Thu, 30 Jan 2020 09:19:19 -0600RAIN netboot, NFS root, etc.

mr's Preposter.us Blog

Last night I did some reading that confirmed that I'm on the right track with getting RAIN to boot over the network.  I found a couple of pages that describe the process in more detail than I've seen before, and one of them even revolves around the Parallella, one of my favorite SBC's

In a nutshell, you take the BOOT and ROOT filesystems that work with the target board, share the BOOT via TFTP, the ROOT via NFS and then tell the target board where to find them.  There's a couple different ways to do this, but my favorite is having the DHCP/BOOTP server handle it, so I'll try that first.

Steps

  1. Select, download & extract a suitable disk image for the SOPINE
  2. Create a subdirectory under the TFTP server root for the disk image BOOT partition contents
  3. Create a subdirectory under /srv/nfs for the ROOT partition contents
  4. Modify the DHCP server config to include the NFS ROOT information option
  5. Setup the NFS server and export the ROOT partition directory

I'd like to use an Armbian image since that will match the head node, but the Armbian images have an initial setup process that makes things a little tricky.  I won't be able to just boot all the nodes from this image w/o going through that setup process at least once, and I'm not sure what will happen if I do it one and then boot the others (there may be some node-specific configuration that happens that I don't want the nodes to share.

I'm also not sure how things like hostnames work if all nodes are using the same ROOT partition?  Maybe that's something I can pass along in the DHCP options as well, but I'd think there are other things (logs?) that the nodes will fight over in the shared NFS ROOT?

I suppose I can just try it with Armbian and see what happens?  If nothing else I should be able to get at least one compute node booting, and that will exercise almost all of the parts of the system.  If I run into trouble when I add another node, I can always swap-out another Linux distro and troubleshoot those specific problems instead of the entire system at once.

Command Log

mkdir -p compute-node-images/armbian-buster

cd compute-node-images/armbian-buster

wget https://dl.armbian.com/pine64so/Buster_current

mv Buster_current Buster_current.7z

sudo apt install p7zip-full

7z e Buster_current.7z

sudo apt install nfs-kernel-server

sudo systemctl restart nfs-kernel-server.service

sudo systemctl status nfs-kernel-server.service

sudo mkdir -p /srv/tftp/armbian-buster

sudo mkdir -p /srv/nfs/armbian-buster

sudo vim /etc/exports

Add the following line to /etc/exports:

/srv/nfs/armbian-buster 10.1.10.11(rw,sync,no_root_squash,no_subtree_check)

sudo exportfs -a

sudo mkdir -p /mnt/img

sudo fdisk -l ./Armbian_20.02.0-rc1_Pine64so_buster_current_5.4.14.img

Calculate the offset by multiplying the block size * the Start block of the partition

sudo mount -t auto -o loop,offset=4194304 ./Armbian_20.02.0-rc1_Pine64so_buster_current_5.4.14.img /mnt/img/

sudo cp -r /mnt/img/boot/* /srv/tftp/armbian-buster/

sudo cp -r /mnt/img/* /srv/nfs/armbian-buster/


That's all for now.  Next we need to modify the files we copied over to the NFS root, modify the DHCP server config to set the NFS root, hostname, etc., verify permissions on all the stuff under /srv and then we might be ready to try booting again!


References


#journal