Recently upgraded to 8.1.7 and set my shared_pool to 600 megs and buffer cache to 2400 megs.
The first strange thing I noticed with 8i was monitoring the Oracle process under task manager showed only the shared_pool + user sessions combined. i.e. 600 megs + sessions.
I don't use MTS and my session levels climb over 600. This combined with a 1 meg sort area size caused me to get a 12500 error when the session level reached approx 524 sessions.
I dropped the buffer cache and hit the problem at similar session level. I was using my old 7.3.4. experience and figured it was the combined settings = limit.
Dropped down the shared pool by 100 megs and this allowed me to take on more sessions. But I hit the problem at 1.2 gb again (user process + shared_pool).
Finally dropped the sort area size to reduce the per session memory requirement. This keeps the process below 1 gb at all times.
I use the boot.ini settings of /pae and /3gb.
Has anyone else hit a problem like this.????


Reply With Quote
Bookmarks