欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

jdbc 多线程公用一个connection PreparedStatement

程序员文章站 2022-06-05 08:51:33
...

 

 最近做一个数据下载接口,功能是这样的:拿n个基金的id,轮训接口下载每一个基金id的详细回来录入自己的数据库。n>10000, 所以采用多线程下载入库。ok。。问题来了。当我采用200根线程同时做这个事情的时候。 调用接口没有问题,在入库的时候connection的创建我有两种想法:

第一:200根线程公用一个connection和PreparedStatement。因为我觉得200根线程的入库操作都是一样的。sql语句都一样。 唯一不同的是给sql参数赋值的过程。(也许会有人说。是否会有线程安全的问题。实践证明并没有引发异常。。。这也是我的一个疑问。 求解。。。。),还有一方面的考虑是,频繁的打开,关闭连接也是会造成了一定的内存消耗。

第二:用线程池管理connection,每根线程拥有自己的connection。(这个不用说,大部分首先都会考虑用线程池,但这里也会有一个问题。当启动200根线程,而连接池*我的连接假设只有50个,那么也就是意味着有150根线程处于等的状态,这就照成整个接口全部执行完消耗的时间增大。)

 

希望有人能为我解答上面两种方案的优劣。!

代码是这样的。

 

public class DownloadTask implements Runnable {
	
	private final Connection con;
	

	public DownloadTask(Connection con) {
	      this.con = con;
	
	}
public void run() {
		
//		Long downloadLong = System.currentTimeMillis();
		List<MonitorEntity> monitorEntitys = new ArrayList<MonitorEntity>();
//调用接口下载详细
		for (String id : idsList) {
			MonitorEntity entity = getMonitorEntity(id);
			monitorEntitys.add(entity);
		}
//		System.out.println("i ="+i+" download api over "+(System.currentTimeMillis()-downloadLong)/1000+" s");
		Long timeLong = System.currentTimeMillis();
	                Connection connection = con;
		PreparedStatement ps = null;
		try {
//			connection.setAutoCommit(false);
			ps = connection.prepareStatement(sqlString);
			for (MonitorEntity entity  : monitorEntitys) {	
			
					ps.setString(1, entity.getApi_name());
					ps.setString(2, entity.getId());
					ps.setString(3, entity.getId_type());
					ps.setString(4, entity.getRequest_url());
					ps.setInt(5, entity.getHttp_code());
					ps.setInt(6, entity.getRes_code());
					ps.setInt(7, entity.getRes_time());
					ps.setString(8, entity.getRes_msg());
					ps.setString(9, entity.getResult());
					ps.setString(10, entity.getLast_update_time());
					ps.setString(11, entity.getRequest_url());
					ps.setInt(12, entity.getHttp_code());
					ps.setInt(13, entity.getRes_code());
					ps.setInt(14, entity.getRes_time());
					ps.setString(15, entity.getRes_msg());
					ps.setString(16, entity.getResult());
					ps.setString(17, entity.getLast_update_time());
					ps.addBatch();
			}
			ps.executeBatch();
//			connection.commit(); 
		} catch (Exception e) {
			log.error("insert or update database error", e);
		} finally {
			jdbcUtil.closePreparedStatement(ps);
					}
		System.out.println("i ="+i+" insertOrUpdate database over "+(System.currentTimeMillis()-timeLong)/1000+" s");
		countdown.countDown();
	}

 

 

一个类实现Runnable接口,在这个这个类构造函数中增加一个参数Connection con,然后再主线程 new 子线程的时候吧Connection 作为参数传递进来。

 这里只是公用一个connection, 效率很慢,一部分线程只用不到1s钟就执行完了。 但越到后面的线程执行的时间却越长,我个人认为是一个connection只能new出一定的PreparedStatement ,所以导致后面的线程必须等前面的线程执行完,释放掉PreparedStatement ,才能创建新的PreparedStatement 。不知道这样认为是否正确。

后来我想,如果是这样干脆把PreparedStatement 作为参数传递进来。是否能快呢?

代码如下:

 

 

public class DownloadTask implements Runnable {
	static Log log = LogFactory.getLog(DownloadTask.class);
	private final JDBCUtil jdbcUtil;
	private final Connection con;
	private final List<String> idsList;
	private final String url_head;
	private final String url_param;
	private final String sqlString;
	private final String api_name;
	private final String id_type;
	CountDownLatch countdown;
	private final PreparedStatement ps;
	private final int i;
	static SimpleDateFormat sf = new SimpleDateFormat("yyyy-MM-dd hh:mm:ss");

	public DownloadTask(JDBCUtil jdbcUtil, Connection con,
			List<String> idsList, String url_head, String url_param,
			String sqlString, String api_name, String id_type,CountDownLatch countdown,int i,PreparedStatement ps) {
		this.jdbcUtil = jdbcUtil;
		this.con = con;
		this.idsList = idsList;
		this.url_head = url_head;
		this.url_param = url_param;
		this.sqlString = sqlString;
		this.api_name = api_name;
		this.id_type = id_type;
		this.countdown = countdown;
		this.i = i ;
		this.ps = ps;
	}

	private MonitorEntity getMonitorEntity(String id) {
		MonitorEntity entity = new MonitorEntity();
		String url = url_head + id + url_param;
		Long startTimeLong = System.currentTimeMillis();
		String[] responseString;
		try {
			responseString = DownloadData.download(url);
		} catch (Exception e) {
			log.error("download error url="+url,e);
			return null;
		}
		Long endTimeLong = System.currentTimeMillis();
		Long restimeLong = endTimeLong - startTimeLong;
		entity.setId(id);
		entity.setId_type(id_type);
		entity.setApi_name(api_name);
		entity.setRequest_url(url);
		entity.setHttp_code(Integer.parseInt(responseString[0]));
		entity.setRes_time(restimeLong.intValue());
		String resultString = responseString[1];
		try {
			
			JSONObject jsonObject = JSONObject.fromObject(resultString);

			JSONObject statusJsonObject = jsonObject.getJSONObject("status");
			entity.setRes_code(statusJsonObject.getInt("code"));
			entity.setRes_msg(statusJsonObject.getString("message"));
			entity.setResult(jsonObject.getJSONObject("data").toString());
		} catch (Exception e) {
			log.error("analysis jsonobject error ", e);
			entity.setRes_code(-1);
			entity.setRes_msg("analysis jsonobject error");
			entity.setResult("error");
		}
		
		
		entity.setLast_update_time(sf.format(new Date()));
		return entity;
	}

	public void run() {
		Long timeLong = System.currentTimeMillis();
//		Long downloadLong = System.currentTimeMillis();
		List<MonitorEntity> monitorEntitys = new ArrayList<MonitorEntity>();
		for (String id : idsList) {
			MonitorEntity entity = getMonitorEntity(id);
			monitorEntitys.add(entity);
		}
//		System.out.println("i ="+i+" download api over "+(System.currentTimeMillis()-downloadLong)/1000+" s");
		
	  //  Connection connection = jdbcUtil.getConnection();
		//PreparedStatement ps = null;
		try {
//			connection.setAutoCommit(false);
			//ps = connection.prepareStatement(sqlString);
			for (MonitorEntity entity  : monitorEntitys) {	
			
					ps.setString(1, entity.getApi_name());
					ps.setString(2, entity.getId());
					ps.setString(3, entity.getId_type());
					ps.setString(4, entity.getRequest_url());
					ps.setInt(5, entity.getHttp_code());
					ps.setInt(6, entity.getRes_code());
					ps.setInt(7, entity.getRes_time());
					ps.setString(8, entity.getRes_msg());
					ps.setString(9, entity.getResult());
					ps.setString(10, entity.getLast_update_time());
					ps.setString(11, entity.getRequest_url());
					ps.setInt(12, entity.getHttp_code());
					ps.setInt(13, entity.getRes_code());
					ps.setInt(14, entity.getRes_time());
					ps.setString(15, entity.getRes_msg());
					ps.setString(16, entity.getResult());
					ps.setString(17, entity.getLast_update_time());
					ps.addBatch();
			}
			ps.executeBatch();
//			connection.commit(); 
		} catch (Exception e) {
			log.error("insert or update database error", e);
		} finally {
			//jdbcUtil.closePreparedStatement(ps);//由主线程控制关闭
			//jdbcUtil.closeConnection(connection);//由主线程控制关闭
		}
		System.out.println("i ="+i+" insertOrUpdate database over "+(System.currentTimeMillis()-timeLong)/1000+" s");
		countdown.countDown();
	}
}

 

 

测试结果是入库的时间快了一倍,每一个线程的入库时间基本相同。(如果用多线程一部分线程回慢,因为要等需要的connection)。。

求解释。。。 多线程下公用一个PreparedStatement ,是否回引发线程安全的问题。理论上上面应该会引发才对。但运行结果。。却没有。。。 还有一个隐形问题。。 多线程公用一个PreparedStatement  ,在赋值时是否会有问题。比如:A线程吧值赋到B线程去了。 反正异常是没有报。

 

主线程代码如下:

package com.morningstar.api;

import java.sql.Connection;
import java.sql.PreparedStatement;
import java.util.List;
import java.util.Map;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.TimeUnit;

import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;

import com.morningstar.api.ApiConfig;
import com.morningstar.api.util.JDBCParameterCallBack;
import com.morningstar.api.util.JDBCUtil;

public class Application {
	static Log log = LogFactory.getLog(Application.class);
	private static String sqlString = "(api_name,id,id_type,request_url,http_code,state_code,response_time,state_msg,result,last_update) "
			+ "value(?,?,?,?,?,?,?,?,?,?)"
			+ " on duplicate key update request_url=?,http_code=?,state_code=?,response_time=?,"
			+ "state_msg=?,result=?,last_update=?";


	public static void main(String[] args) throws Exception {
		ApiConfig apiConfig = new ApiConfig();

		
		String url_param = "?username=" + apiConfig.getUsername()
				+ "&password=" + apiConfig.getPassword() + "&format=json";
		//获取ids集合
		Map<String, List<String>> idTypeMap = apiConfig.getIdsMap();
		//任务数
		int taskCount = apiConfig.getTask_count();

		JDBCUtil jdbcUtil = new JDBCUtil(apiConfig.getJdbc_url(),
				apiConfig.getJdbc_name(), apiConfig.getJdbc_pass(),
				apiConfig.getJdbc_driver(), apiConfig.getMax_conn(),
				apiConfig.getMin_conn());
		Connection con = null;
		//API名字集合
		List<String> apiNames = apiConfig.getApi_names();
		//获取创建表sql
		String createTableSql = apiConfig.getCreateTableSql();
		Connection connection = jdbcUtil.getConnection();
		for (String apiname : apiNames) {
			String url_head = apiConfig.getApi_url() + apiname + "/";
			
			jdbcUtil.insertOrUpdate("CREATE TABLE IF NOT EXISTS t_monitor_"+apiname+createTableSql);
			String sql = "insert t_monitor_"+apiname+sqlString;
			PreparedStatement ps = null;
			ps = connection.prepareStatement(sql);
			for (String idtype : idTypeMap.keySet()) {
				Long starttimeLong = System.currentTimeMillis();
				List<String> idsList = idTypeMap.get(idtype);
				log.info("start download  "+apiname+" " + idtype + " list size:"
						+ idsList.size());
				int number = idsList.size() / taskCount;
				int rem = idsList.size() % taskCount;
				if (rem != 0) {
					taskCount++;
				}
				CountDownLatch countdown = new CountDownLatch(taskCount);
				for (int i = 0; i < taskCount; i++) {

					int startIndex = i * number;
					int endIndex = (i + 1) * number > idsList.size() ? idsList
							.size() : (i + 1) * number;
					List<String> ids = idsList.subList(startIndex, endIndex);
					
					DownloadTask downloadTask = new DownloadTask(jdbcUtil, con,
							ids, url_head + idtype + "/", url_param, sql,
							apiname, idtype, countdown, i,ps);
					Thread thread = new Thread(downloadTask);
					thread.start();
				}
				try {
					long timeout = 60;
					countdown.await(timeout, TimeUnit.MINUTES);
				} catch (InterruptedException e) {
					log.error("thread is interrupted", e);
				}
				Long endtimeLong = System.currentTimeMillis();
				log.info("end download  "+apiname+" "  + idtype + " list size:" + idsList.size()
						+ " over time:" + (endtimeLong - starttimeLong) / 1000
						+ "s");

			}
			jdbcUtil.closePreparedStatement(ps);
		}
		jdbcUtil.closeConnection(connection);
	}
}