Just did some tests to verify failover and I came across two issues: - running FreeRADIUS 1.0.1 - one realm (used with DEFAULT Huntgroup,Proxy-To-Relam in users), entered with two home-servers With config proxy server { synchronous = no retry_delay = 10 retry_count = 3 dead_time = 600 default_fallback = no post_proxy_authorize = no } The first failed server *never* gets marked as "dead" using retry_delay = 5 retry_count = 3 one authentication try fails, then the server is marked as dead and subsequent suceed within dead_time The problem is, I need the larger timeout/retry-delay for lame upstream servers and I know the failover also worked with these settings some time ago, but this might have been with an earlier 0.9x release. Is there a way now (I remember some dicussions years ago it wasn't) to make the failover to multiple proxies work without losing one authentication-try each dead_time ? I checked the changelog but found no hints something was fixed or changed in later versions.. Might it be worth a try to use v1.1.1 ? (although I'd really love to keep my stable Debian-package) Michael
"Michael Markstaller" <mm@elabnet.de> wrote:
Is there a way now (I remember some dicussions years ago it wasn't) to make the failover to multiple proxies work without losing one authentication-try each dead_time ?
Source code changes. The problem is mainly that in order to do it, the server must schedule events to happen in the future, and then wake up to process those events. It's not impossible, it's just more work and more code. Alan DeKok.
participants (2)
-
Alan DeKok -
Michael Markstaller