[Linux-cluster] One big GFS or a handful of smaller ones?

[Patton, Matthew F, CTR, OSD-PA&E]

Classification: UNCLASSIFIED

With CLVM on top of a couple of SAN volumes, I can grow the underlying VG
and LVs. Say I want HOME, VMImage, APPS, and CONTENT all on GFS so every
node can share. Is there a locking benefit or a performance benefit by
having one big LV (and therefore one GFS filesystem) with subdirectories to
organize things versus having multiple LVs each formatted GFS? In other
words, each node can have 1 or 4 GFS mounts. Kicking a node over means 1 or
alternatively all 4 GFS filesystems have to go into recovery. At first blush
the latter doesn't sound like a very good idea with more things to go wrong
and time out etc. Or is having multiple GFS mounts really not so bad and the
ability to selectively grow/shrink and unmount a filesystem a quite useful
benefit?

thoughts?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20060322/176246c9/attachment.htm>