Cluster System Split-Brain Prevention via Path Reliability

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing cluster systems can experience a 'split-brain syndrome' where multiple servers attempt to execute the same operation processing due to incorrect fault detection, leading to clashes in processing content, especially when communication interfaces fail.

Innovation Solution

Incorporating a reliability determination unit within each server to assess the communication path's reliability and stop processing if it's deemed unreliable, preventing unnecessary takeover and clash of operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If fault detection is based on heartbeat reception state and packet reception state, then operations continuity is maintained through takeover, but split-brain syndrome occurs where multiple servers execute same processing

Engineering Contradiction:

Improveoperations continuityVSAvoidsplit-brain syndrome

Core Design Contradiction:

ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The patent introduces a communication path reliability determination unit as an intermediary mechanism that independently assesses the state of communication paths. This unit acts as a mediator between the heartbeat/packet reception detection and the takeover decision, providing additional verification to prevent false fault detection that would cause split-brain syndrome while maintaining operations continuity through valid takeovers

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If server takes over operations upon detecting fault through heartbeat timeout, then operations stoppage is prevented, but unnecessary takeover occurs when communication fault is misidentified

Engineering Contradiction:

Improveoperations availabilityVSAvoidfault detection accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent applies preliminary action by determining communication path reliability before making the takeover decision. The communication path reliability determination unit performs this assessment in advance, checking whether the communication path is actually faulty or if the server is truly down. This preliminary verification prevents unnecessary takeovers while ensuring that genuine faults are properly handled to maintain operations availability

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS10102088B2Cluster system, server device, cluster system management method, and computer-readable recording medium

Publication Date: 2018.10.16 NEC SOLUTION INNOVATORS LTD

AI summary

A cluster system includes server apparatuses (10,20) connected by communication paths. Each server apparatus includes: a signal transmission unit (11) that transmits a signal indicating a self-presence to the other server apparatus via the communication paths (30-50); a reliability determination unit (12) that determines whether the communication paths (30-50) are reliable; and a processing management unit (13) that, in a state where the signal is not receivable from the other server apparatus, checks whether the communication paths to the other server apparatus have been determined to be reliable, and stops processing that is being executed when the result of the check shows that the communication paths have not been determined to be reliable.