Flink 重启策略(Restart Strategy)
1. 重启策略概述
重启策略决定作业失败时是否可以重启以及重启的间隔。
重要说明:
- 未定义重启策略时,使用集群默认策略(通过flink-conf.yaml配置)
- 作业级配置会覆盖集群默认策略
- 未启用检查点时,默认采用"不重启"策略
- 启用检查点且未配置策略时,默认使用FixedDelayRestartStrategy(Integer.MAX_VALUE, 10000)
2. 重启策略类型
2.1 策略类型总览
| 固定延迟策略 | fixed-delay | 临时性故障 | 必须启用 |
| 故障率策略 | failure-rate | 周期性/突发性故障 | 必须启用 |
| 不重启策略 | none | 批处理/无状态任务 | 可选 |
| 回退策略 | fallback | 集群默认策略 | 可选 |
2.2 策略选择决策树
#mermaid-svg-pNikvG4vnPJsWsUI{font-family:\”trebuchet ms\”,verdana,arial,sans-serif;font-size:16px;fill:#333;}@keyframes edge-animation-frame{from{stroke-dashoffset:0;}}@keyframes dash{to{stroke-dashoffset:0;}}#mermaid-svg-pNikvG4vnPJsWsUI .edge-animation-slow{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 50s linear infinite;stroke-linecap:round;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-animation-fast{stroke-dasharray:9,5!important;stroke-dashoffset:900;animation:dash 20s linear infinite;stroke-linecap:round;}#mermaid-svg-pNikvG4vnPJsWsUI .error-icon{fill:#552222;}#mermaid-svg-pNikvG4vnPJsWsUI .error-text{fill:#552222;stroke:#552222;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-thickness-normal{stroke-width:1px;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-thickness-thick{stroke-width:3.5px;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-pattern-solid{stroke-dasharray:0;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-thickness-invisible{stroke-width:0;fill:none;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-pattern-dashed{stroke-dasharray:3;}#mermaid-svg-pNikvG4vnPJsWsUI .edge-pattern-dotted{stroke-dasharray:2;}#mermaid-svg-pNikvG4vnPJsWsUI .marker{fill:#333333;stroke:#333333;}#mermaid-svg-pNikvG4vnPJsWsUI .marker.cross{stroke:#333333;}#mermaid-svg-pNikvG4vnPJsWsUI svg{font-family:\”trebuchet ms\”,verdana,arial,sans-serif;font-size:16px;}#mermaid-svg-pNikvG4vnPJsWsUI p{margin:0;}#mermaid-svg-pNikvG4vnPJsWsUI .label{font-family:\”trebuchet ms\”,verdana,arial,sans-serif;color:#333;}#mermaid-svg-pNikvG4vnPJsWsUI .cluster-label text{fill:#333;}#mermaid-svg-pNikvG4vnPJsWsUI .cluster-label span{color:#333;}#mermaid-svg-pNikvG4vnPJsWsUI .cluster-label span p{background-color:transparent;}#mermaid-svg-pNikvG4vnPJsWsUI .label text,#mermaid-svg-pNikvG4vnPJsWsUI span{fill:#333;color:#333;}#mermaid-svg-pNikvG4vnPJsWsUI .node rect,#mermaid-svg-pNikvG4vnPJsWsUI .node circle,#mermaid-svg-pNikvG4vnPJsWsUI .node ellipse,#mermaid-svg-pNikvG4vnPJsWsUI .node polygon,#mermaid-svg-pNikvG4vnPJsWsUI .node path{fill:#ECECFF;stroke:#9370DB;stroke-width:1px;}#mermaid-svg-pNikvG4vnPJsWsUI .rough-node .label text,#mermaid-svg-pNikvG4vnPJsWsUI .node .label text,#mermaid-svg-pNikvG4vnPJsWsUI .image-shape .label,#mermaid-svg-pNikvG4vnPJsWsUI .icon-shape .label{text-anchor:middle;}#mermaid-svg-pNikvG4vnPJsWsUI .node .katex path{fill:#000;stroke:#000;stroke-width:1px;}#mermaid-svg-pNikvG4vnPJsWsUI .rough-node .label,#mermaid-svg-pNikvG4vnPJsWsUI .node .label,#mermaid-svg-pNikvG4vnPJsWsUI .image-shape .label,#mermaid-svg-pNikvG4vnPJsWsUI .icon-shape .label{text-align:center;}#mermaid-svg-pNikvG4vnPJsWsUI .node.clickable{cursor:pointer;}#mermaid-svg-pNikvG4vnPJsWsUI .root .anchor path{fill:#333333!important;stroke-width:0;stroke:#333333;}#mermaid-svg-pNikvG4vnPJsWsUI .arrowheadPath{fill:#333333;}#mermaid-svg-pNikvG4vnPJsWsUI .edgePath .path{stroke:#333333;stroke-width:2.0px;}#mermaid-svg-pNikvG4vnPJsWsUI .flowchart-link{stroke:#333333;fill:none;}#mermaid-svg-pNikvG4vnPJsWsUI .edgeLabel{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-pNikvG4vnPJsWsUI .edgeLabel p{background-color:rgba(232,232,232, 0.8);}#mermaid-svg-pNikvG4vnPJsWsUI .edgeLabel rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-pNikvG4vnPJsWsUI .labelBkg{background-color:rgba(232, 232, 232, 0.5);}#mermaid-svg-pNikvG4vnPJsWsUI .cluster rect{fill:#ffffde;stroke:#aaaa33;stroke-width:1px;}#mermaid-svg-pNikvG4vnPJsWsUI .cluster text{fill:#333;}#mermaid-svg-pNikvG4vnPJsWsUI .cluster span{color:#333;}#mermaid-svg-pNikvG4vnPJsWsUI div.mermaidTooltip{position:absolute;text-align:center;max-width:200px;padding:2px;font-family:\”trebuchet ms\”,verdana,arial,sans-serif;font-size:12px;background:hsl(80, 100%, 96.2745098039%);border:1px solid #aaaa33;border-radius:2px;pointer-events:none;z-index:100;}#mermaid-svg-pNikvG4vnPJsWsUI .flowchartTitleText{text-anchor:middle;font-size:18px;fill:#333;}#mermaid-svg-pNikvG4vnPJsWsUI rect.text{fill:none;stroke-width:0;}#mermaid-svg-pNikvG4vnPJsWsUI .icon-shape,#mermaid-svg-pNikvG4vnPJsWsUI .image-shape{background-color:rgba(232,232,232, 0.8);text-align:center;}#mermaid-svg-pNikvG4vnPJsWsUI .icon-shape p,#mermaid-svg-pNikvG4vnPJsWsUI .image-shape p{background-color:rgba(232,232,232, 0.8);padding:2px;}#mermaid-svg-pNikvG4vnPJsWsUI .icon-shape rect,#mermaid-svg-pNikvG4vnPJsWsUI .image-shape rect{opacity:0.5;background-color:rgba(232,232,232, 0.8);fill:rgba(232,232,232, 0.8);}#mermaid-svg-pNikvG4vnPJsWsUI .label-icon{display:inline-block;height:1em;overflow:visible;vertical-align:-0.125em;}#mermaid-svg-pNikvG4vnPJsWsUI .node .label-icon path{fill:currentColor;stroke:revert;stroke-width:revert;}#mermaid-svg-pNikvG4vnPJsWsUI :root{–mermaid-font-family:\”trebuchet ms\”,verdana,arial,sans-serif;}
是
否
临时性
周期性/突发性
是否需要状态恢复?
启用检查点
选择no-restart或fallback
故障类型?
fixed-delay
failure-rate
3. 固定延迟重启策略(Fixed Delay Restart Strategy)
3.1 核心原理
- 固定间隔重启:每次失败后等待固定时间再尝试
- 最大尝试次数:达到阈值后停止重启
- 状态恢复:每次重启从最新检查点恢复
3.2 配置参数
3.2.1 代码配置
// 基础配置
env.setRestartStrategy(RestartStrategies.fixedDelayRestart(
3, // attempts
Time.seconds(30) // delay
));
// 指数退避配置
env.setRestartStrategy(RestartStrategies.fixedDelayRestart(
5, // attempts
Time.seconds(10), // initial-delay
Time.seconds(300), // max-delay
2.0 // multiplier
));
3.2.2 配置文件
restart-strategy: fixed–delay
restart-strategy.fixed-delay.attempts: 3
restart-strategy.fixed-delay.delay: 30 s
restart-strategy.fixed-delay.exponential-delay.initial-delay: 10 s
restart-strategy.fixed-delay.exponential-delay.max-delay: 5 min
restart-strategy.fixed-delay.exponential-delay.multiplier: 2.0
3.3 触发条件
| 任务执行失败 | 用户代码异常、反序列化错误 | TaskFailureException |
| 检查点失败 | 超时、存储错误 | Checkpoint expired |
| 资源管理失败 | TM失联、Slot分配失败 | Heartbeat timed out |
3.4 生产建议
- 小状态作业:attempts=3-5, delay=30s, 启用指数退避
- 大状态作业:attempts=5-10, delay=60s, 启用指数退避
- 关键作业:attempts=5, delay=30s, 人工审批流程
4. 故障率重启策略(Failure Rate Restart Strategy)
4.1 核心原理
- 时间窗口统计:在failure-rate-interval内统计失败次数
- 阈值触发:超过max-failures-per-interval后触发重启
- 冷却期:重启后清空窗口统计
4.2 配置参数
4.2.1 代码配置
env.setRestartStrategy(RestartStrategies.failureRateRestart(
3, // maxFailuresPerInterval
Time.minutes(5), // failureRateInterval
Time.seconds(30) // delay
));
4.2.2 配置文件
restart-strategy: failure–rate
restart-strategy.failure-rate.max-failures-per-interval: 3
restart-strategy.failure-rate.failure-rate-interval: 5 min
restart-strategy.failure-rate.delay: 30 s
4.3 触发场景
| 外部系统周期性不稳定 | 5-10分钟 | 2-3次 | 30s |
| 网络风暴 | 1-2分钟 | 3-5次 | 10s |
| 资源竞争 | 3-5分钟 | 2-3次 | 60s |
4.4 生产建议
- 窗口大小应大于检查点间隔
- 避免窗口过短(<1分钟)或阈值过低(<2次)
- 监控failureRate指标
5. 不重启策略(No Restart Strategy)
5.1 适用场景
- 批处理作业
- 无状态任务
- 需要人工干预的严重故障
5.2 配置方式
restart-strategy: none
5.3 检查点关联性
| 未启用 | 不重启 | 安全 |
| 已启用 | 不重启 | 状态丢失风险 |
6. 回退策略(Fallback Restart Strategy)
6.1 行为说明
- 使用集群默认策略
- 通常用于批处理作业
6.2 配置方式
restart-strategy: fallback
7. 生产环境最佳实践
7.1 检查点协同配置
// 检查点间隔应小于重启间隔
env.enableCheckpointing(20000); // 20秒
// 检查点超时时间应大于重启间隔
env.getCheckpointConfig().setCheckpointTimeout(300000); // 5分钟
// 允许并发检查点
env.getCheckpointConfig().setMaxConcurrentCheckpoints(2);
7.2 监控指标
| numberOfRestarts | 重启次数 | > attempts |
| failureRate | 失败率 | > max-failures-per-interval |
| lastCheckpointDuration | 检查点持续时间 | > 检查点间隔 |
7.3 故障排查
- 重启间隔不生效:检查检查点是否启用
- 状态丢失:检查state.checkpoints.dir权限
- 重启风暴:启用指数退避
8. 附录:配置优先级
restart-strategy: fixed–delay
- 启用检查点:FixedDelayRestartStrategy(Integer.MAX_VALUE, 10000)
- 未启用检查点:NoRestartStrategy
最后更新:2025-12-26 参考文档:Flink 1.12 官方文档



