Strange!? Could you post a some sample code or a small test case?
regards, Armin ----- Original Message ----- From: "Chris Lewington" <[EMAIL PROTECTED]> To: "OJB Users List" <[EMAIL PROTECTED]> Sent: Friday, October 18, 2002 3:14 PM Subject: Re: BUG REPORT: OJB 0.9.7 - Commit Failure In Eager Release Mode > Hi Armin, > > OK, tried a number of things: > > 1) left the OJB.properties as is (originally set to use > ConnectionFactoryConPooledImpl) - result: "can't get connection from > ConnectionManager" thrown from StatementsForClass. This is where the default > size of the pool (21) gets exceeded, which seems to relate to the number of PB > instances (ie 21). This happens when trying to delete a number of test objects > which were in fact stored OK (in a sperate method). Setting the connection > pool size to 50 worked, but I worry about such "solutions" as I never know > quite when capacity will run out. > > 2) changed OJB.properties to use ConnectionFactoryDefaultImpl - result: > original tests seemed OK, but a DB fill application which previously worked > threw lots of NullPointerExceptions and inserted only some data. Weird! > > All of the above goes away if I set eager-release to true and test for a live > transaction in the closeResources method of StatementManager - so my > connection pool size stays small and everything else works OK. A nice side > effect was also that for my test cases/application I noticed an improved > performance by a factor of *10* when setting eager-release to true!! Before I > get drunk on excitement though, is this likely to cause problems in a more > multiuser environment? > > cheers, > > Chris > > Armin Waibel wrote: > > > When using the OJB internal connection pooling use the default > > (original, no need to change properties) OJB.properties. > > Your problem seems to be a > > user-do-the-wrong-thing problem ;-) > > Full version of OJB shipped with a stress test were > > 10 threads at the same time handling 2000 objects per thread > > (store, fetch, delete each object) and this test passed > > well (call bin\build.bat performance2). > > > > When using the PB-api don't forget to call PB.close() after using a > > PB instance (close() returns instance to pool). > > > > HTH > > > > regards, > > Armin > > > > > > > > Cheers, > > > > > > Chris > > > > > > > > > > > > -- > > > To unsubscribe, e-mail: > > <mailto:ojb-user-unsubscribe@;jakarta.apache.org> > > > For additional commands, e-mail: > > <mailto:ojb-user-help@;jakarta.apache.org> > > > > > > > > > > > > > -- > > To unsubscribe, e-mail: <mailto:ojb-user-unsubscribe@;jakarta.apache.org> > > For additional commands, e-mail: <mailto:ojb-user-help@;jakarta.apache.org> > > -- > <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< > Dr Chris Lewington tel: +44 (0)1223 492 523 > European Bioinformatics Institute fax: +44 (0)1223 494 468 > Wellcome Trust Genome Campus email: [EMAIL PROTECTED] > Hinxton, Cambridge CB10 1SD (UK) > >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> > > > > -- > To unsubscribe, e-mail: <mailto:ojb-user-unsubscribe@;jakarta.apache.org> > For additional commands, e-mail: <mailto:ojb-user-help@;jakarta.apache.org> > > > -- To unsubscribe, e-mail: <mailto:ojb-user-unsubscribe@;jakarta.apache.org> For additional commands, e-mail: <mailto:ojb-user-help@;jakarta.apache.org>
