Introduction to Apache Spark and Zeppelin on Google Cloud Dataproc

Want to learn more about using Apache Spark and Zeppelin on Dataproc via the Google Cloud Platform? You’ve come to the right place.

Cloud Dataproc is Google’s answer to Amazon EMR (Elastic MapReduce). Like EMR, Cloud Dataproc provisions and manage Compute Engine-based Apache Hadoop and Spark data processing clusters.
If you are not familiar with Amazon EMR, check out my two-part series about using Apache Spark and Zeppelin on EMR – See Part 1 and Part 2. You may find my notes on IAM helpful, too.

First Impressions: The Key Differences between EMR and Cloud Dataproc

Cloud Dataproc is fairly new. It was first released in beta last September, and is now generally available since February this year. If you have previously used EMR, you may find Cloud Dataproc familiar.
An important difference I have observed is this: In EMR, when you create a cluster, you know exactly what you are installing because you are presented with an option to choose from a list of supported Hadoop components. Cloud Dataproc, on the other hand, will just install all the supported components by default.
EMR is a more mature platform. After all, it has been around since 2009. It has support for many applications, including Tez, Ganglia, Presto, HBase, Pig, Hive, Mahout, Sqoop, and Zeppelin. As for Cloud Dataproc, it only supports Hadoop, Spark, Hive, and Pig (see the supported Cloud Dataproc versions page). Fortunately, you can specify initialization actions when creating a Cloud Dataproc cluster so that you can install the additional software you need.

Installing Zeppelin on Cloud Dataproc

We will go through the steps to do exactly that when we set up Zeppelin with Spark on Cloud Dataproc. Why Zeppelin? It’s an innovative web-based notebook that enables interactive data analytics.
With Zeppelin, you can create data-driven documents based on a variety of different backends, including Hadoop. It’s a great starting project, so let’s jump right into it.

Our Assumptions

Creating a Cloud Dataproc Cluster

Google provided a collection of initialization actions that we can use to install additional (but unsupported) Hadoop components when we create a cluster. For this example, we will use the Zeppelin initialization action.
To use an initialization action, we need to access the initialization action script in a Cloud Storage bucket.
We will not use the publicly-accessible gs://dataproc-initialization-actions Cloud Storage bucket as instructed in the README. At the time of writing, the version of the Zeppelin initialization action script is outdated. If we were to create a cluster with it, we would encounter errors. Let’s upload the script to our own Cloud Storage bucket instead.

$ git clone https://github.com/GoogleCloudPlatform/dataproc-initialization-actions.git
Cloning into 'dataproc-initialization-actions'...
remote: Counting objects: 267, done.
remote: Total 267 (delta 0), reused 0 (delta 0), pack-reused 266
Receiving objects: 100% (267/267), 89.24 KiB | 63.00 KiB/s, done.
Resolving deltas: 100% (88/88), done.
Checking connectivity... done.
$ cd dataproc-initialization-actions/apache-zeppelin/
$ gsutil mb gs://cloudacademy/
Creating gs://cloudacademy/...
$ gsutil cp zeppelin.sh gs://cloudacademy/
Copying file://zeppelin.sh [Content-Type=application/x-sh]...
Uploading   gs://cloudacademy/zeppelin.sh:                       4.47 KiB/4.47 KiB

Next, we will issue the gcloud command to set up a Cloud Dataproc cluster.

$ gcloud dataproc clusters create spark-zeppelin \
> --initialization-actions gs://cloudacademy/zeppelin.sh \
> --initialization-action-timeout 15m
Waiting on operation [projects/operating-spot-133003/regions/global/operations/cdf1fadd-032d-4261-9520-c2f55f8c46fa].
Waiting for cluster creation operation...done.
Created [https://dataproc.googleapis.com/v1/projects/operating-spot-133003/regions/global/clusters/spark-zeppelin].
clusterName: spark-zeppelin
clusterUuid: f05a9f22-5ee6-48c8-83d3-7079e2d1d834
config:
  configBucket: dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia
  gceClusterConfig:
    networkUri: https://www.googleapis.com/compute/v1/projects/operating-spot-133003/global/networks/default
    serviceAccountScopes:
    - https://www.googleapis.com/auth/bigquery
    - https://www.googleapis.com/auth/bigtable.admin.table
    - https://www.googleapis.com/auth/bigtable.data
    - https://www.googleapis.com/auth/cloud.useraccounts.readonly
    - https://www.googleapis.com/auth/devstorage.full_control
    - https://www.googleapis.com/auth/devstorage.read_write
    - https://www.googleapis.com/auth/logging.write
    zoneUri: https://www.googleapis.com/compute/v1/projects/operating-spot-133003/zones/asia-east1-a
  initializationActions:
  - executableFile: gs://cloudacademy/zeppelin.sh
    executionTimeout: 900.000s
  masterConfig:
    diskConfig:
      bootDiskSizeGb: 500
    imageUri: https://www.googleapis.com/compute/v1/projects/cloud-dataproc/global/images/dataproc-1-0-20160516-190717
    instanceNames:
    - spark-zeppelin-m
    machineTypeUri: https://www.googleapis.com/compute/v1/projects/operating-spot-133003/zones/asia-east1-a/machineTypes/n1-standard-4
    numInstances: 1
  softwareConfig:
    imageVersion: '1.0'
    properties:
      distcp:mapreduce.map.java.opts: -Xmx2457m
      distcp:mapreduce.map.memory.mb: '3072'
      distcp:mapreduce.reduce.java.opts: -Xmx4915m
      distcp:mapreduce.reduce.memory.mb: '6144'
      mapred:mapreduce.map.cpu.vcores: '1'
      mapred:mapreduce.map.java.opts: -Xmx2457m
      mapred:mapreduce.map.memory.mb: '3072'
      mapred:mapreduce.reduce.cpu.vcores: '2'
      mapred:mapreduce.reduce.java.opts: -Xmx4915m
      mapred:mapreduce.reduce.memory.mb: '6144'
      mapred:yarn.app.mapreduce.am.command-opts: -Xmx4915m
      mapred:yarn.app.mapreduce.am.resource.cpu-vcores: '2'
      mapred:yarn.app.mapreduce.am.resource.mb: '6144'
      spark:spark.driver.maxResultSize: 1920m
      spark:spark.driver.memory: 3840m
      spark:spark.executor.cores: '2'
      spark:spark.executor.memory: 5586m
      spark:spark.yarn.am.memory: 5586m
      spark:spark.yarn.am.memoryOverhead: '558'
      spark:spark.yarn.executor.memoryOverhead: '558'
      yarn:yarn.nodemanager.resource.memory-mb: '12288'
      yarn:yarn.scheduler.maximum-allocation-mb: '12288'
      yarn:yarn.scheduler.minimum-allocation-mb: '1024'
  workerConfig:
    diskConfig:
      bootDiskSizeGb: 500
    imageUri: https://www.googleapis.com/compute/v1/projects/cloud-dataproc/global/images/dataproc-1-0-20160516-190717
    instanceNames:
    - spark-zeppelin-w-0
    - spark-zeppelin-w-1
    machineTypeUri: https://www.googleapis.com/compute/v1/projects/operating-spot-133003/zones/asia-east1-a/machineTypes/n1-standard-4
    numInstances: 2
projectId: operating-spot-133003
status:
  state: RUNNING
  stateStartTime: '2016-06-07T09:35:33.783Z'
statusHistory:
- state: CREATING
  stateStartTime: '2016-06-07T09:32:56.398Z'

By default, Cloud Dataproc clusters use n1-standard-4 machine type for the master and worker nodes. These are the standard instances with 4 virtual CPUs and 15GB of memory. You can change the defaults by specifying the relevant flags. See the gcloud dataproc clusters create documentation.

$ gcloud compute instances list
NAME                ZONE          MACHINE_TYPE   PREEMPTIBLE  INTERNAL_IP  EXTERNAL_IP      STATUS
spark-zeppelin-m    asia-east1-a  n1-standard-4               10.140.0.2   104.199.169.41   RUNNING
spark-zeppelin-w-0  asia-east1-a  n1-standard-4               10.140.0.4   104.199.160.183  RUNNING
spark-zeppelin-w-1  asia-east1-a  n1-standard-4               10.140.0.3   130.211.248.2    RUNNING

SSH to the Master Node

Now we can connect to the master node remotely. Instead of running ssh directly, we can issue the gcloud compute ssh spark-zeppelin-m command.

$ gcloud compute ssh spark-zeppelin-m
WARNING: The private SSH key file for Google Compute Engine does not exist.
WARNING: You do not have an SSH key for Google Compute Engine.
WARNING: [/usr/bin/ssh-keygen] will be executed to generate a key.
This tool needs to create the directory [/Users/eugeneteo/.ssh] before being
able to generate SSH keys.
Do you want to continue (Y/n)?
Generating public/private rsa key pair.
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /Users/eugeneteo/.ssh/google_compute_engine.
Your public key has been saved in /Users/eugeneteo/.ssh/google_compute_engine.pub.
The key fingerprint is:
SHA256:xTjbTqWIfroXr670RWQkscOPfGdY6rzpi14WCoFI5Ko eugeneteo@eugeneteos-MacBook-Pro.local
The key's randomart image is:
+---[RSA 2048]----+
| .o     o..      |
| o . . . *       |
|  o . . * = o    |
| .     + % *     |
|.     o S @ o    |
|.    . ..O +     |
|E     o ooB      |
|     . +.=.o     |
|      +**o=.     |
+----[SHA256]-----+
Updated [https://www.googleapis.com/compute/v1/projects/operating-spot-133003].
Warning: Permanently added 'compute.5792161037633137215' (ECDSA) to the list of known hosts.
The programs included with the Debian GNU/Linux system are free software;
the exact distribution terms for each program are described in the
individual files in /usr/share/doc/*/copyright.
Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent
permitted by applicable law.
eugeneteo@spark-zeppelin-m:~$

Spark’s Scala Shell

We will not cover the Spark programming model in this article, but we will learn just enough to start an interpreter on the command-line and to make sure it works. We will launch spark-shell on YARN.

$ spark-shell --master=yarn
16/06/07 09:53:32 INFO org.spark-project.jetty.server.Server: jetty-8.y.z-SNAPSHOT
16/06/07 09:53:32 INFO org.spark-project.jetty.server.AbstractConnector: Started SocketConnector@0.0.0.0:36325
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /___/ .__/\_,_/_/ /_/\_\   version 1.6.1
      /_/
Using Scala version 2.10.5 (OpenJDK 64-Bit Server VM, Java 1.8.0_72-internal)
Type in expressions to have them evaluated.
Type :help for more information.
16/06/07 09:53:38 INFO akka.event.slf4j.Slf4jLogger: Slf4jLogger started
16/06/07 09:53:38 INFO Remoting: Starting remoting
16/06/07 09:53:38 INFO Remoting: Remoting started; listening on addresses :[akka.tcp://sparkDriverActorSystem@10.140.0.2:40692]
16/06/07 09:53:39 INFO org.spark-project.jetty.server.Server: jetty-8.y.z-SNAPSHOT
16/06/07 09:53:39 INFO org.spark-project.jetty.server.AbstractConnector: Started SelectChannelConnector@0.0.0.0:4040
16/06/07 09:53:39 INFO org.apache.hadoop.yarn.client.RMProxy: Connecting to ResourceManager at spark-zeppelin-m/10.140.0.2:8032
16/06/07 09:53:43 INFO org.apache.hadoop.yarn.client.api.impl.YarnClientImpl: Submitted application application_1465292021581_0001
[...]
scala> val data = sc.parallelize(1 to 100000)
data: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[0] at parallelize at <console>:27
scala> data.filter(_ < 100).collect()
res0: Array[Int] = Array(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99)

It worked!

Access the Zeppelin Notebook

Before we can access the Zeppelin Notebook, we will need to create a SSH tunnel to the master node.

$ gcloud compute ssh --ssh-flag="-D 31337" --ssh-flag="-N" --ssh-flag="-n" spark-zeppelin-m

Configure your web browser to use the SOCKS proxy localhost:31337.
Having done that, we can now access http://localhost:8080/.
zeppelin1
zeppelin2
Great! The notebook works, too!

Terminating the Cloud Dataproc Cluster

Always remember to terminate your cluster after you have completed your work to avoid spending more money than you have planned. We are billed by the minute, based on the size of our cluster and the duration we ran our jobs.

$ gcloud dataproc clusters delete spark-zeppelin
The cluster 'spark-zeppelin' and all attached disks will be deleted.
Do you want to continue (Y/n)?  y
Waiting on operation [projects/operating-spot-133003/regions/global/operations/e2b88d41-e215-480e-a970-e6b49c0de574].
Waiting for cluster deletion operation...done.
Deleted [https://dataproc.googleapis.com/v1/projects/operating-spot-133003/regions/global/clusters/spark-zeppelin].
$ gsutil -m rm -r gs://cloudacademy/ gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/
Removing gs://cloudacademy/zeppelin.sh#1465291266953000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/cluster.properties#1465291982908000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-m/dataproc-initialization-script-0_output#1465292128609000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-m/dataproc-startup-script_SUCCESS#1465292128628000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-m/dataproc-startup-script_output#1465292038813000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-w-0/dataproc-initialization-script-0_output#1465292016529000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-w-0/dataproc-startup-script_SUCCESS#1465292016563000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-w-0/dataproc-startup-script_output#1465292015001000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-w-1/dataproc-initialization-script-0_output#1465292016544000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-w-1/dataproc-startup-script_SUCCESS#1465292016575000...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/google-cloud-dataproc-metainfo/f05a9f22-5ee6-48c8-83d3-7079e2d1d834/spark-zeppelin-w-1/dataproc-startup-script_output#1465292014612000...
Removing gs://cloudacademy/...
Removing gs://dataproc-6dfb6ea8-847d-438c-befd-2883003f61e5-asia/...

What We’ve Learned

In this article, we have learned:

  1. How to set up a Cloud Dataproc cluster with Zeppelin.
  2. We have launched spark-shell on YARN.
  3. We have also set up a SSH tunnel to access the Zeppelin Notebook from the master node.

Related Courses

Data Management on Google Cloud Platform might be a place to start. It’s 26 minutes of data management goodness from David Clinton, an expert Linux Sysadmin.

For a guided instructional experience, check out a Cloud Academy Learning Path. We offer a free 7-day trial subscription with access all video courses, self-test quizzes, and labs. Our labs are a great tool for applying what you learned and testing your understanding in a live environment. Get started today!

Avatar

Written by

Eugene Teo

Eugene Teo is a director of security at a US-based technology company. He is interested in applying machine learning techniques to solve problems in the security domain.

Related Posts

Avatar
Stuart Scott
— June 20, 2019

Working with AWS Networking & Amazon VPC

Being able to architect your own isolated segment of AWS is a simple process using VPCs; understanding how to architect its related networking components and connectivity architecture is key to making it a powerful service.Many services within Amazon Web Services (AWS) require you t...

Read more
  • AWS
  • VPC
Avatar
Stuart Scott
— June 19, 2019

AWS Compute Fundamentals Update

AWS is renowned for the rate at which it reinvents, revolutionizes, and meets customer demands and expectations through its continuous cycle of feature and service updates. With hundreds of updates a month, it can be difficult to stay on top of all the changes made available.  Here ...

Read more
  • AWS
Jeff Hyatt
Jeff Hyatt
— June 18, 2019

10 Steps for an Effective Reserved Instances Strategy

Amazon Web Services (AWS) offers three different ways to pay for EC2 Instances: On-Demand, Reserved Instances, and Spot Instances. This article will focus on effective strategies for purchasing Reserved Instances. While most of the major cloud platforms offer pre-pay and reservation dis...

Read more
  • AWS
  • EC2
Joe Nemer
Joe Nemer
— June 18, 2019

AWS Certification Practice Exam: What to Expect from Test Questions

If you’re building applications on the AWS cloud or looking to get started in cloud computing, certification is a way to build deep knowledge in key services unique to the AWS platform. AWS currently offers 11 certifications that cover major cloud roles including Solutions Architect, De...

Read more
  • AWS
  • AWS Certifications
Avatar
John Chell
— June 13, 2019

AWS Certified Solutions Architect Associate: A Study Guide

The AWS Solutions Architect - Associate Certification (or Sol Arch Associate for short) offers some clear benefits: Increases marketability to employers Provides solid credentials in a growing industry (with projected growth of as much as 70 percent in five years) Market anal...

Read more
  • AWS
  • AWS Certifications
Chris Gambino and Joe Niemiec
Chris Gambino and Joe Niemiec
— June 11, 2019

Moving Data to S3 with Apache NiFi

Moving data to the cloud is one of the cornerstones of any cloud migration. Apache NiFi is an open source tool that enables you to easily move and process data using a graphical user interface (GUI).  In this blog post, we will examine a simple way to move data to the cloud using NiFi c...

Read more
  • AWS
  • S3
Avatar
Chandan Patra
— June 11, 2019

Amazon DynamoDB: 10 Things You Should Know

Amazon DynamoDB is a managed NoSQL service with strong consistency and predictable performance that shields users from the complexities of manual setup.Whether or not you've actually used a NoSQL data store yourself, it's probably a good idea to make sure you fully understand the key ...

Read more
  • AWS
  • DynamoDB
Avatar
Andrew Larkin
— June 6, 2019

The 11 AWS Certifications: Which is Right for You and Your Team?

As companies increasingly shift workloads to the public cloud, cloud computing has moved from a nice-to-have to a core competency in the enterprise. This shift requires a new set of skills to design, deploy, and manage applications in cloud computing.As the market leader and most ma...

Read more
  • AWS
  • AWS Certifications
Sam Ghardashem
Sam Ghardashem
— May 15, 2019

Aviatrix Integration of a NextGen Firewall in AWS Transit Gateway

Learn how Aviatrix’s intelligent orchestration and control eliminates unwanted tradeoffs encountered when deploying Palo Alto Networks VM-Series Firewalls with AWS Transit Gateway.Deploying any next generation firewall in a public cloud environment is challenging, not because of the f...

Read more
  • AWS
Joe Nemer
Joe Nemer
— May 3, 2019

AWS Config Best Practices for Compliance

Use AWS Config the Right Way for Successful ComplianceIt’s well-known that AWS Config is a powerful service for monitoring all changes across your resources. As AWS Config has constantly evolved and improved over the years, it has transformed into a true powerhouse for monitoring your...

Read more
  • AWS
  • Compliance
Avatar
Francesca Vigliani
— April 30, 2019

Cloud Academy is Coming to the AWS Summits in Atlanta, London, and Chicago

Cloud Academy is a proud sponsor of the 2019 AWS Summits in Atlanta, London, and Chicago. We hope you plan to attend these free events that bring the cloud computing community together to connect, collaborate, and learn about AWS. These events are all about learning. You can learn how t...

Read more
  • AWS
  • AWS Summits
Paul Hortop
Paul Hortop
— April 2, 2019

How to Monitor Your AWS Infrastructure

The AWS cloud platform has made it easier than ever to be flexible, efficient, and cost-effective. However, monitoring your AWS infrastructure is the key to getting all of these benefits. Realizing these benefits requires that you follow AWS best practices which constantly change as AWS...

Read more
  • AWS
  • Monitoring