Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: NT striping dangerous with Oracle 7 or 8?

Re: NT striping dangerous with Oracle 7 or 8?

From: Steve Phelan <stevep_at_XXnospamXX.toneline.demon.co.uk>
Date: 1998/03/04
Message-ID: <889025255.12698.0.nnrp-08.c2de712e@news.demon.co.uk>#1/1

The other issue I have with battery backup is that it only covers crashes from the "Whoops, I've lost power..." point of view.

You also have to consider hardware failure inside the server - disks, memory, network cards, disk controllers, CPUs, internal power-supplies and so on. I've put together a few systems for large Banks where we've basically ended up mirroring everything - and I mean everything. Remember, a chain is only as strong as it's weakest link, and that includes a lot more than just the power supply...

It's worth pushing your vendor really hard on these issues, along with potential disaster scenarios - because it's suprising just how many 'misunderstandings' you can manage to pick out of their PR stuff in the sales brochures (and that includes Oracle).

Just my thoughts (again!) :-)

Steve Phelan.

bdurrett_at_ccci.org wrote in message <6djknn$uu$1_at_nnrp1.dejanews.com>...
>I think you are right Steve. Our new building is supposed to have a UPS
 and a
>generator to backup up our system, so that would theoretically prevent the
>battery backup from running out. But if the power went out on a weekend
 and
>things didn't work with the generator we would still be toast when the
 battery
>backup ran out.
>
>I checked into the post about disk drives write caching and I found out it
 was
>also true, but ours are set up with "write-through" type writing by
 default.
>So if we set our raid controller to write through we should be okay.
>
>We can still lose the data currently being written, but with database
 logging
>Oracle can recover it.
>
>- Bobby
>>
>> I sorta lost with the reasoning here, because you'd still be buggered
 when
>> the battery backup ran out... OK, you may have warning alarms, etc., but
 in
>> the middle of the night you might not be there to hear them.
>>
>> Surely a 'write through cache' is the only real solution?
>>
>> Steve Phelan.
>>
>> Eugen Nyffeler wrote in message <34FBDB7C.AE75B140_at_ubs.com>...
>> >bdurrett_at_ccci.org wrote:
>> ><snip>
>> >> Wow, so you are saying that you really can't protect your database
 from
>> >> losing data? If you lose writes to the redo log and to a data file
 wouldn't
>> >> you really be sunk? i.e. You couldn't even recover in that case. If
 you
 were
>> >> doing a lot of heavy updating and commiting transactions and the power
 went
>> >> out you could lose a bunch of stuff.
>> >>
>> >> But this is not NT specific I suppose. A Sun or HP unix box would
 have
 the
>> >> same issues, correct?
>> >
>> >Yeap,
>> >Any system which has no battery buffer will loos data which is only in
>> >the
>> >cache at this moment. Any system i know has this problem (Unix, NT,
>> >etc.).
>>
>>
>
>
>-----== Posted via Deja News, The Leader in Internet Discussion ==-----
>http://www.dejanews.com/ Now offering spam-free web-based newsreading
Received on Wed Mar 04 1998 - 00:00:00 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US