routemaster-drain

A Rack-based event receiver for the Routemaster event bus.

   

routemaster-drain is a collection of Rack middleware to receive and parse Routemaster events, filter them, and preemptively cache the corresponding resources.

It provides prebuilt middleware stacks (Basic, Mapping, and Caching) for typical use cases, illustrated below, or you can easily roll your own by combining middleware.

• Installation
• Illustrated use cases
  • Simply receive events from Routemaster
  • Receive change notifications without duplicates
  • Cache data for all notified resources
• Internals
  • Dirty map
  • Filter
• Contributing

Installation

Add this line to your application's Gemfile:

gem 'routemaster-drain'

Configuration

This gem is configured through the environment, making 12factor compliance easier.

Required:

• ROUTEMASTER_DRAIN_TOKENS: a comma-separated list of valid authentication tokens, used by Routemaster to send you events.

Optional:

• ROUTEMASTER_DRAIN_REDIS: the URL of the Redis instance used for filtering and dirty mapping. Required if you use either feature, ignored otherwise. A namespace can be specified. Example: redis://user:s3cr3t@myhost:1234/12/ns.
• ROUTEMASTER_CACHE_REDIS: the URL of the Redis instance used for caching. Required if you use the feature, ignored otherwise. Formatted like ROUTEMASTER_DRAIN_REDIS.
• ROUTEMASTER_CACHE_EXPIRY: if using the cache, for how long to cache entries, in seconds. Default 1 year (31,536,000).
• ROUTEMASTER_CACHE_AUTH: if using the cache, specifies what username/password pairs to use to fetch resources. The format is a comma-separated list of colon-separate lists of regexp, username, password values. Example: server1:user:p4ss,server2:user:p4ass.
• ROUTEMASTER_QUEUE_NAME: if using the cache, on which Resque queue the cache population jobs should be enqueued.
• ROUTEMASTER_CACHE_TIMEOUT: if using the cache, how long before Faraday will timeout fetching the resource. Defaults to 1 second.
• ROUTEMASTER_CACHE_VERIFY_SSL: if using the cache, whether to verify SSL when fetching the resource. Defaults to false.

Illustrated use cases

Simply receive events from Routemaster

Provide a listener for events:

class Listener
  def on_events_received(batch)
    batch.each do |event|
      puts event.url
    end
  end
end

Each event is a Hashie::Mash and responds to type (one of create, update, delete, or noop), url (the resource), and t (the event timestamp, in milliseconds since the Epoch).

Create the app that will process events:

require 'routemaster/drain/basic'
$app = Routemaster::Drain::Basic.new

Bind the app to your listener:

$app.subscribe(Listener.new, prefix: true)

And finally, mount your app to your subscription path:

# typically in config.ru
map '/events' do
  run $app
end

This relies on the excellent event bus from the wisper gem.

Receive change notifications without duplicates

When reacting to changes of some resource, it's common to want to avoid receiving further change notifications until you've actually processed that resource.

Possibly you'll want to process changes in batches at regular time intervals.

For this purpose, use Routemaster::Drain::Mapping:

require 'routemaster/drain/mapping'
$app = Routemaster::Drain::Mapping.new

And mount it as usual:

# in config.ru
map('/events') { run $app }

Instead of processing events, you'll check for changes in the dirty map:

require 'routemaster/dirty/map'
$map = Routemaster::Dirty::Map.new

every_5_minutes do
  $map.sweep do |url|
    # do something about this changed resource
    true
  end
end

Until you've called #sweep and your block has returned true, you won't be bugged again — the dirty map acts as a buffer of changes (see below for internals).

Notes:

• You can limit the number of resources to be swept ($map.sweep(123) { ... }).
• You can count the number of resources to be swept with $map.count.
• You're not told what is to be swept; entities won't be swept in the order of events received (much like Routemaster does not guarantee ordering).
• If your sweeper fails, the dirty map will not be cleaned, so you can have leftovers. It's good practice to regularly run $map.sweep { ... } and perform cleanup regularly.
• The map won't tell you if the resources has been changed, created, or deleted. You'll have to figure it out with an API call.
• You can still attach a listener to the app to get all events.

Cache data for all notified resources

Another common use case is that you'll actually need the representation of the resources Routemaster tells you about.

The Caching prebuilt app can do that for you, using Resque to populate the cache as events are received.

For this purpose, use Routemaster::Drain::Caching:

require 'routemaster/drain/machine'
$app = Routemaster::Drain::Caching.new

And mount it as usual:

# in config.ru
map('/events') { run $app }

You can still attach a listenenr if you want the incoming events. Typically, what you'll want is the cache:

require 'routemaster/cache'
$cache = Routemaster::Cache.new

response = @cache.fget('https://example.com/widgets/123')
puts response.body.id

In this example, is your app was notified by Routemaster about Widget #123, the cache will be very likely to be hit; and it will be invalidated automatically whenever the drain gets notified about a change on that widget.

Note that Cache#fget is a future, so you can efficiently query many resources and have any HTTP GET requests (and cache queries) happen in parallel.

See rubydoc for more details on Cache.

Internals

The more elaborate drains are built with two components which can also be used independently, Dirty::Map and Dirty::Filter.

Dirty map

A dirty map collects entities that have been created, updated, or deleted (or rather, their URLs). It can be used to delay your service's reaction to events, for instance combined with Resque.

A dirty map map gets marked when an event about en entity gets processed that indicates a state change, and swept to process those changes.

Practically, instances of Routemaster::Dirty::Map will emit a :dirty_entity event when a URL is marked as dirty, and can be swept when an entity is "cleaned". If a URL is marked multiple times before being swept (e.g. for very volatile entities), the event will only by broadcast once.

To sweep the map, you can for instance listen to this event and call #sweep_one.

If you're not in a hurry and would rather run through batches you can call #sweep which will yield URLs until it runs out of dirty resources.

Filter

Routemaster::Dirty::Filter is a simple event filter that performs reordering. It ignores events older than the latest known information on an entity.

It stores transient state in Redis and will emit :entity_changed events whenever an entity has changed. This event can usefully be fed into a dirty map, as in Receiver::Filter for instance.

Contributing

1. Fork it ( http://github.com/HouseTrip/routemaster-drain/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create new Pull Request