wenweihu86/raft-java
Raft Java
Raft-java is a lightweight Java consensus library and replicated state machine engine designed to synchronize transaction logs and maintain fault-tolerant state across distributed server nodes. The framework relies on a leader-based replication model that routes write operations through a designated coordinator node, executing RPC-driven consensus protocols and automatic leader elections to maintain cluster agreement.
The library supports dynamic cluster membership changes, allowing nodes to be added or removed during runtime without downtime through joint consensus configuration entries appended directly into the replicated log stream. State synchronisation is handled by streaming transaction logs from active coordinators to follower nodes, alongside persistent disk storage and pluggable state machine interfaces that execute committed log entries against custom application logic.
To manage storage growth and accelerate node recovery, the engine captures point-in-time application data snapshots to disk, enabling older transaction log entries to be safely truncated and discarded.
Features
- Replicated State Machines - Builds reliable Java applications that synchronize data changes across a cluster by applying committed transaction logs.
- Distributed Consensus Protocols - Provides a lightweight consensus engine that synchronizes transaction logs across multiple server nodes for fault tolerance.
- Dynamic Cluster Membership Management - Supports dynamic cluster membership changes, allowing nodes to be added or removed during runtime without downtime.
- Single-Leader Replication - Routes all write operations through a single designated coordinator node that serialises and streams updates out to follower nodes.
- Joint Consensus Protocols - Reconfigures active cluster nodes on the fly by appending joint consensus configuration entries directly into the replicated log stream.
- Consensus Log Replication - Ensures cluster-wide state consistency by synchronizing transaction logs from the active leader node to all followers.
- Leader Election - Features automatic leader election mechanisms to negotiate and elect an active leader node to coordinate distributed consensus.
- Consensus and Coordination Systems - Manages leader election and log replication across multiple nodes to maintain consistent state in a distributed system.
- Consensus Protocols - Coordinates distributed nodes by exchanging standard request and response messages over network connections to maintain cluster agreement.
- Raft Consensus Implementations - Implements distributed consensus and replicated state machines with leader election and log replication in Java.
- Pluggable Interfaces - Executes committed log entries against custom application logic through standard interfaces to maintain synchronised business data across nodes.
- Point-In-Time Snapshots - Compresses historical application data into storage checkpoints to truncate old log files and speed up node recovery times.
- Local Consensus Snapshots - Captures point-in-time application data snapshots to disk, enabling older transaction log entries to be safely discarded.
- Checkpoints and Recovery - Compresses old log entries into local snapshots and loads them during startup to reduce storage space and speed recovery.
- Persistent Log Buffers - Saves transaction entries directly to disk storage so the system can recover state and catch up lagging nodes after a restart.
- State Machine Integrations - Enables custom state machine integration by executing replicated commands and applying committed log entries directly to application data.
