rman block recovery

[zigz...@yahoo.com]

I have a database which has database block corruption. I had a successful rman backup of Sunday.
Monday and Tuesday rman backup failed saying
ORA-19566: exceeded limit of 0 corrupt blocks for file F:\EM…….

Then I modified rman script to skip up to 5 corrupt blocks..
set maxcorrupt for datafile 'F:\......' to 5;

Now my rman backup is successful on Wednesday…

Now, I want to repair blocks using rman’s block recover command:
recover datafile 19 block 321, 322, 385, 513, 577 restore
My questions are:
a. Will rman automatically start with Sunday’s successful back and then apply archived logs.. or
Will it start with Wednesday’s backups and then it won’t be able to recover corrupted blocks…
b. How to force rman block recover to start from Sunday’s backup … (Do I have to mark Wedensday's backup as invalid in rman catalog ...

[zigz...@yahoo.com]

My version of database is 10.2.0.4 on Windows 2003 R2.

[ddf]

> My version of database is 10.2.0.4 on Windows 2003 R2.- Hide quoted text -
>
> - Show quoted text -

What level backups are these? RMAN will take the most recent level 0
backup, recover the blocks from that then apply changes to those
blocks from the most recent cumulative level 1 backup until the blocks
are considered recovered.


David Fitzjarrell

[zigz...@yahoo.com]

David:
Thanks. I found by trial and error that it picks up the latest backup even though it skipped the corrupt blocks instead of one on Sunday which does not have any corrupted blocks. I have to use rman’s change backup.. Unavailable command to make most recent backup unavailable, then it goes back to Sunday’s backup.

[joel garry]

The 10.2 docs seem to say only full backups and archived logs are
used: http://docs.oracle.com/cd/B19306_01/backup.102/b14191/rcmconc2.htm#BRADV118
I think it is different in 11. I have no real experience with this.

There are examples of, for example, tags or time:
http://docs.oracle.com/cd/B19306_01/backup.102/b14194/rcmsynta010.htm#RCMRF108

jg
--
@home.com is bogus.
http://www.itpro.co.uk/640018/oracle-reveals-raft-of-xeon-e5-based-sun-servers

[zigz...@yahoo.com]

On Wednesday, April 11, 2012 11:44:13 AM UTC-4, zigz...@yahoo.com wrote:

I wonder when block recovery is done will Oracle stop as soon as block is "recoverd" or will it keep applying changes from archived logs to that block until it runs out of archive logs. If a point in time recovery is applied, then that block can be incosistent with all other blocks because of incomplete recovery ...

[Mladen Gogala]

On Wed, 11 Apr 2012 19:05:52 -0700, zigzagdna wrote:


> I wonder when block recovery is done will Oracle stop as soon as block
> is "recoverd" or will it keep applying changes from archived logs to
> that block until it runs out of archive logs. If a point in time
> recovery is applied, then that block can be incosistent with all other
> blocks because of incomplete recovery ...

Sounds like a good thing to test. BTW, you should keep your databases on
SAN, no corrupt blocks there.

--
http://mgogala.byethost5.com

[zigz...@yahoo.com]

My database is alreday on SAN but somehow my data center SAN has issues causing corruption.

[joel garry]

Look again at the link I posted, restrictions, you can't do PITR with
block recovery. But the example shows you can limit the time of the
backup restored _from_. An interesting distinction, eh? It would be
inconsistent, but I'm presuming the inability to write to it once
corrupt would prevent the inconsistency, because the final redo stream
vector would be the last good one. There should be some testing done
around this, presumptions aren't so good, and something is bothering
me about the redo getting ahead of the data file, especially if the
corruption is not detected until backup time.

jg
--
@home.com is bogus.

http://www.utsandiego.com/news/2012/apr/11/developer-bank-apps-fight-usaa-patent-infringement/

[Mladen Gogala]

On Thu, 12 Apr 2012 09:13:19 -0700, joel garry wrote:

> Look again at the link I posted, restrictions, you can't do PITR with
> block recovery. But the example shows you can limit the time of the
> backup restored _from_. An interesting distinction, eh? It would be
> inconsistent, but I'm presuming the inability to write to it once
> corrupt would prevent the inconsistency, because the final redo stream
> vector would be the last good one. There should be some testing done
> around this, presumptions aren't so good, and something is bothering me
> about the redo getting ahead of the data file, especially if the
> corruption is not detected until backup time.

I have never used block recovery with RMAN. Modern SAN equipment is very
reliable, I've never had any need to do so. Have you done it? What were
the circumstances?



--
http://mgogala.byethost5.com

[joel garry]

Haven't done it. The only corruption I've seen is a damager
impatiently removing too many drives on a RAID-5, during testing.
That did not corrupt a few blocks...

jg
--
@home.com is bogus.

http://webdemo.visionobjects.com/equation.html?locale=default

[TheBoss]

joel garry <joel-...@home.com> wrote in
news:cf21d93e-57f1-4339...@x17g2000yqj.googlegroups.com:
<...>

>
> Haven't done it. The only corruption I've seen is a damager
> impatiently removing too many drives on a RAID-5, during testing.
> That did not corrupt a few blocks...
>

You can't remove too many drives on a RAID-5, baarf, baarf, baarf ...

--
Jeroen

[zigz...@yahoo.com]

Joel:

the link I posted, restrictions, you can't do PITR with
> > block recovery. But the example shows you can limit the time of the
> > backup restored _from_. An interesting distinction, eh?

Thanks a lot for pointing this,