Fix will and wildecards subscriptions tests

[robertsLando]

@mcollina Based on my tests the results are:

• memory: :+1: (clusters not supported by memory)
• redis: with clusters :+1: without clasters :+1: (if using mqemitter-redis with my fix)
• mongo: with clusters :-1: (random errors, some messages are not delivered) , without clusters :+1:

[mcollina]

Interesting that Mongo with cluster do not work. I think it might need some time to bring the cluster up?

[robertsLando]

@mcollina In my opinion it could be caused by mongo not ready when aedes start. All mongo persistence methods have once.ready listener (https://github.com/moscajs/aedes-persistence-mongodb/blob/master/persistence.js#L181) if the db is not ready and so it could loose some messages.

I could try to fix aedes process to wait db ready event

[robertsLando]

@mcollina I have removed all timeouts, I have also made aedes process waiting for db ready and a cleandb function to be sure there is not problems caused by someting already present inside dbs.

So everytime a test starts it starts the aedes process using a promise, that promise resolves when in order:

• All clusters are up
• All clusters persistences have been cleaned and trigger ready event
• All clusters servers are listeninig

Even with this checks before run result are the same: everything is :+1: except mongo with clusters that seems to loose some messages randomly. Any ideas?

P.S: Remember to merge and release a new version of mqemitter-redis

[robertsLando]

@gnought Do you have any idea why mongo could cause that kind of errors with clusters?

[mcollina]

no

[robertsLando]

At least, are tests now good after latest fixes?

If tests are ok, after you publish mqemitter-redis on npm, I will bump that here then I merge this?

[mcollina]

the new version of mqemitter-redis is out.

[robertsLando]

@mcollina Ping