Execution health · rule test-failing
dbt tests that are failing
A Metacenta review checks this under the rule Declared tests pass. Everything below applies whether or not you ever commission one.
What this rule checks
From the results of your last dbt run or build, this rule flags every data test that failed or errored. It names how many rows failed, where dbt recorded that.
Why it matters
The project asserts something is true of its data, and it is not. Anything downstream is being served from data the project itself says is wrong. A test left failing also teaches the team to ignore the suite.
How to fix it
Fix or retire the tests that are currently failing. Fix the data, or retire the assertion if it no longer describes what the model should be. Do not leave it failing: a permanently red test hides the next real failure.
When it is fine to leave
A test deliberately set to severity: warn is a signal, not a breach, and dbt records it as a warning rather than a failure. A test that fails on a known, dated issue with a fix in progress is a decision, if it is written down.
What we need to check it
run_results.json from a dbt test or dbt build, alongside manifest.json. When the run reached no verdict on any test, the rule declines rather than report a pass.