savvydatainsights / spark

Apache Spark cluster lab.

Spark

Apache Spark cluster lab.

Setup

vagrant up

WebUIs

Node	URL
master	http://192.168.33.10:8080
slave1	http://192.168.33.20:8081
slave2	http://192.168.33.30:8081

Submitting the sample application

The sample application counts how many times each word appears in a lorem ipsum text.

In order to submit it, execute:

ansible-playbook submit-spark-application.yml

The output then can be viewed under the output folder.

Manually from the master

The application can also be submitted manually from the master host of the cluster.

First, SSH into the master: vagrant ssh master.

Once inside the master, become root: sudo su -.

After that, build the application: mvn install -f /vagrant.

Finally, submit the application:

/opt/spark/bin/spark-submit --master spark://192.168.33.10:7077 \
    --conf spark.driver.host=192.168.33.10 \
    --class uk.co.savvydatainsights.WordCount \
    /vagrant/target/spark-examples-1.0-SNAPSHOT.jar \
    /vagrant/input/lorem-ipsum.txt

Submitting your own Java application

You can also submit to the Spark cluster your own Java application, by setting the parameters repo and class, like in the example:

ansible-playbook submit-spark-application.yml \
    -e "repo=https://github.com/project/repo.git" \
    -e "class=com.domain.spark.JavaApplication"

You will be prompted to inform your repository credentials. Then, the application will be cloned, built and submitted to the Spark master instance.

Requirements:

• The project must be a Maven project;
• It is expected one input file under the src/main/resources folder.