Mathieu,
If you are using makeflow with the -Tslurm option, then submitting makeflow itself as a slurm job may not work for you. The issue is that makeflow needs to generate and submit slurm jobs, and this usually can only be done from the frontend machine you log in to the cluster. Thus, a first step would be to check with your sysadmin if slurm jobs can be submitted from the compute nodes.
When compute nodes cannot submit jobs, then we recommend using makeflow with Work Queue, -Twq. With this, you can submit makeflow as a slurm job. The difference is that you also need to submit as slurm jobs a set of work queue workers. These workers then connect to your running makeflow to execute your workflow. The makefow file may have to be slightly rewritten when switching from -Tslurm to -Twq, as Work Queue does not assume a shared file system.
(You would use slurm_submit_workers)
Ben