Demo: implementing a Twitter-like news feed using Apache Samza
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
LICENSE Fix bad copy and paste Nov 17, 2014
pom.xml Build cleanup Sep 8, 2014

Samza newsfeed demo

This is an experimental sample project for Samza, demonstrating how to implement a Twitter-like real-time news feed. It uses an asymmetric follower model, where each user sees the messages posted by all users they are following.

Each time someone follows someone, that's a follow event. Each time someone posts a message, that's a message event. If you don't have real data, this project includes Samza jobs for generating a random social graph and random messages, respectively.

The fan-out job uses Samza's state management facilities to store the list of followers for each user. Every time that user posts a message, the fan-out job makes a copy of that message for each follower, and sends it to the follower on a separate deliveries stream. A downstream job then groups together all the messages that need to be delivered to the same user (called a home timeline, again implemented using Samza state).

How to run

Start YARN, Zookeeper and Kafka through Hello Samza's bin/grid start all script.

This job depends on Samza 0.8.0, which is not yet released. So you need to build it from source and put it in your local Maven repository, as follows:

git clone samza
cd samza
./gradlew publishToMavenLocal

Then you can build and run the newsfeed jobs as follows:

  • mvn clean package && rm -rf deploy && mkdir -p deploy && tar xzf target/newsfeed-0.0.1-dist.tar.gz -C deploy
  • deploy/bin/ --config-path=file://$PWD/deploy/config/
  • deploy/bin/ --config-path=file://$PWD/deploy/config/
  • deploy/bin/ --config-path=file://$PWD/deploy/config/ (if necessary; automatically terminates after creating a social graph of a certain size)
  • deploy/bin/ --config-path=file://$PWD/deploy/config/ (if necessary; keeps running and continually generates random messages)

You can use Hello Samza's kafka-console-consumer to inspect the output, e.g. the newsfeed-home-timeline-changelog topic.


This is experimental code, not fit for production use. It is released under the terms of the MIT license.