Re: Change block tracking and corruption detection in 10g

On 11/14/06, Ram Raman <veeeraman_at_gmail.com> wrote:
>
> Hi all,
>
> Some time ago there was a discussion on how we would detect the corrupt
> blocks if the blocks are skipped from being read during RMAN backup because
> of change block tracking. I was in an Oracle training class and I asked the
> question to the instructor. His answer was that if the block was not being
> accessed it could not go corrupt. ie. If the block is changed then it is
> tracked by Oracle, if it is not accessed and changed how it could go
> corrupt. Any comments?
>

If the sector of the brown spinning stuff that the un-accessed block lives on
has a defect and causes a physical corruption, the database block may be unaffected.

Or if your SAN/Disk Controller/NAS has a hiccup, something bad might happen. ( It *does* happen)

This can be used to check for corruptions

run {

    allocate channel d1 type disk;
    allocate channel d2 type disk;
    backup check logical validate database;     release channel d1;
    release channel d2;
}

the view V$DATABASE_BLOCK_CORRUPTION will be populated if corrupt blocks are found.

-- 
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist

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