Tariq Rashid <[EMAIL PROTECTED]> wrote: > how do you mean? do mean that large number of active threads will kill my > system?
A large number of "active" threads which are doing nothing other than blocking for 30 seconds is a waste of resources. > i dont see what is wrong with, say, 200 threads for proxying. It's a bad design. See my previous response. > That way, if there are 50 bad requests which cause the home servers > not to reply, the other 150 from the thread pool can continue to > handle requests. It's not needed. See my previous response. > hote that we intend to proxy to DIFFERENT sets of radius servers under the > control of different entities - and this is why we need to be cautious about > non-returning home servers. this is why a sequential proxy service is not a > good idea. Which is why the server is not designed that way. > is it possible to have say 50 threads in 5 pools? the 5 pools corresponding > to 5 different home radius servers. It's not needed. It's a bad design. The threads on the server are unrelated to clients, databases, or to home RADIUS servers. Tying threads to any one of those objects would be a bad design, and would cause problems. The threads on the server are visible only to the server "core". They're not visible to modules, databases, or anything else. The purpose of the threads is to allow the server to parallelize internal operations. The threads do not block for ANYTHING except to serialize access to server modules (rlm_*). Please go READ src/main/radiusd.c, the code there explains how proxying is handled. Alan DeKok. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
