This is why we had to make another abstraction layer that was the "uniqueness
driver" This allowed the admin to configure a device without seial numbers to
be multipathed if they knew it was multipathed by explicitly specifing the
host bus, target, lun of all the paths to a speciic device. We never got the
chance to develop an intellegent way a determining multipathed-ness so we
left it to the user. Not ideal but it worked for our customer. I would like
to think there is a better way to determind uniqueness, but the nice thing
was we made the uniquness driver a module that could be replaced, thus
letting us develop new uniqueness algorithms in the future. I think the
concept is good, maybe not the implimentation. It also allows you to say you
don't like how we do... well you know the rest
Thanks
Brian
On Thursday 19 September 2002 5:16 pm, Bill Davidsen wrote:
> On Wed, 18 Sep 2002, Patrick Mansfield wrote:
> > Devices without serial numbers are treated as if they had different
> > serial numbers, they show up as if there was no multi-path support.
>
> That doesn't solve the problem, does it? If you have two devices w/o
> serial they could look like one with multipath, with the change you note
> that is prevented by making a single multipath device w/o serial look like
> two. I have visions of programs using /dev/st0 and /dev/st1, having used
> backup programs which grabbed every drive with a tape ready.
>
> It is indeed a messy problem.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Sep 23 2002 - 22:00:28 EST