Distributed Erlang direction

[d-ferenczi]

First of all, I wanted to thank you for this project!
I have been doing many measurements and experiments with AtomVM setups in the past months and found the project fun and pleasant to work with.

Do you perhaps have a list of features that you plan to add to Distributed Erlang in the approximate future?
I would be interested to know if monitoring nodes and remote processes from AtomVM nodes can be expected.

[pguyot]

Thank you for your interest in AtomVM's disterl implementation.

Do you have any precise scenario of what is missing for you? Remote links do work, but it's true that monitoring nodes is not implemented and I guess monitoring remote processes isn't either.

[d-ferenczi]

I'm experimenting with leader-election between nodes, and was planning on basing it on monitoring.
I think, that links would work well though as well, thank you for the hint!

[d-ferenczi]

@pguyot are erlang messages between AtomVM nodes functional at the moment?
I though I had a small playground working, but both sender and receiver nodes appear to crash a couple of seconds after they exchange messages.
Sending messages from a BEAM host doesn't seem to trigger this issue.
I'm still working on verifying this behavior.

[pguyot]

@d-ferenczi This is indeed not something I tested much. Happy to have a look if you have a simple scenario that doesn't work.

[d-ferenczi]

@pguyot the bug seems tied to more complicated use-cases than I thought initially: AtomVM nodes crash when they receive messages from both AtomVM and BEAM nodes side by side.
I can't observe any bug, if they only receive messages from sources of the same runtime type.
I've uploaded a test project for easier building of .avm files: https://codeberg.org/d-ferenczi/atomvm-dist-msg.
The NODE macro in the source controls the resulting node's name and purpose. Node "a" will expect 4 messages and print them out, before expecting a quit atom and terminating. Other values for NODE will generate sender nodes, that send 3 messages to node "a".
Nodes are expected to be run on localhost, with an instance of EPMD already running.

I also found some other, possibly related issues:

• crashes seems to propagate between nodes
• senders crash when sending messages to non-existing (or disconnected) targets

Let me know, if I can provide any further details!