Raft Node State Transition for Sub-Cluster Reintegration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In network clusters using Byzantine Fault Tolerance-Raft (BFT-Raft), when a cluster splits into isolated sub-clusters due to network interruptions, the operational reliability is low as node devices in one sub-cluster ignore voting requests and heartbeat information from another, preventing reintegration into a single system.

Innovation Solution

Node devices in a leader state obtain cycle length information and latest log indexes from voting requests, degrading their state to follower or candidate if the information is greater than their own, allowing participation in elections across sub-clusters and reintegration into a single system until a new leader node is established.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a node device in leader state ignores voting requests and heartbeat information from other sub-clusters, then the leader state is maintained, but the system cannot be reintegrated into a single system and operational reliability is low

Engineering Contradiction:
Improveoperational reliabilityVSAvoidability to participate in elections
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The node device dynamically adjusts its state based on received voting requests. When voting requests are received during leader state, the device transitions to follower or candidate state, allowing the system to adapt to network recovery scenarios and enable reintegration of sub-clusters.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The invention changes the operational parameters of the node device by introducing conditions under which a leader node should transition to follower or candidate state. This parameter change enables the system to handle split-brain scenarios and improves operational reliability through state flexibility.

Inventive Principle:
Principle #35Parameter changes

2Stability of the object's composition

If a node device in leader state always ignores voting requests, then the leader state stability is maintained, but the device cannot combine with other sub-clusters into one system

Engineering Contradiction:
Improveleader state stabilityVSAvoidsystem integration capability
Core Design Contradiction:
Stability of the object's compositionVSAdaptability or versatility

Solution Approach 1:

The node device employs dynamic state transition based on received voting requests. Instead of statically ignoring all voting requests, the leader node dynamically evaluates incoming requests and transitions to follower or candidate state when appropriate, balancing stability with adaptability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The invention introduces feedback mechanisms where the leader node responds to voting requests from other sub-clusters. This feedback loop enables the system to detect split-brain scenarios and trigger appropriate state transitions, allowing reintegration while maintaining stability through controlled responses.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3562123B1Node device running method, working state switching device, node device, and medium
Publication Date: 2021.08.11 TENCENT TECHNOLOGY (SHENZHEN) CO LTD
  • EP3562123B1 patent drawingFigure 1A~1B
  • EP3562123B1 patent drawingFigure 2
  • EP3562123B1 patent drawingFigure 3

AI summary

This application discloses a node device operation method, a work status switching apparatus, a node device, and a medium, and relates to the field of network technologies. The method includes: receiving voting requests of a plurality of node devices, a quantity of the plurality of node devices being greater than half a quantity of node devices in a system; obtaining cycle length information and latest log indexes from the voting requests of the plurality of node devices if a current node device operates in a leader state; and switching a current work status from the leader state to a follower state or a candidate state if all the cycle length information in the voting requests of the plurality of node devices is greater than cycle length information of the current node device and all the latest log indexes in the voting requests of the plurality of node devices are not less than a latest log index of the current node device. This application enables that a first sub-cluster and a second sub-cluster can be combined into one system to work together, thereby improving the operational reliability of the system.