Re: another XFS+LVM+SoftwareRAID5 query

[Steve Lord <lord@xxxxxxx>]

Jan-Frode Myklebust wrote:

> And, what consequence does the increased sector size have on the fs?
There are some chunks of metadata at the start of each allocation group
which where originally layed out as being 512 bytes long. The superblock
and some headers for allocation structures. The log is also written
in sector sized chunks (or multiples thereof). All the rest of the
metadata and file data is in filesystem block sized chunks.

Bumping the sector size rounds these up to a larger size, the change
in disk space usage is tiny.

The reason xfs uses two different sizes is that otherwise all I/O would
have had to be submitted in 512 byte chunks and the overhead is horrible.

Steve