On Thu, 27 Dec 2007, Jos Vos wrote:
On Thu, Dec 27, 2007 at 01:45:56PM -0800, Scooter Morris wrote:
Yes, I am aware of that. We're not looking to put things into
production on this cluster for awhile. However, are you telling me that
GFS2 is just flat unstable or not yet thoroughly tested? If the former,
then I won't waste my time even looking at it, if the latter, then I'm
happy to help with the testing by banging on it a little.
I don't have any concrete experience with GFS2, but I think it's the latter.
No, it's the former. If it can be reliably crashed in seconds by intensive
random writes to random files in random directory trees, it's way past
unstable. If it crashed after days of intensive abuse then that might fall
under the "not yet thoroughly tested" category, but that clearly isn't the
case at the moment.
Unfortunately, the GFS performance is for some applications dramatically
bad (I will test some new tuning options that should be in 5.1), so I'd
love to use GFS2, but I can't afford using non-production software for
that particular configuration.
If you expect performance differences of an order of magnitude or more
from GFS2 over GFS1, you are likely to be rather disappointed. But I'll be
quite happy to be proven wrong on that.
Gordan
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
