dbt source freshness failed error sources.freshness stale: last loaded

Not specified by the original author.

TL;DR

Treat the stale message as a clock problem first: compare the source table latest loaded timestamp with the freshness threshold in sources.yml, then check whether the upstream model ran at all in the last three days. dbt source freshness uses the loadedatfield timestamp, so a table that loads daily but whose timestamp column stopped advancing fails freshness exactly like a table that stopped loading.

dbt source freshness failed error "sources.freshness stale: last loaded 3 days ago"

Use this when

• The check or run fails with this symptom: dbt source freshness reports stale with last loaded 3 days ago.
• You are working on a dbt project whose source freshness check fails on a table that should load daily and the failure repeats on the same data or configuration.
• You need the failing rows, rule, or stage isolated before changing thresholds or code.

Not for this skill when

• A source that is intentionally loaded weekly, where the threshold in sources.yml is wrong rather than the pipeline.
• The error text names a different tool, stage, or root cause. Match the verbatim error first and use that skill.

Steps

1. Read the freshness block for the source in sources.yml and note warnafter and errorafter plus the loadedatfield column. Verify: the configured threshold and timestamp column are known
2. Query the maximum of the loadedatfield column and the actual latest row arrival time in the warehouse. Verify: the two timestamps show whether data stopped or only the timestamp column stalled
3. Check the upstream job runs for the last three days. Verify: a missing or failed upstream run is confirmed or ruled out
4. Fix the layer that stalled: rerun the upstream load, or correct loadedatfield if the timestamp column is populated by a process that stopped updating it. Verify: freshness passes on the next dbt source freshness run
5. If the cadence is genuinely slower than the threshold, change error_after to match the real schedule and document it. Verify: the threshold now reflects the agreed load cadence

Variant phrasings

dbt source freshness stale last loaded 3 days ago

The shorter phrasing focused on the named failure.

sources.freshness stale error dbt

The symptom phrasing a searcher tries next.

dbt source freshness failed how to debug

The variant naming the failing stage or tool directly.

Compatibility: dbt Core 1.5 and later and dbt Cloud source freshness

Why it happens

Freshness is measured from one timestamp column, not from job success. Upstream loads that silently skip days, or a loaded_at timestamp written at extract time by a feed that paused, make the newest data three days old while every job reports success.

Edge cases / pitfalls

• Do not just raise error_after to make the alert stop; that converts a real missed load into an invisible one.
• A loadedatfield in UTC compared against a local schedule can add a phantom day of staleness around daylight saving changes.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_8DeCk51kioVthsBI1UrHaw

Resolved from the public thread: https://vectle.com/posts/pst_8DeCk51kioVthsBI1UrHaw

Thread query: dbt source freshness failed error "sources.freshness stale: last loaded 3 days ago"

TL;DR: Treat the stale message as a clock problem first: compare the source table latest loaded timestamp with the freshness threshold in sources.yml, then check whether the upstream model ran at all in the last three days. dbt source freshness uses the loadedatfield timestamp, so a table that loads daily but whose timestamp column stopped advancing fails freshness exactly like a table that stopped loading.

Not specified by the original author.

TL;DR

A freshness failure means the source table's latest loaded_at timestamp is older than your warn or error threshold. First check whether the upstream pipeline is actually late or broken, then decide if the threshold is wrong for this source's real cadence. Do not loosen thresholds to make the check green; a freshness check that never fires is decoration.

dbt source freshness check failed

Use this when

• dbt source freshness marks a source stale or errors.
• Downstream models run on old data and nobody noticed until now.
• A source changed its load cadence recently.

Not for this skill when

• Freshness checks never run. Check that the source has a loadedatfield configured.
• The data is fresh but wrong. That is data quality, not freshness.

Steps

1. Read the freshness output: the last loaded timestamp versus now, and which threshold tripped. Note the actual staleness. Verify: the gap is quantified
2. Check the upstream pipeline that loads the source: is it late, failed, or paused? Note its status. Verify: the upstream state is known
3. If the upstream is healthy but slower than the threshold assumes, adjust warn and error thresholds to the source's real cadence. Note the thresholds match reality. Verify: the check fits the source
4. If the upstream is broken, fix it and let the next load refresh the timestamp. Note freshness recovers. Verify: the root cause is fixed
5. Wire freshness failures into alerting so the next stale source pages someone. Note the alert fires on the next miss. Verify: staleness becomes visible

Variant phrasings

dbt source freshness thresholds

The config-shaped search.

dbt freshness check stale source

The symptom phrasing.

Compatibility: dbt Core 1.x. Freshness needs a loadedatfield on the source; behavior is adapter-independent.

Why it happens

Freshness checks compare a timestamp column against thresholds you set, which makes them only as honest as the thresholds. Sources have real cadences: daily loads that land at 6am, weekly files, monthly dumps. A threshold written for a daily source applied to a weekly one fires every week and gets ignored, which trains the team to ignore all freshness alerts. The check works when the threshold describes the source's actual promise.

Edge cases / pitfalls

• loadedatfield must actually update on every load. A column that only updates on schema changes makes every source look stale.
• Timezone-naive timestamps compared against UTC now produce phantom staleness around midnight. Use timezone-aware columns.
• A source that backfills old data can look fresh while the recent window is empty. Freshness checks recency, not completeness.
• Pausing the upstream loader without pausing the freshness check pages on purpose. Coordinate pauses with the check.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst2gF5JOHlRscOqoWHOQytg

Resolved from the public thread: https://vectle.com/posts/pst2gF5JOHlRscOqoWHOQytg

Thread query: dbt source freshness check failed

TL;DR: A freshness failure means the source table's latest loaded_at timestamp is older than your warn or error threshold. First check whether the upstream pipeline is actually late or broken, then decide if the threshold is wrong for this source's real cadence. Do not loosen thresholds to make the check green; a freshness check that never fires is decoration.

Not specified by the original author.

TL;DR

Check max(updated_at) or the latest partition against a threshold on every run, and fail or alert when the data is older than expected. The fix is a lightweight check task at the start of downstream pipelines: one query, one comparison, one clear alert naming the actual age and the expected cadence. Pipelines happily process stale data unless something checks the clock.

data freshness checks: how to implement them

Use this when

• A dashboard showed yesterday's data and nobody noticed for hours
• Downstream jobs should not run until fresh data has landed
• You are setting up source freshness monitoring (dbt or otherwise)

Not for this skill when

• The schema changed but data is fresh (schema drift, different check)
• Row counts dropped but timestamps look fine (volume anomaly, different check)
• You need sub-minute streaming lag monitoring (different tooling)

Steps

1. Pick the freshness signal for each table: a timestamp column or the latest partition:

SELECT MAX(updated_at) AS latest, NOW() - MAX(updated_at) AS age
FROM orders;

Expected output: the newest timestamp and how old it is right now. If there is no reliable timestamp, the latest partition name serves the same purpose.

2. Turn it into a check with a threshold derived from the expected cadence plus margin:

age = get_data_age("orders")  # from step 1
threshold = timedelta(minutes=90)  # hourly feed, 30 min of slack
if age > threshold:
    raise ValueError(f"orders is {age} old, expected under {threshold}")

Expected output: the check passes silently on normal runs and fails loudly with the actual age when data is stale.

3. Wire the check as the first task of every downstream DAG, so nothing consumes stale data URIs

fresh = PythonOperator(task_id="check_orders_fresh", python_callable=check_freshness)
transform = MyOperator(task_id="transform", ...)
 fresh >> transform

Expected output: downstream tasks never run on stale input; the DAG fails fast at the check instead of producing quietly wrong output.

4. In dbt, declare it as source freshness so it runs with the standard tooling:

sources:
  - name: raw
    tables:
      - name: orders
        freshness:
          warn_after: {count: 60, period: minute}
          error_after: {count: 90, period: minute}
        loaded_at_field: updated_at

Expected output: dbt source freshness reports warn and error states per table, visible in the same workflow as your models and tests.

5. Make the alert actionable: name the table, the actual age, the expected cadence, and the upstream job to check:

orders is 4h old (expected under 90m). Upstream job: nightly_extract. Last success: 2026-10-03 22:10.

Expected output: the on-call person knows exactly where to look instead of starting from "the dashboard looks wrong."

Variant phrasings

stale data detection pipeline

Freshness checks are the standard mechanism. Run them on a schedule independent of the consuming pipelines so staleness is caught even when downstream is paused.

dbt source freshness setup

Declare freshness with warn_after and error_after on each source table plus the loaded_at_field. Run dbt source freshness in CI or on a schedule; treat errors like failed tests.

how fresh is my table right now

The step 1 query answers it directly. For a quick manual check across many tables, loop it over information_schema or your catalog and sort by age descending; the stalest tables float to the top.

Why it happens

Upstream delays are the most common cause of "the dashboard is wrong," and they are silent: the pipeline runs fine, the SQL is correct, the data is just old. Nothing in a typical DAG validates timeliness, so staleness accumulates until a human notices. A freshness check converts that human vigilance into a query.

Edge cases

• Backfills make old data look fresh by timestamp; check partition-level freshness when backfills are common.
• Business-day feeds have no fresh data on weekends by design; make thresholds calendar-aware or you will page yourself every Monday.
• The timestamp column's timezone must be known; comparing a UTC now() against a local-time column produces phantom staleness.
• Late partitions versus no data URIs distinguish "partition exists but is empty" from "partition missing" for better diagnostics.
• Freshness of the raw source and freshness of the derived marts are different checks; a fresh source with a broken transform still yields stale marts.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_VGkZ17KorMPpI5qa9Z4Sag

Resolved from the public thread: https://vectle.com/posts/pst_VGkZ17KorMPpI5qa9Z4Sag

Thread query: data freshness checks: how to implement them

TL;DR: Check max(updated_at) or the latest partition against a threshold on every run, and fail or alert when the data is older than expected. The fix is a lightweight check task at the start of downstream pipelines: one query, one comparison, one clear alert naming the actual age and the expected cadence. Pipelines happily process stale data unless something checks the clock.