Apache Camel SQL批量插入需要很长时间

时间:2016-10-26 12:06:22

标签: java apache-camel batch-processing camel-sql camel-jdbc

我正在使用Apache Camel SQL批量插入过程。

  1. 我的应用程序正在从Active MQ中读取票证,其中包含大约2000张票证。

  2. 我已将批次更新为100。

  3. 我要解雇的查询如下:

    sql.subs.insertCdr= insert into subscription_logs(master_id,request_type,req_desc,msisdn,amount,status,resp_code,resp_desc,channel,transaction_id,se_mode,be_mode,sub_type,sub_timeleft,srv_name,srv_id,start_date,end_date,operator,circle,country,time_offset,retry_count,user_status,previous_state,se_reqrecvtime,se_respsenttime,be_reqsenttime,be_resprecvtime,cp_id,cp_name,sub_srvname,sub_srvid,msg_senderid,msg_text,call_back_url,call_back_resp,client_ip,se_sysIp,language,cp_callbackurlhittime,action,alert,notification_url,notification_resp) values(:#masterId, :#requestType,:#reqDesc,:#msisdnCdr,:#price,:#status,:#responseCode,:#reason,:#channel,:#transactionId,:#seMode,:#beMode,:#subType,:#subTimeLeft,:#serviceName,:#serviceId,:#subStartDate,:#cdrEndDate,:#operator,:#circle,:#country,:#timeOffset,:#retryCount,:#userStatus,:#previousState,:#seReqRecvTime,:#seRespSentTime,:#beReqSentTime,:#beRespRecvTime,:#cpId,:#cpName,:#subServiceName,:#subServiceId,:#shortCode,:#message,:#callBackUrl,:#callBackResp,:#clientIp,:#seSysIp,:#language,:#cpCallbackUrlHitTime,:#action,:#alert,:#notificationUrl,:#notificationResponse)

  4. SQL批处理路由定义如下:

    <pipeline>
       <log message="Going to insert in database"></log>
       <transform>
          <method ref="insertionBean" method="subsBatchInsertion"></method>
       </transform>
       <choice>
           <when>
               <simple>${in.header.subsCount} == ${properties:batch.size}</simple>
               <to uri="sql:{{sql.subs.insertCdr}}?batch=true"></to>
               <log message="Inserted rows ${body}"></log>
           </when>
       </choice>
    </pipeline>
    
  5. 以下是我的java代码:

    public List<Map<String, Object>> subsBatchInsertion(Exchange exchange) {
    if (subsBatchCounter > batchSize) {
        subsPayLoad.clear();
        subsBatchCounter = 1;
    }
    subsPayLoad.add(generateInsert(exchange.getIn().getBody(SubscriptionCdr.class)));
    exchange.getIn().setHeader("subsCount", subsBatchCounter);
    subsBatchCounter++;
    return subsPayLoad;
    }
    
    public Map<String, Object> generateInsert(Cdr cdr) {
    Map<String, Object> insert = new HashMap<String, Object>();
    try {
        insert = BeanUtils.describe(cdr);
    } catch (Exception e) {
        Logger.sysLog(LogValues.error, this.getClass().getName()+" | "+Thread.currentThread().getStackTrace()[1].getMethodName(), coreException.GetStack(e));
    } 
    for (String name : insert.keySet()) {
        Logger.sysLog(LogValues.APP_DEBUG, this.getClass().getName(), name + ":"+ insert.get(name) + "\t");
    }
    return insert;
    }
    
  6. 现在问题是当ActiveMQ中有大约120个票证时,SQL批处理应该已经开始将值插入到数据库中。但它需要花费更多的时间。当ActiveMQ中有大约500张票时,它开始插入过程。 anyboody可以帮助优化插入过程吗? 还是其他方法?

1 个答案:

答案 0 :(得分:6)

问题在于ActiceMQ的消费者数量。

  

当我将消费者数量更改回1时,批次会更新   准时。

实际上,当消费者数量为10时,门票被并行消耗。这意味着,对于有10个消费者的activemq消费的100张票,每个消费者大约有10张票,因此增加了更多的时间。当任何一个消费者获得100张票时,批次就会更新。

  

因此,将消费者数量更改为1会使所有门票成为现实   由单个消费者处理,从而执行批量更新   细