Abstract
Edge cloud environments are often build as virtualized coordinated clusters of possibly heterogeneous devices. In the emerging of cluster platforms like Kubernetes or Docker Swarm, scalability is based on resource utilization. Resource utilization has been used for capacity planning and for forecasting resource demand. However, due to the large scale and complex structure of these architectures, analyzing large amount of monitoring data may cause a huge resource overhead that affects the performance of anomaly detection and the accuracy of anomaly location. To address such challenges, we propose a self-adaptive healing approach that detects, identifies, predicts and recovers anomalies in clustered architectures. The approach will be evaluated to assess the accuracy of the mechanism.