A cluster is unstable when a node has a Status of NotReady. For example:
backend7-123:~ # kubectl get nodes NAME STATUS ROLES AGE VERSION 192.168.123.72 Ready master 2d v1.9.1 192.168.123.73 NotReady worker 2d v1.9.1 192.168.123.74 Ready worker 2d v1.9.1
To debug a cluster, see: https://kubernetes.io/docs/tasks/debug-application-cluster/debug-cluster/.