LWN.net Logo

The case for the /usr merge

The case for the /usr merge

Posted Jan 28, 2012 12:20 UTC (Sat) by rleigh (subscriber, #14622)
In reply to: The case for the /usr merge by deater
Parent article: The case for the /usr merge

"I ran a large cluster where /usr was mounted read-only by all the compute nodes,
which PXE-booted with a bare-minimal / in initrd (just enough to mount /usr). Worked well most of the time except for shutdown, when the initscripts just couldn't handle the idea that /usr would go away at unmount-all time. I suppose this proposal wouldn't necessarily make things worse, but it will make it harder to populate an initrd with a minimal / setup if no one is even pretending to try to have a version of mount that can live w/o /usr anymore. I guess you'd have to make it out of busybox or something instead of just copying over selected files from /bin, /sbin, and /lib."

Exactly, this is essentially describing a modern initramfs. Nowadays, there's no need to mount /usr over the network in a cluster environment--you just NFS mount the rootfs (which can include /usr), making the need for a separate /usr in this context redundant.


(Log in to post comments)

Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds