Repository navigation
Fix AWS Spark Python task acceptance test access mode - #7187
shreyas-goenka wants to merge 3 commits into
Conversation
|
Commit: Regression test reportSelected 1 changed tests (limit=20, 0 not selected) 2 tests: regression=0 unreleased=0 coverage=2 failing=0
|
Integration test reportCommit: 87e25b1
Top 15 slowest tests (at least 2 minutes):
|
denik
left a comment
There was a problem hiding this comment.
Should we change the config? it seems like a backend issue. I have ES ticket with jobs team. We can use known failures for now.
| node_type_id: $NODE_TYPE_ID | ||
| instance_pool_id: $TEST_INSTANCE_POOL_ID | ||
| num_workers: 2 | ||
| data_security_mode: USER_ISOLATION |
There was a problem hiding this comment.
Nothing here says why the mode is pinned: the AWS test account enabled admin protection for no-isolation shared clusters, and the test that actually needs this is run/spark_python_task, which copies this template; deploy/simple itself never runs the job. Without that, a later trim of the "simple" fixture or a switch to $DATA_SECURITY_MODE like the library fixtures use would quietly bring the AWS failure back. Can you add a one-line comment above it with the reason and the dependent test?
The AWS test account enabled admin protection for no-isolation shared clusters on October 22 at 14:43 UTC. The CI principal is a workspace admin, and the test left cluster access mode unspecified. Workspace file uploads and API reads succeed, but this protection prevents the driver from reading the Python file. The original test passed at 12:38 UTC and failed by 19:09 UTC that day; an account audit event confirms the setting enablement between those runs.
Set USER_ISOLATION (standard access mode) in the existing shared cluster template. This restores the task's workspace-file access while retaining the security protection. The final diff is one added line.
The restriction is documented: https://docs.databricks.com/aws/en/admin/account-settings/no-isolation-shared
Validation: