A method, including deploying multiple servers as a distributed computing system (DCS) configured to provide a service to clients. At a first time, a given server is configured as an active node in the DCS, the active node configured to and to process requests for the service received from the clients. At the first time, one or more remaining first servers are configured as one or more semi-active nodes in the DCS, each given semi-active node configured to receive, from the clients, requests for the service, and to forward the received requests to the active node for processing. Each of the servers maintains respective counts of the received requests, and at a second time subsequent to the first time, upon identifying a given remaining server having a highest count the given remaining server is reconfigured as the active node, and the given server is reconfigured as a semi-active node.