Re: recovery issue...

From: Volker Hetzer <firstname.lastname_at_ieee.org>
Date: Thu, 02 Nov 2006 21:39:40 +0100
Message-ID: <eidl2p$q24$1@nntp.fujitsu-siemens.com>

hpuxrac schrieb:
> Volker Hetzer wrote:

>> Volker Hetzer schrieb:
>>> Hi!
>>> (10.2.0.1, Linux RH ES 4)
>>> We've got a tablespace with some logging and some nologging tables.
>>> Now we've got some corruption and I took the tablespace offline for
>>> recovery.
>>> Turned out, the corruption was in the nologging part, recovery failed
>>> and whenever I try to do put the tablespace online I get a message
>>> that it needs recovery.
>>> I'm perfectly happy with losing the nologging tables but how do I
>>> achieve that?
>>> I've got incremental backups until monday and archivelogs until today.
>>> I've opened a SR but so far oracle hasn't responded.
>> Ok, I got to talk to support.
>> They say the backups are indeed corrupt and it has nothing to do with
>> nologging operations.
>> What I see from the rman messages is that the monday backup was just fine,
>> the tuesday backup failed because of a corrupt block in the users tablespace
>> and the archivelogs on tuesday are corrupt.
>> Great.
>>
>> How can such a thing happen?

>
> First you shouldn't mix nologging and logging things in the same
> tablespace. That's beginning to play with fire from my point of view.
Yeah, got that in one.
I'll give some serious thought about splitting the spatial app off but I'll have to check whether spatial (locator) can be taken out of the system tablespace too. That's what prevented a tablespace point in time recovery.

>
> You can try a couple of things but you should probably keep working
> with oracle support.

I do.

>
> You should be able to use rman to fix the block corruption in the users
> tablespace ( except maybe if you can't read your archived logs? ...
> what do you mean they are corrupt ).

rman reports corruptness in the alert.log. What happened was that there was a corrupt block in the database. My normal weekday backup is
backup database incremental level 1 plus archivelogs. The archivelogs were backed up, the datafile backup aborted the job, no restore validate took place and when I tried to use the archivelogs for recovery they were found to be corrupt too.

>
> You can try recovering the "good backup" from monday to a separate
> database and get the data that you want from that tablespace.
I fear that will be the case.

> Block corruption is just something that I encounter in my environment.
> It's a huge danger sign of possible instability in your disk subsystem
> somewhere.

Ok, but what can one do with asm? At least with file based stuff I could run fsck.
I fear that may be an aftershock of a RAID controller crash we had a few weeks ago. We added an archivelog target on our netapp filer in order to cover controller issues. Worked out great. :-(

Volker

-- 
For email replies, please substitute the obvious.

Received on Thu Nov 02 2006 - 14:39:40 CST