Re: Interesting problem

From: John Hurley <hurleyjohnb_at_yahoo.com>
Date: Wed, 25 May 2011 08:27:25 -0700 (PDT)
Message-ID: <253f7f09-6b5d-4081-8dc7-bcc8b89ce526_at_e17g2000prj.googlegroups.com>



Mladen:

# When I come to think of it, I would expect it to use library cache latch, because of the opening and closing the cursor literally thousands of times per second, but not buffer chains latch. Does anyone have an explanation?

Most logical IOs will take 2 cache buffer chains latch gets. One latch get/release for looking up the buffer and pinning the buffer and then later another get for unpinning the buffer.

A common reason for cbc latch contention is excessive logical io's against the same blocks.

If you put a 10046 trace on this code my guess is that you will see a ton of logical io's ... library cache latching from plsql should often be pretty minimal and session_cached_cursors is set above 0 by default in 10g and above. Received on Wed May 25 2011 - 10:27:25 CDT

Original text of this message