Very Computer

Microsoft documentation for Error 625 (which I've been getting) says to make
sure disk controller write caching is not enabled. Is this common knowledge
and practice, to disable disk controller write caching on SQL servers? Or is
it only a "possible" cause for some types of data corruption--to be tried
only if other solutions fail?

TIA

depends what controller it is really, I believe that Compaq supply the
only 'approved' write cache controller

Stu

Mike,

 Check with your raid controller vendor - as long as it doesn't have
any bugs then you should be ok.  Obviously don't run without battery
backup.

 Neil Pike MVP/MCSE
 Protech Computing Ltd

Neil,

Quote:>Obviously don't run without battery
>backup.

Roy

Correct - I'd have the UPS/APS as well though.

 Neil Pike MVP/MCSE
 Protech Computing Ltd

What is the purpose of this?  To get extended battery time beyond what the
UPS provides, or to cover any possible lag time between socket power cutoff
& UPS power starting, or something else?  How would "batteries be mounted on
the disk controller itself"?

Quote:>>Obviously don't run without battery
>>backup.

>Just to emphasize that this is NOT just plugging the server into a
>UPS, this means batteries mounted on the disk controller itself to
>keep the write cache alive.

>Roy

On the compaq smart-array 2 controllers there are what look like "AA"
batteries on the controller card to protect the write cache.

--
Richard Waymire, MCT, MCSE+Internet, MCSD, MVP (SQLServer)
Senior Principal Instructor/Consultant, ARIS Corp.
President, Pacific NW SQLServer Users Group (http://www.pnwssug.org)
Visit us at http://www.aris.com

Jamie,

A UPS can carry you through a power outage - for a while - but it
cannot handle someone removing the power cord or hitting the OFF
switch.  The batteries on the controller keep the memory on the
controller alive for a few days so that when power is restored the
controller can write any unwritten part of the write cache.  Without
such an arrangement you are heading for database corruption.

Roy

One additional thing to consider (?):

Hw crash of the disk controller. Battery backup wouldn't help here.

I've heard that DEC (on some controller) has an option where you use a cord
between the failed controller and a new one to transfer the cache (wouldn't
help if memory on controller crashed)...
--

MCSE, MCSD, MCT, SQL Server MVP
Cornerstone Sweden AB

Quote:> Jamie,

> A UPS can carry you through a power outage - for a while - but it
> cannot handle someone removing the power cord or hitting the OFF
> switch.  The batteries on the controller keep the memory on the
> controller alive for a few days so that when power is restored the
> controller can write any unwritten part of the write cache.  Without
> such an arrangement you are heading for database corruption.

> Roy

Tibor,

Quote:>One additional thing to consider (?):

>Hw crash of the disk controller. Battery backup wouldn't help here.

>I've heard that DEC (on some controller) has an option where you use a cord
>between the failed controller and a new one to transfer the cache (wouldn't
>help if memory on controller crashed)...

Roy