The alert KubeAPIDown is not reflecting what really the alert means. 

In the summary, it's indicated that 

  "Target disappeared from Prometheus target discovery"

but really, the only values are 0 or 1 for this alert, then, if you bring down a kubeapi server, this alert is not triggered.

4.13 and previous

Always. Shutting down a master node

1. Shutdown a master node
2. Wait and check the alert KubeAPIDown never is triggered since it's 1 or 2 and 2 apiserver are up. 

The alert is not reflecting the summary:

  "Target disappeared from Prometheus target discovery"

The alert is triggered if we respect the summary and the alertname. Then 2 possible options that they could go together:

1. Create an alert that reflects the status of the kubeapi server pods. Then, a apiserver pod is down, it's reflected as a degraded status
2. Modify the current alert to the name summary to reflect the really monitored