Terminate supervisor once all workers are done

Why have the additional Supervisor there in the first place?

I am following the example from chapter 2 of Concurrent Data Processing in Elixir: Fast, Resilient Applications with OTP, GenStage, Flow, and Broadway by Svilen Gospodinov and the source code of this is here - concurrent_elixir/jobber at master · imeraj/concurrent_elixir · GitHub

The way it’s presented, it’s supposed to have multiple supervisor -> worker tree under top level DynamicSupervisor. In case, any of the jobs is terminated abnormally with exception, other supervisors or their jobs should not be impacted.

So instead of this supervision tree -

image502×458 20.2 KB

we got this -

image840×482 30.8 KB

tagging @svilen here too!

Sounds like you want to familiarize with the max_restarts, max_seconds setting. Because you don’t need that indirection to have that behaviour.

Thanks for the pointer. I looked into it more and this did the trick -

@impl Supervisor
  def init(args) do
    children = [
      %{
        id: Jobber.Job,
        start: {Jobber.Job, :start_link, [args]},
        significant: true,
        restart: :transient
      }
    ]
    options = [strategy: :one_for_one, max_seconds: 30, auto_shutdown: :all_significant]
    Supervisor.init(children, options)
  end

Looks like auto_shutdown option does it. Not sure if its the ideal solution!