Source monitoring · rule source-freshness-stale
dbt sources past their freshness threshold
A Metacenta review checks this under the rule Monitored sources are arriving on time. Everything below applies whether or not you ever commission one.
What this rule checks
From the results of your last dbt source freshness run, this rule flags each source that was past its warning or error threshold when it was measured.
Why it matters
Every model built from the source since then was computed on data the project itself classifies as too old to rely on. A breached error threshold is the most severe finding in this dimension.
How to fix it
Fix the ingestion behind the sources breaching their freshness thresholds. Trace the loader behind the source. A freshness breach is an ingestion failure, not a dbt one, so the fix is usually in the pipeline that loads the table.
When it is fine to leave
A source that stops arriving on purpose, such as a retired system, should lose its threshold. Past its warning but not its error threshold, it is an early signal, not yet a breach.
What we need to check it
sources.json, which dbt source freshness writes, alongside manifest.json. Without it the rule declines; it never assumes sources are fresh. The result is only as recent as that run.