> Two strange things: > - there seemed to be no performance impact with all the threads hanging around > (maybe we steered clear of this since we regularly restart radiusd)
We didn't see any mayor effects but we did get increasing reports of random connection failures, which I think would be related to the conflicting requests. People would get on later but would have to try 2-3 times to auth. > We use an entirely different setup than Graeme does, we don't use MySQL but a custom > radacct-module that writes to a file, and a Berkeley DB user database. > > We decided to upgrade to 0.9.1 from the official page and the problem vanished. one problem we are currently experiencing with 0.9.1 from CVS is that the system spins out of control after it has been up for around 2 days. The problem happens when a script runs to process radacct data (it just basically moves rows from one table to another). Now this script runs every day at 4am.. but the freeradius daemon only goes nuts on the 3rd day that the script runs, if I stop and start the daemon then it will be fine for 2 days (including the times when this happens). Could this be a sign of something changing in the SQL module? (i'll try a diff later) -- ----- Graeme Hinchliffe (BSc) Core Team Member Zen Internet (http://www.zen.co.uk) ICQ 3842605 (link) Direct: 0845 058 9074 Main : 0845 058 9000 Fax : 0845 058 9005 - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
