Webserver 2.1 Cartrige problems

We had developed a large Web app. using OWS 2.1 on NT with . Netscape or ORACLE HTTP Server
Architecture:
The Cartridges are written in C++ /OCI (9 cartridges) connecting ORACLE 7.3 on a separate machine (we can't use the PL/SQL Agent for this app). During WRB_Init of the cartdriges we connect to the database and we keep the connection until the Server shuts down (WRB_Shutdown). This architecture is not very stable because if someone is reloading a page 10 times(or more) the request broker spawns new instances of the cartridge, but most of them are hanging and we have many dead connections on the database. Also the Webserver is completly hanging from time to time and we can't stop the listener (ORACLE Listener or Netscape). If we do a connect on WRB_Exec
and after all work is done a disconnect, everything works fine.

Is this an OCI/SQL*Net or a Request Broker problem ? We don't want to connect and disconnect each time a request is comming in from a client because of performance issues.

Any ideas?

thanks,
 T.G. Received on Sat Apr 26 1997 - 00:00:00 CDT