mrjob: the Python MapReduce library

mrjob is a Python 2.7/3.4+ package that helps you write and run Hadoop Streaming jobs.

Stable version (v0.7.4) documentation

Development version documentation

mrjob fully supports Amazon's Elastic MapReduce (EMR) service, which allows you to buy time on a Hadoop cluster on an hourly basis. mrjob has basic support for Google Cloud Dataproc (Dataproc) which allows you to buy time on a Hadoop cluster on a minute-by-minute basis. It also works with your own Hadoop cluster.

Some important features:

• Run jobs on EMR, Google Cloud Dataproc, your own Hadoop cluster, or locally (for testing).
• Write multi-step jobs (one map-reduce step feeds into the next)
• Easily launch Spark jobs on EMR or your own Hadoop cluster
• Duplicate your production environment inside Hadoop
  • Upload your source tree and put it in your job's $PYTHONPATH
  • Run make and other setup scripts
  • Set environment variables (e.g. $TZ)
  • Easily install python packages from tarballs (EMR only)
  • Setup handled transparently by mrjob.conf config file
• Automatically interpret error logs
• SSH tunnel to hadoop job tracker (EMR only)
• Minimal setup
  • To run on EMR, set $AWS_ACCESS_KEY_ID and $AWS_SECRET_ACCESS_KEY
  • To run on Dataproc, set $GOOGLE_APPLICATION_CREDENTIALS
  • No setup needed to use mrjob on your own Hadoop cluster

Installation

pip install mrjob

As of v0.7.0, Amazon Web Services and Google Cloud Services are optional depedencies. To use these, install with the aws and google targets, respectively. For example:

pip install mrjob[aws]

A Simple Map Reduce Job

Code for this example and more live in mrjob/examples.

"""The classic MapReduce job: count the frequency of words.
"""
from mrjob.job import MRJob
import re

WORD_RE = re.compile(r"[\w']+")


class MRWordFreqCount(MRJob):

    def mapper(self, _, line):
        for word in WORD_RE.findall(line):
            yield (word.lower(), 1)

    def combiner(self, word, counts):
        yield (word, sum(counts))

    def reducer(self, word, counts):
        yield (word, sum(counts))


if __name__ == '__main__':
     MRWordFreqCount.run()

Try It Out!

# locally
python mrjob/examples/mr_word_freq_count.py README.rst > counts
# on EMR
python mrjob/examples/mr_word_freq_count.py README.rst -r emr > counts
# on Dataproc
python mrjob/examples/mr_word_freq_count.py README.rst -r dataproc > counts
# on your Hadoop cluster
python mrjob/examples/mr_word_freq_count.py README.rst -r hadoop > counts

Setting up EMR on Amazon

• create an Amazon Web Services account
• Get your access and secret keys (click "Security Credentials" on your account page)
• Set the environment variables $AWS_ACCESS_KEY_ID and $AWS_SECRET_ACCESS_KEY accordingly

Setting up Dataproc on Google

• Create a Google Cloud Platform account, see top-right
• Learn about Google Cloud Platform "projects"
• Select or create a Cloud Platform Console project
• Enable billing for your project
• Go to the API Manager and search for / enable the following APIs...
  • Google Cloud Storage
  • Google Cloud Storage JSON API
  • Google Cloud Dataproc API
• Under Credentials, Create Credentials and select Service account key. Then, select New service account, enter a Name and select Key type JSON.
• Install the Google Cloud SDK

Advanced Configuration

To run in other AWS regions, upload your source tree, run make, and use other advanced mrjob features, you'll need to set up mrjob.conf. mrjob looks for its conf file in:

• The contents of $MRJOB_CONF
• ~/.mrjob.conf
• /etc/mrjob.conf

See the mrjob.conf documentation for more information.

Project Links

• Source code
• Documentation
• Discussion group

Reference

• Hadoop Streaming
• Elastic MapReduce
• Google Cloud Dataproc

More Information

• PyCon 2011 mrjob overview
• Introduction to Recommendations and MapReduce with mrjob (source code)
• Social Graph Analysis Using Elastic MapReduce and PyPy

Thanks to Greg Killion (ROMEO ECHO_DELTA) for the logo.