10TB filesystem possible on Red Hat 4 EL System?
Peter Kjellstrom
cap at nsc.liu.se
Thu Aug 16 02:07:00 CDT 2007
On Wednesday 15 August 2007, Sam Flory wrote:
> Peter Kjellstrom wrote:
> > On Wednesday 15 August 2007, Sam Flory wrote:
...
> >> CentOS 5 x86_64 will make things a whole lot easier. You can then
> >> either install the xfs module, or the kernel with the xfsprogs rpm from
> >> the Centos extras tree.
> >
> > Not quite true nor complete, the xfs stuff lives in centosplus
>
> Yup I was confusing Centos terminology with old fedora terminology.
Actually I should check my facts better, it has moved to extras *slaps himself
around a bit*
> > and you will
> > not need Centos-5 for this. Centos-4 would work too.
>
> Yes, and no. As the xfs code doesn't get update in the RHEL/Centos
> kernel. The Centos4 is about 3 years behind on bug fixes,
Not true at all. For both centos-4 and -5 you use the kmod-xfs package with
the normal kernel. Not the ancient lets-compile-2.6.9-with-xfs-enabled.
...
> >> but I've yet to meet
> >> people who are very happy with the results over a TB, or 2.
> >
> > I'm one. Depending on the hardware ext3 can be as good as 10-15% slower
> > than xfs on such a size.
>
> More like 25-40% on RHEL 4. RHEL 5 performs a lot better in most use
> cases. Generally to making ext3 perform well requires writeback
> journaling which sort of throws out most of the argument for ext3.
As I wrote "depending on hardware" (and other thigns too of course). Ext3 is
actually extremely sensitive with which hardware it likes. On some
controllers I've seen, as stated, 10-15. Other controllers, for the same
sequential write, take >100% hit with ext3... (areca, 3ware are examples
where ext3 truely sucks).
> > And that is, IMHO, not worth the xfs hassle (manual
> > intervention on kernel updates, dangerous full fillsystem situations,
> > less mainstream...).
>
> If you run centos you just point your yum config at the centosplus
> repo, and be sure you are running the centosplus kernel.
As said before, use the normal kernel and the kmod-xfs package. But even now,
you have to manually handle stuff at upgrade (kmod-xfs lags the normal kernel
release, kmod-xfs needs to be matched to the kernel and yum does not do this
automatically...).
/Peter
> Ext3 people
> wave their hands and talk about the dangers of xfs, and reiserfs, but
> I've used them for years in production, and never had an issue. Now yes
> if you are using a laptop, or crappy ide/sata drives. Ext3 is a safer
> choice. When your drive goes horribly wrong ext3 is going to gives you
> the best chance of recovery. In general if you're running a 10TB
> filesystem you're on a solid raid array, and run regular backups. (If
> you have 10TB of data and don't have backups you are just plain insane.)
...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part.
Url : http://lists.us.dell.com/pipermail/linux-poweredge/attachments/20070816/1dabb90e/attachment.sig
More information about the Linux-PowerEdge
mailing list