diff --git a/cluster.go b/cluster.go
index d5c3d7b..30271e7 100644
--- a/cluster.go
+++ b/cluster.go
@@ -2,6 +2,7 @@ package timerx
 
 import (
 	"context"
+	"encoding/json"
 	"errors"
 	"fmt"
 	"runtime/debug"
@@ -335,7 +336,14 @@ func (c *Cluster) watch() {
 			_, ok := clusterWorkerList.Load(keys[1])
 			if !ok {
 				c.logger.Errorf(c.ctx, "watch timer:任务不存在%+v", keys[1])
-				c.redis.SAdd(c.ctx, c.setKey, keys[1])
+
+				rd := ReJobData{
+					TaskId: keys[1],
+					Times:  1,
+				}
+				rdb, _ := json.Marshal(rd)
+
+				c.redis.SAdd(c.ctx, c.setKey, string(rdb))
 				continue
 			}
 			go c.doTask(c.ctx, keys[1])
@@ -345,7 +353,7 @@ func (c *Cluster) watch() {
 	// 处理重入任务
 	go func() {
 		for {
-			taskId, err := c.redis.SPop(c.ctx, c.setKey).Result()
+			res, err := c.redis.SPop(c.ctx, c.setKey).Result()
 			if err != nil {
 				if err == redis.Nil {
 					// 已经是空了就不要浪费资源了
@@ -355,18 +363,40 @@ func (c *Cluster) watch() {
 				}
 				continue
 			}
-			_, ok := clusterWorkerList.Load(taskId)
-			if !ok {
-				c.logger.Errorf(c.ctx, "watch timer:任务不存在%+v", taskId)
-				c.redis.SAdd(c.ctx, c.setKey, taskId)
+
+			var rd ReJobData
+			err = json.Unmarshal([]byte(res), &rd)
+			if err != nil {
+				c.logger.Errorf(c.ctx, "json.Unmarshal err:%+v", err)
 				continue
 			}
-			go c.doTask(c.ctx, taskId)
+
+			_, ok := clusterWorkerList.Load(rd.TaskId)
+			if !ok {
+				c.logger.Errorf(c.ctx, "watch timer:任务不存在%+v", rd.TaskId)
+
+				if rd.Times >= 3 {
+					// 重试3次还是失败就不执行了
+					continue
+				}
+				rd.Times++
+
+				rdb, _ := json.Marshal(rd)
+
+				c.redis.SAdd(c.ctx, c.setKey, string(rdb))
+				continue
+			}
+			go c.doTask(c.ctx, rd.TaskId)
 		}
 	}()
 
 }
 
+type ReJobData struct {
+	TaskId string
+	Times  int
+}
+
 // 执行任务
 func (c *Cluster) doTask(ctx context.Context, taskId string) {