Subject: Re: Another changer, another changer problem
To: Roger Brooks <R.S.Brooks@liverpool.ac.uk>
From: Giles Lean <giles@nemeton.com.au>
List: current-users
Date: 10/03/1998 08:13:01
On Fri, 2 Oct 1998 17:00:40 +0100 (BST)  Roger Brooks wrote:

> On Fri, 2 Oct 1998, David Maxwell wrote:
> 
> >Is there space in the disklabel to insert (or is there one already there?)
> >a unique name for a drive?
> >
> >What are people's thoughts about that, and teaching mount to handle this:
> >
> >mount -l dm_root / 
> >mount -l dm_usr /usr 
> >mount -l dm_var /var 
> >
> >Where -l is meant to mean 'name from diskLabel'. If mount knew this, and
> >fstab's format allowed, you could 'name' your disks, and then ignore
> >changes of Scsi ID etc...
> 
> Duh... You mean mount wanders off, groping all the disks it can find until
> it gets one with the right label?  What if it finds more than one with the
> required label?  I don't see how you could enforce a requirement that the
> label names be unique (particularly if you have removeable media).

Well, the logical volume managers I'm familiar with do label disks,
and do provide facilities to work out what is out there.  (vgscan on
HP-UX, dunno the name on AIX.) This is at the disk level, not
partition level.

Identification of disks is pretty much *required* in this world of
high end disk enclosures with alternate paths.  Then there are the
high availability setups where these alternate paths are connected to
different systems, which may or may not be using them simultaneously.

Drifting from the topic, I know, but if any redesign is going on might
as well think about the high end as well.

Giles