In article <41294N189@web2news.com>,
Stefan Coors <scoors.news.invalid@web2news.net> wrote:
>Chris,
>
>I'm curious to know why you don't see this as a feasable solution...
>...instead of buying some hughly overpriced external storage arrays
>(like D1000/A1000 or Hitachi FCAL SA).
>We're about to try this "unsupported" feature with the V880
>and I'd like to get some insight from people that - Successfully - tried
>this.

Reason why this does not seem to be a "good" solution (not saying it's
technically impossible): if you share internal disks between the systems
and are trying to produce a highly available configuration, what happens
when you have a hardware issue on one of the nodes that requires that the
entire node be brought offline, including the internal disks?

About the only way that this is a tennable solution for HA is if:
- OS partition(s) are mirrored on two locally internal drives
- internal disks are fenced off into local and shared disks
- local internal disks comprise one half of mirrored volume(s)
- remote shared disks comprise other half of mirrored volume

That way, if you lose an entire node, you still have all of your data
available (even if in a non-redundant state).

Of course, this means that all of your RAIDing is done in software, which
may be sub-optimal for your application.

-tom

--

"You can only be -so- accurate with a claw-hammer." --me