Missing option flags in sbatch command of remote desktop?

Get Help
1

Hello!

We’d like to deploy open ondemand on our cluster (which uses slurm), and we are running some tests beforehand on virtual machines with vagrant.

So far, we’ve been able to install open ondemand on ubuntu 20.04 with the ansible role (GitHub - OSC/ood-ansible: An ansible role for Open Ondemand), and we are able to identify, then use file management and shell access.

The problem I face is with remote desktop in interactive apps. For example, if I launch with default parameters (empty account and partition, 1 in hours and in nodes) I get the following error:

Failed to submit session with the following error:
sbatch: error: Unable to open file 1

After some digging, the 1 is in fact the 1 in requested nodes, if I request 5 nodes I will get ‘Unable to open file 5’.

If I am not mistaken, the command behind this is the following:

/usr/local/bin/sbatch -D /home/vagrant/ondemand/data/sys/dashboard/batch_connect/sys/bc_desktop/democluster/output/b8518b93-0f41-4e9c-a8bb-54048109a78e -J sys/dashboard/sys/bc_desktop/democluster -o /home/vagrant/ondemand/data/sys/dashboard/batch_connect/sys/bc_desktop/democluster/output/b8518b93-0f41-4e9c-a8bb-54048109a78e/output.log -p cpupart -A demoaccount -t 01:00:00 --export NONE 1 1 --parsable -M democluster

If I use this command directly on my front node, it returns the same error:

sbatch: error: Unable to open file 1

It seems the 1 after ‘–export NONE’ should (?) have the respective flags -N -n as in the following

/usr/local/bin/sbatch -D /home/vagrant/ondemand/data/sys/dashboard/batch_connect/sys/bc_desktop/democluster/output/b8518b93-0f41-4e9c-a8bb-54048109a78e -J sys/dashboard/sys/bc_desktop/democluster -o /home/vagrant/ondemand/data/sys/dashboard/batch_connect/sys/bc_desktop/democluster/output/b8518b93-0f41-4e9c-a8bb-54048109a78e/output.log -p cpupart -A demoaccount -t 01:00:00 --export NONE -N 1 -n 1 --parsable -M democluster

but I don’t understand what could cause the flags to be missing, if that is indeed the problem.

Has anyone already encountered this problem and can help us solve it?

2

The submit line that you’re showing there comes out of the submit.yml.erb that you’ve supplied from the application. Can you share that file? It seems you missed the -N arg in that file.

Here’s ours for reference where we specify --nodes

3

Oh thank you very much!! I overlooked that submit part in the yml file, so it was still the default:

submit: |
        ---
        script:
          native:
            - "<%= bc_num_slots.blank? ? 1 : bc_num_slots.to_i %>"
            - "1"