A Highly Available Lock Manager For HA-NFS

Abstract

This paper presents the design and implementation of a highly available lock manager for highly available NFS (HA-NFS). HA-NFS provides highly available network file service to NFS clients and can be used by any NFS client without modification. This is provided by having two servers share dual-ported disks so that one server can take over the other's disks and file systems if it fails. Making the NFS service highly available is not enough since many applications that use NFS also use other services provided with NFS such as the network lock manager. We describe a scheme whereby each server transfers enough of its lock state to the other so that if it fails, the other server can go through a lock recovery protocol. Our design goal was to make the overhead of transferring the state during failure-free operation as low as possible.