System and method for failure recovery in a shared resource system having a moving write lock

Claim 1

In a computer network that includes a plurality of processing nodes, a method for distributing control of a write lock for a shared data object from, a master processing node currently in control of the write lock, to one of the plurality of processing nodes requesting use of the write lock upon failure of the master processing node wherein the master processing node fails to transfer the write lock to the processing node requesting use of the write lock, the method comprising the following steps in the sequence set forth: detecting by the processing node requesting use of the write lock, when the master processing node fails to transfer the write lock to the processing node requesting use of the write lock; determining, by the processing node requesting use of the write lock, whether the processing node requesting use of the write lock holds the write lock for the shared data object; when the processing node requesting use of the write lock holds the write lock, requesting that the processing node requesting use of the write lock assume control over the write lock for that shared data object from the master processing node failing to transfer use of the write lock for that shared data object to establish a new master processing node for that shared data object; when the processing node requesting use of the write lock does not hold the write lock for the shared data object, determining whether another processing node of the plurality of processing nodes holds the write lock for the shared data object; when the another processing node holds the write lock for the shared data object, requesting that the another processing node assume control over the write lock for that shared data object from the master processing node failing to transfer use of the write lock for that shared data object to establish a new master processing node for that shared data object; and when the another processing node fails to assume control of the write lock, executing a race among the plurality of processing nodes for control of the write lock for that shared data object from the master processing node failing to transfer use of the write lock for that shared data object to establish a new master processing node for that shared data object.