Skip to content

Remove stale live_test_only markers from azure-ai-ml e2e tests - #48494

Draft
Chakradhar886 with Copilot wants to merge 4 commits into
mainfrom
copilot/re-record-enable-azure-ai-ml-tests
Draft

Remove stale live_test_only markers from azure-ai-ml e2e tests#48494
Chakradhar886 with Copilot wants to merge 4 commits into
mainfrom
copilot/re-record-enable-azure-ai-ml-tests

Conversation

Copilot AI commented Aug 7, 2026

Copy link
Copy Markdown
Contributor

Several azure-ai-ml e2e tests were marked live_test_only citing outdated sanitizer/recording issues, excluding them from playback CI even though valid recordings already exist for them.

Changes

  • Removed 11 stale live_test_only markers across e2e test files, re-enabling these tests in playback CI:
    • tests/component/e2etests/test_component.py::test_load_component_from_flow_in_registry
    • tests/internal/e2etests/test_pipeline_job.py::test_pipeline_with_setting_node_output_mode
    • tests/connection/e2etests/test_connections.py::test_secret_population, test_workspace_connections_create_update_and_delete_snowflake_user_pwd, test_container_registry_managed_id_crud
    • tests/datastore/e2etests/test_datastore.py::test_blob_store, test_blob_store_credential_less, test_file_store, test_credential_less_adls_gen2_store
    • tests/environment/e2etests/test_environment.py::test_environment_create_or_update_docker_context, test_registry_environment_create_conda_and_get
  • Left 2 markers untouched in tests/test_datastore_operations_gaps.py — these are genuinely live-only (depend on compute-instance polling and a fuse subprocess), not stale sanitizer issues.
  • Left unrelated skip/skipif markers as-is elsewhere in these files (e.g. "Disable until preview release", "Backend behavior has shifted") since they reflect real, non-stale constraints.

Verification

Confirmed recordings for all 11 re-enabled tests exist in the azure-sdk-assets repo at the package's pinned tag, with substantive interaction counts (4–35 entries each), consistent with these tests having already been re-recorded after the sanitizer changes that originally motivated the markers.

Playback CI should be run on this branch to confirm the recordings match current request/response shapes; any residual sanitizer mismatch would need a targeted re-record for that specific test.

Co-authored-by: Chakradhar886 <259224138+Chakradhar886@users.noreply.github.com>
@azure-pipelines

Copy link
Copy Markdown
Azure Pipelines:
Successfully started running 1 pipeline(s).
10 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

- Point assets.json at new recording tag python/ml/azure-ai-ml_2b6826af41

- Fix malformed sanitized ML_SUBSCRIPTION_ID (missing trailing zeros)

- Make datastore playback assertions sanitizer-aware and reconstruct account-key credentials for the update path

- Switch registry component test to a supported command component (prompt-flow FlowDefinitionDataUri is rejected by the service)
…ion ID

Previous commit changed sanitized_environment_variables['ML_SUBSCRIPTION_ID'] to the canonical 12-zero GUID form, but 700+ pre-existing recording assets in the tagged bundle use the 9-zero form. That mismatch caused hundreds of playback URI-mismatch failures across the ml test matrix.

- Revert the ML_SUBSCRIPTION_ID fixture back to the 9-zero form used by every historical recording.

- Republish assets tag python/ml/azure-ai-ml_111e4464ce with the 10 newly-recorded fixtures normalized to the 9-zero form so they match the reverted fixture.

- Re-add live_test_only to test_pipeline_with_setting_node_output_mode: its recording relies on an anonymous component hash that cannot be reconstructed with a playback-only edit; it will need an actual live re-record.

The other 10 previously-gated tests remain enabled and pass in playback against the new tag.
Rewrite anonymous-component version hashes in the recording to match the client's deterministic playback hashes (train/score/eval), and remove the live_test_only marker.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants