RE: value from index block or table block

From: Mark W. Farnham <mwf_at_rsiz.com>
Date: Sat, 26 Nov 2011 13:53:33 -0500
Message-ID: <005101ccac6c$b342b2c0$19c81840$_at_rsiz.com>

That makes sense. When you get a chance, I recommend you get a dump of that index block with your favorite block editor to see if there are any more entries in it. Also, I worry a bit about whether the index structure to that block or just the leaf block is corrupt. I believe you may be still hanging fire. So you might want to try validate index to see how bad it is. I guess I'd make a test physical clone of the minimal set to test that validate on, because I'm not sure whether your release might mark the index invalid or dump core if it fails.

Is transportable tablespace (TTS) instead of a complete reload a possibility? Perhaps you could make a Segway (stupid word autocorrects Segway to a leading capital, as if it was only a dangerous personal transportation device and not a normal noun, just like apple. Hmm, MS Word treats apple normally) a tablespace at a time patching up with dblinks to the new place until it is all there. Even though your total size is pretty big ( was it 15 TB?) if you dictionary is reasonably modest TTS shouldn't take too long.

Good luck. That sure is nasty.

(In 1988 I had the similar experience of support telling me to export/import. But my database was 7 GB.

I know, you can hardly build an empty dictionary that small now. But with 25Mhz chips and 5600 RPM nearly no cache drives, an export was a 7 day operation. No one will ever know how long the import would have taken because it never completed.

Regards,

mwf

-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Grzegorz Goryszewski
Sent: Saturday, November 26, 2011 12:58 PM To: oracle-l_at_freelists.org
Subject: Re: value from index block or table block

On 2011-11-26 18:14, Mark W. Farnham wrote:
> Goodness. You dropped a data file out of the SYSTEM tablespace?
No I didnt :). Please see below explanation .

>
> Or are you writing that you dropped the tablespace containing the
> object referenced by the corrupt I_OBJ2 index? (All the indexes of
> sys.obj$ are born in SYSTEM unless you have an altered sql.bsq file.)
Exactly, that was the only table in that tablespace ,when we tried drop that table Oracle said object does not exists (because of corruption). So we offline dropped the only datafile for that tablespace .

>
> If that is the case, did the offline drop clean up the entries in I_OBJ2?
> Then I'm surprised just dropping the offending referenced object
> didn't do the trick.
> If no clean up took place, then I would think you still have a corrupt
> block in I_OBJ2.

Well not sure about that. Ill try to check this on monday . Thats realy interesting whats Oracle doing with objects (in dictionary terms) when You offline drop datafile they belong to .
Hope that explains Your doubts .
Regards
GregG

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


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

Received on Sat Nov 26 2011 - 12:53:33 CST