Cluster Fencing via Application-Level Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In high-availability cluster systems, existing methods struggle to securely and efficiently isolate malfunctioning or inactive physical host nodes and their associated application programs without affecting other resources, leading to potential data corruption or system crashes.
Innovation Solution
A fencing method that allows for targeted isolation of specific application programs or virtual machines within a cluster by using fence operation commands that include information about both the target network device and application program, enabling precise control through designated fence devices, and utilizing STONITH techniques to execute fence operations locally or remotely when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional fencing methods are used to isolate malfunctioning nodes, then system security is improved, but other application programs on the same node are also affected and shut down
Solution Approach 1:
The patent segments the fencing operation from the node level to the application program level. Instead of shutting down the entire node, the system identifies and fences only specific application programs running on the malfunctioning node. This is achieved by including application program identification information in the fence operation command, allowing selective isolation of problematic applications while preserving other applications on the same node.
Solution Approach 2:
The patent applies local quality by making the fencing action targeted and specific rather than blanket. The fence operation command includes specific identification information for the target application program, enabling the system to apply fencing only to the local level of the problematic application rather than the entire node, thus differentiating treatment between healthy and unhealthy applications.
2Reliability
If node-level fencing is implemented, then data protection is improved, but system resource utilization deteriorates due to unnecessary shutdown of healthy applications
Solution Approach 1:
The patent segments the impact scope from node-level to application-level. By including application program identification information in the fence operation command, the system can protect data associated with specific malfunctioning applications without shutting down the entire node or other healthy applications, thereby maintaining resource utilization for still-functional services.
Solution Approach 2:
The patent applies partial action by implementing fencing only to the extent necessary - specifically targeting only the malfunctioning application programs rather than applying blanket node-level fencing. This partial approach prevents excessive action (shutting down healthy applications) while still achieving the necessary data protection for the problematic applications.
3Adaptability or versatility
If remote fence operations are executed, then operational flexibility is improved, but execution reliability deteriorates due to network dependency
Solution Approach 1:
The patent introduces a fence device as an intermediary component that facilitates reliable fence operation execution. The fence device is configured with identification information matching the target application program and receives fence operation commands. This intermediary mechanism ensures that whether the operation is local or remote, the command is properly routed and executed through the designated fence device, maintaining reliability while preserving operational flexibility.
Data Source
AI summary
A method of fencing in a cluster system including network devices and a management network device is disclosed. Each network device may run an application program and comprises a fence device. A first network device receives a fence operation command from the management network device. The fence operation command contains information of a target network device and information of a target application. The first network device may determine a fence device on the first network device that corresponds to the fence operation command and activates the determined fence device such that the fence device performs a fence operation on the target application program according to the fence operation command.