resque-aps
Resque-Aps is an extension to Resque that adds shortcuts to send notifications to Apple's Push System.
Installation and integration with callbacks
To install:
gem install resque-aps
You'll need to add this to your Rails rakefile to see the queue lengths:
require 'resque/plugins/aps/tasks'
task "resque:setup" => :environment
$ rake resque:aps:queue_lengths
I use this to monitor the system in nagios.
To extend the system create an initializer:
require 'resque_aps'
Resque.aps_gateway_host = AppConfig.apn_gateway_host
Resque.aps_gateway_port = AppConfig.apn_gateway_port
Resque.aps_feedback_host = AppConfig.apn_feedback_host
Resque.aps_feedback_port = AppConfig.apn_feedback_port
module Resque::Plugins::Aps
class Application
def after_aps_write(notification)
logger.info("Sent Notification [#{notification.application_name}] [#{notification.device_token}] [#{notification.payload}]") if logger
end
# It probably failed because there is something wrong with the device token
# do not requeue.
def failed_aps_write(notification, exception)
logger.error "failed_aps_write: #{notification.inspect} - #{exception}"
logger.error exception.backtrace.join("\n")
end
# Something is probably wrong with the certificate
def notify_aps_admin(exception)
# Notify once a minute
unless Rails.cache.read("push_certificate:notify_admin:#{name}")
logger.error("notification: #{exception.}")
# Email error to someone
Rails.cache.write("push_certificate:notify_admin:#{name}", "true", :expires_in => 60)
end
rescue
logger.error("#{$!} (#{name}): #{$!.backtrace.join("\n")}")
end
def aps_nil_notification_retry?(sent_count, start_time)
# Live forever
sleep 1
true
# Or not
#false
end
def after_aps_read(feedback)
# Remove the device token from the system for this application so we don't send to it anymore
# feedback.application_name, feedback.device_token
end
end
end
Plagurism alert
This was intended to be an extension to resque and was based heavily on resque-scheduler, which resulted in a lot of the code looking very similar. One massive departure is the use of logging which resque and resque-scheduler do not use but I do.
Contributing
For bugs or suggestions, please just open an issue in github.