Abstract: A computing network provides a database service and implements a caching mechanism for processing database queries. The database services are provided using a primary replica of data and computing resources, and a secondary replica of the data and computing resources. The primary and secondary replicas are synchronized so that the secondary replica can provide the database services in place of the primary replica in response to a failover from the primary replica to the secondary replica. In response to initiating, by the database service, a failover from the primary replica to the secondary replica, the cached plans compiled by the secondary replica are used to process matching query batches by the secondary replica. The secondary replica processes the matching query batches without accessing plans compiled by the primary replica.