Workers

Table of Contents

  1. README.md
  2. src/dependencies.litcoffee
  3. src/index.litcoffee
  4. src/overseer.litcoffee
  5. src/prototypes/parser_proto.litcoffee
  6. src/prototypes/service_proto.litcoffee
  7. src/services.litcoffee
  8. src/workers.litcoffee
  9. CONTRIBUTING.md

Overview

This module acts as the base template for all fork()s by the Overseer.

Source-through

Required modules.

cluster = require 'cluster'
domain  = require 'domain'

Figure out this worker's role from its environment, and require the module for that role.

role = process.env.role

worker = require "./#{role}"

Create a new domain within which the worker's code will be executed.

workerDomain = domain.create()

When an error in the worker process bubbles up all the way to the event loop, it is dangerous to continue running it, since it is technically in an undefined state. To overcome this, we use the errorHandler to essentially catch the error, call the worker's cleanup() method, and disconnect it from the cluster. This emits the cluster's exit event, which is caught by the master to replace this worker.

.unref() which will allow you to create a timer that is active but if it is the only item left in the event loop won't keep the program running.

errorHandler = (error) ->
    console.error 
    try
        deathTimer = setTimeout (-> process.exit 1), 15000
        deathTimer.unref()
        worker.cleanup()
        cluster.worker.disconnect()
    catch deadlyError
        console.error "Error Handler Errored!"

workerDomain.on 'error', errorHandler

workerDomain.run worker.runner
h