当前位置：首页 > news >正文

基于MySQL实现分布式调度系统的选举算法

news 2025/7/22 13:56:44

基于MySQL实现分布式调度系统的选举算法，可通过基于超时机制的数据库表驱动选主方案实现，利用数据库表协调节点间的状态，以下是Java实现的核心步骤和代码：

1. 数据库表设计

创建选主表，用于记录leader节点和状态：

CREATE TABLE leader_election (service_id varchar(128) NOT NULL,leader_id varchar(128) NOT NULL,last_seen_active timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,PRIMARY KEY (service_id)
) ENGINE=InnoDB

2. Java实现选举机制

import java.sql.*;
import java.util.concurrent.*;public class LeaderElection {private final String serviceId;      // 服务标识（如"distributed-scheduler"）private final String nodeId;         // 当前节点ID（如hostname_pid）private final int electionTimeout;   // 选举超时时间（秒）private final ScheduledExecutorService scheduler;private volatile boolean isLeader = false;private Connection conn;             // MySQL连接public LeaderElection(String serviceId, String nodeId, int timeout) {this.serviceId = serviceId;this.nodeId = nodeId;this.electionTimeout = timeout;this.scheduler = Executors.newSingleThreadScheduledExecutor();// 初始化数据库连接this.conn = DriverManager.getConnection("jdbc:mysql://mysql-host:3306/db", "user", "pwd");}// 启动周期性选举public void start() {scheduler.scheduleAtFixedRate(this::attemptLeadership, 0, 1, TimeUnit.SECONDS);}// 尝试获取领导权private void attemptLeadership() {String sql = "INSERT INTO leader_election (service_id, leader_id, last_seen_active) " +"VALUES (?, ?, NOW()) " +"ON DUPLICATE KEY UPDATE " +"leader_id = IF(last_seen_active < NOW() - INTERVAL ? SECOND, VALUES(leader_id), leader_id), " +"last_seen_active = IF(leader_id = VALUES(leader_id), NOW(), last_seen_active)";try (PreparedStatement stmt = conn.prepareStatement(sql)) {stmt.setString(1, serviceId);stmt.setString(2, nodeId);stmt.setInt(3, electionTimeout);  // 超时时间（如20秒）stmt.executeUpdate();updateLeaderStatus();  // 更新当前节点状态} catch (SQLException e) {e.printStackTrace();}}// 检查当前节点是否Leaderprivate void updateLeaderStatus() throws SQLException {String query = "SELECT COUNT(*) AS is_leader FROM leader_election " +"WHERE service_id=? AND leader_id=?";try (PreparedStatement stmt = conn.prepareStatement(query)) {stmt.setString(1, serviceId);stmt.setString(2, nodeId);ResultSet rs = stmt.executeQuery();if (rs.next()) {isLeader = rs.getInt("is_leader") > 0;if (isLeader) System.out.println("✅ Current node is LEADER");}}}// 关闭资源public void shutdown() {scheduler.shutdown();if (conn != null) try { conn.close(); } catch (SQLException ignored) {}}
}

3. 核心逻辑说明

节点初始化
每个节点启动时传入唯一nodeId（如hostname_pid），并连接到MySQL。
领导权竞争逻辑
- 使用ON DUPLICATE KEY UPDATE原子操作保证竞争安全：
  若当前无Leader或Leader超时（last_seen_active < NOW() - 20s），则抢占为Leader。
- 节点周期性（如每秒）尝试更新状态，确保活跃Leader持续续期。
状态更新
- Leader节点：last_seen_active字段被更新为当前时间。
- Follower节点：未更新字段，仅检测自身是否为Leader 3。

4. 高级功能扩展

// 强制指定Leader（管理员操作）
public void forceLeadership() throws SQLException {String sql = "REPLACE INTO leader_election (service_id, leader_id, last_seen_active) VALUES (?, ?, NOW())";try (PreparedStatement stmt = conn.prepareStatement(sql)) {stmt.setString(1, serviceId);stmt.setString(2, nodeId);stmt.executeUpdate();}
}// 查询当前Leader
public String getCurrentLeader() throws SQLException {String sql = "SELECT leader_id FROM leader_election WHERE service_id=?";try (PreparedStatement stmt = conn.prepareStatement(sql)) {stmt.setString(1, serviceId);ResultSet rs = stmt.executeQuery();return rs.next() ? rs.getString("leader_id") : null;}
}