Bifurcated Key Set Registration for Process-Level Fencing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current distributed computing systems face service interruptions due to the inability to selectively isolate and reset failed processes within a node, leading to unnecessary shutdowns of healthy processes, as existing technologies rely exclusively on node-level heartbeat messages for failure detection and resource reassignment.
Innovation Solution
Implementing a bifurcated key set registration system that allows for separate detection and enforcement of node and process failures, using heartbeat signals and process failure indicators to de-register resources at the process level, thereby isolating only the failed processes while keeping healthy processes operational.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If node-level heartbeat messages are used for failure detection, then system simplicity is maintained, but service interruptions occur due to inability to selectively isolate failed processes
Solution Approach 1:
The patent segments the failure detection mechanism into two independent components: node-level heartbeat messages for detecting node failures, and process-level failure indicators for detecting individual process failures. This segmentation allows the system to selectively isolate and reset only the failed process while keeping other healthy processes on the same node operational, thereby resolving the contradiction between maintaining system simplicity and ensuring service continuity.
2Ease of operation
If all processes on a node are shut down upon node failure detection, then resource reassignment is simplified, but healthy processes are unnecessarily interrupted
Solution Approach 1:
The patent extracts the process failure detection and isolation capability from the node-level failure handling mechanism. By introducing process-level failure indicators and associated processing logic, the system can now separately identify and isolate only the failed process (taking it out from the node's process set) while leaving other healthy processes unaffected. This resolves the contradiction by maintaining simplified resource reassignment at the node level while preventing unnecessary interruption of healthy processes through process-level granularity.
3Reliability
If process-level failure detection is implemented, then selective isolation of failed processes is enabled, but system complexity increases
Solution Approach 1:
The patent merges the process-level failure detection and isolation functionality with the existing node-level heartbeat mechanism. The process failure indicators are integrated into the same fencing framework and resource management infrastructure that handles node-level failures. This merging approach enables selective isolation of failed processes while reusing existing system components, thereby reducing the overall complexity increase that would result from implementing a completely separate process-level detection system.
Data Source
AI summary
A key set is registered. The registering the key set includes registering a first shared data resource key. The first shared data resource key includes a first identifier associating a first process with a first shared data resource. The registering the key set further includes registering a second shared data resource key, and the second shared data resource key includes a second identifier associating a second process with a second shared data resource. A failure of a first process is detected, and in response to the detecting the failure of the first process, the first shared data resource key is de-registered. The second shared data resource key remains registered after the de-registering the first shared data resource key.