Add helm chart parameters to customise http probe paths for webserver / api server

[stephen-bracken]

Description

The probe urls for the api server and webserver templates default to {webserver.base_url}/health values in the Http probes, but this causes issues with multi cluster deployments.

Use case/motivation

We are rolling out airflow deployments that are running active-active on multiple kubernetes clusters using a shared database that is also active-active, but the liveness probes on the pods point to the webserver base url, which a common cname which we use for the webservers from both clusters.

This means that if the primary webserver stops responding and fails the liveness / readiness probes then both webservers will register as unhealthy. To resolve this I would like to reroute the Http probes to localhost/health or the local pod url (e.g. airflow-webserver-abcde.airflow-webserver.airflow.svc.cluster.local/health) using the kubernetes dns.

i.e. the webserver liveness probes point to https://airflow-deployment.example.com/health, which both webservers are using as the base_url. This would allow us to use a load balancer to route traffic to both webservers, whilst accurately marking webserver pods as unhealthy if they fail to respond to a local liveness probe.

Related issues

No response

Are you willing to submit a PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[devguoo]

This sounds like a useful feature for multi-cluster deployments! Using localhost or the pod's local DNS name for health probes makes sense to avoid false negatives when the shared base_url becomes unavailable.

Looking forward to your PR! It would be great if the implementation could support both localhost and the pod DNS name as options, giving users flexibility based on their network setup.