FaaS Orchestrator

Building orchestration functions easy as 1, 2, 3

What is this?

FaaSOrchestrator allows you to easily buid Orchestration Functions for OpenFaaS. Imagine wanting to just call two functions, pipeline style:

Orchestrator.new.
  first('get_weather_data').
  finally('post_to_slack')

This code will call the first function, then call the second function with output from the first and then return the result of the second function.

Orchestration?

Well, if you want to have multiple functions work together and not couple them incredibly tightly (by calling B directly from A), you will want a manager function (orchestrator) that calls the functions for you :)

So what...

Now there's a bunch of stuff we have to consider here... configuration, retries, maybe something like calling multiple functions in parallel... Suddenly, you have to write a bunch of code! :( But FaaS Orchestrator alleviates this pain and allows for easily building complex orchestrations!

For example, you might need to modify data in between functions:

Orchestrator.new.
  first('get_weather_data').
  modify do |data|
    data = JSON.parse(data)
    { current_temp: data['Berlin']['now']['celsius'] }
  end.
  finally('post_to_slack')

You can easily use modify blocks for that. The result will be passed on, the rest thrown out! 🎉

You can also fan out and have multiple functions run on the same data, combining the result!

Orchestrator.new.
  first('fetch_performance_analysis').
  then(multiple: [['calc_avg'], ['calc_median']]).
  modify do |data|
    "Current Server performance:\n"\
    "- average response time #{data[0]}\n"\
    "- median response time #{data[1]}."
  end.
  finally(multiple: [['update_status_page'], ['post_to_slack']])

What if I want to keep old data when fanning out?

No problem! Use Orchestrator::RETAIN to move data along:

Orchestrator.new.
  first('get_weather_data').
  then(multiple: [['make_prediction'], [Orchestrator::RETAIN]]).
  modify do |data|
    {
      current_temp: data[1]['current']['celsius'],
      prediction: data[0]['future']['celsius']
    }
  end.
  finally('post_to_slack')

Code is hard, can't I just write a yaml definition?

Sure! If you don't need in-between modification steps, you can also create a yaml file, like:

steps:
  - get_weather_data:
    - method: GET
  - multiple:
    - calc_daily_average:
      - method: POST
    - Orchestrator::RETAIN
  - multiple:
    - post_to_slack
    - post_to_twitter

and then in your function just do (the file will always be in the function folder):

Orchestrator.execute_from_yaml_file('function/name_of_file.yml')

Ok, you got me, how do I use this?

Check out the example folder for a real world use case. First you'd want to clone this repo, then you can copy the required files (see dep.sh) and finally require the orchestrator in your function.