On Tue, 2002-01-29 at 16:13, Vincent Danen wrote: [ ... ] > Anyways, I agree with some of the limitations of reiserfs... I don't > care too much about the lack of dump/restore... I never use it. NFS > over reiserfs works great now... I have no issues with NFS+reiserfs > anymore since 8.1.
dump/restore is a very high priority. In fact... dump/restore works better in XFS than in ext2. The dump in ext2 will hang if the dump partition is mounted and has any dirty inodes in ram (and not committed to disk). This requires the ext2 FS to be unmounted for dumps to complete. (another sysadmin talked to linus about the problem and he said dump was foobarred in 2.4 and no fix was expected till 2.5 when they have a real API to handle the problem...). I guess some people hav'nt had this problem... or maybe they have a workaround but we had many many many many failed hung dumps with ext2... We are required todo FS dumps and keep the backups for at least 1 year for data. With ext2/3, RFS, JFS, they all lose on this requirment. :( In addition we use NFS heavily. We have files that range from very small... to well over 2GB. I dunno what the deal is... But as of mandrake 7.x days NFS just sucked with RFS... So we were stuck with ext2. (since ext2 dump stopped working in 2.4 on any activaly mounted fs we had to rsync to a sun, then do a ufsdump) When 8.1 came out we switched over to XFS for everything except /boot. We've never had a problem with the FS at all. No OOPS... no corruption. Nothing. Just rock solid. That's with NFS mounts from Linux+SUN clients, >2GB files, and heavy database pounding (pgsql). The biggest plus was dump started working... SGI did a good job on making sure highly needed utilities like this worked. (I know... maybe not _everyone_ needs it... but many do...) > I'm not even too concerned about speed... I mean, if xfs is blazing > fast compared to reiserfs, I'd switch in a heartbeat, but if it's > marginally faster, then I wouldn't worry. However, my biggest beef > with reiserfs was the lack of quota support. We have some files >2GB and the speed factor does come in... XFS is faster. > So two questions since I'm too busy to go hunting myself right now and > you seem to know quite a bit about xfs: > > 1) Does it support f/s quotas? Yes, in fact it does better than any other FS in linux... no quotacheck needs to be run! (quota info is in journel) > 2) Is it now in the stock kernel (ie. if I download from kernel.org, > do I need to do any patches to make it work with xfs)? No. They keep submitting the patches to Linus and Alan but they don't even get any feedback... It might be because there is some VM optimizations SGI has added... The FAQ on SGI's webpage doesn't have an answer... they are pretty sure it wil be in 2.5 eventually.. and then in 2.4 later.... ( http://oss.sgi.com/projects/xfs/faq.html#linuskernel ) > I've heard about XFS acl's and such that are supposed to make using > Samba really sweet... do these same acl's (of which I know nothing) > work on the local f/s? Ie. can I use xfs acl's to arbitrarily make a > file read-only even from root? Can't help you there... Our Samba stuff is running on a Sun/Solaris 8 server... You should take a look at http://www.mandrakeforum.org/article.php?sid=1212 for some more stuff on FS's. "Well, if you are new to linux, the best bet is probably XFS." -- Bryan Whitehead SysAdmin - JPL - Interferometry Systems and Technology Phone: 818 354 2903 [EMAIL PROTECTED]
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
