your test application doesn't really necessarily reproduce the problem
since that is actually expected behavior, you could get the same
results by just opening 20 connections in a loop and then ending the
loop without closing them. Can you make connections again as expected
if you issue a GC.Collect() ? The reason I ask is we encountered a very
similar problem that wasn't in the managed datalayer but instead in the
OCI layer itself. There was a critical section orphaned and we couldn't
make any other connections once this happened until we restarted the
parent process (bounced IIS). The problem turned out to be something
involving the use of Oracle Namesserver for name resolution (refer to
bug #4484336 in metalink) and the workaround was to simply use TNSNames
and we haven't seen the problem since. If this isn't your exact problem
I would highly recommend deploying the Microsoft Debugging Tools for
Windows and then using ADPlus to take a memory snapshot to send off to
Oracle Support.
Received on Mon Feb 13 2006 - 16:18:24 CST
