On Fri, May 23, 2008 at 8:01 AM, John Summerfield < [EMAIL PROTECTED]> wrote:
> Matthias Saou wrote: > >> Sandor W. Sklar wrote : >> >> (Not to hijack this subject, but I wish there was some resource or >>> community that had experience in this area. I'm liberal in politics, but >>> uber-conservative at work. I don't want to be the "first" to do something >>> weird, like 8 TB filesystems. I want to learn from the pain and experience >>> of others. :-) >>> >> >> I've got a few production 12+ TB filesystems which are working just >> great. I'm using RHEL5 with XFS. The custom dkms-xfs package and rebuilt >> xfsprogs I use can be found here : >> >> http://ftp.freshrpms.net/pub/freshrpms/redhat/testing/EL5/xfs/ >> >> Note that you should install the proper kernel devel package for your >> system for the XFS module to be able to rebuild. >> >> Note also that this is completely unsupported by Red Hat or by me ;-) >> >> Matthias >> >> Before I used a filesystem not officially supported by RH, I would > clarify with RH what it does to our support agreement. > > In Red Hat's shoes, I might well say, "Go away or pay lots more dollars." > I'd rather use CentOS than pay money and find there's no support when I need > it. > > As to the size of filesystems, I've not tried it myself, but I saw one > report of unacceptable e2fsck times on a filesystem less than one TB (IE one > disk) in size. It's something I intend to try, I don't know whether it's the > software or the user that's broken. > > > > Thanks to everyone for their replies. To answer someone's point about backups, this is for a disk based backup system. :) It has the ability to replication at the application level to other systems. I also agree that I want a filesystem that Red Hat will support. GFS2 is a possibility, I just have no experience with it. On the surface, I am not sure it was designed to be the home of over 150 million files rather a smaller number of large files (databases). To someone else's point, checking that filesystem, journaling, etc. are all greatly affected by these large numbers. I am going to hammer out with the software vendor why the single volume. I have my doubts on this requirement. It centers around DR/restore flexibility if I understand it correctly which doesn't necessarily apply given it is a replicated system. I would really just prefer to do multiple 4TB ext3 volumes on top of LVM. :) Just to throw out another thought on fsck and performance, do the user space tools for GFS2 behave similarly as those of ext2/3 in that fsck/healing processes can only occur while the volume is unmounted? Is there a way to tune this like with tune2fs?
_______________________________________________ rhelv5-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/rhelv5-list
