Reverse Notification for health check failures, when used as a side car for an orchestrator?

[rmohta]

New Feature Request
Envoy being a popular choice as a side-car in microservice architecture - I was wondering if it's possible to send reverse notification to an app when a pod/instance fails health check and is removed from the cluster? Is it possible to achieve this via extension?

I'll expand with below scenarios below

Scenario: When Envoy is used as a side car for an Orchestrator (like Netflix Conductor)

Orchestrator and envoy are deployed as a single pod in Kubernetes. Envoy is configured to perform health check of services (all the services are deployed in same k8s cluster). In a situation, a particular core service starts failing envoy health check (or k8s readiness probe), then the platform might want the orchestrator to take itself out of rotation for the load-balancer (F5 or AWS Load Balancer).

A workaround I use, is to read failed health check logs + poll Envoy admin API for cluster health and act upon it.

What I would expect?

When envoy has less than N healthy connections for a cluster, it would call a POST endpoint (could do gRPC too) with details

• cluster info + any other metadata we have configured
• number of healthy connections envoy has
• min number of configured healthy connections
• total number of connections envoy has (healthy + unhealthy)
• Last Health Check Failure Time

With this info, the destination server can act upon it.

• Google Forum Discussion: https://groups.google.com/forum/#!topic/envoy-dev/eTnDGHzp5rU

[mattklein123]

I think this is now a dup of #8970?

[rmohta]

Agree. Will close this issue, and follow #8970
Thank You @mattklein123!