This gem uses Redis and Resque to allow simple asynchronous communication between apps.
Application A can publish an event
# config
ResqueBus.redis = "192.168.1.1:6379"
# business logic
ResqueBus.publish("user_created", "id" => 42, "first_name" => "John", "last_name" => "Smith")
# or do it later
ResqueBus.publish_at(1.hour.from_now, "user_created", "id" => 42, "first_name" => "John", "last_name" => "Smith")
Application B is subscribed to events
# config
ResqueBus.redis = "192.168.1.1:6379"
# initializer
ResqueBus.dispatch("app_b") do
# processes event on app_b_default queue
# subscribe is short-hand to subscribe to your 'default' queue and this block with process events with the name "user_created"
subscribe "user_created" do |attributes|
NameCount.find_or_create_by_name(attributes["last_name"]).increment!
end
# processes event on app_b_critical queue
# critical is short-hand to subscribe to your 'critical' queue and this block with process events with the name "user_paid"
critical "user_paid" do |attributes|
CreditCard.charge!(attributes)
end
# you can pass any queue name you would like to process from as well IE: `banana "peeled" do |attributes|`
# and regexes work as well. note that with the above configuration along with this regex,
# the following as well as the corresponding block above would both be executed
subscribe /^user_/ do |attributes|
Metrics.record_user_action(attributes["bus_event_type"], attributes["id"])
end
# the above all filter on just the event_type, but you can filter on anything
# this would be _any_ event that has a user_id and the page value of homepage regardless of bus_event_type
subscribe "my_key", { "user_id" => :present, "page" => "homepage"} do
Mixpanel.homepage_action!(attributes["action"])
end
end
Applications can also subscribe within classes using the provided Subscriber
module.
class SimpleSubscriber
include ResqueBus::Subscriber
subscribe :my_method
def my_method(attributes)
# heavy lifting
end
end
The following is equivalent to the original initializer and shows more options:
class OtherSubscriber
include ResqueBus::Subscriber
application :app_b
subscribe :user_created
subscribe_queue :app_b_critical, :user_paid
subscribe_queue :app_b_default, :user_action, :bus_event_type => /^user_/
subscribe :homepage_method, :user_id => :present, :page => "homepage"
def user_created(attributes)
NameCount.find_or_create_by_name(attributes["last_name"]).increment!
end
def user_paid(attributes)
CreditCard.charge!(attributes)
end
def user_action(attributes)
Metrics.record_user_action(attributes["bus_event_type"], attributes["id"])
end
def homepage_method
Mixpanel.homepage_action!(attributes["action"])
end
end
Note: This subscribes when this class is loaded, so it needs to be in your load or otherwise referenced/required during app initialization to work properly.
Each app needs to tell Redis about its subscriptions:
$ rake resquebus:subscribe
The subscription block is run inside a Resque worker which needs to be started for each app.
$ rake resquebus:setup resque:work
The incoming queue also needs to be processed on a dedicated or all the app servers.
$ rake resquebus:driver resque:work
If you want retry to work for subscribing apps, you should run resque-scheduler
$ rake resque:scheduler
We've found it useful to have the bus act like cron
, triggering timed jobs throughout the system. Resque Bus calls this a heartbeat.
It uses resque-scheduler to trigger the events. You can enable it in your Rakefile.
# resque.rake
namespace :resque do
task :setup => [:environment] do
ResqueBus.heartbeat!
end
end
Or add it to your schedule.yml
directly
resquebus_heartbeat:
cron: "* * * * *"
class: "::ResqueBus::Heartbeat"
queue: resquebus_incoming
description: "I publish a heartbeat_minutes event every minute"
It is the equivalent of doing this every minute
seconds = minutes * (60)
hours = minutes / (60)
days = minutes / (60*24)
now = Time.at(seconds)
attributes = {}
now = Time.now
seconds = now.to_i
ResqueBus.publish("hearbeat_minutes", {
"epoch_seconds" => seconds,
"epoch_minutes" => seconds / 1.minute,
"epoch_hours" => seconds / 1.hour,
"epoch_days" => seconds / 1.day,
"minute" => now.min
"hour" => now.hour
"day" => now.day
"month" => now.month
"year" => now.year
"yday" => now.yday
"wday" => now.wday
})
This allows you do something like this:
ResqueBus.dispatch("app_c") do
# runs at 10:20, 11:20, etc
subscribe "once_an_hour", 'bus_event_type' => 'heartbeat_minutes', 'minute' => 20 do |attributes|
Sitemap.generate!
end
# runs every five minutes
subscribe "every_five_minutes", 'bus_event_type' => 'heartbeat_minutes' do |attributes|
next unless attributes["epoch_minutes"] % 5 == 0
HealthCheck.run!
end
# runs at 8am on the first of every month
subscribe "new_month_morning", 'bus_event_type' => 'heartbeat_minutes', 'day' => 1, hour' => 8, 'minute' => 0, do |attributes|
next unless attributes["epoch_minutes"] % 5 == 0
Token.old.expire!
end
end
ResqueBus can live along side another instance of Resque that points at a different Redis server.
# config
Resque.redis = "192.168.1.0:6379"
ResqueBus.redis = "192.168.1.1:6379"
If no Redis instance is given specifically, ResqueBus will use the Resque one.
# config
Resque.redis = "192.168.1.0:6379"
That will use the default (resque) namespace which can be helpful for using the tooling. Conflict with queue names are unlikely. You can change the namespace if you like though.
# config
Resque.redis = "192.168.1.0:6379"
ResqusBus.redis.namespace = :get_on_the_bus
For development, a local mode is also provided and is specified in the configuration.
# config
ResqueBus.local_mode = :standalone
or
ResqueBus.local_mode = :inline
Standalone mode does not require a separate resquebus:driver task to be running to process the incoming queue. Simply publishing to the bus will distribute the incoming events to the appropriate application specific queue. A separate resquebus:work task does still need to be run to process these events
Inline mode skips queue processing entirely and directly dispatches the event to the appropriate code block.
- There are a few spots in the code with TODO notes
- Make this not freak out in development without Redis or when Redis is down
- We might not actually need to publish in tests
- Add some rspec helpers for the apps to use: should_ post an event_publish or something along those lines
- Allow calling resquebus:setup and resquebus:driver together (append to ENV['QUEUES'], don't replace it)
Copyright (c) 2011 Brian Leonard, released under the MIT license