Unexpected behaviour of net_kernel

[Lorenzo Cian]

Dear all,

I am a beginner in Erlang and I'm working on a project.
I am working with nodes that need to connect and disconnect frequently both in "visible" and "hidden" mode.

I wonder if there is a bug in net_kernel, because I observe the following unexpected behaviour and I'm unable to link it to any action caused by my code.  

What happens in the execution of my program is roughly the following:
initial state: node A is connected (in visible mode) to node B,

                   node D is not connected to any other node

then:

1) node A disconnects from B (using net_kernel:disconnect/1)

2) node A connects to node D (using net_kernel:connect_node/1)

result: somehow D receives nodeup messages and connects to both A and B

expected behaviour: node D should connect to just A, while B should be connected to no one

Note that steps 1 and 2 are executed sequentially in the same function so there's no way they could be concurrent.

Could this be a bug in net_kernel? 

I thought that maybe, somehow A has some messages related to net_kernel that hang up in a way such that when step 2 happens, node A is not yet disconnected completely from B?

Is it possible that something like that happens? If yes, how can I avoid this?  

Even if I'm pretty confident that my code works as intended and I tried to debug and check it several times, of course the other option is that this could be caused by a bug in my code.

Of course there are a lot of other processes involved that I can't explain here, but let's suppose that they don't cause any interference with the example.

I hope someone who knows Erlang better than me can help me clarify my doubts.

Thank you,
Lorenzo.

[Dieter Schön]

Hi Lorenzo,


I played around with nodes in the last days, my experiments showed
nominal behaviour.

So, no spurious nodeup/nodedown messages for not connected nodes..

Did you read the fine print in the documentation of
net_kernel:monitor_nodes() ?

I would suspect some kind of race condition in your code.

Do you have timestamped logs which show when a node is shut down, and
when nodeup/down messages

arrive?


Kind regards,

Dieter