Hi everyone, I’m relatively new to Elixir/Erlang, and I’m working on a demo project. The setup involves a master supervisor, a supervisor, and a worker, and I need to run each of them on separate devices. I’m unsure how these components will communicate when they’re on different machines. To tackl…

This is of course possible, here is a crude example in Erlang: -module(dsup). -export([ main/0 ]). -export([ child_init/1, child_loop/1 ]). -behaviour(supervisor). -export([ init/1 ]). % Demo entry point main() -> {ok, _, Node} = peer:start_link(), supervisor:start_link…

What is your intention with this supervision?

[image] saleh-tekm: To tackle this, I created separate projects for each component and connected them using Node.connect() and Phoenix.PubSub. However, I realized that this approach doesn’t handle supervision properly, right? While process links can span across nodes usually supervision and s…

To have a control over the child node and restart the child it if it dies. To ensure fault tolerance…But I dont know if it is the thing that is possible if project is setup at multiple nodes / interproject, like different projects at different nodes. Im sorry if Im wrong at fundamentals itself.

Why not keep it local? Supervision on the node?

What you need, I am not sure even Kubernetes has. It has some capabilities of distributed orchestration but I don’t think it can do OTP-like supervision. It helps to zoom out and ask yourself why do you need something that almost no technology is offering (not stably and bulletproof anyway).

Cool, I am grateful that you showed this. I stand corrected. I still would not want to have this in the split-brain scenarios but then again, I am not sure any other software stack would deal better with it than Erlang.

My sense is that a distributed application that is fault tolerant is a different thing from a distributed fault tolerance mechanism. A distributed application that is fault tolerant seems a reasonable and achievable goal (with some limitation), but a distributed fault tolerance mechanism I feel lik…

Does the child have to detect this and die (self-supervision)? That’s what linking does, the linking is bidirectional, so if the link is severed between the processes, both processes receive an exit signal. What makes supervisors special is that they trap_exit. In this scenario it will mean that …

Don’t do that. If your supervisor starts a child, the child will belong to the supervisor’s node for process groups. For instance it’s IO.puts and other IO will go through the supervisor node. You will have to do a lot of stuff just to get to a normally operating system. And if you turn off your w…

All great points related to the technical capabilities of OTP supervisors, including points I haven’t internalized yet. So we agree that the technical objective of running cross node supervision trees exists; your detailed earlier post also made that clear. My larger point was that just because yo…

Is inter-project supervision possible?

Questions & Help Questions

LostKobrakai December 19, 2024, 8:14am 3

While process links can span across nodes usually supervision and supervision trees are a thing local to a single node.