Re: [OT] VERY STUPID RAID QUESTION!

["matt_miles_uk" <m_miles@xxxxxxxxxxxxxxxx>]

Thank you all for your help - I had no idea this was such a
minefield!  Taking on-board what Phil says my server will only be
serving copies of DVD's I already own so ripping them again if it
breaks isn't a problem.  All my important files (e-mail etc) is
already part of a backed-up system.

So - taking what Ian has said onboard:

1) use standard IDE interfaces - cheap (best part - on a budget!)

2) let Windows configure it all - any specific version? Server or XP

3) be sure to check health of drives etc regularly...

hope I haven't missed anything out!

cheers,

Matt

--- In ukha_d@xxxxxxx, "Ian Lowe" <ian@w...> wrote:
> >What I'm trying to say is weigh up your needs ... If the risks of
the
> >storage system you're looking at are too great then use something
else
> >but there are risks associated with anything.
>
> Yes, I know that. I helped build a datacentre for Deutsche Bank,
and we
> spent a ridiculous amount of money on fault tolerant equipment -
each
> time you do, all you do is push things back to the next most likely
> failure mode, and chalk an extra zero on the end of the bill.
>
> To get 99.9% uptime costs ten times as much as getting 99%, to get
> 99.99% costs as much again. "three nines" reliability,
99.999% is
what
> Compaq were pushing with the high end Proliants, redundant PSUs,
disks,
> controllers, nics, running in clusters with active failover etc
etc.
>
> However, I'm keen to make sure this isn't seen as a debate about
value
> for money - it's not. I am not attacking Highpoint controllers
because
> they are cheap, or because higher priced raid controllers are more
> expensive.
>
> I am attacking highpoint controllers because they have a horrific
track
> record of causing data loss, and worse silent corruption of a file
> system. I'm not knocking IDE, or IDE RAID, or even Hardware RAID.
I'm
> warning anyone who will listen not to use controllers manufactured
by
> this company, because in my professional capacity I have witnessed
them
> going ass-up on a regular basis.
>
> And if you think this is just me - do a google:
>
> http://www.google.com/search?q=HPT%20corruption
>
> Hundreds of complaints both from highpoint cards, and from
highpoint
> chipsets integrated into motherboards, running hard raid, soft
raid, no
> raid at all, single disks, multiple disks, and always, worryingly,
the
> corruption happens without an OS complaint. It's not until a
critical
> file gets hosed that people start to notice.
>
> Of course, do a search for promise + "data corruption" and
you
will get
> a large number of hits too - but in this case, the vast majority
have
> lots of entries in the system log indicating that the disks are
having
> problems. It's less of a problem if you know it's happening.
>
> Another related feature of the Highpoint controllers when running
in
> RAID mode is to conceal the SMART info from the disks - running a
tool
> like AIDA, you get pre-failure diagnostics that can warn you if a
disk
> is looking a bit dodgy - run in a HPT array, none of that info is
passed
> on by the drivers, so you can have a disk squawking it's head off
that
> it's about to die, and you won't know until it does... (and takes
your
> entire array with it, of course ;) )
>
> >For my domestic use there was no other solution to getting 8
ATA133
> >drives hardware RAIDed
>
> Why use hardware raid at all?
>
> A perfectly adequate solution is to simply install regular IDE
> controllers from any vendor you like (although, I'd check that it
isn't
> using an HPT chip!) hook up your disks as regular IDE, and let
windows
> handle the volume configuration.
>
> When I had my Linux based mini-disaster, a lot of the guys said
that
> it's better to use software RAID under Linux too.
>
> >I hope I don't see my DVD server array go tits up but if it does
it
> does
> >so after I had to weigh up the pros and cons ... IDE RAID isn't
evil.
>
> Absolutely - IDE Raid isn't evil. But Highpoint is.
>
> >Just evaluate realistically what you *NEED* and what you are
prepared
> to
> >spend achieving it.
>
> Well, imo (and after all that's all it is, my opinion) the best
solution
> at this price point is simply to use regular IDE controllers (and I
> don't have any great thing for Promise - SIIG, Via and Intel all do
> chipsets that can be found in cheap IDE Controllers) and allow the
> Operating System whether that's windows or linux to handle the disk
> volume configuration.
>
> It's my opinion that this is also more recoverable - in most
cases, the
> disks can be attached to *any* controller in the event of a
hardware
> issue, and windows will sort it out - if using hardware RAID, you
need
> to get another identical (and I mean IDENTICAL - down to the
firmware
> version) raid controller.
>
> I.