What Happens to My Apps if the Master Node Goes Down in a Kubernetes Cluster?

You need to keep in mind that orchestration stops completely if the master is down. Existing workloads on the worker nodes will still run, but since there's no master to control them, any rescheduling or healing won't occur. Your ingress and load balancing may also face issues depending on your setup.

It's important to have a highly available setup with multiple master nodes—usually at least 3—if you want true high availability. With only one master, if it goes down, you potentially lose control over your cluster, which is essential for tasks like managing network rules.

If your master node goes down, your apps that are already running on the worker nodes will still be accessible. They keep running as long as the worker nodes are up and everything is configured correctly. However, without a master, you won't be able to manage or make changes to those workloads; things like scaling up or rescheduling won't happen until the master is back online.

To clarify, the master node is crucial as it manages core components of the control plane. If it goes down and your etcd isn't corrupted, you might not notice issues immediately, but you won't be able to perform any management tasks. Essentially, your apps might continue running but access becomes risky if you need to change anything.