This document describes an active/active high availability solution for central agents in OpenStack ceilometer. It discusses using taskflow to manage the discovery and polling tasks performed by central agents to provide high availability. Discovery agents monitor the sources of telemetry data and polling agents collect the data. Taskflow would store the tasks in a job table coordinated by Zookeeper to ensure tasks are rescheduled if an agent fails. This provides a more robust solution than a simple homebrew approach using RabbitMQ.