Eric Sandeen wrote:
On Mon, 22 Dec 2003, Stephan L Jansen wrote:
Hi,
Back in October Steve Lord said in a post to this group to make sure
and use
Hm, October of which year?
"-i size=512" when making a xfs filesystem > 1 TB. I have four
filesystems
larger than 1 TB that were created without using this switch. Should
I be
worried? Is this a performance issue or is there a possibility for
data
corruption? Thanks.
He was probably referring to problems with inode numbers > 64 bits,
which could be alleviated by making the inode size greater, IIRC.
The Linux XFS code has been restricting inode numbers by other means
for quite some time now, so as long as you don't have ancient code
you're probably fine.
Dredge up that post again, if it was from this year please post
the URL and I'll take a look, maybe Steve was talking about another
issue that I'm not remembering at the moment...
-Eric
You mean 32 bits ;-)
With the default parameters, inodes above the first 1 Tbyte of the
fs will take 33 bits to address. There are mount options (default on
linux), to keep inodes down in the first 1 Tbyte to avoid
this. However, things will behave a little better with the larger
inode size.
I cannot remember another reason for suggesting this configuration.
Steve
|