Importance

Loading Importance Indicators...

Abstract

Apparatus and accompanying methods for use in preferably a multi-system shared data (sysplex (5)) environment which quickly and efficiently isolates (fences), through a pre-defined hierarchical order, failed sysplex components from accessing shared data in order to protect data integrity. Specifically, by dividing a sysplex workload into specified fence groups (FG A, FG B) and providing appropriate software and hardware fence support, fencing can occur at various distinct levels: a member-to-member level, i.e. to allow any member (220, 225, 230, 233,237) of a fence group to fully isolate any other ('target') member of that same group; a fence group level, i.e. to isolate all members of a fence group that execute on a 'target' system (200.sub.1, 200.sub.2, 200.sub.3); and a system level, i.e. to fully isolate an entire 'target' system. Through pre-defined escalation rules (630), fencing can be escalated from a lower member level to a higher, group or system, level in the event a lower level fence can not be successfully imposed. Member level fencing is accomplished in software (1300, 1500); group and system level fencing is accomplished through dedicated hardware fencing facilities (44). An identifier (444, 464) uniquely designates each different fence group existing on a computer processing complex (CPC) (40.sub.1, 40.sub.2, 40.sub.3, 40.sub.M) in the sysplex over the life of that CPC. Advantageously, this technique eliminates erroneous back level fencing, significantly expedites fence processing and also greatly reduces a need for human intervention.

Loading the Abstract Image...

First Claim