Anthony Baxter wrote:
We don't see this. We run on Solaris, talking to Oracle 8.1.7, with a couple of dozen ZEO clients across the entire installation - they all have at least one connection to an Oracle database - in some cases, they have several different connections (for instance, SQLSession works far better if you give it it's own DB connection). The primary cluster has something like 9 ZEO clients, and it gets hammered 24x7. I cannot recall the last time we had problems with the Zope/Oracle interaction, it's at least a couple of years ago. This has been in use for something like 5+ years now.
this is certainly encouraging, in terms of opening the possibility of errors on our behalf that we can fix. We had *terrible* performace with Solaris in general with python and Zope so we reluctantly abondened Solaris about 2 years ago. This was probably due to not fully understanding the implications of the GIL at the time. On Linux we have had problems when running long, slow queries, and having several instances of DCOracel DA. We have expericend loss of connectivity and having 3 of 4 of zope's threads being completly non-responsive, even when the site is idle. -- -- ------------------------------------------------------------------- Dario Lopez-Kästen, IT Systems & Services Chalmers University of Tech.