Forwarding this, as it contains some info useful for other folks.
Steve
-----Forwarded Message-----
From: Steve Lord <lord@xxxxxxx>
To: Andy Arvai <arvai@xxxxxxxxxxx>
Subject: Re: optimizing raid performance with xfs
Date: 19 Mar 2003 20:54:32 -0600
On Wed, 2003-03-19 at 20:42, Andy Arvai wrote:
> Hi,
>
> Thanks for the information. I looked at the man page for mkfs.xfs and
> there is no "-f" option. Is this something I need?
-f is needed to overwrite an existing filesystem on a disk.
>
> Also, according to the man page
>
> If an inode size is specified, or if a filesystem is
> sufficiently large, mkfs.xfs will warn if this will create inode
> numbers > 32 significant bits.
>
> What are the ramifications if I get a warning that inode numbers
> are > 32 bits? Will this cause a problem as the disk gets full?
An xfs inode number is an encoding of a disk address. The larger the
filesystem, the larger the inode numbers can potentially be. With
the default inode size of 256 bytes, it is possible for the inode
numbers to overflow 32 bits at 1 Tbyte. This is a problem for linux
where only 32 bits of inode space is available in system calls
and in the vfs layer. XFS has an option to restrict inode placement
to parts of the disk which will avoid overflowing 32 bits, this is
on by default on linux. This policy can affect layout and performance.
To avoid xfs operating in this mode, specifying inodes as being 512
bytes in size defers this policy switch over to a 2 Tbyte filesystem.
XFS inodes contain a fixed component and a variable component, extent
information can be held in the variable component, as can directory
entries of small directories. This is one reason xfs supports
different sizes of inodes.
So there is no danger of overflowing 32 bits of inode space on xfs,
but using this mkfs option may give better performance.
Steve
|