ snakemake --profile profiles/work_profile  --snakefile Snakefile
Building DAG of jobs...
SLURM run ID: 952ada6a-70b3-44b8-b6d9-30c2ee2fc421
Using shell: /usr/bin/bash
Provided remote nodes: 100
Job stats:
job      count
-----  -------
all          1
test         1
total        2

Select jobs to execute...
Execute 1 jobs...
...
...
Finished job 0.
2 of 2 steps (100%) done
Complete log: .snakemake/log/2024-06-28T103831.904550.snakemake.log

After execution, your directory will look like:

test_run/
├── one.txt
├── profiles/
│   └── work_profile/
│       └── config.v8+.yaml
├── Snakefile
├── logs/
│   ├── smk.e512n47.1118341.err
│   └── smk.e512n47.1118341.out
├── test.log
└── done.txt

Snakemake and plugin versions

snakemake==8.14.0
snakemake-executor-plugin-slurm==0.7.0
snakemake-executor-plugin-slurm-jobstep==0.2.1
snakemake-interface-common==1.17.2
snakemake-interface-executor-plugins==9.1.1
snakemake-interface-report-plugins==1.0.0
snakemake-interface-storage-plugins==3.2.2

Notes

In Snakemake (v8.14.0), SLURM jobs are named using UUIDs. To view the rule name/wildcards in your SLURM queue, Snakemake now adds that information to the --comments section.

References:

• Snakemake SLURM plugin source
• Discussion in Snakemake issue #2494

Use the following squeue command to display comments: We need to add %k to the squeue command to see the --comments section

squeue -u $(whoami) -o "%.14i %.8u %.20k %.20j"

That’s it!

You now have a working Snakemake setup with SLURM executor integration.