Spring Boot集成Liquibase：K8s环境下长耗时DB脚本管理方案咨询

核心思路：拆分变更执行阶段 #

将变更明确分为启动前快速执行和运行时异步执行两类，这和你提出的思路匹配，以下是几种可落地的成熟方案：

方案1：Liquibase标签拆分 + K8s初始化容器 #

把表结构调整、小数据插入这类快速变更放在初始化容器中执行，加索引、大数据迁移这类长耗时变更打专属标签，留到应用运行时触发执行。

示例步骤：

1. 拆分变更集：在changelog里为长耗时变更标记标签：

<!-- 启动前执行的快速变更 -->
<changeSet id="20240520-01" author="dev">
  <createTable tableName="user">
    <column name="id" type="BIGINT" autoIncrement="true">
      <constraints primaryKey="true"/>
    </column>
    <column name="name" type="VARCHAR(50)"/>
  </createTable>
</changeSet>

<!-- 长耗时变更，标记标签：long-running -->
<changeSet id="20240520-02" author="dev" labels="long-running">
  <createIndex indexName="idx_user_name" tableName="user">
    <column name="name"/>
  </createIndex>
</changeSet>

2. K8s初始化容器配置：初始化容器仅执行不带long-running标签的变更，确保快速完成不阻塞应用启动：

apiVersion: apps/v1
kind: Deployment
metadata:
  name: my-app
spec:
  replicas: 1
  template:
    spec:
      initContainers:
      - name: liquibase-fast
        image: my-app-image:latest
        command: ["java", "-jar", "app.jar", "--liquibase.contexts=!long-running"]
        env:
        - name: SPRING_DATASOURCE_URL
          value: jdbc:mysql://db:3306/mydb
        - name: SPRING_DATASOURCE_USERNAME
          value: root
        - name: SPRING_DATASOURCE_PASSWORD
          value: password
      containers:
      - name: app
        image: my-app-image:latest
        ports:
        - containerPort: 8080
        livenessProbe:
          httpGet:
            path: /actuator/health/liveness
            port: 8080
          initialDelaySeconds: 30
          periodSeconds: 10
        readinessProbe:
          httpGet:
            path: /actuator/health/readiness
            port: 8080
          initialDelaySeconds: 15
          periodSeconds: 5

3. 运行时触发长耗时变更：在应用内添加带权限控制的接口，调用Liquibase API执行标记的长耗时变更：

@RestController
@RequestMapping("/db")
public class LiquibaseController {

    private final SpringLiquibase liquibase;

    public LiquibaseController(SpringLiquibase liquibase) {
        this.liquibase = liquibase;
    }

    @PostMapping("/execute-long-running")
    public String executeLongRunningChanges() throws LiquibaseException {
        liquibase.setLabels("long-running");
        liquibase.afterPropertiesSet(); // 触发变更执行
        return "Long-running changes executed successfully";
    }
}

方案2：异步执行 + 分布式锁 #

若不想拆分容器，可在应用启动时仅标记长耗时变更为待执行，通过Spring异步任务结合分布式锁（如Redis锁）执行，避免多实例重复操作：

示例代码：

1. 配置异步执行线程池：

@Configuration
@EnableAsync
public class AsyncConfig {
    @Bean
    public Executor taskExecutor() {
        ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();
        executor.setCorePoolSize(1);
        executor.setMaxPoolSize(1);
        executor.setQueueCapacity(10);
        executor.setThreadNamePrefix("LiquibaseAsync-");
        executor.initialize();
        return executor;
    }
}

2. 异步执行长耗时变更：

@Component
public class LiquibaseAsyncService {

    private final SpringLiquibase liquibase;
    private final StringRedisTemplate redisTemplate;

    public LiquibaseAsyncService(SpringLiquibase liquibase, StringRedisTemplate redisTemplate) {
        this.liquibase = liquibase;
        this.redisTemplate = redisTemplate;
    }

    @Async
    @PostConstruct
    public void executeLongRunningChanges() throws InterruptedException, LiquibaseException {
        // 分布式锁确保仅单个实例执行
        Boolean locked = redisTemplate.opsForValue().setIfAbsent("liquibase:long-running:lock", "locked", 1, TimeUnit.HOURS);
        if (locked != null && locked) {
            try {
                liquibase.setLabels("long-running");
                liquibase.afterPropertiesSet();
            } finally {
                redisTemplate.delete("liquibase:long-running:lock");
            }
        }
    }
}

方案3：独立K8s Job执行长耗时变更 #

把长耗时变更完全剥离出应用，用K8s Job单独执行，完成后再启动应用，适合超大规模数据迁移这类极端耗时场景：

示例Job配置：

apiVersion: batch/v1
kind: Job
metadata:
  name: liquibase-long-running
spec:
  template:
    spec:
      containers:
      - name: liquibase
        image: my-app-image:latest
        command: ["java", "-jar", "app.jar", "--liquibase.labels=long-running"]
        env:
        - name: SPRING_DATASOURCE_URL
          value: jdbc:mysql://db:3306/mydb
        - name: SPRING_DATASOURCE_USERNAME
          value: root
        - name: SPRING_DATASOURCE_PASSWORD
          value: password
      restartPolicy: OnFailure

在Deployment中配置初始化容器等待Job完成：

initContainers:
- name: wait-for-liquibase-job
  image: busybox:latest
  command: ['sh', '-c', 'until kubectl get jobs liquibase-long-running -o jsonpath="{.status.succeeded}" | grep -q 1; do sleep 10; done']

关键注意事项 #

• 所有长耗时变更必须保证幂等性，Liquibase会记录已执行的变更集，只要id和author不变，不会重复执行，但仍需避免非预期副作用
• 合理设置K8s存活/就绪检查的initialDelaySeconds，给启动前的快速变更留足执行时间
• 运行时触发的变更需添加权限控制，防止误操作

内容的提问来源于stack exchange，提问作者lunanigra