Feasability of VERY large ext3 file system?

Eric Swenson eric at reaction-eng.com
Wed Dec 4 15:49:00 CST 2002


On Wednesday 04 December 2002 01:35 pm, Seth Mos wrote:
> At 12:56 4-12-2002 -0700, Eric Swenson wrote:
> >NOT true.  I'm setting up something very similar right now with a
> >brand-spanking new Poweredge 4600, external 2TB u160 storage array, and
> >Redhat 8.0.  I'm mounted it as both ext3 and reiserfs so far, and they can
> >both use the full 2TB (minus a couple of small quorum partitions for using
> >kimberlite or redhat AS clustering later).
>
> And have you actually used more then 1TB of the available space. Some
> things only pop up after you actually start to use it.

Ah.  Good point- I've been stress testing it so far, but I haven't filled it 
up yet until I'm satisfied that everything is stable- a catch-22.  I'll split 
it in half before I use it just to be sure.

> >BTW, has anyone here gone through the trouble of recompiling a redhat 8.0
> >2.4.18-18 kernel with XFS, and if so, what version of XFS?  Can you share
> > out stuff with NFS fine after that?
>
> There is a 1.2pre3 release available on the FTP site.
> ftp://oss.sgi.com/projects/xfs/Release-1.2pre3/
>
> 1.2 is just around the door since no large problems have been found since
> pre3. Watch that space or the XFS website for a release. I'll forward the
> release message here when it is.

Unfortunately, there is a *big* nasty bug with file corruption when using 
NFSv3 with udp to export XFS v1.2pre3 filesystems- see 

http://oss.sgi.com/bugzilla/show_bug.cgi?id=198

Since I have a fairly controlled environment, I can try force the clients to 
mount with nfsv2, or maybe using tcp (if that's enabled in this rh8 kernel- 
haven't tried that yet).  Anyone here using 2.4.17-2.4.18 + xfs 1.2pre3 and 
having problems on their servers?




More information about the Linux-PowerEdge mailing list