rlinks: duplicate events after reconnect

[asimfarooq5]

If the connection between instance A and B is lost, On connection resume/resubscribe the event causes duplicate results for number of times it reconnects.

The same configuration can be used to reproduce the bug #1909

• Start both Instances A and B
• Subscribe event on instance A and B
• Publish event to instance A and B
• get the same event on both instances
• stop/restart instance B then resubscribe
• Now, when the event is published to instance A, this will result in duplicate events on instance B

[om26er]

This issue is real, I can reproduce that. It seems the local leg has the subscriptions even when the remote end disconnected/stopped. In such event the following code is not called https://github.com/crossbario/crossbar/blob/01ec53a100b08a4b78adabe26ab0abb12ba21d5a/crossbar/worker/rlink.py#L167

So we probably need to update on_remote_leave to remove to the remote events subscriptions from the local leg. https://github.com/crossbario/crossbar/blob/01ec53a100b08a4b78adabe26ab0abb12ba21d5a/crossbar/worker/rlink.py#L204

[oberstet]

As a note, it can be daunting/confusing to think about and keep all message interactions for a certain scenario in mind with this rlink stuff. In particular, since those WAMP messages happen at the meta level, and at the app level.

• pls consider https://github.com/crossbario/crossbar/blob/master/docs-cfx/_static/crossbar-rlinks.pdf - I tried to make this stuff approachable, and ..
• This is supposed to intro one graphical notation for documenting what is supposed to happen with rlinks at the WAMP message level in an accurate way.
• If that notation makes sense and is welcome, we could add all message interactions and scenarios

[asimfarooq5]

It seems to be fixed with proposed branch #1919.