using slurm_singularity backend: "sbatch: command not found" on IHEC test data

[paulstretenowich]

Hi @sitag,

I'm trying to run the pipeline on the IHEC test data but using the slurm_singularity backend instead of Local but I habe an sbatch not found error.
Here is my log file: ihec_test.log.txt.
Inside the stderr.submit file the error is /long_path/script.submit: line 3: sbatch: command not found and the script.submit.txt file looks like that.

Thanks in advance for the help,
Paul

[sitag]

As it says in the readme, we only support Local mode.
Supporting other backends will require more bandwidth than we have.
That aside, running anything other Local mode from the container will cause problems. Essentially you will ask singularity task to run singularity from the inside the container.

[dbujold]

Hi Sita, someone from our group could look at that integration on HPCs. Would you be opened to a PR if we had a fix to propose?

[sitag]

@dbujold @paulstretenowich

The way to run on other backends is to run cromwell outside of the container, unlike what we do now - so you just follow the ENCODE docs.

Getting it run should not be hard in sane setups. But keep in mind that "officially" supporting it means you have to be willing to help people set it up and run it in whatever environment they have. Other than that commitment, I dont think it's too much work to get it to run on a cluster.

[paulstretenowich]

Hi @sitag,

To have slurm_singularity as backend I ran directly piperunner.sh using:
$ ./piperunner.sh path/for/analysis ./v2/ihec/cemt0007_h3k4me3.json slurm_singularity

I just needed to have a java installed to be able to run java -jar -Dconfig.file=$BACKEND_CONF -Dbackend.default=$BACKEND cromwell-34.jar run $CHIP -i $jobFile -o $WORKFLOW_OPT from inside piperunner.sh.

[sitag]

@paulstretenowich Awesome.. lets keep this ticket open till you have looked through the results.