Suppose for a scenario, when the consul agent (client) pod is deleted from a k8s worker node, we see that the application pod gets de-registered.
Now, when the consul agent comes back up and running, we noticed that the application pod is not getting re-registered.
This also means, the application pod will not be ready to except any packets until it gets re-registered to consul
While this would work for applications running will multiple replicas, it will be a down time for the one running with only 1 replica.
This becomes a hard stop for us until we know the right way to mitigate this issue.
If you’re using the consul-k8s official helm chart, we should re-register the service when agent pod goes down. Which version of the helm chart are you using? Do you see that the re-registration doesn’t happen every time client agent pod goes down or does it happen intermittently?
Hi @ashwinkupatkar could you file an issue on the consul-k8s repo and provide details on how to exactly reproduce the issue? We would need the helm config and deployment manifests to take a look.