I have a Patroni cluster with Spilo that I deploy with a Kubernetes manifest file like this.
Ad you can see the manifest has a ClusterIP service with no selectors. There is an Endpoint associated with this service here. This endpoint is updated via Kubernetes API by Patroni at startup or switch node throught callbacks.
My problem is that it happened that I needed to run two consecutive deploys where the second one there is no pod version update. In this scenario, no callbacks are activated and the Endpoint is set to None and remain in this status until I don restart manually the leader pod.
I tried several changes to fix this scenario (remove the endpoint, use selectors, etc.) but so far no luck. Can anyone help me on this?