Delayed Job + Octopus 0.10.2 issues on Rails 5.2 (undefined method `any?' for nil:NilClass)

[lunaru]

It looks like Delayed Job and Octopus are not happy with each other out of the box. I know both are fairly common gems used by the Rails community so I thought I'd go ahead and log an issue here with a collection of workarounds cobbled together from various other documented issues.

In particular, it looks like forking in general is a problem and documented here: https://github.com/thiagopradi/octopus/issues/489 and here: https://github.com/thiagopradi/octopus/issues/511 but these are documented as Puma issues, when in fact it's true for delayed_job as well.

In addition, there is also a shard selection issue documented here: https://github.com/tchandy/octopus/issues/241

In case anyone wanders into this Issue, here's initializer that I've put together that seems to fix things:

# from this: https://github.com/tchandy/octopus/issues/241
# this prevents delayed job from pulling from the wrong shard
module Delayed
  module Backend
    module ActiveRecord
      class Job < ::ActiveRecord::Base
        class << self
          alias_method :reserve_without_octopus, :reserve

          def reserve(worker, max_run_time = Worker.max_run_time)
            Octopus.using(:master) do
              reserve_without_octopus(worker, max_run_time)
            end
          end
        end
      end
    end
  end
end

# from this: https://github.com/thiagopradi/octopus/issues/489
# and also this: https://github.com/thiagopradi/octopus/issues/511
# Forking in DJ seems to cause issues with the connections
module Octopus
  class Proxy
    alias_method :safe_connection_without_fork_check, :safe_connection

    def safe_connection_with_fork_check(connection_pool)
      retries ||= 0
      safe_connection_without_fork_check(connection_pool)
    rescue NoMethodError => e
      ActiveRecord::Base.establish_connection
      ActiveRecord::Base.connection.initialize_shards(Octopus.config)
      retry if (retries += 1) < 2
    end

    alias_method :safe_connection, :safe_connection_with_fork_check

    def connected?
      retries ||= 0
      shards.any? { |_k, v| v.connected? }
    rescue NoMethodError
      proxy_config.reinitialize_shards
      retry if (retries += 1) < 2
    end
  end
end

[brentkearney]

Upgrading a Rails 5.1 system to 5.2, and encountered this issue. Switching to @mikwat's fork seems to solve it.

[jakeonfire]

i was having an issue with unicorn which also uses forking. a much simpler solution in that case is to use its after_fork hook to re-establish octopus connections:

# config/unicorn.rb
after_fork do |server, worker|
  if defined?(ActiveRecord::Base)
    ActiveRecord::Base.establish_connection
    ActiveRecord::Base.connection.initialize_shards(Octopus.config) if Octopus.enabled?
  end
end