Described is a technology by which load balancing is accomplished by HTTP redirect responses. A web application (rather than a server) receives an HTTP request, such as identifying a database that is available on one or more servers. From among those servers having the database, the web application selects a server based upon server load information collected from each of the servers. Load balancing is accomplished by selecting a server based upon the load information and which servers have the most recent data, and returning an HTTP redirect response that identifies the selected server. Also described is using redirection to refresh a database on a server without affecting any other databases on that server, by removing that server from those available while the database is refreshing, whereby requests for that database will not be redirected to that server while being refreshed.