Re: Persistent device names

[Sven Luther <sven.luther@wanadoo.fr>]

On Wed, May 17, 2006 at 01:07:29PM +0900, Junichi Uekawa wrote:
> Hi,
> 
> > > - Are there likely to be architecture specific issues with persistent
> > >   device naming?
> > > - Or with filesystems other than ext2/ext3?
> > Nothing I know about. "tree /dev/disk/" should give you some ideas.
> > by-label links are user friendly, but may become ambiguous if users
> > carelessly duplicate file systems using dd. by-uuid links have the same
> > problem *and* need a suitable identifier in the file system metadata.
> > by-id links are more unique but linked to the physical hardware (you
> > can't have it both ways...) and are a bit ugly.
> > by-path links are very descriptive and really unique, but are *really*
> > unique and will change if a disk is moved to a different bus position
> > or something like this.
> 
> 
> Some enterprise-level storage systems have disk-level duplication
> system, as such, filesystem-label-based mount-point would not be the
> end-all way of approaching the problem.
> 
> It might be less error-prone if bus/card info is used in addition to
> the filesystem UUID.

I was thinking the same thing. UUID + by-path if two UUIDs are the same, with
the first of them responding to UUID and the subsequent to a UUID-# kind of
scheme or something, where # is ordered by growing path or something.

Friendly,

Sven Luther