test(dotnet): enable agentless EVP exposure egress

[leoromanovsky]

Motivation

We are shipping agentless CDN-delivered configuration to make customer deployments simpler. The SDK's core Feature Flags telemetry still needs a reliable network path out of the process: automatically use a compatible running Agent or serverless-init relay when available, and fall back to direct EVP intake when no relay exists.

FFLSDK-195 activates .NET against the shared cross-language contract in #7299 so successful construction cannot mask missing configuration polling or exposure delivery.

Changes

Capability	Direct intake	serverless-init	Agent
Exposures	Enabled; exact candidate currently fails	Enabled; exact candidate currently fails	Existing coverage
Flag evaluations	Not emitted yet	Not emitted yet	Not emitted yet
Shutdown exposure flush	Enabled; blocked before flush	—	—

• Activate the poc weblog for the .NET 3.54 development line.
• Make the .NET runtime PID 1 only for the direct shutdown contract so Docker's SIGTERM reaches the ordinary application lifecycle.
• Emit the shared structured server-close marker after ASP.NET Core reports ApplicationStopped.

Decisions

• Target test(ffe): add direct agentless EVP contract #7299 directly; do not depend on another language activation.
• Keep this draft as an executable product gate. Do not weaken or xfail the contracts: the exact SDK artifact exposes a real missing production dependency.
• Scope activation to poc. The uds weblog configures an Agent socket at runtime and therefore cannot prove a no-Agent direct topology.
• Keep this PR exposure-only. The separate flag-evaluation emitter remains owned by FFLSDK-28.
• Keep the PID 1 and shutdown-marker behavior test-only and opt-in through SYSTEM_TESTS_FFE_SHUTDOWN_FLUSH_ENABLED; all other .NET scenarios retain the existing crash-dump recovery wrapper.
• Require dd-trace-dotnet#9044 to land or be combined with #9235: #9235 constructs the fallback transport, but current master has no production caller that starts AgentlessConfigurationSource.
• Keep OTLP out of scope: these contracts validate EVP egress.

Validation

System-tests base: 9c753fbcbdcc6122eed15fd5af827671fbc0771c

System-tests candidate: ccf370de4c3e4dc62aaf09977d2727d063449b05

.NET SDK candidate: 0ac1c3cca2c7423e41bd15b17ab2f543a9c261b3 from dd-trace-dotnet#9235

Candidate artifact: datadog-dotnet-apm-3.54.0.tar.gz, SHA-256 4ad48d2d6890682c4da6a86bd02555f22b93390b425a739177b26c88de3f9c3f, from Azure build 208990.

./format.sh
passed

./run.sh TEST_THE_TEST \
  tests/test_the_test/test_dotnet_direct_evp_activation.py \
  tests/test_the_test/test_manifest.py
34 passed

./build.sh dotnet -w poc
passed
Library: dotnet@3.54.0
Weblog variant: poc

Exact-artifact E2E currently fails before event routing can be exercised:

Direct exposure only:
Timed out waiting for Feature Flags EVP captures: /api/v2/exposures

serverless-init exposure only:
Timed out waiting for Feature Flags EVP captures: /api/v2/exposures

Diagnostic evidence:

• All five /ffe evaluations returned {"reason":"DEFAULT","value":null,"count":1}.
• The Agentless UFC mock recorded requests_total: 0; polling never started.
• Direct intake captured 19 APM telemetry requests and zero exposure requests.
• The runtime topology was correct: source agentless, API key present, no Agent variables, and dotnet app.dll was PID 1.
• FeatureFlagsModule constructs FeatureFlagsEvpTransport and ExposureApi, but subscribes only to RCM. AgentlessConfigurationSource.Create(...).Start() has no production caller on the exact candidate's base.

This is a product-stack dependency failure, not a system-test teardown failure. Rebuild the exact combined #9044 + #9235 candidate and rerun direct exposure/shutdown and serverless-init exposure before calling this activation green.

Both system-test commits are GitHub verified=true, reason=valid.

Reviewer checklist

• Anything but tests/ or manifests/ is modified: request R&P review for the test-only .NET lifecycle hook.
• No Docker base image is modified.
• No scenario is added, removed, or renamed.

[github-actions]

CODEOWNERS have been resolved as:

tests/test_the_test/test_dotnet_direct_evp_activation.py                @DataDog/system-tests-reviewers
manifests/dotnet.yml                                                    @DataDog/system-tests-reviewers
utils/build/docker/dotnet/weblog/Program.cs                             @DataDog/system-tests-reviewers
utils/build/docker/dotnet/weblog/app.sh                                 @DataDog/system-tests-reviewers

[datadog-prod-us1-3]

 

⚠️ Warnings

❌ Your PR has failed checks. Please review the issues below and take necessary action before merging.

🚦 24 Pipeline jobs failed

Testing the test | System Tests (dotnet, dev) / End-to-end #13 / poc 13 — ❌ 1 test failed · 🔧 Needs a code fix, caused by this PR

View more details · View in GitHub Actions

❌ pytest.internal[poc] from system_tests_suite

internal error

Traceback (most recent call last):
  File "/home/runner/work/system-tests/system-tests/venv/lib/python3.12/site-packages/_pytest/main.py", line 268, in wrap_session
    session.exitstatus = doit(config, session) or 0
                         ^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/system-tests/system-tests/venv/lib/python3.12/site-packages/_pytest/main.py", line 321, in _main
    config.hook.pytest_collection(session=session)
  File "/home/runner/work/system-tests/system-tests/venv/lib/python3.12/site-packages/pluggy/_hooks.py", line 265, in __call__
    return self._hookexec(self.name, self.get_hookimpls(), kwargs, firstresult)
...

DataDog/system-tests | Amazon_Linux_2_arm64.CON: [test-app-java-container]

View more details · View in GitLab

DataDog/system-tests | Amazon_Linux_2_arm64.HOS: [test-app-java]

View more details · View in GitLab

View all 24 failed jobs.

ℹ️ Info

No other issues found (see more)

❄️ No new flaky tests detected

Useful? React with 👍 / 👎

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: ccf370d | Docs | View more details | Give us feedback!}