Imagine that I have test/unit/... which are safe to run in parallel and test/functional/... which cannot be run in parallel yet.

Is there an easy way to convince pytest to run the functional ones sequentially? Consider that we are talking about a big number of tests so altering each test function/method would be very noisy.

At this moment we run tests with marker filters so mainly we run them separated. Still, I am looking for a solution for removing the need to run them separated.

You can implement your own scheduler that acts like load or loadscope, depending on what module the test is defined in. Example:

from xdist.scheduler.loadscope import LoadScopeSchedulingclass MyScheduler(LoadScopeScheduling):def _split_scope(self, nodeid):if 'test/functional' in nodeid:return 'functional-tests'return nodeiddef pytest_xdist_make_scheduler(config, log):return MyScheduler(config, log)

All tests under test/functional will be grouped under the same node functional-tests (and thus run in the same worker), the rest will be run in parallel as usual.