Guilherme Franco wrote:
As everything was good before and now it's breaking, the most probable cause is the increase in the number of auth users, which brings lots of acct (0 users in September 2006 and now with 4000 online users pumping radacct). The oracle tables are well indexed so the response time is low. What comes to my mind is that the driver is having trouble to work with high acct throughput under peak time, starving all the 32 threads.
The problems shown by valgrind are there independent of load. The problems SHOULD be fixed!
I've considered radrelay/sqllog before, but wouldn't that break the Simultaneous-Use functionality?
Not if the accounting inserts are done quickly. Separating the authentication request from the accounting may increase the uptime of the server. Alan DeKok. -- http://deployingradius.com - The web site of the book http://deployingradius.com/blog/ - The blog