Health check

[steven-sheehy]

The helm chart does not have a liveness or readiness probe for the loadbalancer-keepalived container. We've had cases where kube-keepalived-vip logs would show everything fine with one in master state and others in backup, but there were no VIPs created on any network interfaces. kube-keepalived-vip should expose a /health endpoint for use in probes. Additionally, that health endpoint should interrogate the status of the VIPs and only respond successfully (200) if the VIPs exist on the interface.