wasabi

Wasabi A/B Testing service is an open source project that is no longer under active development or being supported

Wasabi - A/B Testing Platform

Support:

Documentation: User Guide, JavaDocs

A/B Testing Overview:

Continuous Integration: License:

Project

Wasabi A/B Testing Service is a real-time, enterprise-grade, 100% API driven project. Users are empowered to own their own data, and run experiments across web, mobile, and desktop. It’s fast, easy to use, it’s chock full of features, and instrumentation is minimal.

Learn more about how Wasabi can empower your team to move from hunches to actionable, data-driven user insights with our simple, flexible, and scalable experimentation platform.

Features

• Own your own data - Wasabi runs on your servers, in the cloud or on-premise, where you have complete control over your data.
• Proven - Wasabi is battle-tested in production at Intuit, Inc., a financial technology company. Wasabi is the experimentation platform for TurboTax, QuickBooks, Mint.com, and other Intuit offerings.
• High Performance - Consistent server-side response times for user assignments within 30ms.
• 100% API-Driven - The Wasabi REST API is compatible with any language and environment.
• Platform Agnostic - Uniform, consistent testing across Web, mobile, desktop. Also supports front-end, back-end integrations.
• Real-time user assignments - Assign users into experiments in real time, to preserve traffic for other parallel A/B tests.
• Cloud and on-premise - Designed to live in the cloud or in your own data center.
• Analytics - Core experiment analytics and metrics visualization out of the box, as well as the ability to send data to your existing analytics infrastructure.
• Pluggable - Well-defined interfaces for plugging in your own access control, sending data to data pipelines, and providing fully custom bucket allocations.
• Experiment Management UI - Setup and manage experiments via a modern Web interface. Management via REST API is also possible.
• Dockerized - Spin up a Wasabi Docker instance in 5 minutes and be in production with the platform, instrumentation, and experiments within a day.

User Interface

• Create an experiment and its buckets:

• Filter which customers are considered for your experiment:

• Follow your currently running experiments:

• Track your experiment results in real-time:

Get Started

The following steps will help you install the needed tools, then build and run a complete Wasabi stack.

Bootstrap Your Environment

Mac OS

% /usr/bin/ruby \
  -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
% brew install git
% git clone https://github.com/intuit/wasabi.git
% cd wasabi
% ./bin/wasabi.sh bootstrap

Installed tools include: homebrew 0.9, git 2, maven 3, java 1.8, docker 1.12, node 6 and python 2.7.

Ubuntu

Bootstrapping on Ubuntu requires sudo privileges to install all the required dependencies. You will be prompted to enter your password. Currently only 16.04 (x64) is supported.

% sudo apt-get install git
% git clone https://github.com/intuit/wasabi.git
% cd wasabi
% ./bin/wasabi.sh bootstrap
% sudo reboot

NOTE: A reboot is required after running the bootstrap command on Ubuntu.

For all other processes (build, start etc.) the commands are same for Ubuntu and Mac OS.

Installed tools include: git 2, maven 3, OpenJdk 8, docker 1.12, node 6 and python 2.7

Similar tooling will work for Windows. Contribute a patch :)

Start Wasabi

Now that we have the necessary tools in place, let's move on to build and start Wasabi, followed by issuing a ping command to verify the build:

…

Congratulations! You are the proud owner of a newly minted Wasabi instance. :)

Running Wasabi with remote storage

Set Mysql and Cassandra credentials

• Modify /pom.xml to set the values that apply to your environment

Download Cassandra migration tool https://oss.sonatype.org/content/repositories/public/com/builtamont/cassandra-migration/0.9/cassandra-migration-0.9-jar-with-dependencies.jar

Set up your environment variables

• Set location of the migration tool

export CASSANDRA_MIGRATION=/location/of/cassandra-migration-0.9-jar-with-dependencies.jar

• Set location of migration scripts within your project

export MIGRATION_SCRIPT=/location/of/modules/repository-datastax/src/main/resources/com/intuit/wasabi/repository/impl/cassandra/migration

Set up Cassandra tables

CQLSH_VERSION= CQLSH_USERNAME= CQLSH_PASSWORD= CQLSH_HOST= bin/docker/migration.sh

Run Wasabi with env variables for remote storage hosts

MYSQL_HOST= NODE_HOST= ./bin/wasabi.sh start:wasabi

Run Wasabi outside of docker with WASABI_CONFIGURATION for remote storage hosts

WASABI_CONFIGURATION="
  -Ddatabase.url.host=$MYSQL_HOST\
  -Ddatabase.url.port=$MYSQL_PORT\
  -Ddatabase.url.dbname=$MYSQL_DATABASE\
  -Ddatabase.user=$MYSQL_USER\
  -Ddatabase.password=$MYSQL_PASSWORD\
  -Ddatabase.pool.connections.min=$MYSQL_MIN_CONNECTIONS\
  -Ddatabase.pool.connections.max=$MYSQL_MAX_CONNECTIONS\
  -Dusername=$CASSANDRA_USER\
  -Dpassword=$CASSANDRA_PASSWORD\
  -DnodeHosts=$CASSANDRA_HOST\
  -DtokenAwareLoadBalancingLocalDC=$CASSANDRA_DATACENTER\
  -Dapplication.http.port=$PORT" bash usr/local/wasabi-main-*/bin/run

Troubleshooting

• While starting Wasabi, if you see an error when the docker containers are starting up, you could do the following:

  • Look at the current docker containers that have been successfully started.

…

• The above shell output shows a successful start of 3 docker containers needed by Wasabi: wasabi-main (the Wasabi server), wasabi-mysql, and wasabi-cassandra. If any of these are not running, try starting them individually. For example, if the MySQL container is running, but Cassandra and Wasabi containers failed to start (perhaps due to a network timeout docker could not download the Cassandra image), do the following:

% ./bin/wasabi.sh start:cassandra

% ./bin/wasabi.sh start:wasabi

Call Wasabi

These are the 3 common REST endpoints that you will use to instrument your client application with Wasabi.

Let's assume that you've created and started an experiment, 'BuyButton,' in the 'Demo_App' application with the following buckets:

• 'BucketA': green button (control bucket)
• 'BucketB': orange button bucket

You can assign a user with a unique ID (e.g. 'userID1') to the experiment by making this HTTP request:

Assign a user to experiment and bucket:

% curl -H "Content-Type: application/json" \
    http://localhost:8080/api/v1/assignments/applications/Demo_App/experiments/BuyButton/users/userID1

{  
   "cache":true,
   "payload":"green",
   "assignment":"BucketA",
   "context":"PROD",
   "status":"NEW_ASSIGNMENT"
}

Now the 'userID1' user is assigned into the 'BucketA' bucket. Let's further record an impression, meaning the user has seen a given experience:

Record an impression:

% curl -H "Content-Type: application/json" \
    -d "{\"events\":[{\"name\":\"IMPRESSION\"}]}" \
    http://localhost:8080/api/v1/events/applications/Demo_App/experiments/BuyButton/users/userID1

If the 'userID1' user performs an action such as clicking the Buy button, you'd record that action with the following request:

Record an action:

% curl -H "Content-Type: application/json" \
    -d "{\"events\":[{\"name\":\"BuyClicked\"}]}" \
    http://localhost:8080/api/v1/events/applications/Demo_App/experiments/BuyButton/users/userID1

Explore Various Resources

The following developer resources are available:

API: Swagger API playground

% ./bin/wasabi.sh resource:api

Javadoc

% ./bin/wasabi.sh resource:doc

Wasabi UI

% ./bin/wasabi.sh resource:ui

Cassandra: cqlsh shell

% ./bin/wasabi.sh resource:cassandra

MySQL: mysql shell

% ./bin/wasabi.sh resource:mysql

Java Debugger: Remote attach configuration

-agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=8180

Stop Wasabi

Alas, all good things must come to an end. Let's clean things up a bit stop the newly created Wasabi stack:

At this point in time, we now have all the requisite tools installed, and subsequent invocations of Wasabi will start up much more quickly.

% ./bin/wasabi.sh stop

Get Familiar with wasabi.sh

Further, there are a number of additional wasabi.sh options available you should become familiar with:

…

Develop

Build and Run Wasabi Server

% mvn package
% ./bin/wasabi.sh start:cassandra,mysql
% (cd modules/main/target; \
    WASABI_CONFIGURATION="-DnodeHosts=localhost -Ddatabase.url.host=localhost" ./wasabi-main-*-SNAPSHOT-development/bin/run) &
% curl -i http://localhost:8080/api/v1/ping
...

The runtime logs can be accessed executing the following command in a another shell:

Viewing runtime logs:

% tail -f modules/main/target/wasabi-main-*-SNAPSHOT-development/logs/wasabi-main-*-SNAPSHOT-development.log

Build and Run Wasabi UI

% cd modules/ui
% grunt build

% grunt serve

Stop Wasabi

% ./bin/wasabi.sh stop

Now while that was fun, in all likelihood you will be using an IDE to work on Wasabi. In doing so, you need only add the configuration information above to the JVM commandline prior to startup:

Wasabi runtime configuration:

-DnodeHosts=localhost -Ddatabase.url.host=localhost

Run Integration Tests

Code changes can readily be verified by running the growing collection of included integration tests:

% ./bin/wasabi.sh start test stop

Troubleshooting

Integration tests might fail intermittently due to a time drift issue in docker containers on Mac OSX.

When the Mac sleeps and wakes back up, there is a lag created between the clock in the Mac vs the running docker containers. This is a known issue in Docker for Mac.

This can be fixed by running the following command:

% docker run --rm --privileged alpine hwclock -s

The above command will need to be run every time when there is a time drift.

To automatically run this command and update the time each time the Mac wakes up, you could install the following agent:

% curl https://raw.githubusercontent.com/arunvelsriram/docker-time-sync-agent/master/install.sh | bash

You can read more about this at: quick-tip-fixing-time-drift-issue-on-docker-for-mac

Package and Deploy at Scale

Wasabi can readily be packaged as installable rpm or deb distributions and deployed at scale as follows:

Package by running integration tests 1st:

% ./bin/wasabi.sh start package
``