Thanks. I found this article also. I was concerned about it using driver mode and blocking all worker nodes. This sounds quite bad if you have many concurrent jobs running or need to call stored procs frequently. Are you still using this approach or did you find another approach?