mperham--sidekiq/lib/sidekiq/processor.rb

require 'sidekiq/util'
require 'sidekiq/fetch'
require 'thread'
require 'concurrent'

module Sidekiq
  ##
  # The Processor receives a message from the Manager and actually
  # processes it.  It instantiates the worker, runs the middleware
  # chain and then calls Sidekiq::Worker#perform.
  class Processor

    # To prevent a memory leak, ensure that stats expire. However, they
    # should take up a minimal amount of storage so keep them around
    # for a long time.
    STATS_TIMEOUT = 24 * 60 * 60 * 365 * 5

    include Util

    attr_reader :thread
    attr_accessor :job

    def initialize(mgr, options)
      @mgr = mgr
      @down = false
      @done = false
      @job = nil
      @strategy = (options[:fetch] || Sidekiq::BasicFetch).new(options)
    end

    def terminate(wait=false)
      @done = true
      @thread.value if wait
    end

    def kill(wait=false)
      # unlike the other actors, terminate does not wait
      # for the thread to finish because we don't know how
      # long the job will take to finish.  Instead we
      # provide a `kill` method to call after the shutdown
      # timeout passes.
      @thread.raise ::Sidekiq::Shutdown
      @thread.value if wait
    end

    def start
      @thread ||= safe_thread("processor", &method(:run))
    end

    private unless $TESTING

    def run
      begin
        while !@done
          self.job = fetch
          process(job) if job
          self.job = nil
        end
      rescue Exception => ex
        @mgr.processor_died(self, ex)
      end
    end

    def get_one
      begin
        work = @strategy.retrieve_work
        (logger.info("Redis is online, #{Time.now - @down} sec downtime"); @down = nil) if @down
        work
      rescue => ex
        handle_fetch_exception(ex)
      end
    end

    def fetch
      j = get_one
      if j && @done
        j.requeue
        nil
      else
        j
      end
    end

    def handle_fetch_exception(ex)
      if !@down
        @down = Time.now
        logger.error("Error fetching message: #{ex}")
        ex.backtrace.each do |bt|
          logger.error(bt)
        end
      end
      sleep(1)
    end

    def process(work)
      msgstr = work.message
      queue = work.queue_name

      ack = false
      begin
        msg = Sidekiq.load_json(msgstr)
        klass  = msg['class'.freeze].constantize
        worker = klass.new
        worker.jid = msg['jid'.freeze]

        stats(worker, msg, queue) do
          Sidekiq.server_middleware.invoke(worker, msg, queue) do
            # Only ack if we either attempted to start this job or
            # successfully completed it. This prevents us from
            # losing jobs if a middleware raises an exception before yielding
            ack = true
            execute_job(worker, cloned(msg['args'.freeze]))
          end
        end
        ack = true
      rescue Sidekiq::Shutdown
        # Had to force kill this job because it didn't finish
        # within the timeout.  Don't acknowledge the work since
        # we didn't properly finish it.
        ack = false
      rescue Exception => ex
        handle_exception(ex, msg || { :message => msgstr })
        raise
      ensure
        work.acknowledge if ack
      end
    end

    def execute_job(worker, cloned_args)
      worker.perform(*cloned_args)
    end

    def thread_identity
      @str ||= Thread.current.object_id.to_s(36)
    end

    WORKER_STATE = Concurrent::Map.new
    PROCESSED = Concurrent::AtomicFixnum.new
    FAILURE = Concurrent::AtomicFixnum.new

    def stats(worker, msg, queue)
      # Do not conflate errors from the job with errors caused by updating
      # stats so calling code can react appropriately
      tid = thread_identity
      WORKER_STATE[tid] = {:queue => queue, :payload => msg, :run_at => Time.now.to_i }

      begin
        yield
      rescue Exception
        FAILURE.increment
        raise
      ensure
        WORKER_STATE.delete(tid)
        PROCESSED.increment
      end
    end

    # Deep clone the arguments passed to the worker so that if
    # the message fails, what is pushed back onto Redis hasn't
    # been mutated by the worker.
    def cloned(ary)
      Marshal.load(Marshal.dump(ary))
    end

    # If an exception occurs in the block passed to this method, that block will be retried up to max_retries times.
    # All exceptions will be swallowed and logged.
    def retry_and_suppress_exceptions(max_retries = 5)
      retry_count = 0
      begin
        yield
      rescue => e
        retry_count += 1
        if retry_count <= max_retries
          Sidekiq.logger.info {"Suppressing and retrying error: #{e.inspect}"}
          pause_for_recovery(retry_count)
          retry
        else
          handle_exception(e, { :message => "Exhausted #{max_retries} retries"})
        end
      end
    end

    def pause_for_recovery(retry_count)
      sleep(retry_count)
    end
  end
end