Spark-Job-Server-Client

Backgroud

People always use curl or HUE to upload jar and run spark job in Spark Job Server. But the Spark Job Server official only presents the rest apis to upload job jar and run a job, doesn't give client lib with any language implementation.

Now there is another option to communicate with spark job server in Java, that is Spark-Job-Server-Client, the Java Client of the Spark Job Server implementing the arranged Rest APIs.

Spark-Job-Server-Client is a open-source program of com.bluebreezecf under Apache License v2. It aims to make the java applications use the spark more easily.

How to compile

You can execute the following commands to compile this client:

git clone https://github.com/bluebreezecf/SparkJobServerClient.git
cd SparkJobServerClient
mvn clean package

Then you can findspark-job-server-client-1.0.0.jarin SparkJobServerClient/target, it is the main jar of spark-job-server-client. Besides, spark-job-server-client-1.0.0-sources.jaris the java source jar, and spark-job-server-client-1.0.0-javadoc.jar is the java doc api jar.

How to set dependency

Except the DIY way of getting the latest codes and compiling it, now the real and original spark-job-server has been published to http://repo1.maven.org/maven2/. So everyone, who has set the url as one of the remote maven repository, can download or get it via mvn package after setting the following dependency into your pom.xml.

<dependency>
    <groupId>com.bluebreezecf</groupId>
    <artifactId>spark-job-server-client</artifactId>
    <version>1.0.0</version>
</dependency>

One can find all the releases of spark-job-server-client via search.maven.org.

How to use

The following sample codes shows how to use spark-job-server-client:

import java.io.File;
import java.util.HashMap;
import java.util.List;
import java.util.Map;

import com.bluebreezecf.tools.sparkjobserver.api.ISparkJobServerClient;
import com.bluebreezecf.tools.sparkjobserver.api.ISparkJobServerClientConstants;
import com.bluebreezecf.tools.sparkjobserver.api.SparkJobConfig;
import com.bluebreezecf.tools.sparkjobserver.api.SparkJobInfo;
import com.bluebreezecf.tools.sparkjobserver.api.SparkJobJarInfo;
import com.bluebreezecf.tools.sparkjobserver.api.SparkJobResult;
import com.bluebreezecf.tools.sparkjobserver.api.SparkJobServerClientException;
import com.bluebreezecf.tools.sparkjobserver.api.SparkJobServerClientFactory;

/**
 * A sample shows how to use spark-job-server-client.
 * 
 * @author bluebreezecf
 * @since 2014-09-16
 *
 */
public class SparkJobServerClientTest {
	
	public static void main(String[] args) {
		ISparkJobServerClient client = null;
		try {
			client = SparkJobServerClientFactory.getInstance().createSparkJobServerClient("http://localhost:8090/");
			//GET /jars
			List<SparkJobJarInfo> jarInfos = client.getJars();
			for (SparkJobJarInfo jarInfo: jarInfos) {
				System.out.println(jarInfo.toString());
			}

			//POST /jars/<appName>
			client.uploadSparkJobJar(new File("d:\\spark-examples_2.10-1.0.2.jar"), "spark-test");
			
			//GET /contexts
			List<String> contexts = client.getContexts();
			System.out.println("Current contexts:");
			for (String cxt: contexts) {
				System.out.println(cxt);
			}
			
			//POST /contexts/<name>--Create context with name ctxTest and null parameter
			client.createContext("ctxTest", null);
			//POST /contexts/<name>--Create context with parameters
			Map<String, String> params = new HashMap<String, String>();
			params.put(ISparkJobServerClientConstants.PARAM_MEM_PER_NODE, "512m");
			params.put(ISparkJobServerClientConstants.PARAM_NUM_CPU_CORES, "10");
			client.createContext("cxtTest2", params);
			
			//DELETE /contexts/<name>
			client.deleteContext("ctxTest");
			
			//GET /jobs
			List<SparkJobInfo> jobInfos = client.getJobs();
			System.out.println("Current jobs:");
			for (SparkJobInfo jobInfo: jobInfos) {
				System.out.println(jobInfo);
			}
			
			//Post /jobs---Create a new job
			params.put(ISparkJobServerClientConstants.PARAM_APP_NAME, "spark-test");
			params.put(ISparkJobServerClientConstants.PARAM_CLASS_PATH, "spark.jobserver.WordCountExample");
			//1.start a spark job asynchronously and just get the status information
			SparkJobResult result = client.startJob("input.string= fdsafd dfsf blullkfdsoflaw fsdfs", params);
			System.out.println(result);
			
			//2.start a spark job synchronously and wait until the result
			params.put(ISparkJobServerClientConstants.PARAM_CONTEXT, "cxtTest2");
			params.put(ISparkJobServerClientConstants.PARAM_SYNC, "true");
			result = client.startJob("input.string= fdsafd dfsf blullkfdsoflaw fsdffdsfsfs", params);
			System.out.println(result);
			
			//GET /jobs/<jobId>---Gets the result or status of a specific job
			result = client.getJobResult("fdsfsfdfwfef");
			System.out.println(result);
			
			//GET /jobs/<jobId>/config - Gets the job configuration
			SparkJobConfig jobConfig = client.getConfig("fdsfsfdfwfef");
			System.out.println(jobConfig);
		} catch (SparkJobServerClientException e1) {
			e1.printStackTrace();
		} catch (Exception e) {
			e.printStackTrace();
		}
	}
}

How to contribute

Anyone interested in this program can do the following things:

1. Fork it to your own git repository.
2. Create a new branch for your feature via git checkout -b your-new-feature.
3. Add or modify new codes.
4. Commit the modifications through git commit -am 'add your new feature'.
5. Push the new branch by git push origin your-new-feature.
6. Create a new pull request.

Any questions and discussions can be added in [SparkJobServerClient/issues] (https://github.com/bluebreezecf/SparkJobServerClient/issues)