#amazonemrcluster

Amazon EMR Notebooks

EMR Notebooks: AWS Simplifies Spark Cluster Data Analysis

Amazon Web Services (AWS) makes big data management more flexible and integrated for data scientists and analysts. Amazon EMR Notebooks offer a familiar interactive interface that connects Apache Spark-powered Amazon EMR clusters. The new feature streamlines data searches, model creation, and result visualisation.

Amazon EMR users can access EMR Notebooks as EMR Studio Workspaces. The console interface's “Create Workspace” button simplifies notebook creation. Users need extra IAM role permissions to create or access these Workspaces.

EMR notebooks are “serverless” interfaces. The equations, queries, models, code, and narrative text you write are client-side in the notebook interface, while a kernel on the Amazon EMR cluster executes your commands. This configuration directly uses your EMR system's scalable computing capability for interactive analysis sessions.

Designing to protect your valuable work from computing cluster transience is crucial. EMR notebook contents are automatically stored on Amazon S3. Your notes, code, and analysis are separated from the cluster's data, allowing flexible notebook reuse and durability (your work continues even if the cluster is shut down).

The flexibility of laptop cluster connections is a major benefit. Users can establish an EMR cluster, connect their notebook for analysis, then shutdown the cluster when they're done for cost-effective, on-demand computing. Closing a notebook connected to one cluster to another lets you shift environments or work with data on another cluster fast.

Multiple users can connect their notebooks to the same EMR cluster at once, and notebook files are hosted on Amazon S3, making sharing easy. It is stated that these features will reduce notebook reset time for diverse datasets and clusters.

Interactive console or programmatic use of EMR Notebooks. Headless execution lets users run an EMR notebook over the Amazon EMR API without using the UI. This involves marking a cell in the EMR notebook with “parameters” to enable. When an external script is launched programmatically, this cell acts as a gateway to feed the notebook new data.

This is useful when creating parameterised notebooks that can be reused with different input values without requiring extra copies. Every time an API-executed parameterised notebook is launched, Amazon EMR generates and stores the output notebook on S3. This functionality can be developed using example API instructions.

EMR Notebooks support 5.18.0 and newer clusters. AWS recommends EMR Notebooks with Amazon EMR 5.30.0, 5.32.0, or 6.2.0 clusters for optimum performance. In these latter versions, the Jupyter kernels that run your code run directly on the cluster, thus this guidance is crucial. Direct cluster execution is said to boost performance and kernel and library modification.

Customers considering Amazon EMR Notebooks should consider the costs. As expected, Amazon S3 storage for notebook data will cost. Standard fees will also apply to connected Amazon EMR clusters utilised for notebook instructions.

Minutes to Scalable EMR Cluster on AWS

AWS EMR cluster

Spark helps you easily build up an Amazon EMR cluster to process and analyse data. This page covers Plan and Configure, Manage, and Clean Up.

This detailed guide to cluster setup:

Amazon EMR Cluster Configuration

Spark is used to launch an example cluster and run a PySpark script in the course. You must complete the “Before you set up Amazon EMR” exercises before starting.

While functioning live, the sample cluster will incur small per-second charges under Amazon EMR pricing, which varies per location. To avoid further expenses, complete the tutorial’s final cleaning steps.

The setup procedure has numerous steps:

Amazon EMR Cluster and Data Resources Configuration

This initial stage prepares your application and input data, creates your data storage location, and starts the cluster.

Setting Up Amazon EMR Storage:

Amazon EMR supports several file systems, but this article uses EMRFS to store data in an S3 bucket. EMRFS reads and writes to Amazon S3 in Hadoop.

This lesson requires a specific S3 bucket. Follow the Amazon Simple Storage Service Console User Guide to create a bucket.

You must create the bucket in the same AWS region as your Amazon EMR cluster launch. Consider US West (Oregon) us-west-2.

Amazon EMR bucket and folder names are limited. Lowercase letters, numerals, periods (.), and hyphens (-) can be used, but bucket names cannot end in numbers and must be unique across AWS accounts.

The bucket output folder must be empty.

Small Amazon S3 files may incur modest costs, but if you’re within the AWS Free Tier consumption limitations, they may be free.

Create an Amazon EMR app using input data:

Standard preparation involves uploading an application and its input data to Amazon S3. Submit work with S3 locations.

The PySpark script examines 2006–2020 King County, Washington food business inspection data to identify the top ten restaurants with the most “Red” infractions. Sample rows of the dataset are presented.

Create a new file called health_violations.py and copy the source code to prepare the PySpark script. Next, add this file to your new S3 bucket. Uploading instructions are in Amazon Simple Storage Service’s Getting Started Guide.

Download and unzip the food_establishment_data.zip file, save the CSV file to your computer as food_establishment_data.csv, then upload it to the same S3 bucket to create the example input data. Again, see the Amazon Simple Storage Service Getting Started Guide for uploading instructions.

“Prepare input data for processing with Amazon EMR” explains EMR data configuration.

Create an Amazon EMR Cluster:

Apache Spark and the latest Amazon EMR release allow you to launch the example cluster after setting up storage and your application. This may be done with the AWS Management Console or CLI.

Console Launch:

Launch Amazon EMR after login into AWS Management Console.

Start with “EMR on EC2” > “Clusters” > “Create cluster”. Note the default options for “Release,” “Instance type,” “Number of instances,” and “Permissions”.

Enter a unique “Cluster name” without <, >, $, |, or `. Install Spark from “Applications” by selecting “Spark”. Note: Applications must be chosen before launching the cluster. Check “Cluster logs” to publish cluster-specific logs to Amazon S3. The default destination is s3://amzn-s3-demo-bucket/logs. Replace with S3 bucket. A new ‘logs’ subfolder is created for log files.

Select your two EC2 keys under “Security configuration and permissions”. For the instance profile, choose “EMR_DefaultRole” for Service and “EMR_EC2_DefaultRole” for IAM.

Choose “Create cluster”.

The cluster information page appears. As the EMR fills the cluster, its “Status” changes from “Starting” to “Running” to “Waiting”. Console view may require refreshing. Status switches to “Waiting” when cluster is ready to work.

AWS CLI’s aws emr create-default-roles command generates IAM default roles.

Create a Spark cluster with aws emr create-cluster. Name your EC2 key pair –name, set –instance-type, –instance-count, and –use-default-roles. The sample command’s Linux line continuation characters () may need Windows modifications.

Output will include ClusterId and ClusterArn. Remember your ClusterId for later.

Check your cluster status using aws emr describe-cluster –cluster-id myClusterId>.

The result shows the Status object with State. As EMR deployed the cluster, the State changed from STARTING to RUNNING to WAITING. When ready, operational, and up, the cluster becomes WAITING.

Open SSH Connections

Before connecting to your operating cluster via SSH, update your cluster security groups to enable incoming connections. Amazon EC2 security groups are virtual firewalls. At cluster startup, EMR created default security groups: ElasticMapReduce-slave for core and task nodes and ElasticMapReduce-master for main.

Console-based SSH authorisation:

Authorisation is needed to manage cluster VPC security groups.

Launch Amazon EMR after login into AWS Management Console.

Select the updateable cluster under “Clusters”. The “Properties” tab must be selected.

Choose “Networking” and “EC2 security groups (firewall)” from the “Properties” tab. Select the security group link under “Primary node”.

EC2 console is open. Select “Edit inbound rules” after choosing “Inbound rules”.

Find and delete any public access inbound rule (Type: SSH, Port: 22, Source: Custom 0.0.0.0/0). Warning: The ElasticMapReduce-master group’s pre-configured rule that allowed public access and limited traffic to reputable sources should be removed.

Scroll down and click “Add Rule”.

Choose “SSH” for “Type” to set Port Range to 22 and Protocol to TCP.

Enter “My IP” for “Source” or a range of “Custom” trustworthy client IP addresses. Remember that dynamic IPs may need updating. Select “Save.”

When you return to the EMR console, choose “Core and task nodes” and repeat these steps to provide SSH access to those nodes.

Connecting with AWS CLI:

SSH connections may be made using the AWS CLI on any operating system.

Use the command: AWS emr ssh –cluster-id –key-pair-file <~/mykeypair.key>. Replace with your ClusterId and the full path to your key pair file.

After connecting, visit /mnt/var/log/spark to examine master node Spark logs.