Hi Luca, What is your ingress controller to handle the TCP route? We’ve the same setup but with Traefik in front of FreeRADIUS which also solves your problem since Traefik will have the port open even if there is no radius server behind it. And in regards to the startup-probe for the container, you can add the node IP (where the probes will be send from) as environment variable to allow connections from it. Besides that I would use the K8S functionalities to ensure that the service is healthy like PDB, horizontal pod autoscaling etc. BR, Lineconnect
On 7. Feb 2025, at 09:20, Luca Borruto via Freeradius-Users <freeradius-users@lists.freeradius.org> wrote:
Hello everyone,
I am currently running FreeRADIUS v3.2.6 on K8S behind an Azure Load Balancer, serving RadSec (TLS on TCP 2083) for wifi EAP authentication.
The load balancer is configured with a TCP health probe on port 2083 to verify the service’s availability (that's the way Azure LB works), the issue is that FreeRADIUS does not seem to accept these health probe requests. In the logs, I see messages like:
Ignoring request to auth+acct proto tcp address * port 2083 (TLS) bound to server default from unknown client 10.0.2.4 port 3286 proto tcp
The health probe originates from internal ALB IPs (e.g., 10.0.2.4, 10.0.2.33). FreeRADIUS rejects them as unknown clients and as a result, the load balancer marks the service as unhealthy and so, the traffic is not achieved to the freeradius pods.
I am looking for guidance on the recommended way to configure FreeRADIUS to work behind an Azure Load Balancer:
What is the best practice for handling this scenario? Any official recommendations or insights from the community would be greatly appreciated.
Best regards, Luca Borruto IT System Administrator - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html