Class: BackRun::Pubsub

Inherits:
Object
  • Object
show all
Defined in:
lib/back_run/pubsub.rb

Instance Method Summary collapse

Constructor Details

#initialize ⇒ Pubsub

Returns a new instance of Pubsub.



6
7
8
9
10
11
 
# File 'lib/back_run/pubsub.rb', line 6

def initialize
  @pubsub = Google::Cloud::Pubsub.new(
    project_id: Rails.application.config.back_run[:project_id],
    credentials: Rails.application.config.back_run[:credentials_path]
  )
end

Instance Method Details

#kill_job(job) ⇒ Object 



36
37
38
39
 
# File 'lib/back_run/pubsub.rb', line 36

def kill_job(job)
  BackRun.logger.info("Moving to morgue queue #{job.klass}")
  publish(job, 'morgue')
end

#publish(job, topic = nil) ⇒ Object 



13
14
15
16
17
 
# File 'lib/back_run/pubsub.rb', line 13

def publish(job, topic = nil)
  topic_name = topic || job.queue_name
  topic = fetch_topic(topic_name)
  topic.publish(job.to_json)
end

#subscribe(worker) ⇒ Object

The gem is using Concurrent::CachedThreadPool which overrides the max_threads, so it’s ignoring the stream configuration used in the listen method. It may cause that lot’s of threads are created when the traffic increases. This PR github.com/googleapis/google-cloud-ruby/pull/3682 fixes that but it’s not released yet.



24
25
26
27
28
29
30
31
32
33
34
 
# File 'lib/back_run/pubsub.rb', line 24

def subscribe(worker)
  worker.queues.each do |queue|
    subscription = subscription_for(queue)
    subscriber = subscription.listen do |message|
      job = Job.from_json(message.data)
      worker.message_received(job, ack_callback(job, message), modify_ack_callback(message))
    end
    subscriber.on_error { |error| BackRun.logger.error(error) }
    subscriber.start
  end
end