On Sun, 2005-09-11 at 12:20 -0400, Peter Arremann wrote: > All that said, I too would recommend going with Reiser or XFS. I can't recommend ReiserFS because it lacks interfaces and compatibility, ones that are at the heart of most Red Hat distribution deployments. Red Hat will never support ReiserFS for this reason, and not because of some Tweedie v. Reiser debate. Tweedie has his focus because of Red Hat's focus. SuSE continues to try to hack more and more interface/compatibility support for ReiserFS, including Extended Attributes (EAs). This is no different than when they original did various NFS support hacks. Frankly, I wish SuSE (let alone Red Hat) would put that effort into XFS instead. But SuSE is dedicated to ReiserFS, and Red Hat seems unwilling (often using incorrect assumptions on XFS' interfaces/compatibility, and absolute limitations in Ext3, in many replies) to join SGI in making Red Hat the absolute ultimate enterprise distro (IMHO) with the Ext3/XFS combination. > I once had a ext3 filesystem that had one damaged sector in the > journal... of course it fell back to ext2 behavior and the fs check > took all weekend :-) But it recovered. I'll take that level of trust any day over recovery time. But that's just me. > XFS will handle a defective journal much better - so the chance that > you ever encounter a situation where you have to do a full fs check is > much lower. Assuming the XFS kernel build is complete. The only XFS kernel build I have extensively tested as complete are the official SGI XFS releases. I had a really horrendous experience with the kernel 2.4 backport, and I would never touch a 3rd party rebuild. So far the XFS kernel build in newer Fedora Core 3 seems to be usable, but I haven't put a full load on it. I don't know how it compares to the CentOS Plus kernel at all. There are a few issues with NFS though, and that has been a disappointment. All my attempts to use the XFS from the SGI cvs tree atop of RHEL/FC has also resulted in additional issues. The reason why I adopted the official SGI XFS releases back in the early/mid-2.4 kernel was because of solid NFS capability as well as POSIX EA/ACL support, plus the xfsdump and other user-space utilities (which Ext3 still lacks). If I had to deploy a serious NFS server today (1+TB), I would deploy Solaris/Opteron instead, hands down, no hesitation. Otherwise, RHEL4/FC3 with Ext3 is fine for my typical deployments with filesystems commonly no larger than 100GB -- 1TB is an absolute maximum for myself and Ext3. I will not consider larger because 1TB is the "common denominator" of Ext3 filesystem size on _any_ kernel, _any_ hardware. -- Bryan J. Smith b.j.smith at ieee.org http://thebs413.blogspot.com ---------------------------------------------------------------------- The best things in life are NOT free - which is why life is easiest if you save all the bills until you can share them with the perfect woman