Oracle 9IAS (9.0.3) lockup

[W.C. Epperson]

We have a server running Oracle Internet Application server V.9.0.3 and 9.0.1.0.20. When contacting the 9.0.3 based Apache server, some clients cause the server to throw an SSL encryption error. The socket ends up in close_wait, but the server goes wild trying to read from it even after getting a 0 byte read() return. System calls quickly escalate to 300,000 or more per second, and that Apache instance goes brain dead--nonresponsive.

It happens only with connections from certain outside networks, and appears to have something to do with the way the client closes the connection when the SSL error is thrown. It looks like the client may be closing the connection twice, and may be resulting in a time_wait assassination situation.

Any insights out there?

[Steven E. Protter]

Do you have the required OS patches in? There should be a list on metalink.

Next, I would look at shared memory allocation in the kernel. If there is any tightness there all kinds of wierdness can happen.

Does the error_log in apache give you any clues?

I'd like to see what the apache error log has to say about all this.

SEP

[Stanley_8]

Outside networks?
Does it connect to server through firewall?
If yes, you need to make some ports accessible ex. SSL use port 443 and 9iAS default port.
Please try it!