Re: cache buffers chain latch contention

From: Mladen Gogala <mladen_at_wangtrading.com>
Date: Thu, 25 Mar 2004 11:43:09 -0500
Message-ID: <20040325164309.GA3499@mladen.wangtrading.com>

On 03/25/2004 11:10:17 AM, "Koivu, Lisa" wrote:

> Investigation of v$system_event indicated latch free wait. OK, fine, so
> I look at v$latch and find I have enormous misses and sleeps on cache
> buffers chains. I chased down the child latches and pinpointed the
> affected objects. I had several hash partitioned indexes (4096
> partitions, yes I went crazy, hey why not?) that were on the latches. I
> rebuilt them range partitioned, tried the mv build again, same problem.
> OK, fine, then what?

Lisa, did you check alert.log for ORA-600 errors? Sounds like you might have encountered some.

>
> My impression of this latch, and what happens, is this (PLEASE correct
> me if I am wrong): A db block can house records from multiple tables.

A DB block can not house records from multiple tables, unless the tables are clustered (not in the RAC sense). Are the tables in question clustered? Also, I'm not sure how would one created clustered partitioned tables. I've never tried it and I'm not at all sure that it would work. A DB block can belong to only one extent, which in turn, can belong to only one segment. Table segment is synonymous with the table it belongs to. If that wasn't true, half of my allocation computing scripts would fall apart in a very ugly manner indeed.

Please see the official ORACLE-L FAQ: http://www.orafaq.com

To unsubscribe send email to: oracle-l-request_at_freelists.org put 'unsubscribe' in the subject line.

--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------

Received on Thu Mar 25 2004 - 10:58:56 CST