[Tfug] cluster node syncing using rsync and NFS?

[Jeremy D Rogers]

Hi all,
I have another one run by you people with cluster experience. I
greatly appreciated all the advice and decided to have local installs
on all the nodes. Now I'm looking at how to sync all the nodes when I
install new software and such. I'm currently planning to have a chroot
on the masternode that can be NFS mounted from a script on the nodes
and rsynced to the rest of the node's filesystem. Something like:

rsync -plarv --progress                  \
          --exclude /etc/hostname     \
          --exclude /etc/hosts           \
          --exclude /home                 \
          --exclude /dev                    \
          --exclude /proc                  \
          --exclude /sys                   \
          --exclude /tmp
          --exclude /nfsmount           \
           /nfsmount/chroot/   /

I might also want to run update-grub at the end of the script too.

The other option that I can see is to setup a PXE boot that installs
locally on the node and updates software from a chroot on the master.
This seems excessive and bandwidth hungry for just installing
software.

Are there any reasons not to use rsync the way I'm planning? The only
thing I see so far is it requires nfs to be no_root_squash, but since
this is a local internal network, I guess that should be ok.

I realize this might not take care of everything, but I can setup PXE
boot to actually reinstall the whole local image when anything goes
wrong, which should be pretty infrequent. I just thought rsync across
NFS would be more efficient for simple software updates and other
small changes.

Any thoughts?
Thanks,
JDR



--
Jeremy D. Rogers, Ph.D.
Postdoctoral Fellow
Biomedical Engineering
Northwestern University