Managing the failover operations of a storage unit in a cluster of computers

Disclosed is a method of managing failover operations on a cluster of computers. The method starts with a failover hold module identifying a failure to access data storage in the cluster of computers. Then all read operations directed to the data storage that were received after the failure to access data storage was identified are prevented from execution. All the write operations directed to the data storage that were received after the failure to access data storage was identified are executed, including writing data to a cache. When a failover to an alternative data storage is complete, the held read operations are executed, including reading data from the alternative data storage and copying, from cache to the alternative data storage, the data written to the cache as part of the write operations. The failure to access may be a failed read operation, write operation, access to a file system or failure to access a database.