deadlock detection takes 58 seconds with latch freessage

Hello,

We have a realtime application on Oracle 8.0.5 and Solaris 2.5.1 which experienced a deadlock between two stored procedures. The application has been corrected to avoid this condition in the future but the production application was hung up for about 6 minutes during the deadlock. We reproduced the deadlock in our QA database under controlled conditions and the time was 58 seconds. I traced the execution of the two procedures during this deadlock and the breakdown of the wait times was:

"latch freessage to client" 24.83 seconds
"latch freessage from client" 24.80 seconds
"enqueue to/from client" 6.03 seconds

Additionally, when we reproduced the problem on our Development database (same Oracle level and identical application code) the wait time was only 3 seconds. There is something going on in our QA and Production server to cause this wait that is different from the development server.

My questions are: has anyone experienced the same problem and found the solution or does anyone know what a "latch freessage to client" is and what causes it? Thanks in advance for any help you can give.

Mark Ediger
Chicago Mercantile Exchange

-----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own Received on Sat Feb 27 1999 - 09:15:28 CST