fix(sj_1.1.0): 优化工作流全局上下文逻辑

snail-job-client/snail-job-client-job-core/src/main/java/com/aizuda/snailjob/client/job/core/client/JobEndPoint.java

@@ -1,5 +1,6 @@
 package com.aizuda.snailjob.client.job.core.client;
 
+import cn.hutool.core.util.StrUtil;
 import com.aizuda.snailjob.client.common.annotation.Mapping;
 import com.aizuda.snailjob.client.common.annotation.SnailEndPoint;
 import com.aizuda.snailjob.client.common.log.support.SnailJobLogManager;
@@ -19,12 +20,15 @@ import com.aizuda.snailjob.common.core.context.SpringContext;
 import com.aizuda.snailjob.common.core.enums.JobTaskTypeEnum;
 import com.aizuda.snailjob.common.core.model.JobContext;
 import com.aizuda.snailjob.common.core.model.Result;
+import com.aizuda.snailjob.common.core.util.JsonUtil;
 import com.aizuda.snailjob.common.log.SnailJobLog;
 import com.aizuda.snailjob.common.log.enums.LogTypeEnum;
+import com.google.common.collect.Maps;
 import jakarta.validation.Valid;
 import org.springframework.validation.annotation.Validated;
 
 import java.util.Objects;
+import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ThreadPoolExecutor;
 
 import static com.aizuda.snailjob.common.core.constant.SystemConstants.HTTP_PATH.JOB_DISPATCH;
@@ -116,6 +120,18 @@ public class JobEndPoint {
         jobContext.setRetryScene(dispatchJob.getRetryScene());
         jobContext.setTaskName(dispatchJob.getTaskName());
         jobContext.setMrStage(dispatchJob.getMrStage());
+
+        String wfContext = dispatchJob.getWfContext();
+        if (StrUtil.isNotBlank(wfContext)) {
+            try {
+                jobContext.setWfContext(JsonUtil.parseConcurrentHashMap(wfContext));
+            } catch (Exception e) {
+                SnailJobLog.REMOTE.warn("workflow context parse error", e);
+            }
+        } else {
+            jobContext.setWfContext(Maps.newConcurrentMap());
+        }
+
         return jobContext;
     }
 

snail-job-client/snail-job-client-job-core/src/main/java/com/aizuda/snailjob/client/job/core/dto/JobArgs.java

@@ -1,7 +1,11 @@
 package com.aizuda.snailjob.client.job.core.dto;
 
+import cn.hutool.core.util.StrUtil;
 import lombok.Data;
 
+import java.util.Map;
+import java.util.Objects;
+
 /**
  * @author: opensnail
  * @date : 2023-10-18 16:53
@@ -15,4 +19,15 @@ public class JobArgs {
     private String executorInfo;
 
     private Long taskBatchId;
+
+    private Map<String, Object> wfContext;
+
+    public void appendContext(String key, Object value) {
+        if (Objects.isNull(wfContext) || StrUtil.isBlank(key) || Objects.isNull(value)) {
+            return;
+        }
+
+        wfContext.put(key, value);
+    }
+
 }

snail-job-client/snail-job-client-job-core/src/main/java/com/aizuda/snailjob/client/job/core/executor/AbstractJobExecutor.java

@@ -1,5 +1,6 @@
 package com.aizuda.snailjob.client.job.core.executor;
 
+import cn.hutool.core.util.StrUtil;
 import com.aizuda.snailjob.client.common.log.support.SnailJobLogManager;
 import com.aizuda.snailjob.client.job.core.IJobExecutor;
 import com.aizuda.snailjob.client.job.core.cache.FutureCache;
@@ -12,6 +13,7 @@ import com.aizuda.snailjob.client.model.ExecuteResult;
 import com.aizuda.snailjob.common.core.enums.JobTaskTypeEnum;
 import com.aizuda.snailjob.common.core.enums.MapReduceStageEnum;
 import com.aizuda.snailjob.common.core.model.JobContext;
+import com.aizuda.snailjob.common.core.util.JsonUtil;
 import com.aizuda.snailjob.common.log.enums.LogTypeEnum;
 import com.google.common.collect.Lists;
 import com.google.common.util.concurrent.Futures;
@@ -60,6 +62,8 @@ public abstract class AbstractJobExecutor implements IJobExecutor {
                 jobArgs = buildJobArgs(jobContext);
             }
 
+            jobArgs.setWfContext(jobContext.getWfContext());
+
             try {
                 // 初始化调度信息（日志上报LogUtil）
                 initLogContext(jobContext);
@@ -117,6 +121,7 @@ public abstract class AbstractJobExecutor implements IJobExecutor {
         jobArgs.setArgsStr(jobContext.getArgsStr());
         jobArgs.setExecutorInfo(jobContext.getExecutorInfo());
         jobArgs.setTaskBatchId(jobContext.getTaskBatchId());
+        jobArgs.setWfContext(jobContext.getWfContext());
         return jobArgs;
     }
 

snail-job-client/snail-job-client-job-core/src/main/java/com/aizuda/snailjob/client/job/core/executor/JobExecutorFutureCallback.java

@@ -1,5 +1,6 @@
 package com.aizuda.snailjob.client.job.core.executor;
 
+import cn.hutool.core.collection.CollUtil;
 import cn.hutool.core.date.DatePattern;
 import com.aizuda.snailjob.client.common.cache.GroupVersionCache;
 import com.aizuda.snailjob.client.common.config.SnailJobProperties;
@@ -32,6 +33,7 @@ import lombok.extern.slf4j.Slf4j;
 
 import java.time.LocalDateTime;
 import java.util.List;
+import java.util.Map;
 import java.util.Objects;
 import java.util.Optional;
 import java.util.concurrent.CancellationException;
@@ -160,6 +162,12 @@ public class JobExecutorFutureCallback implements FutureCallback<ExecuteResult>
         dispatchJobRequest.setTaskStatus(status);
         dispatchJobRequest.setRetry(jobContext.isRetry());
         dispatchJobRequest.setRetryScene(jobContext.getRetryScene());
+        // 传递上下文
+        Map<String, Object> wfContext = jobContext.getWfContext();
+        if (CollUtil.isNotEmpty(wfContext)) {
+            dispatchJobRequest.setWfContext(JsonUtil.toJsonString(wfContext));
+        }
+
         return dispatchJobRequest;
     }
 

snail-job-common/snail-job-common-client-api/src/main/java/com/aizuda/snailjob/client/model/request/DispatchJobRequest.java

@@ -73,5 +73,5 @@ public class DispatchJobRequest {
     /**
      * 工作流上下文
      */
-    private String wkContext;
+    private String wfContext;
 }

snail-job-common/snail-job-common-client-api/src/main/java/com/aizuda/snailjob/client/model/request/DispatchJobResultRequest.java

@@ -44,5 +44,5 @@ public class DispatchJobResultRequest {
     /**
      * 工作流上下文
      */
-    private String wkContext;
+    private String wfContext;
 }

snail-job-common/snail-job-common-core/src/main/java/com/aizuda/snailjob/common/core/model/JobContext.java

@@ -3,6 +3,7 @@ package com.aizuda.snailjob.common.core.model;
 import lombok.Data;
 
 import java.util.List;
+import java.util.Map;
 
 /**
  * @author: opensnail
@@ -59,5 +60,15 @@ public class JobContext {
      */
     private String taskName;
 
+    /**
+     * 动态分片所处的阶段
+     */
     private Integer mrStage;
+
+    /**
+     * 工作流全局上下文
+     */
+    private Map<String, Object> wfContext;
+
+
 }

snail-job-common/snail-job-common-core/src/main/java/com/aizuda/snailjob/common/core/util/JsonUtil.java

@@ -17,6 +17,7 @@ import java.util.ArrayList;
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.concurrent.ConcurrentHashMap;
 
 import static com.aizuda.snailjob.common.core.constant.SystemConstants.YYYY_MM_DD;
 import static com.aizuda.snailjob.common.core.constant.SystemConstants.YYYY_MM_DD_HH_MM_SS;
@@ -72,6 +73,16 @@ public class JsonUtil {
         return JsonMapper.toJavaObject(jsonString, HashMap.class);
     }
 
+    /**
+     * 将JSON字符串转ConcurrentHashMap 对象
+     *
+     * @param jsonString
+     * @return
+     */
+    public static <K, V> Map<K, V> parseConcurrentHashMap(String jsonString) {
+        return JsonMapper.toJavaObject(jsonString, ConcurrentHashMap.class);
+    }
+
     /**
      * 将JSON字符串转Map 对象
      *

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/dto/RealJobExecutorDTO.java

@@ -102,5 +102,5 @@ public class RealJobExecutorDTO extends BaseDTO {
     /**
      * 工作流上下文
      */
-    private String wkContext;
+    private String wfContext;
 }

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/JobTaskConverter.java

@@ -106,7 +106,8 @@ public interface JobTaskConverter {
             @Mapping(source = "jobTask.extAttrs", target = "extAttrs"),
             @Mapping(source = "jobTask.namespaceId", target = "namespaceId"),
             @Mapping(source = "jobTask.taskName", target = "taskName"),
-            @Mapping(source = "jobTask.mrStage", target = "mrStage")
+            @Mapping(source = "jobTask.mrStage", target = "mrStage"),
+            @Mapping(source = "context.wfContext", target = "wfContext")
     })
     RealJobExecutorDTO toRealJobExecutorDTO(JobExecutorContext context, JobTask jobTask);
 

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/callback/ClientCallbackContext.java

@@ -50,5 +50,5 @@ public class ClientCallbackContext {
     /**
      * 工作流上下文
      */
-    private String wkContext;
+    private String wfContext;
 }

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/dispatch/JobExecutorActor.java

@@ -198,7 +198,7 @@ public class JobExecutorActor extends AbstractActor {
         context.setWorkflowTaskBatchId(taskExecute.getWorkflowTaskBatchId());
         context.setWorkflowNodeId(taskExecute.getWorkflowNodeId());
         if (Objects.nonNull(workflowTaskBatch)) {
-            context.setWkContext(workflowTaskBatch.getWfContext());
+            context.setWfContext(workflowTaskBatch.getWfContext());
         }
         return context;
     }

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/dispatch/JobExecutorResultActor.java

@@ -41,7 +41,6 @@ public class JobExecutorResultActor extends AbstractActor {
     private static final String KEY = "job_complete_{0}_{1}";
     private final JobTaskMapper jobTaskMapper;
     private final JobTaskBatchHandler jobTaskBatchHandler;
-    private final WorkflowBatchHandler workflowBatchHandler;
     private final DistributedLockHandler distributedLockHandler;
 
     @Override

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/dispatch/WorkflowExecutorActor.java

@@ -118,7 +118,7 @@ public class WorkflowExecutorActor extends AbstractActor {
         // 添加父节点，为了判断父节点的处理状态
         List<JobTaskBatch> allJobTaskBatchList = jobTaskBatchMapper.selectList(new LambdaQueryWrapper<JobTaskBatch>()
                 .select(JobTaskBatch::getWorkflowTaskBatchId, JobTaskBatch::getWorkflowNodeId,
-                        JobTaskBatch::getTaskBatchStatus, JobTaskBatch::getOperationReason)
+                        JobTaskBatch::getTaskBatchStatus, JobTaskBatch::getOperationReason, JobTaskBatch::getId)
                 .eq(JobTaskBatch::getWorkflowTaskBatchId, workflowTaskBatch.getId())
                 .in(JobTaskBatch::getWorkflowNodeId,
                         Sets.union(brotherNode, Sets.newHashSet(taskExecute.getParentId())))
@@ -167,6 +167,10 @@ public class WorkflowExecutorActor extends AbstractActor {
         List<Job> jobs = jobMapper.selectBatchIds(StreamUtils.toSet(workflowNodes, WorkflowNode::getJobId));
         Map<Long, Job> jobMap = StreamUtils.toIdentityMap(jobs, Job::getId);
 
+        // TODO 合并job task的结果到全局上下文中
+         workflowBatchHandler.mergeWorkflowContextAndRetry(workflowTaskBatch,
+            StreamUtils.toSet(allJobTaskBatchList, JobTaskBatch::getId));
+
         // 只会条件节点会使用
         Object evaluationResult = null;
         for (WorkflowNode workflowNode : workflowNodes) {

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/executor/job/JobExecutorContext.java

@@ -93,6 +93,6 @@ public class JobExecutorContext {
     /**
      * 工作流上下文
      */
-    private String wkContext;
+    private String wfContext;
 
 }

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/executor/workflow/AbstractWorkflowExecutor.java

@@ -5,6 +5,7 @@ import cn.hutool.core.lang.Assert;
 import cn.hutool.core.util.StrUtil;
 import com.aizuda.snailjob.common.core.enums.JobArgsTypeEnum;
 import com.aizuda.snailjob.common.core.enums.WorkflowNodeTypeEnum;
+import com.aizuda.snailjob.common.log.SnailJobLog;
 import com.aizuda.snailjob.server.common.exception.SnailJobServerException;
 import com.aizuda.snailjob.server.job.task.dto.WorkflowNodeTaskExecuteDTO;
 import com.aizuda.snailjob.server.job.task.support.WorkflowExecutor;
@@ -56,6 +57,9 @@ public abstract class AbstractWorkflowExecutor implements WorkflowExecutor, Init
 
     @Override
     public void execute(WorkflowExecutorContext context) {
+
+        // 若多个兄弟节点的情况下，同时处理完成则每个节点都有可能来执行后继节点，
+        // 因此这里这里添加分布式锁
         distributedLockHandler.lockWithDisposableAndRetry(
                 () -> {
                     long total = 0;
@@ -69,6 +73,7 @@ public abstract class AbstractWorkflowExecutor implements WorkflowExecutor, Init
 
                         if (CollUtil.isNotEmpty(jobTaskBatches)) {
                             total = jobTaskBatches.size();
+                            // ToDo
                             JobTaskBatch jobTaskBatch = jobTaskBatches.get(0);
                             if (WORKFLOW_SUCCESSOR_SKIP_EXECUTION.contains(jobTaskBatch.getOperationReason())) {
                                 context.setEvaluationResult(Boolean.FALSE);
@@ -89,9 +94,6 @@ public abstract class AbstractWorkflowExecutor implements WorkflowExecutor, Init
                         return;
                     }
 
-                    // 合并job task的结果到全局上下文中
-                    workflowBatchHandler.mergeAllWorkflowContext(context.getWorkflowTaskBatchId(), context.getTaskBatchId());
-
                     transactionTemplate.execute(new TransactionCallbackWithoutResult() {
                         @Override
                         protected void doInTransactionWithoutResult(final TransactionStatus status) {

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/handler/JobTaskBatchHandler.java

@@ -40,6 +40,7 @@ import java.util.Map;
 import java.util.Objects;
 import java.util.stream.Collectors;
 
+import static com.aizuda.snailjob.common.core.enums.JobTaskBatchStatusEnum.COMPLETED;
 import static com.aizuda.snailjob.common.core.enums.MapReduceStageEnum.*;
 
 /**
@@ -59,13 +60,23 @@ public class JobTaskBatchHandler {
     @Transactional
     public boolean complete(CompleteJobBatchDTO completeJobBatchDTO) {
 
+        // 幂等处理
+        Long countJobTaskBatch = jobTaskBatchMapper.selectCount(new LambdaQueryWrapper<JobTaskBatch>()
+            .eq(JobTaskBatch::getId, completeJobBatchDTO.getTaskBatchId())
+            .in(JobTaskBatch::getTaskBatchStatus, COMPLETED)
+        );
+        if (countJobTaskBatch > 0) {
+            // 批次已经完成了，不需要重复更新
+            return true;
+        }
+
         List<JobTask> jobTasks = jobTaskMapper.selectList(
-                new LambdaQueryWrapper<JobTask>()
+            new LambdaQueryWrapper<JobTask>()
-                        .select(JobTask::getTaskStatus, JobTask::getMrStage)
+                .select(JobTask::getTaskStatus, JobTask::getMrStage)
-                        .eq(JobTask::getTaskBatchId, completeJobBatchDTO.getTaskBatchId()));
+                .eq(JobTask::getTaskBatchId, completeJobBatchDTO.getTaskBatchId()));
 
         if (CollUtil.isEmpty(jobTasks) ||
-                jobTasks.stream().anyMatch(jobTask -> JobTaskStatusEnum.NOT_COMPLETE.contains(jobTask.getTaskStatus()))) {
+            jobTasks.stream().anyMatch(jobTask -> JobTaskStatusEnum.NOT_COMPLETE.contains(jobTask.getTaskStatus()))) {
             return false;
         }
 
@@ -73,7 +84,7 @@ public class JobTaskBatchHandler {
         jobTaskBatch.setId(completeJobBatchDTO.getTaskBatchId());
 
         Map<Integer, Long> statusCountMap = jobTasks.stream()
-                .collect(Collectors.groupingBy(JobTask::getTaskStatus, Collectors.counting()));
+            .collect(Collectors.groupingBy(JobTask::getTaskStatus, Collectors.counting()));
 
         long failCount = statusCountMap.getOrDefault(JobTaskBatchStatusEnum.FAIL.getStatus(), 0L);
         long stopCount = statusCountMap.getOrDefault(JobTaskBatchStatusEnum.STOP.getStatus(), 0L);
@@ -85,7 +96,8 @@ public class JobTaskBatchHandler {
             jobTaskBatch.setTaskBatchStatus(JobTaskBatchStatusEnum.STOP.getStatus());
         } else {
             // todo 调试完成删除
-            SnailJobLog.LOCAL.info("尝试完成任务. taskBatchId:[{}] [{}]", completeJobBatchDTO.getTaskBatchId(), JsonUtil.toJsonString(jobTasks));
+            SnailJobLog.LOCAL.info("尝试完成任务. taskBatchId:[{}] [{}]", completeJobBatchDTO.getTaskBatchId(),
+                JsonUtil.toJsonString(jobTasks));
 
             jobTaskBatch.setTaskBatchStatus(JobTaskBatchStatusEnum.SUCCESS.getStatus());
             if (needReduceTask(completeJobBatchDTO, jobTasks)) {
@@ -107,9 +119,9 @@ public class JobTaskBatchHandler {
 
         jobTaskBatch.setUpdateDt(LocalDateTime.now());
         return 1 == jobTaskBatchMapper.update(jobTaskBatch,
-                new LambdaUpdateWrapper<JobTaskBatch>()
+            new LambdaUpdateWrapper<JobTaskBatch>()
-                        .eq(JobTaskBatch::getId, completeJobBatchDTO.getTaskBatchId())
+                .eq(JobTaskBatch::getId, completeJobBatchDTO.getTaskBatchId())
-                        .in(JobTaskBatch::getTaskBatchStatus, JobTaskBatchStatusEnum.NOT_COMPLETE)
+                .in(JobTaskBatch::getTaskBatchStatus, JobTaskBatchStatusEnum.NOT_COMPLETE)
         );
 
     }
@@ -149,14 +161,15 @@ public class JobTaskBatchHandler {
 
     private static boolean isAllMapTask(final List<JobTask> jobTasks) {
         return jobTasks.size() == jobTasks.stream()
-                .filter(jobTask -> Objects.nonNull(jobTask.getMrStage()) && MAP.getStage() == jobTask.getMrStage())
+            .filter(jobTask -> Objects.nonNull(jobTask.getMrStage()) && MAP.getStage() == jobTask.getMrStage())
-                .count();
+            .count();
     }
 
     private static boolean isALeastOneReduceTask(final List<JobTask> jobTasks) {
         return jobTasks.stream()
-                .filter(jobTask -> Objects.nonNull(jobTask.getMrStage()) && REDUCE.getStage() == jobTask.getMrStage())
+                   .filter(
-                .count() > 1;
+                       jobTask -> Objects.nonNull(jobTask.getMrStage()) && REDUCE.getStage() == jobTask.getMrStage())
+                   .count() > 1;
     }
 
     /**
@@ -167,21 +180,21 @@ public class JobTaskBatchHandler {
      */
     public void openResidentTask(Job job, TaskExecuteDTO taskExecuteDTO) {
         if (Objects.isNull(job)
-                || JobTaskExecutorSceneEnum.MANUAL_JOB.getType().equals(taskExecuteDTO.getTaskExecutorScene())
+            || JobTaskExecutorSceneEnum.MANUAL_JOB.getType().equals(taskExecuteDTO.getTaskExecutorScene())
-                || JobTaskExecutorSceneEnum.AUTO_WORKFLOW.getType().equals(taskExecuteDTO.getTaskExecutorScene())
+            || JobTaskExecutorSceneEnum.AUTO_WORKFLOW.getType().equals(taskExecuteDTO.getTaskExecutorScene())
-                || JobTaskExecutorSceneEnum.MANUAL_WORKFLOW.getType().equals(taskExecuteDTO.getTaskExecutorScene())
+            || JobTaskExecutorSceneEnum.MANUAL_WORKFLOW.getType().equals(taskExecuteDTO.getTaskExecutorScene())
-                // 是否是常驻任务
+            // 是否是常驻任务
-                || Objects.equals(StatusEnum.NO.getStatus(), job.getResident())
+            || Objects.equals(StatusEnum.NO.getStatus(), job.getResident())
-                // 防止任务已经分配到其他节点导致的任务重复执行
+            // 防止任务已经分配到其他节点导致的任务重复执行
-                || !DistributeInstance.INSTANCE.getConsumerBucket().contains(job.getBucketIndex())
+            || !DistributeInstance.INSTANCE.getConsumerBucket().contains(job.getBucketIndex())
         ) {
             return;
         }
 
         long count = groupConfigMapper.selectCount(new LambdaQueryWrapper<GroupConfig>()
-                .eq(GroupConfig::getNamespaceId, job.getNamespaceId())
+            .eq(GroupConfig::getNamespaceId, job.getNamespaceId())
-                .eq(GroupConfig::getGroupName, job.getGroupName())
+            .eq(GroupConfig::getGroupName, job.getGroupName())
-                .eq(GroupConfig::getGroupStatus, StatusEnum.YES.getStatus()));
+            .eq(GroupConfig::getGroupStatus, StatusEnum.YES.getStatus()));
         if (count == 0) {
             return;
         }
@@ -208,7 +221,7 @@ public class JobTaskBatchHandler {
         Duration duration = Duration.ofMillis(milliseconds - DateUtils.toNowMilli() % 1000);
 
         log.debug("常驻任务监控. [{}] 任务时间差:[{}] 取余:[{}]", duration, milliseconds,
-                DateUtils.toNowMilli() % 1000);
+            DateUtils.toNowMilli() % 1000);
         job.setNextTriggerAt(nextTriggerAt);
         JobTimerWheel.registerWithJob(() -> new ResidentJobTimerTask(jobTimerTaskDTO, job), duration);
         ResidentTaskCache.refresh(job.getId(), nextTriggerAt);

snail-job-server/snail-job-server-job-task/src/main/java/com/aizuda/snailjob/server/job/task/support/handler/WorkflowBatchHandler.java

@@ -61,7 +61,7 @@ import static com.aizuda.snailjob.common.core.enums.JobTaskBatchStatusEnum.NOT_C
 @Component
 @RequiredArgsConstructor
 public class WorkflowBatchHandler {
-    private static final String KEY = "update_wf_context_{}";
+    private static final String KEY = "update_wf_context_{0}";
 
     private final DistributedLockHandler distributedLockHandler;
     private final WorkflowTaskBatchMapper workflowTaskBatchMapper;
@@ -320,8 +320,14 @@ public class WorkflowBatchHandler {
         }
     }
 
-    public void mergeWorkflowContextAndRetry(Long workflowTaskBatchId, String waitMergeContext) {
+    /**
-        if (StrUtil.isBlank(waitMergeContext) || Objects.isNull(workflowTaskBatchId)) {
+     * 合并工作流上下文若合并失败先自旋3次1.5s, 若失败了升级到悲观锁
+     *
+     * @param workflowTaskBatch 工作流批次
+     * @param taskBatchIds 批次列表
+     */
+    public void mergeWorkflowContextAndRetry(WorkflowTaskBatch workflowTaskBatch, Set<Long> taskBatchIds) {
+        if (CollUtil.isEmpty(taskBatchIds)) {
             return;
         }
 
@@ -349,42 +355,62 @@ public class WorkflowBatchHandler {
                 }).build();
 
         try {
-            retryer.call(() -> mergeWorkflowContext(workflowTaskBatchId, JsonUtil.parseHashMap(waitMergeContext, Object.class)));
+            retryer.call(() -> mergeAllWorkflowContext(workflowTaskBatch, taskBatchIds));
         } catch (Exception e) {
-            SnailJobLog.LOCAL.warn("update workflow global context error. workflowTaskBatchId:[{}] waitMergeContext:[{}]",
+            SnailJobLog.LOCAL.warn("update workflow global context error. workflowTaskBatchId:[{}] taskBatchIds:[{}]",
-                    workflowTaskBatchId, waitMergeContext, e);
+                workflowTaskBatch.getId(), taskBatchIds, e);
             if (e.getClass().isAssignableFrom(RetryException.class)) {
                 // 如果自旋失败，就使用悲观锁
                 distributedLockHandler.lockWithDisposableAndRetry(() -> {
-                    mergeWorkflowContext(workflowTaskBatchId, JsonUtil.parseHashMap(waitMergeContext, Object.class));
+                    mergeAllWorkflowContext(workflowTaskBatch, taskBatchIds);
-                }, MessageFormat.format(KEY, workflowTaskBatchId), Duration.ofSeconds(1), Duration.ofSeconds(1), 3);
+                }, MessageFormat.format(KEY, workflowTaskBatch.getId()), Duration.ofSeconds(1), Duration.ofSeconds(1), 3);
             }
         }
     }
 
-    public boolean mergeAllWorkflowContext(Long workflowTaskBatchId, Long taskBatchId) {
+    public boolean mergeAllWorkflowContext(WorkflowTaskBatch workflowTaskBatch, Set<Long> taskBatchIds) {
+        if (CollUtil.isEmpty(taskBatchIds)) {
+            return true;
+        }
+
         List<JobTask> jobTasks = jobTaskMapper.selectList(new LambdaQueryWrapper<JobTask>()
                 .select(JobTask::getWfContext, JobTask::getId)
-                .eq(JobTask::getTaskBatchId, taskBatchId));
+                .in(JobTask::getTaskBatchId, taskBatchIds)
+        );
         if (CollUtil.isEmpty(jobTasks)) {
             return true;
         }
 
         Set<Map<String, Object>> maps = jobTasks.stream().map(r -> {
             try {
-                return JsonUtil.parseHashMap(r.getWfContext(), Object.class);
+                if (StrUtil.isNotBlank(r.getWfContext())) {
+                    return JsonUtil.parseHashMap(r.getWfContext(), Object.class);
+                }
             } catch (Exception e) {
-                SnailJobLog.LOCAL.warn("taskId:[{}] result value 不是一个json对象. result:[{}]", r.getId(), r.getResultMessage());
+                SnailJobLog.LOCAL.warn("taskId:[{}] result value is not a JSON object. result:[{}]", r.getId(), r.getResultMessage());
             }
             return new HashMap<String, Object>();
         }).collect(Collectors.toSet());
 
-        Map<String, Object> mergeMap = Maps.newHashMap();
+        Map<String, Object> mergeMap;
+        if (StrUtil.isBlank(workflowTaskBatch.getWfContext())) {
+            mergeMap = Maps.newHashMap();
+        } else {
+            mergeMap = JsonUtil.parseHashMap(workflowTaskBatch.getWfContext());
+        }
+
         for (Map<String, Object> map : maps) {
             mergeMaps(mergeMap, map);
         }
 
-        return mergeWorkflowContext(workflowTaskBatchId, mergeMap);
+        WorkflowTaskBatch waitUpdateWorkflowTaskBatch = new WorkflowTaskBatch();
+        waitUpdateWorkflowTaskBatch.setId(workflowTaskBatch.getId());
+        waitUpdateWorkflowTaskBatch.setWfContext(JsonUtil.toJsonString(mergeMap));
+        waitUpdateWorkflowTaskBatch.setVersion(1);
+        return 1 == workflowTaskBatchMapper.update(waitUpdateWorkflowTaskBatch, new LambdaQueryWrapper<WorkflowTaskBatch>()
+            .eq(WorkflowTaskBatch::getId, workflowTaskBatch.getId())
+            .eq(WorkflowTaskBatch::getVersion, workflowTaskBatch.getVersion())
+        );
     }
 
     /**