Cleanup and Parallel Runs

TestShed removes what it starts, when the fixture that started it goes out of scope, and again at the end of the session. Each session gets its own network and labels, so parallel runs don't collide.

At a Glance

• Nothing to set up: cleanup and xdist support are always on.
• --shed-container-logs prints container logs when a test fails.

What Gets Removed

• Containers: removed with their anonymous volumes when their fixture ends: after the test for shed with markers, docker_sidecar and shed_deferred; after the module or the session for the wider-scoped fixtures.
• Volume files: the temporary files behind InlineVolume and RemoteVolume.
• The network: removed at the end of the session.

At the end of the session, TestShed also removes any container still carrying the session's label, so a test that crashed mid-setup doesn't leave one behind. Containers you attach with shed_attach are never removed.

Every container carries two labels, for finding them while a session runs:

Label	Value
com.kloudkit.testshed	The session, such as testshed-app-4821.
com.kloudkit.testshed.test	The test that started a shed container (its pytest node ID, or shed_default).

docker ps --filter label=com.kloudkit.testshed

Pressing Ctrl-C stops the session and still runs the cleanup; it can take a moment.

Parallel Runs

pytest -n auto (pytest-xdist) works without extra setup:

pytest -n auto

The network name and labels include the xdist worker (gw0, gw1, …), so workers don't share containers, and each worker cleans up its own.

Container Logs on Failure

--shed-container-logs prints a container's logs when a test fails because of it: a command that failed, or a probe that timed out.

pytest --shed-container-logs

Next Steps

• Command-Line Options: every flag and its default.
• Readiness Probes: fail fast when a container isn't ready.