Kubernetes is the leading cluster management platform, and within Kubernetes, an operator is an application-specific program that leverages the Kubernetes API to automate operation tasks for man- aging an application deployed on a Kubernetes cluster. Users can declare a desired state for the managed cluster, specifying their configuration preferences, and the operator program is responsible for reconciling the cluster’s actual state to align with the desired state. However, intricate bugs in Kubernetes operators can have severe impacts on the operation and stability of both applications and the underlying Kubernetes cluster.

In this paper, we conduct an empirical study on 208 operator bugs from 36 Kubernetes operators. We investigate the root causes, manifestations, impacts and fixing strategies for these bugs. Our study reveals many interesting findings that can guide the detection and testing of operator bugs, as well as the development of more reliable operators.