On Jul 24, 2015, at 7:56 PM, Geaaru <geaaru@gmail.com> wrote:
I see freeradius documentation but I have some doubt about how load-balancing/fail-over it works when is enable status-server or not.
1) From documentation I see that an home_server_pool with load-balancing split requests between all home server. But FWIK or I understand if request doesn't receive a response from first home_server is not send to second like for fail-over. Is this correct ?
No. The proxy fails over when: a) the home server is marked dead, and then b) the NAS sends another packet. If the NAS doesn't send another packet, the proxy will not be able to fail over *for that packet* to another home server. The *next packet* will see that the home server is down, and the proxy will choose another home server
3) About status-server check... In my configuration I set status_server to yes under security section but If I haven't enable site status (on port 18121 for example) I can receive however Status-Server request only on port 1812 and 1813. Is this correct?
Yes.
4) home_server gone on alive status only if at least one request is been received from server else home_server is leave on "unknown" status. Is this correct ?
Yes. "Unknown" is equivalent to "alive". The only difference is that it's a bit more information for the administrator.
5) Is there a way to set alive status of all home_server automatically at bootstrap ?
No.
I see that with fail-over if I haven't received request from startup (status of home_server is unknown) then on a failure of first home_server, request is not sent to second home_server (at least on 3.0.8 version). Instead, if I have both home_server in alive status and I power off first home_server, than request is correctly sent to second on failure of first.
It's more complicated than that. The NAS has to retransmit. Use radclient, with "-r 20", so it sends 20 retransmits. You WILL see the proxy fail over from one home server to another.
6) If I haven't active status-server is there a way to handle home_server only alive to avoid zombie period if a request go to failure / timeout ?
I don't know what that means. Alan DeKok.