Shared, distributed lock manager for loosely coupled processing systems

Abstract

Apparatus and process for managing shared, distributed locks in a multiprocessing complex. The manager operates using a partitionable lock space with logical processor connection. Logically connected processors are subject to validation and disconnection due to failure. The locks synchronize data access to identifiable subunits of DASD. Denied lock requests are queued for servicing when the lock becomes available. Lock partitions are used to speed DASD to DASD copying without halting processing on the first DASD. A special partition is assigned to the copy task and the processors writing to the DASD can determine copy status with a single read or test. Operations requiring multilateral agreement of processors, such as rebuilding locks or moving locks, are protected by fencing any nonresponsive processor. A special queue partition is designated for master/slave control point designation. All processors contend for the master lock and losing contenders are queue. Queuing provides automatic fallback in case of a failing processor.