一、总体流程

• 黄色： Job 部分的执行阶段
• 蓝色： Task 部分的执行阶段
• 绿色：框架执行阶段。

二、程序入口

1.datax.py

ENGINE_COMMAND = "java -server ${jvm} %s -classpath %s  ${params} com.alibaba.datax.core.Engine -mode ${mode} -jobid ${jobid} -job ${job}" % (DEFAULT_PROPERTY_CONF, CLASS_PATH)

• 从这里看出来，java的入口是：com.alibaba.datax.core.Engine

2.com.alibaba.datax.core.Engine.java

• 路径是：datax/core/src/main/java/com/alibaba/datax/core/Engine.java

参数加载、容器创建、容器启动

public void start(Configuration allConf) {

      // 绑定column转换信息
      ColumnCast.bind(allConf);

      /**
       * 初始化PluginLoader，可以获取各种插件配置
       */
      LoadUtil.bind(allConf);

      boolean isJob = !("taskGroup".equalsIgnoreCase(allConf
              .getString(CoreConstant.DATAX_CORE_CONTAINER_MODEL)));
      //JobContainer会在schedule后再行进行设置和调整值
      int channelNumber =0;
      AbstractContainer container;
      long instanceId;
      int taskGroupId = -1;
      if (isJob) {
          // 如果是作业，创建作业容器
          allConf.set(CoreConstant.DATAX_CORE_CONTAINER_JOB_MODE, RUNTIME_MODE);
          container = new JobContainer(allConf);
          instanceId = allConf.getLong(
                  CoreConstant.DATAX_CORE_CONTAINER_JOB_ID, 0);

      } else {
      	   // 如果不是作业容器，创建作业组容器
          container = new TaskGroupContainer(allConf);
          instanceId = allConf.getLong(
                  CoreConstant.DATAX_CORE_CONTAINER_JOB_ID);
          taskGroupId = allConf.getInt(
                  CoreConstant.DATAX_CORE_CONTAINER_TASKGROUP_ID);
          channelNumber = allConf.getInt(
                  CoreConstant.DATAX_CORE_CONTAINER_TASKGROUP_CHANNEL);
      }

      //缺省打开perfTrace
      boolean traceEnable = allConf.getBool(CoreConstant.DATAX_CORE_CONTAINER_TRACE_ENABLE, true);
      boolean perfReportEnable = allConf.getBool(CoreConstant.DATAX_CORE_REPORT_DATAX_PERFLOG, true);

      //standalone模式的 datax shell任务不进行汇报
      if(instanceId == -1){
          perfReportEnable = false;
      }

      Configuration jobInfoConfig = allConf.getConfiguration(CoreConstant.DATAX_JOB_JOBINFO);
      //初始化PerfTrace
      PerfTrace perfTrace = PerfTrace.getInstance(isJob, instanceId, taskGroupId, traceEnable);
      perfTrace.setJobInfo(jobInfoConfig,perfReportEnable,channelNumber);
      container.start();
 }

容器启动的执行过程

@Override
public void start() {
    LOG.info("DataX jobContainer starts job.");

    boolean hasException = false;
    boolean isDryRun = false;
    try {
        this.startTimeStamp = System.currentTimeMillis();
        isDryRun = configuration.getBool(CoreConstant.DATAX_JOB_SETTING_DRYRUN, false);
        if(isDryRun) {
            LOG.info("jobContainer starts to do preCheck ...");
            this.preCheck();
        } else {
            userConf = configuration.clone();
            LOG.debug("jobContainer starts to do preHandle ...");
            //Job 前置操作
            this.preHandle();

            LOG.debug("jobContainer starts to do init ...");
            //初始化 reader 和 writer
            this.init();
            LOG.info("jobContainer starts to do prepare ...");
            //全局准备工作，比如 odpswriter 清空目标表
            this.prepare();
            LOG.info("jobContainer starts to do split ...");
            // 拆分task是重点要看的
            this.totalStage = this.split();
            LOG.info("jobContainer starts to do schedule ...");
            // 调度是重点要看的
            this.schedule();
            LOG.debug("jobContainer starts to do post ...");
            this.post();

            LOG.debug("jobContainer starts to do postHandle ...");
            this.postHandle();
            LOG.info("DataX jobId [{}] completed successfully.", this.jobId);

            this.invokeHooks();
        }
    } catch (Throwable e) {
        ...
    } finally {
        ...
    }
}

3.切分的逻辑

private int split() {
	// 调整channel数量
       this.adjustChannelNumber();

       if (this.needChannelNumber <= 0) {
           this.needChannelNumber = 1;
       }

       // 切分逻辑：读和写的数量必须要对应。自己写插件的时候需要注意
       List<Configuration> readerTaskConfigs = this
               .doReaderSplit(this.needChannelNumber);
       int taskNumber = readerTaskConfigs.size();
       List<Configuration> writerTaskConfigs = this
               .doWriterSplit(taskNumber);

       List<Configuration> transformerList = this.configuration.getListConfiguration(CoreConstant.DATAX_JOB_CONTENT_TRANSFORMER);

       LOG.debug("transformer configuration: "+ JSON.toJSONString(transformerList));
       /**
        * 输入是reader和writer的parameter list，输出是content下面元素的list
        */
       List<Configuration> contentConfig = mergeReaderAndWriterTaskConfigs(
               readerTaskConfigs, writerTaskConfigs, transformerList);


       LOG.debug("contentConfig configuration: "+ JSON.toJSONString(contentConfig));

       this.configuration.set(CoreConstant.DATAX_JOB_CONTENT, contentConfig);

       return contentConfig.size();
   }

并发数的确认

private void adjustChannelNumber() {
        int needChannelNumberByByte = Integer.MAX_VALUE;
        int needChannelNumberByRecord = Integer.MAX_VALUE;
        
        // 每秒传输的字节数的上限
        // 配置在json文件的 job.setting.speed.byte
        boolean isByteLimit = (this.configuration.getInt(
                CoreConstant.DATAX_JOB_SETTING_SPEED_BYTE, 0) > 0);
        if (isByteLimit) {
            long globalLimitedByteSpeed = this.configuration.getInt(
                    CoreConstant.DATAX_JOB_SETTING_SPEED_BYTE, 10 * 1024 * 1024);

            // 在byte流控情况下，单个Channel流量最大值必须设置，否则报错！
            Long channelLimitedByteSpeed = this.configuration
                    .getLong(CoreConstant.DATAX_CORE_TRANSPORT_CHANNEL_SPEED_BYTE);
            if (channelLimitedByteSpeed == null || channelLimitedByteSpeed <= 0) {
                throw DataXException.asDataXException(
                        FrameworkErrorCode.CONFIG_ERROR,
                        "在有总bps限速条件下，单个channel的bps值不能为空，也不能为非正数");
            }
			
            needChannelNumberByByte =
                    (int) (globalLimitedByteSpeed / channelLimitedByteSpeed);
            needChannelNumberByByte =
                    needChannelNumberByByte > 0 ? needChannelNumberByByte : 1;
            LOG.info("Job set Max-Byte-Speed to " + globalLimitedByteSpeed + " bytes.");
        }
		
		//这个参数用于设置总TPS（记录每秒）限速。
        // 配置在json文件的 job.setting.speed.record
        boolean isRecordLimit = (this.configuration.getInt(
                CoreConstant.DATAX_JOB_SETTING_SPEED_RECORD, 0)) > 0;
        if (isRecordLimit) {
            long globalLimitedRecordSpeed = this.configuration.getInt(
                    CoreConstant.DATAX_JOB_SETTING_SPEED_RECORD, 100000);

            Long channelLimitedRecordSpeed = this.configuration.getLong(
                    CoreConstant.DATAX_CORE_TRANSPORT_CHANNEL_SPEED_RECORD);
            if (channelLimitedRecordSpeed == null || channelLimitedRecordSpeed <= 0) {
                throw DataXException.asDataXException(FrameworkErrorCode.CONFIG_ERROR,
                        "在有总tps限速条件下，单个channel的tps值不能为空，也不能为非正数");
            }

            needChannelNumberByRecord =
                    (int) (globalLimitedRecordSpeed / channelLimitedRecordSpeed);
            needChannelNumberByRecord =
                    needChannelNumberByRecord > 0 ? needChannelNumberByRecord : 1;
            LOG.info("Job set Max-Record-Speed to " + globalLimitedRecordSpeed + " records.");
        }

        // 取较小值
        this.needChannelNumber = needChannelNumberByByte < needChannelNumberByRecord ?
                needChannelNumberByByte : needChannelNumberByRecord;

        // 如果从byte或record上设置了needChannelNumber则退出
        if (this.needChannelNumber < Integer.MAX_VALUE) {
            return;
        }

                 
        // 这个参数用于设置DataX Job内Channel的并发数。
        // 配置在json文件的 job.setting.speed.channel
        boolean isChannelLimit = (this.configuration.getInt(
                CoreConstant.DATAX_JOB_SETTING_SPEED_CHANNEL, 0) > 0);
        if (isChannelLimit) {
            this.needChannelNumber = this.configuration.getInt(
                    CoreConstant.DATAX_JOB_SETTING_SPEED_CHANNEL);

            LOG.info("Job set Channel-Number to " + this.needChannelNumber
                    + " channels.");

            return;
        }

        throw DataXException.asDataXException(
                FrameworkErrorCode.CONFIG_ERROR,
                "Job运行速度必须设置");
    }

3.调度

3.1 确定组数和分组算法

• datax/core/src/main/java/com/alibaba/datax/core/job/JobContainer.java

private void schedule() {
     /**
      * 这里的全局speed和每个channel的速度设置为B/s
      */
     int channelsPerTaskGroup = this.configuration.getInt(
             CoreConstant.DATAX_CORE_CONTAINER_TASKGROUP_CHANNEL, 5);
     int taskNumber = this.configuration.getList(
             CoreConstant.DATAX_JOB_CONTENT).size();
	
	// this.needChannelNumber参数在split里面计算出来的，和taskNumber任务数 取最小值
     this.needChannelNumber = Math.min(this.needChannelNumber, taskNumber);
     PerfTrace.getInstance().setChannelNumber(needChannelNumber);

     /**
      * 通过获取配置信息得到每个taskGroup需要运行哪些tasks任务
      */
	// 公平的分配task到对应的taskGroup中
     List<Configuration> taskGroupConfigs = JobAssignUtil.assignFairly(this.configuration,
             this.needChannelNumber, channelsPerTaskGroup);

     LOG.info("Scheduler starts [{}] taskGroups.", taskGroupConfigs.size());

     ExecuteMode executeMode = null;
     AbstractScheduler scheduler;
     try {
     	executeMode = ExecuteMode.STANDALONE;
         scheduler = initStandaloneScheduler(this.configuration);

         //设置 executeMode
         for (Configuration taskGroupConfig : taskGroupConfigs) {
             taskGroupConfig.set(CoreConstant.DATAX_CORE_CONTAINER_JOB_MODE, executeMode.getValue());
         }

         if (executeMode == ExecuteMode.LOCAL || executeMode == ExecuteMode.DISTRIBUTE) {
             if (this.jobId <= 0) {
                 throw DataXException.asDataXException(FrameworkErrorCode.RUNTIME_ERROR,
                         "在[ local | distribute ]模式下必须设置jobId，并且其值 > 0 .");
             }
         }

         LOG.info("Running by {} Mode.", executeMode);

         this.startTransferTimeStamp = System.currentTimeMillis();
         
         // 这里调用了schedule
         scheduler.schedule(taskGroupConfigs);

         this.endTransferTimeStamp = System.currentTimeMillis();
     } catch (Exception e) {
         LOG.error("运行scheduler 模式[{}]出错.", executeMode);
         this.endTransferTimeStamp = System.currentTimeMillis();
         throw DataXException.asDataXException(
                 FrameworkErrorCode.RUNTIME_ERROR, e);
     }

     /**
      * 检查任务执行情况
      */
     this.checkLimit();
 }

• 如果100个Task和20个Channel，需要几个TaskGroup？
  • 每个TaskGroup默认5个channel，那么需要4个组

public final class JobAssignUtil {
    private JobAssignUtil() {
    }

    public static List<Configuration> assignFairly(Configuration configuration, int channelNumber, int channelsPerTaskGroup) {
        ...
        // 计算需要多少个TaskGrouop组的核心逻辑
        int taskGroupNumber = (int) Math.ceil(1.0 * channelNumber / channelsPerTaskGroup);
        ...
        // 计算N个Task如何分到这些TaskGroup组
        List<Configuration> taskGroupConfig = doAssign(resourceMarkAndTaskIdMap, configuration, taskGroupNumber);
    }
}

分组算法

	/**
     * /**
     * 需要实现的效果通过例子来说是：
     * <pre>
     * a 库上有表：0, 1, 2
     * b 库上有表：3, 4
     * c 库上有表：5, 6, 7
     *
     * 如果有 4个 taskGroup
     * 则 assign 后的结果为：
     * taskGroup-0: 0,  4,
     * taskGroup-1: 3,  6,
     * taskGroup-2: 5,  2,
     * taskGroup-3: 1,  7
     *
     * </pre>
     */
    private static List<Configuration> doAssign(LinkedHashMap<String, List<Integer>> resourceMarkAndTaskIdMap, Configuration jobConfiguration, int taskGroupNumber) {
        List<Configuration> contentConfig = jobConfiguration.getListConfiguration(CoreConstant.DATAX_JOB_CONTENT);

        Configuration taskGroupTemplate = jobConfiguration.clone();
        taskGroupTemplate.remove(CoreConstant.DATAX_JOB_CONTENT);

        List<Configuration> result = new LinkedList<Configuration>();

        List<List<Configuration>> taskGroupConfigList = new ArrayList<List<Configuration>>(taskGroupNumber);
        for (int i = 0; i < taskGroupNumber; i++) {
            taskGroupConfigList.add(new LinkedList<Configuration>());
        }

        int mapValueMaxLength = -1;

        List<String> resourceMarks = new ArrayList<String>();
        for (Map.Entry<String, List<Integer>> entry : resourceMarkAndTaskIdMap.entrySet()) {
            resourceMarks.add(entry.getKey());
            if (entry.getValue().size() > mapValueMaxLength) {
                mapValueMaxLength = entry.getValue().size();
            }
        }

        int taskGroupIndex = 0;
        for (int i = 0; i < mapValueMaxLength; i++) {
            for (String resourceMark : resourceMarks) {
                if (resourceMarkAndTaskIdMap.get(resourceMark).size() > 0) {
                    int taskId = resourceMarkAndTaskIdMap.get(resourceMark).get(0);
                    taskGroupConfigList.get(taskGroupIndex % taskGroupNumber).add(contentConfig.get(taskId));
                    taskGroupIndex++;

                    resourceMarkAndTaskIdMap.get(resourceMark).remove(0);
                }
            }
        }

        Configuration tempTaskGroupConfig;
        for (int i = 0; i < taskGroupNumber; i++) {
            tempTaskGroupConfig = taskGroupTemplate.clone();
            tempTaskGroupConfig.set(CoreConstant.DATAX_JOB_CONTENT, taskGroupConfigList.get(i));
            tempTaskGroupConfig.set(CoreConstant.DATAX_CORE_CONTAINER_TASKGROUP_ID, i);

            result.add(tempTaskGroupConfig);
        }

        return result;
    }

调度核心代码实现

• 多线程池
• datax/core/src/main/java/com/alibaba/datax/core/job/scheduler/AbstractScheduler.java

public abstract class AbstractScheduler {
    private static final Logger LOG = LoggerFactory
            .getLogger(AbstractScheduler.class);

    private ErrorRecordChecker errorLimit;

    private AbstractContainerCommunicator containerCommunicator;

    private Long jobId;

    public Long getJobId() {
        return jobId;
    }

    public AbstractScheduler(AbstractContainerCommunicator containerCommunicator) {
        this.containerCommunicator = containerCommunicator;
    }

    public void schedule(List<Configuration> configurations) {
        ...
        // 核心代码
        startAllTaskGroup(configurations);
        ...
    }
}

• datax/core/src/main/java/com/alibaba/datax/core/job/scheduler/processinner/ProcessInnerScheduler.java

public abstract class ProcessInnerScheduler extends AbstractScheduler {

    private ExecutorService taskGroupContainerExecutorService;

    public ProcessInnerScheduler(AbstractContainerCommunicator containerCommunicator) {
        super(containerCommunicator);
    }

    @Override
    public void startAllTaskGroup(List<Configuration> configurations) {
        // 使用了线程池
        this.taskGroupContainerExecutorService = Executors
                .newFixedThreadPool(configurations.size());

        for (Configuration taskGroupConfiguration : configurations) {
            TaskGroupContainerRunner taskGroupContainerRunner = newTaskGroupContainerRunner(taskGroupConfiguration);
            this.taskGroupContainerExecutorService.execute(taskGroupContainerRunner);
        }

        this.taskGroupContainerExecutorService.shutdown();
    }
}

3.2 数据传输

• 线程池执行了TaskGroupContainerRunner对象
• datax/core/src/main/java/com/alibaba/datax/core/taskgroup/runner/TaskGroupContainerRunner.java

public class TaskGroupContainerRunner implements Runnable {
	@Override
	public void run() {
		try {
            ...
            // 启动了这个组容器
            this.taskGroupContainer.start();
			...
		} catch (Throwable e) {
			...
		}
	}
}

• 查看一下实现了什么run方法
• datax/core/src/main/java/com/alibaba/datax/core/taskgroup/TaskGroupContainer.java

public class TaskGroupContainer extends AbstractContainer {
@Override
    public void start() {
        try {
            ...
            while (true) {
            	...
            	
                while(iterator.hasNext() && runTasks.size() < channelNumber){
                    ...
                	TaskExecutor taskExecutor = new TaskExecutor(taskConfigForRun, attemptCount);
                    taskStartTimeMap.put(taskId, System.currentTimeMillis());
                    // 这里是真正的执行逻辑
                	taskExecutor.doStart();
                    ...
                }
                ...
            }
            ...
        } catch (Throwable e) {
            ...
        }finally {
            ...
        }
    }

	public void doStart() {
	        // 写的线程启动
            this.writerThread.start();

            // reader没有起来，writer不可能结束
            if (!this.writerThread.isAlive() || this.taskCommunication.getState() == State.FAILED) {
                throw DataXException.asDataXException(
                        FrameworkErrorCode.RUNTIME_ERROR,
                        this.taskCommunication.getThrowable());
            }
			// 读的线程启动
            this.readerThread.start();

            // 这里reader可能很快结束
            if (!this.readerThread.isAlive() && this.taskCommunication.getState() == State.FAILED) {
                // 这里有可能出现Reader线上启动即挂情况 对于这类情况 需要立刻抛出异常
                throw DataXException.asDataXException(
                        FrameworkErrorCode.RUNTIME_ERROR,
                        this.taskCommunication.getThrowable());
            }

        }
}

• writerThread来自WriterRunner，readerThread来自ReaderRunner
  • ReaderRunner跟限速有关系
• datax/core/src/main/java/com/alibaba/datax/core/taskgroup/runner/ReaderRunner.java

public class ReaderRunner extends AbstractRunner implements Runnable {

   private static final Logger LOG = LoggerFactory
           .getLogger(ReaderRunner.class);

   private RecordSender recordSender;

   public void setRecordSender(RecordSender recordSender) {
       this.recordSender = recordSender;
   }

   public ReaderRunner(AbstractTaskPlugin abstractTaskPlugin) {
       super(abstractTaskPlugin);
   }

   @Override
   public void run() {
       assert null != this.recordSender;

       Reader.Task taskReader = (Reader.Task) this.getPlugin();

       //统计waitWriterTime，并且在finally才end。
       PerfRecord channelWaitWrite = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.WAIT_WRITE_TIME);
       try {
           channelWaitWrite.start();

           LOG.debug("task reader starts to do init ...");
           PerfRecord initPerfRecord = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.READ_TASK_INIT);
           initPerfRecord.start();
           taskReader.init();
           initPerfRecord.end();

           LOG.debug("task reader starts to do prepare ...");
           PerfRecord preparePerfRecord = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.READ_TASK_PREPARE);
           preparePerfRecord.start();
           taskReader.prepare();
           preparePerfRecord.end();

           LOG.debug("task reader starts to read ...");
           PerfRecord dataPerfRecord = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.READ_TASK_DATA);
           // 最核心
           dataPerfRecord.start();
           taskReader.startRead(recordSender);
           recordSender.terminate();

           dataPerfRecord.addCount(CommunicationTool.getTotalReadRecords(super.getRunnerCommunication()));
           dataPerfRecord.addSize(CommunicationTool.getTotalReadBytes(super.getRunnerCommunication()));
           dataPerfRecord.end();

           LOG.debug("task reader starts to do post ...");
           PerfRecord postPerfRecord = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.READ_TASK_POST);
           postPerfRecord.start();
           taskReader.post();
           postPerfRecord.end();
           // automatic flush
           // super.markSuccess(); 这里不能标记为成功，成功的标志由 writerRunner 来标志（否则可能导致 reader 先结束，而 writer 还没有结束的严重 bug）
       } catch (Throwable e) {
           LOG.error("Reader runner Received Exceptions:", e);
           super.markFail(e);
       } finally {
           LOG.debug("task reader starts to do destroy ...");
           PerfRecord desPerfRecord = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.READ_TASK_DESTROY);
           desPerfRecord.start();
           super.destroy();
           desPerfRecord.end();

           channelWaitWrite.end(super.getRunnerCommunication().getLongCounter(CommunicationTool.WAIT_WRITER_TIME));

           long transformerUsedTime = super.getRunnerCommunication().getLongCounter(CommunicationTool.TRANSFORMER_USED_TIME);
           if (transformerUsedTime > 0) {
               PerfRecord transformerRecord = new PerfRecord(getTaskGroupId(), getTaskId(), PerfRecord.PHASE.TRANSFORMER_TIME);
               transformerRecord.start();
               transformerRecord.end(transformerUsedTime);
           }
       }
   }

   public void shutdown(){
       recordSender.shutdown();
   }
}

• 最核心的是 dataPerfRecord.start()
• 拿一个MySQL的reader案例看：MysqlReader.java

public class MysqlReader extends Reader {

    @Override
        public void startRead(RecordSender recordSender) {
            int fetchSize = this.readerSliceConfig.getInt(Constant.FETCH_SIZE);
			
		    // 核心方法
            this.commonRdbmsReaderTask.startRead(this.readerSliceConfig, recordSender,
                    super.getTaskPluginCollector(), fetchSize);
        }
}

• 查看 CommonRdbmsReaderTask的startRead函数

public class CommonRdbmsReader {
	public static class Task {
        ...
        public void startRead(Configuration readerSliceConfig,
                              RecordSender recordSender,
                              TaskPluginCollector taskPluginCollector, int fetchSize) {
            ...
            try {
                ...
                while (rs.next()) {
                    rsNextUsedTime += (System.nanoTime() - lastTime);
                    // 核心逻辑：对单个数据的处理逻辑
                    this.transportOneRecord(recordSender, rs,
                            metaData, columnNumber, mandatoryEncoding, taskPluginCollector);
                    lastTime = System.nanoTime();
                }
                ...
            }catch (Exception e) {
                throw RdbmsException.asQueryException(this.dataBaseType, e, querySql, table, username);
            } finally {
                DBUtil.closeDBResources(null, conn);
            }
        }
    }

	protected Record transportOneRecord(RecordSender recordSender, ResultSet rs, 
                ResultSetMetaData metaData, int columnNumber, String mandatoryEncoding, 
                TaskPluginCollector taskPluginCollector) {
            Record record = buildRecord(recordSender,rs,metaData,columnNumber,mandatoryEncoding,taskPluginCollector); 
            // 每次处理完成，都会发送到writer
            recordSender.sendToWriter(record);
            return record;
        }
}

• 如何将处理好的数据发送到writer的？
• datax/core/src/main/java/com/alibaba/datax/core/transport/exchanger/BufferedRecordExchanger.java

public class BufferedRecordExchanger implements RecordSender, RecordReceiver {
	@Override
	public void sendToWriter(Record record) {
		....
		if (isFull) {
		    // 核心代码
			flush();
		}
        ...
	}
   
    @Override
	public void flush() {
		if(shutdown){
			throw DataXException.asDataXException(CommonErrorCode.SHUT_DOWN_TASK, "");
		}
		// 核心代码，将数据通过channel推送给writer
		this.channel.pushAll(this.buffer);
		this.buffer.clear();
		this.bufferIndex = 0;
		this.memoryBytes.set(0);
	}
}

• 查看channel的pushAll
• datax/core/src/main/java/com/alibaba/datax/core/transport/channel/Channel.java

public abstract class Channel {
	public void pushAll(final Collection<Record> rs) {
        Validate.notNull(rs);
        Validate.noNullElements(rs);
        this.doPushAll(rs);
        // rs.size()：数据条数
        // this.getByteSize(rs)：数据量
        this.statPush(rs.size(), this.getByteSize(rs));
    }

	private void statPush(long recordSize, long byteSize) {
        ...
        if (interval - this.flowControlInterval >= 0) {
            ...

            // 如果是通过速率数限速的
            if (isChannelRecordSpeedLimit) {
                long currentRecordSpeed = (CommunicationTool.getTotalReadRecords(currentCommunication) -
                        CommunicationTool.getTotalReadRecords(lastCommunication)) * 1000 / interval;
                // 当前的速率大于限速的速率，会指定一个睡眠时间
                if (currentRecordSpeed > this.recordSpeed) {
                    // 计算根据recordLimit得到的休眠时间
                    recordLimitSleepTime = currentRecordSpeed * interval / this.recordSpeed
                            - interval;
                }
            }

            // 休眠时间取较大值
            long sleepTime = byteLimitSleepTime < recordLimitSleepTime ?
                    recordLimitSleepTime : byteLimitSleepTime;
            if (sleepTime > 0) {
                try {
                    Thread.sleep(sleepTime);
                } catch (InterruptedException e) {
                    Thread.currentThread().interrupt();
                }
            }
            ...
        }
    }
}

三、DataX性能优化

1.关键参数

• job.setting.speed.channel : channel 并发数
• job.setting.speed.record : 全局配置 channel 的 record 限速
  • 最终的单个channel的限速 = 全局配置channel的record限速 / channel并发数
• job.setting.speed.byte：全局配置 channel 的 byte 限速
  • 最终的单个channel的限速 = 全局配置channel的byte限速 / channel并发数
• core.transport.channel.speed.record：单个 channel 的 record 限速
• core.transport.channel.speed.byte：单个 channel 的 byte 限速

2.优化：提升每个 channel 的速度

• 在 DataX 内部对每个 Channel 会有严格的速度控制，分两种，一种是控制每秒同步的记录数，另外一种是每秒同步的字节数，默认的速度限制是 1MB/s，可以根据具体硬件情况设置这个 byte 速度或者 record 速度，一般设置 byte 速度，比如：我们可以把单个 Channel 的速度上限配置为 5MB

3.优化：提升 DataX Job 内 Channel 并发数

• 并发数 = taskGroup 的数量 * 每个 TaskGroup 并发执行的 Task 数 (默认为 5)。

3.1 配置全局 Byte 限速以及单 Channel Byte 限速

• Channel 个数 = 全局 Byte 限速 / 单 Channel Byte 限速

{
    "core": {
        "transport": {
            "channel": {
                "speed": {
                    "byte": 1048576
                }
            }
        }
    },
    "job": {
        "setting": {
            "speed": {
                "byte": 5242880
            }
        },
	 ...
    }
}

• core.transport.channel.speed.byte=1048576，job.setting.speed.byte=5242880，所以 Channel个数 = 全局 Byte 限速 / 单 Channel Byte 限速=5242880/1048576=5 个

3.2 配置全局 Record 限速以及单 Channel Record 限速

• Channel 个数 = 全局 Record 限速 / 单 Channel Record 限速

{
    "core": {
        "transport": {
            "channel": {
                "speed": {
                    "record": 100
                }
            }
        }
    },
    "job": {
        "setting": {
            "speed": {
                "record": 500
            }
        },
 	...
    }
}

• core.transport.channel.speed.record=100 ， job.setting.speed.record=500, 所 以 配 置 全 局Record 限速以及单 Channel Record 限速，Channel 个数 = 全局 Record 限速 / 单 Channel Record 限速=500/100=5

3.3 直接配置 Channel 个数

• 只有在上面两种未设置才生效，上面两个同时设置是取值小的作为最终的 channel 数。

{
    "job": {
        "setting": {
            "speed": {
                "channel": 5
            }
        },
 	...
    }
}

• 直接配置 job.setting.speed.channel=5，所以 job 内 Channel 并发=5 个

3.提高 JVM 堆内存

• 当提升 DataX Job 内 Channel 并发数时，内存的占用会显著增加，因为 DataX 作为数据交换通道，在内存中会缓存较多的数据。例如 Channel 中会有一个 Buffer，作为临时的数据交换的缓冲区，而在部分 Reader 和 Writer 的中，也会存在一些 Buffer，为了防止 OOM 等错误，调大 JVM 的堆内存。
• 建议将内存设置为 4G 或者 8G，这个也可以根据实际情况来调整。
• 调整 JVM xms xmx 参数的两种方式：一种是直接更改 datax.py 脚本；另一种是在启动的时候，加上对应的参数，如下：

python datax/bin/datax.py --jvm="-Xms8G -Xmx8G" XXX.json