RabbitMQ Test Suites

Useful targets

make unit  # runs the Erlang unit tests
make lite  # runs the Erlang unit tests and the Java client / functional tests
make full  # runs both the above plus the QPid test suite
make tests # runs the Erlang multi-node integration tests + all of the above

The multi-node tests take a long time, so you might want to run a subset:

make standalone-tests FILTER=dynamic_ha               # <- run just one suite
make standalone-tests FILTER=dynamic_ha:change_policy # <- run just one test

The multi-node tests also default to coverage on, to turn it off:

make standalone-tests COVER=false

tests-with-broker target also supports FILTER and COVER arguments, but instead runs modern single-node test - where all suites are executed inside single broker instance.

This repository is not related to plugin tests; run "make tests" in a plugin directory to test that plugin.

Kinds of tests

There are different kinds of tests located inside tests/src directory:

• Legacy test suite that can be only run as a whole using rabbit_tests:all_tests/0. Tests are run inside a running broker, so they could perform testing by doing direct calls to modules under the test.
• Modern tests that are run in a running broker. This tests are auto-discovered by function name, which should end with _test. As tests are being run inside a running broker, testing can be performed by directly calling modules under the test. But as tests are run independently by eunit, some care should be taken to preserve broker in a good state. (It's not a big problem for legacy test set as they stop after detecting the first failure).
• Multi-node tests, where node(s) are being setup from clean slate for every test case. This test are auto-disovered by function name, which should end with _with. Those _with functions provide meta-information about the way that nodes(s) should be setup, actual testing code should be placed in the function without _with postfix.

Writing multi-node tests

Multi-node test-case is represented by 2 functions: test itself with arbitary name, and setup metadata in a function named like the test function plus suffix _with. So, e.g. test named abc should be represented by: abc/1 and abc_with/0.

rabbit_test_runner calls XXX_with/0 and performs all initializations mentioned there, accumulating some config data while doing this. Possible return values for _with functions are of type rabbit_test_runner:initializers():

• atom referencing predefined function from rabbit_test_configs.
• function reference which should accept single parameter of type rabbit_test_configs:config().
• list of things mentioned above.

In the end we will have a proplist (or list of proplists for each node), that will be passed to test case itself.