Module: TcpUserTimeout::ActiveJob

Extended by:
ActiveSupport::Concern
Defined in:
lib/tcp_user_timeout/active_job.rb

Overview

Mix into ActiveJob jobs to enforce a per-job deadline:

class FetchUpstreamJob < ApplicationJob
  include TcpUserTimeout::ActiveJob
  self.max_execution_time = 30.seconds
  self.tcp_user_timeout_keepalive = true   # opt in for read-side coverage

  def perform(url)
    Net::HTTP.get(URI(url))
  end
end

Sockets opened during ‘perform` get TCP_USER_TIMEOUT bounded by the max_execution_time (with a small headroom so the kernel fires before any outer guard). Pre-existing connections (the AR pool, persistent HTTP pools) are not touched — only what the job opens itself.

Class Method Summary collapse

Class Method Details

.headroom_seconds(max) ⇒ Object

Reserve a small buffer below max_execution_time so the kernel closes the socket before any outer guard (Sidekiq’s job-killer, SolidQueue’s watchdog, etc.). 5s headroom at production scales (>=10s); 90% of max below that so short timeouts in tests still land below the deadline.



45
46
47
48
 
# File 'lib/tcp_user_timeout/active_job.rb', line 45

def self.headroom_seconds(max)
  max_f = max.to_f
  max_f >= 10 ? max_f - 5 : max_f * 0.9
end