pytest-xdist

The pytest-xdist plugin extends pytest with new test execution modes, the most used being distributing tests across multiple CPUs to speed up test execution:

pytest -n auto

With this call, pytest will spawn a number of workers processes equal to the number of available CPUs, and distribute the tests randomly across them. There is also a number of distribution modes to choose from.

NOTE: due to how pytest-xdist is implemented, the -s/--capture=no option does not work.

Installation

Install the plugin with:

pip install pytest-xdist

To use psutil for detection of the number of CPUs available, install the psutil extra:

pip install pytest-xdist[psutil]

Features

• Test run parallelization: tests can be executed across multiple CPUs or hosts. This allows to speed up development or to use special resources of remote machines.

• --looponfail: run your tests repeatedly in a subprocess. After each run pytest waits until a file in your project changes and then re-runs the previously failing tests. This is repeated until all tests pass after which again a full run is performed.

• Multi-Platform coverage: you can specify different Python interpreters or different platforms and run tests in parallel on all of them.

  Before running tests remotely, pytest efficiently "rsyncs" your program source code to the remote place. You may specify different Python versions and interpreters. It does not installs/synchronize dependencies however.

  Note: this mode exists mostly for backward compatibility, as modern development relies on continuous integration for multi-platform testing.

Running tests across multiple CPUs

To send tests to multiple CPUs, use the -n (or --numprocesses) option:

pytest -n 8

Pass -n auto to use as many processes as your computer has CPU cores. This can lead to considerable speed ups, especially if your test suite takes a noticeable amount of time.

The test distribution algorithm is configured with the --dist command-line option:

• --dist load (default): Sends pending tests to any worker that is available, without any guaranteed order.

• --dist loadscope: Tests are grouped by module for test functions and by class for test methods. Groups are distributed to available workers as whole units. This guarantees that all tests in a group run in the same process. This can be useful if you have expensive module-level or class-level fixtures. Grouping by class takes priority over grouping by module.

• --dist loadfile: Tests are grouped by their containing file. Groups are distributed to available workers as whole units. This guarantees that all tests in a file run in the same worker.

• --dist loadgroup: Tests are grouped by the xdist_group mark. Groups are distributed to available workers as whole units. This guarantees that all tests with same xdist_group name run in the same worker.

  @pytest.mark.xdist_group(name="group1")
  def test1():
      pass
  
  class TestA:
      @pytest.mark.xdist_group("group1")
      def test2():
          pass

  This will make sure test1 and TestA::test2 will run in the same worker. Tests without the xdist_group mark are distributed normally as in the --dist=load mode.

• --dist no: The normal pytest execution mode, runs one test at a time (no distribution at all).

Running tests in a Python subprocess

To instantiate a python3.9 subprocess and send tests to it, you may type:

pytest -d --tx popen//python=python3.9

This will start a subprocess which is run with the python3.9 Python interpreter, found in your system binary lookup path.

If you prefix the --tx option value like this:

--tx 3*popen//python=python3.9

then three subprocesses would be created and tests will be load-balanced across these three processes.

Running tests in a boxed subprocess

This functionality has been moved to the pytest-forked plugin, but the --boxed option is still kept for backward compatibility.

Sending tests to remote SSH accounts

Suppose you have a package mypkg which contains some tests that you can successfully run locally. And you have a ssh-reachable machine myhost. Then you can ad-hoc distribute your tests by typing:

pytest -d --tx ssh=myhostpopen --rsyncdir mypkg mypkg

This will synchronize your mypkg package directory to a remote ssh account and then locally collect tests and send them to remote places for execution.

You can specify multiple --rsyncdir directories to be sent to the remote side.

You can specify multiple --rsyncignore glob patterns to be ignored when file are sent to the remote side. There are also internal ignores: .*, *.pyc, *.pyo, *~ Those you cannot override using rsyncignore command-line or ini-file option(s).

Sending tests to remote Socket Servers

Download the single-module socketserver.py Python program and run it like this:

python socketserver.py

It will tell you that it starts listening on the default port. You can now on your home machine specify this new socket host with something like this:

pytest -d --tx socket=192.168.1.102:8888 --rsyncdir mypkg mypkg

Running tests on many platforms at once

The basic command to run tests on multiple platforms is:

pytest --dist=each --tx=spec1 --tx=spec2

If you specify a windows host, an OSX host and a Linux environment this command will send each tests to all platforms - and report back failures from all platforms at once. The specifications strings use the xspec syntax.

When tests crash

If a test crashes a worker, pytest-xdist will automatically restart that worker and report the test’s failure. You can use the --max-worker-restart option to limit the number of worker restarts that are allowed, or disable restarting altogether using --max-worker-restart 0.

How-tos

Identifying the worker process during a test

New in version 1.15.

If you need to determine the identity of a worker process in a test or fixture, you may use the worker_id fixture to do so:

@pytest.fixture()
def user_account(worker_id):
    """ use a different account in each xdist worker """
    return "account_%s" % worker_id

When xdist is disabled (running with -n0 for example), then worker_id will return "master".

Worker processes also have the following environment variables defined:

• PYTEST_XDIST_WORKER: the name of the worker, e.g., "gw2".
• PYTEST_XDIST_WORKER_COUNT: the total number of workers in this session, e.g., "4" when -n 4 is given in the command-line.

The information about the worker_id in a test is stored in the TestReport as well, under the worker_id attribute.

Since version 2.0, the following functions are also available in the xdist module:

def is_xdist_worker(request_or_session) -> bool:
    """Return `True` if this is an xdist worker, `False` otherwise

    :param request_or_session: the `pytest` `request` or `session` object
    """

 def is_xdist_controller(request_or_session) -> bool:
    """Return `True` if this is the xdist controller, `False` otherwise

    Note: this method also returns `False` when distribution has not been
    activated at all.

    :param request_or_session: the `pytest` `request` or `session` object
    """

def is_xdist_master(request_or_session) -> bool:
    """Deprecated alias for is_xdist_controller."""

def get_xdist_worker_id(request_or_session) -> str:
    """Return the id of the current worker ('gw0', 'gw1', etc) or 'master'
    if running on the controller node.

    If not distributing tests (for example passing `-n0` or not passing `-n` at all)
    also return 'master'.

    :param request_or_session: the `pytest` `request` or `session` object
    """

Identifying workers from the system environment

New in version 2.4

If the setproctitle package is installed, pytest-xdist will use it to update the process title (command line) on its workers to show their current state. The titles used are [pytest-xdist running] file.py/node::id and [pytest-xdist idle], visible in standard tools like ps and top on Linux, Mac OS X and BSD systems. For Windows, please follow setproctitle's pointer regarding the Process Explorer tool.

This is intended purely as an UX enhancement, e.g. to track down issues with long-running or CPU intensive tests. Errors in changing the title are ignored silently. Please try not to rely on the title format or title changes in external scripts.

Uniquely identifying the current test run

New in version 1.32.

If you need to globally distinguish one test run from others in your workers, you can use the testrun_uid fixture. For instance, let's say you wanted to create a separate database for each test run:

import pytest
from posix_ipc import Semaphore, O_CREAT

@pytest.fixture(scope="session", autouse=True)
def create_unique_database(testrun_uid):
    """ create a unique database for this particular test run """
    database_url = f"psql://myapp-{testrun_uid}"

    with Semaphore(f"/{testrun_uid}-lock", flags=O_CREAT, initial_value=1):
        if not database_exists(database_url):
            create_database(database_url)

@pytest.fixture()
def db(testrun_uid):
    """ retrieve unique database """
    database_url = f"psql://myapp-{testrun_uid}"
    return database_get_instance(database_url)

Additionally, during a test run, the following environment variable is defined:

• PYTEST_XDIST_TESTRUNUID: the unique id of the test run.

Accessing sys.argv from the controller node in workers

To access the sys.argv passed to the command-line of the controller node, use request.config.workerinput["mainargv"].

Specifying test exec environments in an ini file

You can use pytest's ini file configuration to avoid typing common options. You can for example make running with three subprocesses your default like this:

[pytest]
addopts = -n3

You can also add default environments like this:

[pytest]
addopts = --tx ssh=myhost//python=python3.9 --tx ssh=myhost//python=python3.6

and then just type:

pytest --dist=each

to run tests in each of the environments.

Specifying "rsync" dirs in an ini-file

In a tox.ini or setup.cfg file in your root project directory you may specify directories to include or to exclude in synchronisation:

[pytest]
rsyncdirs = . mypkg helperpkg
rsyncignore = .hg

These directory specifications are relative to the directory where the configuration file was found.

Making session-scoped fixtures execute only once

pytest-xdist is designed so that each worker process will perform its own collection and execute a subset of all tests. This means that tests in different processes requesting a high-level scoped fixture (for example session) will execute the fixture code more than once, which breaks expectations and might be undesired in certain situations.

While pytest-xdist does not have a builtin support for ensuring a session-scoped fixture is executed exactly once, this can be achieved by using a lock file for inter-process communication.

The example below needs to execute the fixture session_data only once (because it is resource intensive, or needs to execute only once to define configuration options, etc), so it makes use of a FileLock to produce the fixture data only once when the first process requests the fixture, while the other processes will then read the data from a file.

Here is the code:

import json

import pytest
from filelock import FileLock


@pytest.fixture(scope="session")
def session_data(tmp_path_factory, worker_id):
    if worker_id == "master":
        # not executing in with multiple workers, just produce the data and let
        # pytest's fixture caching do its job
        return produce_expensive_data()

    # get the temp directory shared by all workers
    root_tmp_dir = tmp_path_factory.getbasetemp().parent

    fn = root_tmp_dir / "data.json"
    with FileLock(str(fn) + ".lock"):
        if fn.is_file():
            data = json.loads(fn.read_text())
        else:
            data = produce_expensive_data()
            fn.write_text(json.dumps(data))
    return data

The example above can also be use in cases a fixture needs to execute exactly once per test session, like initializing a database service and populating initial tables.

This technique might not work for every case, but should be a starting point for many situations where executing a high-scope fixture exactly once is important.

How does xdist work?

xdist works by spawning one or more workers, which are controlled by the controller. Each worker is responsible for performing a full test collection and afterwards running tests as dictated by the controller.

The execution flow is:

1. controller spawns one or more workers at the beginning of the test session. The communication between controller and worker nodes makes use of execnet and its gateways. The actual interpreters executing the code for the workers might be remote or local.
2. Each worker itself is a mini pytest runner. workers at this point perform a full test collection, sending back the collected test-ids back to the controller which does not perform any collection itself.
3. The controller receives the result of the collection from all nodes. At this point the controller performs some sanity check to ensure that all workers collected the same tests (including order), bailing out otherwise. If all is well, it converts the list of test-ids into a list of simple indexes, where each index corresponds to the position of that test in the original collection list. This works because all nodes have the same collection list, and saves bandwidth because the controller can now tell one of the workers to just execute test index 3 index of passing the full test id.
4. If dist-mode is each: the controller just sends the full list of test indexes to each node at this moment.
5. If dist-mode is load: the controller takes around 25% of the tests and sends them one by one to each worker in a round robin fashion. The rest of the tests will be distributed later as workers finish tests (see below).
6. Note that pytest_xdist_make_scheduler hook can be used to implement custom tests distribution logic.
7. workers re-implement pytest_runtestloop: pytest’s default implementation basically loops over all collected items in the session object and executes the pytest_runtest_protocol for each test item, but in xdist workers sit idly waiting for controller to send tests for execution. As tests are received by workers, pytest_runtest_protocol is executed for each test. Here it worth noting an implementation detail: workers always must keep at least one test item on their queue due to how the pytest_runtest_protocol(item, nextitem) hook is defined: in order to pass the nextitem to the hook, the worker must wait for more instructions from controller before executing that remaining test. If it receives more tests, then it can safely call pytest_runtest_protocol because it knows what the nextitem parameter will be. If it receives a “shutdown” signal, then it can execute the hook passing nextitem as None.
8. As tests are started and completed at the workers, the results are sent back to the controller, which then just forwards the results to the appropriate pytest hooks: pytest_runtest_logstart and pytest_runtest_logreport. This way other plugins (for example junitxml) can work normally. The controller (when in dist-mode load) decides to send more tests to a node when a test completes, using some heuristics such as test durations and how many tests each worker still has to run.
9. When the controller has no more pending tests it will send a “shutdown” signal to all workers, which will then run their remaining tests to completion and shut down. At this point the controller will sit waiting for workers to shut down, still processing events such as pytest_runtest_logreport.

FAQ

Question: Why does each worker do its own collection, as opposed to having the controller collect once and distribute from that collection to the workers?

If collection was performed by controller then it would have to serialize collected items to send them through the wire, as workers live in another process. The problem is that test items are not easily (impossible?) to serialize, as they contain references to the test functions, fixture managers, config objects, etc. Even if one manages to serialize it, it seems it would be very hard to get it right and easy to break by any small change in pytest.