I tuned my Kernel and PostgreSQl installation so that I can open 200 db connections.
This removed the "JDBCExceptionReporter:101 - FATAL: sorry, too many clients already" problem.
However, now I get the original "too many file open" problem back in GF V2 server relatively early in the stress test :-(
It does not seem to be a temporary congestion problem and appears to be a file descriptor leak in LDAP code use in GF V2. The reason I suspect that is that even after I stop pumping requests to my server and let the system quite down the server remains in a state that is no longer accepting new requests.
Shall I file an issue? Is there any other tests the dev team would like me to conduct?
TIA for your help.
[Message sent by forum member 'najmi' (farrukh_at_wellfleetsoftware.com)]
http://forums.java.net/jive/thread.jspa?messageID=367440