[Linux-cluster] other cluster FS

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

Classification: UNCLASSIFIED

with the latest community kernel release I was looking at oracle's FS and
then SGI's CXFS. I liked the CXFS paper and wondered if GFS was working in
similar fashion where meta-data is stored by a 'master' and then pieces of
it farmed out to nodes as required. So essentially all meta-data resides on
the first one to mount GFS I guess and it handles the periodic flushing of
changes? Is there a machanism to designate "backup" masters or does this
arise implicitly by virtue of a neighboring node(s)recovering a peer's
journals after an election?

CXFS is/was available on seemingly a number of different OS's. Anyone
actually using it? Is GFS ported to anything besides Linux? I want and
intend to standardize on Linux as the bare-metal OS, but being able to play
nice with winblows would be really cool. Does CXFS use DLM and other pieces
of the Linux Clustering package set? I was really not impressed by the claim
that OCFS2 panics the machine on error. Is that for real?

Anyone try IBM's GPFS?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20060320/2be04833/attachment.htm>