Fix integration-test hang and stale proxy.conf directive

[pawelchcki]

Three fixes that together unwedge the integration suite and make it CI-visible:

• AioHTTPServer / TestAgent cleanup: make them context managers, daemonize their background threads, bound the stop-time join, fix _app.cleanup() → runner.cleanup(), and set a TCPSite shutdown_timeout. Without these, a failed assertion in test_http_proxy kept non-daemon threads alive and wedged the process in __futex_wait until the container was SIGKILL'd.
• proxy.conf: drop the orphan <Location "/delegate"> block left behind by 470db65 (deprecate sampling delegation); it broke apachectl with "Invalid command" for every proxy test.
• CI: add test-integration:$ARCH jobs running the same run-integration-tests.sh entrypoint as make test-integration, scoped to the matching devcontainer-image matrix slot. The existing test-rum job only covered requires_rum tests — the rest of the suite was never run in CI.

[pawelchcki]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• Rewrite readiness-probe comment to be bug-agnostic #64
• Fix integration-test hang and stale proxy.conf directive #63 👈 (View in Graphite)
• Add devcontainer workflow for local builds and integration tests #60
• Respect child-scope override for tracing directives #62
• Switch inject-browser-sdk to submodule; consolidate deps #61
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[Copilot]

Pull request overview

Fixes integration-test shutdown hangs and stale Apache proxy configuration by improving teardown behavior, reducing noisy readiness probes, and ensuring CI runs the full integration suite.

Changes:

• Add context-manager support and safer thread settings for the temporary aiohttp server used in test_http_proxy.
• Fix TestAgent teardown to properly close aiohttp runner/sites, bound shutdown time, and avoid readiness HTTP probes that create traces.
• Remove orphan /delegate Apache config and add CI jobs to run the full integration-test suite.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
test/integration-test/scenarios/test_proxy.py	Add context-manager lifecycle to prevent test hangs from leaked background threads.
test/integration-test/scenarios/conf/proxy.conf	Remove stale <Location "/delegate"> block that breaks apache config validation.
test/integration-test/conftest.py	Use TCP readiness probe, fix aiohttp cleanup, bound shutdown, and improve thread behavior for the test agent.
.gitlab-ci.yml	Add CI jobs to run the full integration-test suite using the same entrypoint as local devs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

AioHTTPServer.internal_run() still calls app.cleanup() and never calls runner.cleanup() / site.stop(). That can leave the listening socket open and may prevent the thread from exiting cleanly (same teardown issue fixed for TestAgent in conftest.py). Consider switching to runner.cleanup() (and/or explicitly stopping the site) before closing the loop so the context-manager shutdown reliably releases the TCP listener.

[Copilot]

stop() joins the server thread with a timeout but then silently proceeds even if the thread is still alive. That can mask teardown failures and leave a background aiohttp server bound to the port, making later tests flaky. After join(timeout=…), consider checking is_alive() and surfacing a clear failure (or at least logging a warning) so a stuck shutdown doesn’t go unnoticed.

Suggested change

	self._thread.join(timeout=3.0)
	self._thread.join(timeout=3.0)
	if self._thread.is_alive():
	raise RuntimeError(
	f"AioHTTPServer failed to stop within 3.0 seconds for "
	f"{self._host}:{self._port}; the server thread is still running."
	)

[Copilot]

TestAgent.stop() now uses join(timeout=3.0) but doesn’t check whether the thread actually stopped. If the aiohttp loop is stuck (the scenario this change is trying to mitigate), the test run will continue while the agent thread is still running, which can lead to port-binding flakiness or a noisy shutdown. Consider checking self._thread.is_alive() after the timeout and surfacing a clear error/warning, and optionally guard call_soon_threadsafe() in case the loop has already been closed due to an internal_run error.

[datadog-prod-us1-4]

🎯 Code Coverage (details)
• Patch Coverage: 100.00%
• Overall Coverage: 50.93% (+0.00%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: c5d42da | Docs | Datadog PR Page | Give us feedback!}

[xlamorlette-datadog]

Thanks a lot for fixing these tests and adding the missing tests in the CI, this is great!

[xlamorlette-datadog]

This can be factorized:

test-integration:
  extends: .test-integration-template
  parallel:
    matrix:
      - ARCH: amd64
        TOOLCHAIN_ARCH: x86_64
      - ARCH: arm64
        TOOLCHAIN_ARCH: aarch64
  needs:
    - job: devcontainer-image
      parallel:
        matrix:
          - ARCH: $ARCH
            TOOLCHAIN_ARCH: $TOOLCHAIN_ARCH
      artifacts: true
    - job: "build:$ARCH"
  tags: ["arch:$ARCH"]

Note that this needs to update the needs in .upload-template (to be tested):

• test-integration:amd64 → test-integration:[amd64]
• test-integration:arm64 → test-integration:[arm64]

Similarly, we should try to factorize build:amd64 / build:arm64 and test-rum:amd64 / test-rum:arm64.