So, I'm also just going to put in a few safeguards against bad connections, like using ifup.
Also, I will assume that at somepoint I will get NFS automounting but in order to forge ahead, I am going to be writing a little script at the head of the cluster. One of the things it will do is pass commands thru the ssh login. Here's an example:
ssh pi@bramble-1`sudo mount /mnt/nfs`
That of course, will mount the SD card I shared out from the head, but the command will be run as if it were pi and inside the compute node. And I JUST got done testing, and the NFS connection is persistent even after logging off.
Just so you know, stuff like the above is what can allow you to rapidly deploy and test multiple systems at once. Well, almost at once. AND WITHOUT USING PARALLEL PROCESSING. There's a lot more involved with this, i.e. coordinating variables, checking status, etc. I did this exact thing years ago and loved every minute of it. I think I still have some old CDs around that had the beginnings of it.