Re: Intermittent Oracle client errors

1. better diagnostics available in 9.2.0 client (vs. 8.1.7 client).
2. 8.1.7 is no longer officially supported (but some security patches do appear)
3. host-based firewalls may be in the picture, I know that I run ipfilter on my linux boxes, blocking almost all ports beneath 1024 (tcp). perhaps your sysadmin has done the same, although this would not cause intermittent failures, unless he has only opened a subset of the high tcp ports for dedicated server processes (can check this with sysctl on Linux, I'm not sure how to check this on Solaris).

How about installing a 9.2 home on a web server, applying whatever patchset is deemed appropriate (e.g. 9.2.0.7) and configuring the web server service to use that Oracle Home?

Then enable network tracing at admin level on both servers, web and oracle for a short test period, turn off tracing and check the logs.

It is entirely possible that the load on the Sun box is sufficient that the listener is not able to spawn a dedicated server process within the timeout period. That condition would have to be tested under load - which - while having network tracing enabled at ADMIN level would be a bad idea in general, but sounds as if it has to be done.

Are the web servers creating dedicated server sessions for each web app user, or are they pooling connections? Creating oracle dedicated server sessions is an expensive operation in terms of CPU. Try to avoid it.

you might also consider creating pre-spawned processes (sessions).

hth.

-bdbafh Received on Fri Oct 14 2005 - 15:41:57 CDT