Fix flake on rabbitmq_mqtt auth_SUITE

[MarcialRosales]

Proposed Changes

Separate invalid client test from the valid one onto separate group and this way they use different SSL certs

Types of Changes

What types of changes does your code introduce to this project?
Put an x in the boxes that apply

• Bug fix (non-breaking change which fixes issue #NNNN)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause an observable behavior change in existing systems)
• Documentation improvements (corrections, new content, etc)
• Cosmetic change (whitespace, formatting, etc)
• Build system and/or CI

[ansd]

CI fails

[MarcialRosales]

This fix was only for the flake on the mqtt auth_SUITE ... I cannot see any test suite related to mqtt or mqtt's auth_SUITE failing besides the dialyzer suite.
I have noticed though that the selenium test is failing. The change done on this PR is totally unrelated. I am looking at this selenium failure now locally. I have run it two times in CI and it is failing in the same test.

[ansd]

This PR doesn't fix the flake. I can repro the flake locally as follows:

    {error, {client_identifier_not_valid, _}} = emqtt:connect(C),
    timer:sleep(1),
    unlink(C).

which makes the test fail with:

=== === Reason: {'EXIT',{shutdown,client_identifier_not_valid}}

That's the flake which we also observed in CI.

The correct fix is to unlink first, then connect.

[MarcialRosales]

i cannot reproduce it locally and i dont see CI failing because of this . I can modify the test as you suggest though.

[MarcialRosales]

I still dont understand what you are suggesting. I am expecting that when the user connects it fails because the client_id is not valid. That is exactly what it is doing line 594.

[MarcialRosales]

Can you send me the link to the failed job in CI where this test case is failing ?

[ansd]

Can you send me the link to the failed job in CI where this test case is failing ?

Yes, here are two flakes failing due to {'EXIT',{shutdown,client_identifier_not_valid}}:

• https://github.com/rabbitmq/rabbitmq-server/actions/runs/12769170662/job/35591148832
• https://github.com/rabbitmq-pro/tanzu-rabbitmq/actions/runs/12768915497/artifacts/2428404146

[ansd]

i cannot reproduce it locally and i dont see CI failing because of this .

The CT test case process runs concurrently with the MQTT connection process. CI can fail if the CT test case process executes unlink(C) after receiving the EXIT signal.

[ansd]

I still dont understand what you are suggesting.

I'm suggesting to flip the two lines, i.e. to unlink before the expected failure as done in many other places in this test suite, e.g. in:

• rabbitmq-server/deps/rabbitmq_mqtt/test/auth_SUITE.erl

  Lines 810 to 812 in 3a61274

  	unlink(C2),
  	?assertMatch({error, _},
  	emqtt:connect(C2)),

• rabbitmq-server/deps/rabbitmq_mqtt/test/auth_SUITE.erl

  Lines 945 to 947 in 3a61274

  	unlink(C2),
  	?assertMatch({error, _},
  	emqtt:connect(C2)),

• rabbitmq-server/deps/rabbitmq_mqtt/test/auth_SUITE.erl

  Lines 1227 to 1228 in 3a61274

  	unlink(C),
  	?assertMatch({error, _}, emqtt:connect(C)),

[MarcialRosales]

those ci failures are because this change has not been merged yet

[MarcialRosales]

Regardless i have committed that change.

[MarcialRosales]

but in this PR, with just my changes you dont see the flake. You will see the flake in other CI which do not have this fix.