Health awareness for downstream services

[bmarlow]

As feature growth and adoption of GatewayAPI continue to increase, more people are increasing looking to it as a potential for replacing load balancing services that may have been previously done by external hardware load balancers.

Currently GatewayAPI has no concept of the health of any downstream services (or available APIs to query as a result).

One particular area of concern is integration with MetalLB.

When MetalLB is used to provide a load balancer, the only thing that MetalLB can check for is the status of the gateway itself. As a result, this can leave MetalLB in a state where it is advertising addresses for services that are unhealthy.

The first step to remedying this would be to have some sort of 'health awareness' API available.

Health checks should still be performed via liveliness and readiness probes on the respective pods, but it would be good for the gateway to understand if that service is healthy or not, so that it can be integrated with for other services.

This will also provide a quicker path for engineers doing application troubleshooting, as they can quickly take a peak at a gateway object and get a relatively broad understanding of overall health (gateway configured, deployed, and relevant health of downstream services).