On 05/02/2005 09:38:16 PM, Wolfgang Breitling wrote: > Mladen, >=20 > I have to second Yongping's sentiment. The required size of the shared po= ol=20 > depends on a lot of factors. I am running several Peoplesoft databases wi= th=20 > a shared pool of 80-100M each and none is getting 4031 errors. Granted,=20 > they are development instances. Wolfgang, I am managing an OLTP database accessed by a homegrown applicatio= ns, written in Java and relying heavily on PL/SQL. DBMS_LOB is used on a regula= r basis. The company that worked for previously was using Oracle*Text and also requi= red large shared pools. I monitored V$SGASTAT from OEM and it used to have approximat= ely 30% of free space, after everything that was used regularly was pinned. My numb= er may have=20 been influenced by recent upgrade of 10.1.0.3 database to 10.1.0.4 which fa= ils if you=20 don't have at least 150MB of shared pool, but not by much. > I agreee that Oracle 9i probably needs a bit more shared pool than 8i, al= l=20 > else being equal. Yes, but not that much more. The problem is with abundant new features, lik= e DBMS_METADATA which is really expensive. When people start using those, it starts to get = expensive. After=20 switching from an austere version 8i, 9.2 is like a candy store. >=20 > Laura, >=20 > do you have any non-default cursor parameters? Such as "open_cursors",=20 > "session_cached_cursors", or "cursor_space_for_time"? hat is the exact=20 > error you are getting. Is it always the same routine that is getting the=20 > error. Have you tried "keep"ing the "STANDARD" (and others) PL/SQL proced= ures? Ufff, I forgot about "cursor_space_for_time". I haven't used in ages, ever = since=20 the early versions of 8i. I should be consulting more. --=20 Mladen Gogala Oracle DBA -- //www.freelists.org/webpage/oracle-l