org.apache.spark.SparkException: Job aborted due to stage failure: Task from application

Question

I have a problem with running spark application on standalone cluster. (I use spark 1.1.0 version). I succesfully run master server by command:

bash start-master         


        

Answer 1

Found a way to run it using IDE / Maven

1. Create a Fat Jar ( One which includes all dependencies ). Use Shade Plugin for this. Example pom :

<plugin>
    <groupId>org.apache.maven.plugins</groupId>
    <artifactId>maven-shade-plugin</artifactId>
    <version>2.2</version>
    <configuration>
        <filters>
            <filter>
                <artifact>*:*</artifact>
                <excludes>
                    <exclude>META-INF/*.SF</exclude>
                    <exclude>META-INF/*.DSA</exclude>
                    <exclude>META-INF/*.RSA</exclude>
                </excludes>
            </filter>
        </filters>
    </configuration>
    <executions>
        <execution>
            <id>job-driver-jar</id>
            <phase>package</phase>
            <goals>
                <goal>shade</goal>
            </goals>
            <configuration>
                <shadedArtifactAttached>true</shadedArtifactAttached>
                <shadedClassifierName>driver</shadedClassifierName>
                <transformers>
                    <transformer implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer"/>
                    <!--
                    Some care is required:
                    http://doc.akka.io/docs/akka/snapshot/general/configuration.html
                    -->
                    <transformer implementation="org.apache.maven.plugins.shade.resource.AppendingTransformer">
                        <resource>reference.conf</resource>
                    </transformer>
                    <transformer implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer">
                        <mainClass>mainClass</mainClass>
                    </transformer>
                </transformers>
            </configuration>
        </execution>
        <execution>
            <id>worker-library-jar</id>
            <phase>package</phase>
            <goals>
                <goal>shade</goal>
            </goals>
            <configuration>
                <shadedArtifactAttached>true</shadedArtifactAttached>
                <shadedClassifierName>worker</shadedClassifierName>
                <transformers>
                    <transformer implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer"/>
                </transformers>
            </configuration>
        </execution>
    </executions>
</plugin>

2. Now we have to send the compiled jar file to the cluster. For this, specify the jar file in the spark config like this :

SparkConf conf = new SparkConf().setAppName("appName").setMaster("spark://machineName:7077").setJars(new String[] {"target/appName-1.0-SNAPSHOT-driver.jar"});

3. Run mvn clean package to create the Jar file. It will be created in your target folder.

4. Run using your IDE or using maven command :

mvn exec:java -Dexec.mainClass="className"

This does not require spark-submit. Just remember to package file before running

If you don't want to hardcode the jar path, you can do this :

1. In the config, write :

SparkConf conf = new SparkConf() .setAppName("appName") .setMaster("spark://machineName:7077") .setJars(JavaSparkContext.jarOfClass(this.getClass()));

2. Create the fat jar ( as above ) and run using maven after running package command :

java -jar target/application-1.0-SNAPSHOT-driver.jar

This will take the jar from the jar the class was loaded.

Answer 2

For the benefit of others running into this problem:

I faced an identical issue due to a mismatch between the spark connector and spark version being used. Spark was 1.3.1 and the connector was 1.3.0, an identical error message appeared:

org.apache.spark.SparkException: Job aborted due to stage failure:
  Task 2 in stage 0.0 failed 4 times, most recent failure: Lost 
  task 2.3 in stage 0.0

Updating the dependancy in SBT solved the problem.