Layered Cluster Fencing Coordinator for Data Availability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In layered cluster systems, concurrent fencing actions at different layers (application and operating system layers) can lead to data unavailability due to conflicting high-availability mechanisms, where the base operating system and application layer independently execute their own failover logic, causing issues like data unavailability and service disruption.
Innovation Solution
Implementing a layered cluster system architecture where the base operating system keeps track of both application and operating system states, allowing it to return success for fencing requests if the peer application is already down, and coordinating fencing requests to prevent concurrent fencing, using soft and hard fencing mechanisms to ensure successful failover and recovery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If independent failover logic is executed at both application and operating system layers, then high-availability capabilities are enhanced, but data unavailability and service disruption occur due to conflicting fencing actions
Solution Approach 1:
The patent introduces a fencing coordinator as an intermediary component that mediates between the application layer fencing logic and operating system layer fencing logic. The coordinator receives fencing requests from both layers, coordinates their execution to prevent conflicts, and ensures that fencing actions are performed in a controlled sequence. This intermediary structure allows independent failover logic to operate at both layers while preventing the harmful effect of conflicting fencing actions that would cause data unavailability.
2Reliability
If redundant capabilities are provided across multiple computers, then application continuity is improved during failures, but system complexity increases due to coordination requirements
Solution Approach 1:
The patent merges the fencing coordination functionality into a centralized component that handles both application layer and operating system layer fencing requests. Instead of having separate coordination mechanisms for each layer, the patent combines them into a unified fencing coordinator that manages redundant capabilities across multiple computers. This merging approach maintains application continuity through redundancy while reducing overall system complexity by consolidating coordination logic.
3Speed
If concurrent fencing actions are allowed at different layers, then failover responsiveness is improved, but conflicting fencing actions cause service disruption
Solution Approach 1:
The patent implements preliminary action by having the fencing coordinator evaluate and prepare fencing requests before actual fencing actions are executed. When fencing requests arrive from both application and operating system layers, the coordinator预先 (in advance) assesses their compatibility, determines the correct execution sequence, and prevents conflicting actions. This preliminary coordination maintains fast failover responsiveness while ensuring service continuity by avoiding conflicting fencing actions.
Data Source
AI summary
Techniques are provided for high availability events in a layered architecture. In an example two computing nodes coordinate to provide a computing service, where each node has a base operating system configured to fence the other base operating system, and an application configured to fence the other application. In some examples, fencing requests by an application are routed through its base operating system, which coordinates application-level fencing requests and operating system-level fencing requests.


