[Pierre Dumonceau] > Thank you very much for your eye-opening answer. We thought that there > could have been a memory leak because the memory was not returned to the > OS. That degraded the performances of our application. After some search > on the Web, we: > * applied the patch from Evan Jones to the interpreter > (obmalloc-free-arenas). That proved useful in our case. > * reduced the size of the ZODB connection pool as you advised to > one of your colleague > Thank you very much for your time.
WRT #1, I'm afraid I'm on the critical path to getting Evan's patch into Python, and I haven't been able to make enough spare time for it (it's a non-trivial reviewing task). I still hope it can get into Python 2.5, anyway. WRT #2, sorry, I'm not even sure what it means ;-) You mean you passed a value smaller than 7 as DB()'s ``pool_size`` argument? _______________________________________________ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org http://mail.zope.org/mailman/listinfo/zodb-dev