On 17 Jul 2014, at 15:39, Arnaud . <arn0@live.fr> wrote:
Hello,
When I read the explanations of "max-request" must change this number to the number of clients. But what that"clients": the number of public IP that connect different or the number of NAS in the database?
Doing it by clients is only a very rough way of determining the correct queue size.
I have a Wifi Hotspot system with terminals for customers. There are about 274 clients (routers connected to theADSL Box customers).
It controls number of requests the server holds state for. You should determine the normal number of requests received per second, then multiple it by cleanup_delay, and double it to give you a bit of a buffer. It's mainly useful for keeping the server from melting down completely in overload situations. If you keep reaching the limit of the buffer and the server starts discarding requests, one of your backend databases or processes is too slow, or the server is processing a very high number of packets per second and you've left the queue value at default. Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS development team FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2