NEGOTIATED GRACEFUL TAKEOVER IN A NODE CLUSTER

The invention is a method for operating a file server system in a cluster mode that provides for relatively rapid and reliable takeover of a failed file server in the cluster by a partner file server when the failed file server has detected a fault that will cause it to shut down. This is done by the failed file server requesting the one file server to take over its file services after detecting the fault in its operations, the one file server letting the failed file server complete existing file service requests from clients, refusing further file service requests addressed to the failed file server, and the one file server taking over by transferring file service requests to the one file server. As part of this takeover the one file server takes on the identity of the failed filer and activates network interfaces and network addresses that replicate the failed server's network addresses.