ElasticSearch 应用开发（四）Java Client 连接ElasticSearch集群方法的对比与总结

ElasticSearch 应用开发（二）Java Client 连接ElasticSearch集群

ElasticSearch 应用开发（三）基于commons-pool2和TransportClient的ElasticSearch连接池

上述内容，介绍java client连接Elasticsearch的方法，简单做一个对比和总结

---

1.对比

1.1单线程测试获取连接地址

（a）TransportClient

/**
 * 单线程 创建一个简单的TransportClient连接
 */
@Test
public void testTransportClient() {
	for (int i = 0; i < 10; i++) {
		Client client = TransportClientBuild.getClient();
		System.out.println(("连接对象————地址:" + client));

	}
}

结果如下： 

 （b）Rest Client

@Test
public void testLowLevelClient() {
	for (int i = 0; i < 10; i++) {
		System.out.println(LowLevelClient.getClient().toString());
	}
}

@Test
public void testHighLevelClient() {
	try {
		for (int i = 0; i < 10; i++) {
			System.out.println(HighLevelClient.getClient().toString());
		}
	} catch (Exception e) {
		e.printStackTrace();
	}
}

结果如下： 

（c）分析

简单for循环获取连接地址，经过简单对比，可知：

1.每次获取TransportClient地址是同一个，针对其做池化有一定的意义，高并发环境下，提高系统稳定性；

2.每次获取Rest Low Level Client和Rest High Level Client地址不同，针对其做做池化技术，意义并不是很大； 

1.2多线程条件下 ，测试获取连接地址

（a）TransportClient

public void testTransprotClient() {

	//System.out.println(TransportClientBuild.getClient());
	System.setProperty("es.set.netty.runtime.available.processors", "false");

	ExecutorService executorService = new ThreadPoolExecutor(10, 10,
			0L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue<Runnable>());

	for (int i = 0; i < 10; i++) {
		int index = i;

		if (!executorService.isShutdown()) {
			executorService.execute(new Runnable() {
				@Override
				public void run() {
					Client client = TransportClientBuild.getClient();
					System.out.println(("第" + index + "次获取到了连接对象————地址:" + client));
				}
			});
		}

	}
	executorService.shutdown();

	try {
		while (!executorService.awaitTermination(10000, TimeUnit.MILLISECONDS)) {
			System.out.println("10秒没有执行完，强制关闭线程池");
			executorService.shutdownNow();
		}
	} catch (Exception e) {
		e.printStackTrace();
	}

}

结果：

 （b）Rest Client

@Test
public void testRestClient() {

	//System.out.println(TransportClientBuild.getClient());
	System.setProperty("es.set.netty.runtime.available.processors", "false");

	ExecutorService executorService = new ThreadPoolExecutor(10, 10,
			0L, TimeUnit.MILLISECONDS, new LinkedBlockingQueue<Runnable>());

	for (int i = 0; i < 10; i++) {
		int index = i;

		if (!executorService.isShutdown()) {
			executorService.execute(new Runnable() {
				@Override
				public void run() {
					System.out.println(("第" + index + "次获取到了连接对象————地址:" + HighLevelClient.getClient().toString()));
				}
			});
		}

	}
	executorService.shutdown();

	try {
		while (!executorService.awaitTermination(10000, TimeUnit.MILLISECONDS)) {
			System.out.println("10秒没有执行完，强制关闭线程池");
			executorService.shutdownNow();
		}
	} catch (Exception e) {
		e.printStackTrace();
	}

}

结果：

（c）分析

使用多线程条件下，获取TransportClient和Rest Client 地址，每次地址都不同。由此，可以判断多线程条件下两个连接方式都能够胜任稍微复杂的业务。

1.3 多线程条件下，测试连接池

针对 ElasticSearch 应用开发（三）基于commons-pool2和TransportClient的ElasticSearch连接池构建的连接池，在多线程环境下进行测试。

代码：

/**
 * 多线程线程池测试TransportClient连接池
 */
@Test
public void testESPoolClient() {

	/*配置节点熟悉*/
	Set<TransportClientNodeConfig> nodeSets = TransportClientPoolConfig.getNodesConfig();

	AtomicReference<Set<TransportClientNodeConfig>> nodesReference = new AtomicReference<Set<TransportClientNodeConfig>>();
	nodesReference.set(nodeSets);

	/*配置线程池熟悉*/
	TransportClientPoolConfig transportClientPoolConfig = new TransportClientPoolConfig();

	final TransportClientPool pool = new TransportClientPool(nodesReference, transportClientPoolConfig);

	System.setProperty("es.set.netty.runtime.available.processors", "false");

	LinkedBlockingQueue<Runnable> linkBlockingQuene = new LinkedBlockingQueue<Runnable>();
	ExecutorService cachedThreadPool = new ThreadPoolExecutor(10, 10, 0L, TimeUnit.MILLISECONDS, linkBlockingQuene);
	for (int i = 0; i < 15; i++) {
		cachedThreadPool.execute(new Runnable() {

			@Override
			public void run() {
				TransportClient client = null;
				try {
					client = pool.getResource();
					System.out.println("当前线获取线程地址:" + client.toString());
				} catch (Exception e) {
					e.printStackTrace();
				} finally {
					pool.returnResource(client);
				}
			}
		});
	}
	cachedThreadPool.shutdown();

	try {
		while (!cachedThreadPool.awaitTermination(10000, TimeUnit.MILLISECONDS)) {
			System.out.println("10秒没有执行完，强制关闭线程池");
			cachedThreadPool.shutdownNow();
		}
	} catch (InterruptedException e) {
		e.printStackTrace();
	}
}

结果：

 分析：

多线程条件下，使用线程池，测试获取连接地址，每次地址都不一样.

2.总结

2.1针对TransportClient

若业务比较简单，连接ElasticSearch集群需求比较少，可以直接连接；

业务复杂，情况下，务必注意线程的使用，尽量在不同的线程中调用client；

当然，业务的增加，也可以使用连接池的形式，可以应对高并发的需求；

2.2针对Rest Client

连接就行了，建议Elasticsearch6.x之后，开始使用，应该能够解决绝大部分业务需求；

 

说明，以上代码都可以在github上找得到：

https://github.com/codersfarm/elasticsearch/tree/master/elasticsearch5.x

https://github.com/codersfarm/elasticsearch/tree/master/elasticsearch6.x