AWS Info - gt-big-data/wiki GitHub Wiki

Log in with your credentials for GT big data here. If you don't have them, ask^^

Using EMR with MRJob

First, install MRJob pip install mrjob

Then, Create a key pair that will let you spin up ec2 instances

Next, create a config file for MRJob. Save it as ~/.mrjob.conf

    aws_access_key_id: << YOUR ACCESS KEY >>>
    aws_secret_access_key: << YOUR SECRET KEY >>
    aws_region: us-east-1
    ec2_key_pair: << The name of the ec2 key pair you created in the previous step (keypair) >>
    ec2_key_pair_file: << The absolute path + filename of the keypair on your local machine (/home/you/keypair.pem)>>
    ec2_instance_type: m1.medium
    num_ec2_instances: 10

    s3_log_uri: s3://gt-big-data-club/logs
    s3_scratch_uri: s3://gt-big-data-club/tmp/

    # Less important
    base_tmp_dir: /tmp
      TZ: America/Chicago

    base_tmp_dir: /tmp

Then, to run the job in EMR with input and output from s3, you can do this:

python <> -r emr s3://gt-big-data-club/<path/to/myInputFile.txt> --output-dir=s3://gt-big-data-club/<my-output-dir> --no-output

The official library for using MRJob is here

Uploading Data to Amazon S3

Web UI

The easiest way to upload data is from the S3 console.

Command line: S3CMD

We can use s3cmd to use s3 from the command line.

First install with sudo apt-get install s3cmd , or follow directions on their site.

Then, run s3cmd --configure and input your AWS access keys.

Here's some sample interaction

Showing all buckets

patrick@pv--hpm6:~$ s3cmd ls
2014-01-25 19:36  s3://gt-big-data
2014-02-12 18:15  s3://gt-big-data-club

List a specific bucket

patrick@pv--hpm6:~$ s3cmd ls s3://gt-big-data-club/
                       DIR   s3://gt-big-data-club/logs/
                       DIR   s3://gt-big-data-club/movie-lens-dir/
                       DIR   s3://gt-big-data-club/tmp/
                       DIR   s3://gt-big-data-club/twitter-superbowl/

Upload a file to a bucket

Note: The / at the end of your bucket name is important! Without the slash, it will upload to the parent bucket of what you intended with the filename of the bucket you intended.

patrick@pv--hpm6:~/proj/twitter-superbowl/analysis$ s3cmd put top-ngrams.json s3://gt-big-data-club/twitter-superbowl/
WARNING: Module python-magic is not available. Guessing MIME types based on file extensions.
top-ngrams.json -> s3://gt-big-data-club/twitter-superbowl/top-ngrams.json  [1 of 1]
 8302786 of 8302786   100% in    0s     9.55 MB/s  done

Download a file from a bucket

patrick@pv--hpm6:~/proj/twitter-superbowl/analysis$ s3cmd get s3://gt-big-data-club/twitter-superbowl/top-ngrams.json top-ngrams-from-s3.json
s3://gt-big-data-club/twitter-superbowl/top-ngrams.json -> top-ngrams-from-s3.json  [1 of 1]
8302786 of 8302786   100% in    1s     5.16 MB/s  done

Download from a bucket recursively

You can use this to download the output from a large MRJob

patrick@pv--hpm6:~/proj/MovieLens/mrjob$ s3cmd get --recursive s3://gt-big-data-club/movie-lens-dir/out
s3://gt-big-data-club/movie-lens-dir/out/_SUCCESS -> ./out/_SUCCESS  [1 of 16]
 0 of 0     0% in    0s     0.00 B/s  done
s3://gt-big-data-club/movie-lens-dir/out/part-00000 -> ./out/part-00000  [2 of 16]
 25021 of 25021   100% in    0s   106.80 kB/s  done
s3://gt-big-data-club/movie-lens-dir/out/part-00014 -> ./out/part-00014  [16 of 16]
 25882 of 25882   100% in    0s   133.07 kB/s  done

Using EMR Command Line Interface

EMR stands for Elastic MapReduce. We use EMR to run hadoop jobs on large datasets.

It is actually pretty simple. Follow instructions here:

  1. The EMR CLI requires Ruby 1.8.7. This is an old version (current is 2.1). For easy installation and management of these different versions, you can use Ruby Version Manager. See

To install RVM, run this in your terminal:

$ curl -sSL | bash -s stable

Afterwards, add the following line to your .bashrc or .zshrc to activate rvm

source ~/.rvm/scripts/rvm

Open a new terminal and type in rvm to make sure everything is working correctly.

To install / use ruby 1.8.7 run

$ rvm install 1.8.7-p374
$ rvm use 1.8.7-p374@emr-cli --create

Then if you type ruby -v you should get 1.8.7-p374

  1. You must fill in credentials.json for requires several fields.
  • access_id / private_key are in a credentials.csv file. Either you or Patrick have it. Ask if you can't find it.
  • key-pair / key-pair-file: create a key-pair file via (note: you probably already have openssl if you have linux). You must '''give the pem file to Patrick''' so we can associate it with your user account.
  • Use "log_uri": "s3://gt-big-data-club/logs",
  • Use "region": "us-east-1"

A good way to test out if EMR will work for you is to do this tutorial with Google NGrams.

⚠️ ** Fallback** ⚠️