Literally, I was correct in saying that a gen_server can’t handle multiple clients concurrently. But normally this isn’t an issue since casting is so quick (since its a one-way message). If you want a gen_server to begin processing a message and be able to send back to the caller, I would spawn off the long running process, store the From parameter of gen_server:call and manually use gen_server:reply to reply via the spawned process.

To put this into a clearer example (I hope), imagine having a math server (based on gen_server) and there is a call which calculates all the primes up to n. For certain large n, this process can take quite awhile, so instead of doing all the calculations in the math server process, I would do something like the following:

handle_call({primes, N}, From, State) ->
  spawn(?MODULE, calculate_primes, [From, N]),
  {noreply, State}.

calculate_primes(From, N) ->
  Primes = [],
  % Use some method of calculating prime numbers...
  gen_server:reply(From, {primes, Primes}).

Hopefully the example above poses a fairly obvious solution. As you can see in handle_call, since it spawns off the calculation work, it is able to become ready to handle more calls/casts/etc. right away. Then, when the calculate_primes method ends, it explicitly sends the respond back to the caller.

Mitchell

First, thank you for the kind words. Next, to answer your question, I have taken a very brief look at Disco but I haven’t had enough experience using it in practice to come to a valuable conclusion about it.

Maybe I will, but at the moment I don’t have a need for disco