DynamiSupervisor not restart child

Questions / Help
1

if child stop, DynamicSupervisor not restart child.

    def add_child(camera) do
        root_dir = "/home/gdeon/page"
        spec = %{
            id: String.to_atom(camera.id),  
            start: {VideoVisor, :start_link, [camera.url, camera.id, root_dir]},
            restart: :permanent,
            type: :worker
        }
        DynamicSupervisor.start_child(__MODULE__, spec)
    end
2

Hello welcome to the forum, this code looks good, and it does not seem to have error in it. Do You have more info? Like your worker start code?

Just a side note… generating atom dynamically is highly discouraged, because they are limited in size and not garbage collected.

String.to_atom(camera.id)
3

This is all the code for my dynamic supervisor:

def start_link(_init_arg) do
    DynamicSupervisor.start_link(__MODULE__, :ok, name: __MODULE__)
end

@impl true
def init(:ok) do
    DynamicSupervisor.init(strategy: :one_for_one)
end


def add_child(camera) do
    root_dir = "/home/gdeon/vidos"
    spec = %{
        id: String.to_atom(camera.id),  
        start: {VideoVisor, :start_link, [camera.url, camera.id, root_dir]},
        restart: :permanent,
        type: :worker
    }
    DynamicSupervisor.start_child(__MODULE__, spec)
end

And the application runs this dynamic supervisor

4

I meant the worker, not the supervisor

5 
defmodule Archive.VideoVisor do
    use GenServer, restart: :permanent

    def start_link(camera_url, camera_id, root_dir) do
      GenServer.start(__MODULE__, [camera_url, camera_id, root_dir], name: {:global, "camera:#{camera_id}"})
    end

    def init([camera_url, camera_id, root_dir]) do
      Process.flag(:trap_exit, true)
      ....
      {:ok, initial_state}
    end

    def terminate(_reason, _state) do
      Logger.error("Test")
      :normal
    end

    def handle_info({:EXIT, exit_pid, :normal} = msg, %{ffmpeg_worker: %{pid: worker_pid}} = state) when exit_pid == worker_pid do
      {:stop, :normal, state}
    end

    def handle_info({:EXIT, exit_pid, {:exit_status, 256}} = msg, %{ffmpeg_worker: %{pid: worker_pid}} = state) when exit_pid == worker_pid do
      {:stop, :normal, state}
    end

    def handle_info(msg, state) do
      Logger.warn("Unhandled info for call session... MSG: #{inspect(msg)}, STATE: #{inspect(state)}")
      {:noreply, state}
    end
end
6

Your code seems ok, it should restart as set as permanent, but it looks your workers have also a supervisor role… they seem to be in charge of some ffmpeg sub workers.

Did You try to comment out code related to sub workers and see if it restarts?

I would choose a solution where ffmpeg_workers have their own supervisor tho.

7

ffmpeg_worker it
defp spawn_ffmpeg(camera_url, video_dir) do
script_exec = “…”
{:ok, pid, os_pid} = Exexec.run_link(script_exec, stdout: true)
%{pid: pid, os_pid: os_pid}
end

8

the problem was:
GenServer.start(MODULE, [camera_url, camera_id, root_dir], name: {:global, “camera:#{camera_id}”})
need to
GenServer.start_link(MODULE, [camera_url, camera_id, root_dir], name: {:global, “camera:#{camera_id}”})