[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: ICP Vortex GDT6513RS
- To: misc_(_at_)_openbsd_(_dot_)_org
- Subject: RE: ICP Vortex GDT6513RS
- From: "Lawrence W. Smith" <lws_(_at_)_juiceco_(_dot_)_com>
- Date: Sun, 22 Apr 2001 00:11:12 +0100
> Hi Nick,
> --- Nick Holland <nick_(_at_)_holland-consulting_(_dot_)_net> wrote:
> > I would politely disagree.
> > I do not view RAID as a "guaranteed keep running"
> > solution, but rather
> > as part of a "rapid repair" solution.
> I have never seen a RAID controller manufacturer offer
> a guarantee to "keep running" however live replacement
> of a failed drive is an advertised and proven feature
> of SCSI RAID controllers. If it were not I would not
> think all SCSI RAID controllers would have it or spend
> resources maintaining the software which performs it.
> > There are just too many times when one drive fails
> > and "shuts down"
> > another drive on the same SCSI channel or jams the
> > entire channel.
> Oof. I suppose this happens but I have not been a
> victim of it. I have seen my share of drives fail but
> none have been so bad as to take down a whole channel.
If you make a point of using cards with a significant
(to application) cache size and on board battery backup,
such cards seem to very reliably cope with multiple
failures (1 dead drive / locked bus) with minimal loss
of data due to reset / rollback.
I've seen the onboard firmware pull some 30 odd Mb of
data off the cache and reconcile it with the state of
the good, but locked bus drives and do so in reasonable
time. (mostly on compaq / dec hardware but I doubt others
differ so much)
Of course if you lose a second drive you're stuffed anyway ...