Configure YARP for Blocking Traffic When Unhealthy

[roneytham-ideagen]

I’m using YARP to manage reverse proxy traffic for my .NET services and have configured active health checks for my cluster (as shown in https://microsoft.github.io/reverse-proxy/articles/dests-health-checks.html). My goal is to ensure that if the health check endpoint (/healthz/ready) returns a failure status (e.g., HTTP 503), YARP completely blocks traffic to that endpoint and responds with HTTP 503 to clients attempting to access it.

However, even when the health check marks a destination as unhealthy, clients can still access the endpoint, and traffic is not blocked as expected.

Steps:

1. Configure YARP with the following setup:

{ "ReverseProxy": { "Clusters": { "MyCluster": { "Destinations": { "Destination1": { "Address": "http://service1/" } }, "HealthCheck": { "Active": { "Enabled": true, "Path": "/healthz/ready", "Interval": "00:00:10", "Timeout": "00:00:05", "Policy": "ConsecutiveFailures" } } } } } }

2. Start the application.
3. Simulate an unhealthy service by having /healthz/ready return an HTTP 503 status.
4. Attempt to access the endpoint (http://service1/) through the reverse proxy, e.g. http://localhost/service1.

Result:
I gotten this log which is returned from the active health check:
Probing destination "Destination1" on cluster "MyCluster" completed with the response code "503".
However, YARP continues forwarding traffic to the destination, e.g. calling an endpoint from http://localhost/service1/, even though the health check has detected it as unhealthy.

I would like to achieve the following behavior:

• Once the health check detects /healthz/ready as unhealthy:
  • YARP should stop forwarding traffic to the destination http://localhost/service1/.
  • YARP should return an HTTP 503 status for all requests to that endpoint.

Questions:

1. How can I configure YARP to block traffic on http://localhost/service1/ and return HTTP 503 when the health check destination is marked as unhealthy by the active health checks?
2. Does YARP require additional middleware or a custom policy to enforce this behavior?
3. How can I debug the health-check mechanism and verify if destinations are being properly excluded when unhealthy?

Thanks.

[MihaZupan]

What YARP does when you run out of healthy destinations is up to the IAvailableDestinationsPolicy.
There are two such policies built-in: HealthyAndUnknown and HealthyOrPanic.

We currently default to HealthyOrPanic, which will "panic" and allow traffic to all destinations if you run out of healthy ones.
If you switch to HealthyAndUnknown, you will see the behavior you wanted, where YARP will respond with a 503.

"HealthCheck": {
  "AvailableDestinationsPolicy": "HealthyAndUnknown",
  "Active": {
    // ...
  }
}

[roneytham-ideagen]

It works, I am able to see response with HTTP status code 503, thank you.