The rise of AI powered computer vision algorithms offers the possibility of intelligent data processing, and decision making based on streaming data from video cameras. Applications such as Smart Cities could potentially use these video cameras for a variety of use cases such as pedestrian detection, public safety, and traffic monitoring. The requirements for low latency for real-time decision making, and the privacy needs of video data, leads to use of edge computing to process raw video frames. However, unlike cloud computing with almost unbounded resources, the edge is characterized by compute nodes of limited capacity and power budget. Additionally, fault tolerance is limited due to replication costs at the edge.In this thesis, we investigate the design of a fault tolerant edge cluster consisting of low power ARM based Raspberry Pi 4 nodes. In the cloud, Kubernetes is used as a system orchestrator for large clusters. An edge tailored version of Kubernetes, K3s has recently been made available. However, prior research has not characterized the resource consumption and latency impact of K3s on realistic edge clusters. In this thesis we fill the gap by an extensive evaluation of K3s at the edge on our Raspberry Pi 5 cluster. Our results indicates that while K3s does add significant resource and latency overhead to edge applications, it still delivers on fault tolerance at the edge.