alexandrubagu

alexandrubagu

Hi,

I want to handle the following exit example ( String.to_integer(input) is jus an example and can be replaced with raise). This happen when I call My.GenServer.some_method(My.GenServer, "test"). I will be grateful if someone has a suggestion for this.

10:26:42.321 [error] GenServer My.GenServer terminating
** (stop) exited in: Task.await(%Task{owner: #PID<0.124.0>, pid: #PID<0.128.0>, ref: #Reference<0.4181578440.4035444737.94083>}, 5000)
    ** (EXIT) an exception was raised:
        ** (ArgumentError) argument error
            :erlang.binary_to_integer("test")

Here’s the code:

defmodule My.Application do
  use Application

  def start(_type, _args) do
    import Supervisor.Spec, warn: false

    children = [
      worker(My.GenServer, []),
      supervisor(Task.Supervisor, [[name: My.Task.Supervisor, restart: :transient]]),
    ]
    
    opts = [strategy: :one_for_one, name: Test.Supervisor]
    Supervisor.start_link(children, opts)
  end
end

defmodule My.GenServer do
    def start_link do
      GenServer.start_link(__MODULE__, :ok, name: __MODULE__)
    end

    def init(:ok) do
      # Process.flag(:trap_exit, true)
      {:ok, []}
    end

    def some_method(pid, input) do
        GenServer.call(pid, {:some_method, input})
    end

    def handle_call({:some_method, input}, _from, state) do
      # because we don't want to terminate GenServer use Task.async_nolink
      task = Task.Supervisor.async_nolink(My.Task.Supervisor, fn ->
          # raise argument error when passing binary
          String.to_integer(input)
      end)
      {:reply, Task.await(task), state}
    end

    def handle_info({:EXIT, from, reason}, state) do
      IO.inspect "handle_info::exit"
      IO.inspect reason
      {:noreply, state}
    end

    def handle_info({_ref, result}, state) do
      IO.inspect "handle_info::result"
      IO.inspect result
      {:noreply, state}
    end

    def handle_info({:DOWN, _ref, :process, _pid, reason}, state) do
      IO.inspect "handle_info::down"
      IO.inspect reason
      {:noreply, state}
    end

    def handle_info(msg, state) do
      IO.inspect "handle_info"
      IO.inspect msg
      {:noreply, state}
    end
end

I’ve read the docs from here regarding Task.Supervisor.async_link: if you create a task using async_nolink inside an OTP behaviour like GenServer, you should match on the message coming from the task inside your GenServer.handle_info/2 callback.

My problem is that handle_info is not called, I try to debug using observer attaching a trace to My.GenServer and I see that My.GenServer is receiving a :DOWN message:

Any solution to this problem ?
Thanks

Showing Posts 1 to 10

dom

dom

You’re not getting a result because there is no result: the task failed. It never got a chance to reply. :DOWN is the right thing to handle in that case.

NobbZ

NobbZ

When using Task.await/1 it will handle the :DOWN-message from the Task. It will handle it by calling exit/1. So the GenServer will be exited because of that.

If you want to be more failsafe, you have to do the waiting for the result and the :DOWN-message completely on your own.

alexandrubagu

alexandrubagu OP

I don’t get a :DOWN message in handle_info({:DOWN, _ref, :process, _pid, reason}, state)

If I remove Task.await(task) and replace {:reply, Task.await(task), state} with {:reply, :ok, state} I got that a :DOWN message but I need to wait for response from task.

dom

dom

Oh, I see what you mean. Instead of using await you can store the task’s ref and the caller in your GenServer state, then use GenServer.reply in your handle_info to send back the response to the caller.

NobbZ

NobbZ

As it is now you gain nothing from the Task except for the timeout, you could do the work in the GenServer directly and it wouldn’t matter (as long as no timout would happen in the Task)

The more correct way were to put the returned task into your state, alongside the from and return a :noreturn tuple in handle_call/3.

Some time later you will receive a message with either the result of the computation done in the Task, which you then can GenServer(from, answer) to your caller.

Or you might receive a :DOWN, which you will need to find ways to tell your caller about. Most idiomatic way were to use an :error-tuple I think.

alexandrubagu

alexandrubagu OP

Thanks a lot for your help @NobbZ and @dom , right now works.

defmodule My.GenServer do
    def start_link do
      GenServer.start_link(__MODULE__, :ok, name: __MODULE__)
    end

    def init(:ok) do
      {:ok, []}
    end

    def some_method(pid, input) do
        GenServer.call(pid, {:some_method, input})
    end

    def handle_call({:some_method, input}, from, _state) do
      Task.Supervisor.async_nolink(My.Task.Supervisor, fn ->
          # raise argument error when passing binary that not represent an integer
          String.to_integer(input)
      end)
      {:noreply, from}
    end

    def handle_info({:DOWN, _ref, :process, _pid, reason}, state) do
      GenServer.reply(state, {:error, reason})
      {:noreply, state}
    end

    def handle_info({_ref, result}, state) do
      GenServer.reply(state, {:ok, result})
      {:noreply, state}
    end

    def handle_info(_msg, state) do
      {:noreply, state}
    end
end
iex(1)> My.GenServer.some_method(My.GenServer, "test")
{:error,
 {:badarg,
  [{:erlang, :binary_to_integer, ["test"], []},
   {My.GenServer, :"-handle_call/3-fun-0-", 1, [file: 'lib/test.ex', line: 17]},
   {Task.Supervised, :do_apply, 2, [file: 'lib/task/supervised.ex', line: 85]},
   {Task.Supervised, :reply, 5, [file: 'lib/task/supervised.ex', line: 36]},
   {:proc_lib, :init_p_do_apply, 3, [file: 'proc_lib.erl', line: 247]}]}}
iex(2)> My.GenServer.some_method(My.GenServer, "1")   
{:ok, 1}
NobbZ

NobbZ

This will break when your GenServer will be hit by multiple calls before the result of the spawned task is sent back!

Each time you end up in your handle_cast you are overwriting the receiver of the answer.

Once you get an answer back, you will send it to the last one who asked for the info, letting all other processes starve. Also when subsequent answers from all the other tasks come in, you will send them into the mailbox of the same process (unless another one has asked in the meantime), spamming that processes mailbox with stuff that will probably never be read again.

Please make sure, that you map a task to a caller, and only send the answers/results back correspondingly.

alexandrubagu

alexandrubagu OP

Something like this ?

defmodule My.GenServer do
    def start_link do
      GenServer.start_link(__MODULE__, :ok, name: __MODULE__)
    end

    def init(:ok) do
      {:ok, %{}}
    end

    def some_method(pid, input) do
        GenServer.call(pid, {:some_method, input})
    end

    def handle_call({:some_method, input}, from, state) do
      task = Task.Supervisor.async_nolink(My.Task.Supervisor, fn ->
        # raise argument error when passing binary
        String.to_integer(input)
      end)

      {:noreply, Map.put(state, Map.get(task, :ref), from)}
    end

    def handle_info({:DOWN, ref, :process, _pid, reason}, state) do
      new_state = reply_to_caller(ref, {:error, reason}, state)

      {:noreply, new_state}
    end

    def handle_info({ref, result}, state) do
      new_state = reply_to_caller(ref, {:ok, result}, state)

      {:noreply, new_state}
    end

    def handle_info(_msg, state) do
      {:noreply, state}
    end

    def reply_to_caller(ref, message, state) do
      case Map.get(state, ref) do
        nil -> state
        from  -> 
          GenServer.reply(from, message)
          Map.delete(state, ref)
      end
    end
end

Thanks again, @NobbZ

NobbZ

NobbZ

Looks good on a first glance. I’d prefer to not call functions from inside the result tuple, but thats more of a personal and stylistic issue.

Where Next? Top

Trending in Questions Top

stjefim
Hello! Suppose you are building workflow (order / task / payment) processing system with the following requirements: Each workflow con...
New
water
I’m aware of 2 such projects, are there others? git: GitHub - team-alembic/realworld: A fullstack Phoenix LiveView application with back...
New
peck
Hi everyone. My team and I have been working on a fairly modest app based around video streaming and chat, but we’ve landed a customer t...
New
bradley
I really like the adapter patterns that ecto, nebulex, waffle, etc. use and would love find something similar for a key management servic...
New
unaware8150
Hello folks! So at work, we are seeing some situations where we have to define some “fixed” strings that are used across the codebase in...
New
Alvinkariuki
How Can I Optimise Compile Time Dependencies I have been building an elixir application for about 2 years now. Many modules and files ha...
New
leoschweizer
Hi, we run different LLM models across a cluster, with each node loading a configurable subset. Model-specific queues let nodes consume ...
New

Other Trending Topics Top

mudasobwa
I fully migrated to my own harness from Anthropic/Gemini and I think it’s time to share it. Welcome DSH, the DeepSeek Harness, fully writ...
New
KristerV
Hey. Is there anyone here who creates agents in their apps? Not talking about using agents, but creating them. I’m finding it pretty diff...
New
jimsynz
Beam Bots (or just BB for short) is a framework for building fault-tolerant robotics applications in Elixir using familiar OTP patterns. ...
New
juhalehtonen
There has been a thread to discuss the Stack Overflow Developer Survey on this forum every year since 2018, so here’s yet another one for...
New
webofbits
With AI doing more of the implementation work, I’ve been wondering how much coding I should deliberately keep doing myself. My main conc...
#ai
New
budgie
I love Elixir. It’s one of 2 programming languages I’ve ever fallen in love with. But I don’t use it anymore. Serverless was the promis...
New

We're in Beta

About us Mission Statement

Options

Thread Display Mode




Thread Preview

Skip Thread Previews