We use the same sort of setup, have a look through your radius.log at the same time that the unclosed accounting session went bad, you'll probably find that you have a lot of duplicate discarded records. We're looking into some solutions ourselves, as mikrotik will often take requests from faulty clients and forward hundreds of requests directly to the radius server without waiting for a response from the server (though maybe this is just a weakness of UDP protocols). In mysql cluster situations, this can be painful, as cluster sql queries take a bit longer to complete than non cluster queries. Null length Stop requests also cause this sort of problem too. For the moment, we're going limit the amount of connections a single NAS can make, until we find a better solution.
Hi list, I have an installation with freeradius 2.1.9, postgresql like db and pppoe server more than one mikrotik routerboards. Like now I'm making some test and, a part that all seem to work well (users can connected to the chain pppoe-server mikrotik -> freeradius -> db), I see a strange thing inside the radacct table: I have a session that aren't "closed" (acctstoptime field are NULL) and I have another session, with the _same_ framedipaddress, opened after the previous one. The nas value are different. Looking into both nas, the first that as into the db the session opened, haven't the pppoe session active and radwho say me that there is no session for that framedipaddress. I thought that "someone", before open a session with a specific framedipaddress see if it's already assigned, but it's not the truth. I think that I'm in wrong in somewhere, but I don't know.
The question: who forgot to close the session? Or better, do I have to say to freeradius to close it after a "timeout" or something other?
Thanks, Michele - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html