Re: Hash Table Infrastructure ran out of memory

From: Greg Rahn <greg_at_structureddata.org>
Date: Thu, 27 Aug 2009 07:14:09 -0700
Message-ID: <a9c093440908270714h6c57bf86t62a7b31606a0fe8c_at_mail.gmail.com>



As I mentioned, my guess is that the execution plan has a HASH GROUP BY operation and you are hitting a bug and hence your most productive means of resolution is to work with Oracle support. They will work with you to diagnose the issue to verify it is indeed the bug I mentioned.

As a means of "guess" debugging on your own you can try to disable the hash aggregation by using the hint
/*+ NO_USE_HASH_AGGREGATION */
and see if that avoids the bug, but the bug I mentioned has already been fixed and is part of 10.2.0.4 as well as available as a one off patch for 10.2.0.3.

I'm now sure how else I could help you dig into the issue.

On Thu, Aug 27, 2009 at 5:35 AM, Taral Desai<taral.desai_at_gmail.com> wrote:
> Hi Greg,
> Here is little more information i forgot to put
> We are doing kind of
>
> select x,y,z from abc partition(Q) group by x,y,z;
>
> Where Partition is hash partition.
>
> So, can you help me how to dig into this isssue
> On Thu, Aug 27, 2009 at 12:31 AM, Greg Rahn <greg_at_structureddata.org> wrote:
>>
>> This sounds like a bug with a hash group by.  I would recommend
>> working with Oracle Support.  Bug/patch 6471770 would be where I would
>> start.
>>

-- 
Regards,
Greg Rahn
http://structureddata.org
--
http://www.freelists.org/webpage/oracle-l
Received on Thu Aug 27 2009 - 09:14:09 CDT

Original text of this message