RE: How does one manually corrupt a database (database file, redo logs and backups)

Charles

> Funny topic, eh? I am practicing backup and recovery procedures and
> finding that recovery is quite easy since I never have corrupted
> blocks to worry about. So, how do I practice for a "worse case"
> scenarios?

IMHO you already get excellent advices. I would just add that ideally you should not only check physical corruptions (RMAN reports such blocks as "FRACTURED" in v$database_block_corruption) but logical corruptions as well.

About logical corruptions consider the following: - RMAN, during backups, doesn't check them per default... CHECK LOGICAL must be added.
- The only way to produce them, if you are not lucky with random changes ;-), is to delve in the structure of the blocks, typically with BBED, and carefully change the blocks in a way that they are not recognized as physical corrupted.

Another point is that RMAN doesn't detect inconsistencies (e.g. between DATA and INDEX segments). Only ANALYZE do it.

HTH
Chris

> Are there white papers out there that go over this kind
> of stuff? I found several "Best Practices" in metalink (look Mladen,
> not plural), but that is not exactly what I am looking for as they
> tell you how to fix the problem, not make the problem. I have tried
> overwriting the binary files, but being simpled-minded I was using
> vi. I thought about using the unix dd utility - unfortuantely I do
> not have enough experience to be surgical (easy to mess up the first
> few blocks, but what about the middle?). Can I use the same method
> to isolate corruption to various parts of the database (ie, database
> files, redo logs or rman backups)?
>
> I have a gut feeling that someone has already documented this, so if
> this is the case, please point me there.
> Thanks,
>
> PS - I have been practicing on 9iR2. Yet anything above 9iR1 would
> be fine.
>
> --
> Charles Schultz

--
http://www.freelists.org/webpage/oracle-l