RE: Help me solve this cardinality puzzle
Date: Sun, 20 Apr 2008 22:30:43 +0200
You can not detect from the "explain plan for .. "whether bind peekig occurs or not. In order to determine whether bind peekig is occuring or not you should trace your session with one of the events 10053, 10132 or 10046.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of 조동욱 Sent: Friday, April 18, 2008 2:08 AM
Cc: kutrovsky.oracle_at_gmail.com; oracle-l Subject: Re: Help me solve this cardinality puzzle
2008/4/18, Milen Kulev <makulev_at_gmx.net>:
Be aware of the fact, that on 10.2.0.3 there is bug related to bind peeking. In the secription of Bug 5082178 you will read:
In some situations bind peeking can occur when it should not eg: Bind peeking can occur for user binds even if
"_optim_peek_user_binds" is set to FALSE.
This can cause binds to be marked "unsafe" leading to cursors not being shared when they should be.
Full bug description you will find in ML:5082178.8. Oracle 10.2.0.3 is definitely one of the affected versions (although I have noticed the same Problem on 10.2.0.2 several times).
The only relief is to go on 10.2.0.4 ...
Good info, but i don't believe this bug would show itself in "explain plan". Chriso, can you show us the DDL script for your table and index?
I believe that i've seen similar cases a couple of months ago, but didn't have time to investigate. I tried to replay them but with no success.
http://www.freelists.org/webpage/oracle-l Received on Sun Apr 20 2008 - 15:30:43 CDT