Re: Oracle DataGuard to prevent corruption?

"Vladimir M. Zakharychev" <vladimir.zakharychev_at_gmail.com> wrote in message news:1172642803.350050.323460_at_8g2000cwh.googlegroups.com...
> On Feb 28, 7:13 am, Magnus Warker <mag..._at_warker.co> wrote:
>> Dear group,
>>
>> we encountered a serious problem with our Oracle database cluster. Data
>> corruption occurred and a fixed set of records is not readable anymore.
>> It
>> seems that this was caused by heavy loads, but there still is no
>> clarification of what really happened.
>>
>> Our consultant made the advice to setup a new installation with a product
>> named DataGuard. This would operate with two copies of the same database,
>> and, when data corruption occurs on one database, we still would have the
>> other one.
>>
>> What do you think of this kind of workaround, especially from the point
>> of
>> view that we are already operating a high available database cluster,
>> which
>> would become needless in this case?
>>
>> Thank you
>> Magnus
>
> With shared storage cluster, like RAC, storage is still a single point
> of failure , as demonstrated by your case: although physical storage
> is still available the data is corrupt and unusable. Clustering
> provides for high availability, but not necessarily for high
> durability - there are bugs in software, there are glitches in
> hardware... That you have a clustered database doesn't mean it can't
> fail. It can and it will, some day. Proper backups are still essential
> for clustered databases, and Data Guard (which is actually a new name
> for technology behind standby databases,) can greatly simplify things.
> For example, if you had a standby database, and assuming that the
> corruption was not caused by a bug (which could trigger on standby
> while rolling it forward and introduce the same corruption into it,)
> you could simply switch to that standby db and then rebuild the
> primary (for example, making it standby of the new primary db and then
> switching back to it,) without compromising availability a bit.
>
> Regards,
> Vladimir M. Zakharychev
> N-Networks, makers of Dynamic PSP(tm)
> http://www.dynamicpsp.com
>

I would add for clarity that most Data Guard configurations include at least one standby database located on hardware separate from that of production and that this hardware platform need not be identical to production. Even if production is RAC/OPS, the standby system can still be single-instance non-RAC/OPS.

-klb Received on Wed Feb 28 2007 - 00:44:18 CST