[Gllug] Maximum size of NFS filesystem under 2.6 kernel

[Steve Nelson]

On 11/26/05, Bruce Richardson <itsbruce at uklinux.net> wrote:

> What's the maximum size of filesystem that can be exported via NFS under
> a 2.6 kernel (assuming 2.6 kernel on both server and client)?

Assuming you're using NFSv3, which uses a 64 bit field for file size
and offset, the theoretical limit, regardless of other determining
steps is going to be:

$ echo "2^64" | bc -l
18446744073709551616

ie 18 exabytes - this is substantially larger than the largest
filesystem available under LVM or XFS.

How will the filesystem be presented beneath NFS?  Are we talking
aboout a massive LUN presented as one huge block device, which we're
then going to carve up?

The next limiting step will be the combination of underlying
filesystem and architecture.  If you have 64 bit architecture, which I
recall you do, and, for some reason, you wanted to use ext3, then with
an 8k blocksize you could theoretically approach 32TB.  However, you'd
waste a lot of space with superblocks and the like, and would have to
deal with the irritation of periodic fsck-ing, unless you disabled it.

If you used XFS you could reach a filesystem of 9EB.

I'm not sure of the figures for LVM - I'm sure others will be able to advise.

> Bruce

Steve
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug