Thanks! This is what I'm already doing. It works fine for normal compute clusters, but it doesn't work for job clusters and gives the error mentioned above.