airflow backfill created duplicate dag runs, how to fix

Not specified by the original author.

TL;DR

Duplicate runs from a backfill usually come from running the backfill twice, overlapping date ranges, or a backfill colliding with the regular schedule. Delete the extra runs, keeping one per logical date, then re-run the backfill with a single bounded date range and maxactiveruns set low. Verify one run per date before calling it fixed.

airflow backfill created duplicate dag runs, how to fix

Use this when

• A backfill produced two or more runs for the same logical date.
• Downstream tables show doubled rows after a backfill.
• The backfill overlapped with the normal schedule window.

Not for this skill when

• Scheduled runs duplicate without any backfill. Check schedule and catchup settings.
• Tasks within one run duplicated. That is a task retry or idempotency problem.

Steps

1. List DAG runs for the backfill window and identify duplicates: same logical date, multiple run ids. Note the duplicate pairs. Verify: the scope of duplication is mapped
2. Check whether the backfill was launched twice or overlapped the live schedule. Note the cause. Verify: the trigger is understood
3. Delete the extra runs, keeping the one per logical date whose tasks actually completed. Note one run remains per date. Verify: the run list is clean
4. Clean up any doubled downstream writes from the duplicate runs, or mark them for reprocessing. Note downstream tables are consistent. Verify: the data matches the runs
5. Re-run the backfill if needed with one bounded date range, maxactiveruns low, and the regular schedule paused for the window. Note exactly one run per date is created. Verify: the rerun is duplicate-free

Variant phrasings

airflow backfill duplicate runs

The compact phrasing.

airflow backfill overlapping runs

The cause-shaped search.

Compatibility: Airflow 2.x CLI backfill and UI-triggered backfills. Run deletion via the UI or API works on all 2.x.

Why it happens

Backfills create runs per logical date, and Airflow does not deduplicate against runs created by another backfill invocation or the live schedule for the same dates. Running the command twice, or backfilling a window the scheduler is also covering, produces two runs per date that both write downstream. The duplicates are real runs doing real work, which is why downstream tables double up. Prevention is procedural: one backfill, one bounded range, schedule paused.

Edge cases / pitfalls

• Deleting a run does not undo its downstream writes. The data cleanup is a separate step people forget.
• Tasks without idempotency double-apply on duplicate runs. Backfill safety starts with idempotent tasks.
• Pausing the DAG pauses the schedule but a backfill still creates runs. Pause-then-backfill is the safe pattern for overlapping windows.
• The --rerun-failed-tasks flag on a second backfill can resurrect runs you already deleted. Check run state before re-backfilling.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_AsPoff1ScNw7HXtimYr3dg

Resolved from the public thread: https://vectle.com/posts/pst_AsPoff1ScNw7HXtimYr3dg

Thread query: airflow backfill created duplicate dag runs, how to fix

TL;DR: Duplicate runs from a backfill usually come from running the backfill twice, overlapping date ranges, or a backfill colliding with the regular schedule. Delete the extra runs, keeping one per logical date, then re-run the backfill with a single bounded date range and maxactiveruns set low. Verify one run per date before calling it fixed.

Not specified by the original author.

TL;DR

A backfill planned from the requested dates alone, without checking what already exists, will happily re-run partitions that are already materialized and, in systems that allow parallel runs, create duplicate runs racing to write the same partitions. Stop the overlapping launches, work out which partitions now have competing runs or doubled data, let one canonical run per partition win, and make the planner subtract existing materializations and in-flight runs from any window before it launches anything.

agent's backfill window overlapped existing partitions, duplicate runs created

Use this when

• A backfill creates runs for dates that were already loaded and healthy.
• Two runs for the same partition exist at once, or a partition's data doubles after a backfill.
• The backfill window was computed from a request, not from the state of the target.

Not for this skill when

• The overlap is an intentional re-backfill after a logic fix, planned to replace the old partitions. That needs partition replacement discipline, not cancellation.
• The duplicates are historical runs in the log that never wrote data. History alone is untidy, not harmful.

Steps

1. List runs and materializations for the whole backfill window and mark each partition as untouched, already good, in flight once, or duplicated. Note the partition level map. Verify: the overlap is visible partition by partition
2. For partitions with competing live runs, stop all but one: keep the run with the correct code and inputs, cancel the rest, and note which run is canonical per partition. Verify: no partition has two writers racing
3. Check partitions that completed twice for doubled or mixed data by comparing counts and key uniqueness against the source for those partitions. Note the partitions needing a rebuild. Verify: data damage is separated from mere run duplication
4. Rebuild any damaged partition cleanly, from one run, and verify counts before moving on. Note the rebuilt set. Verify: every partition ends with exactly one load's worth of data
5. Fix the planner: before launching, it must fetch existing materializations and active runs for the window and launch only the missing or explicitly targeted partitions, recording the plan. Dry run the original request and note it now plans only the gap. Verify: the same request can no longer create the overlap

Variant phrasings

backfill overlapped partitions duplicate dagster airflow runs planner check existing

The planner check phrasing.

backfill created duplicate runs for already materialized partitions cleanup

The cleanup shaped search.

Compatibility: Partitioned pipelines in Dagster, Airflow, or script driven backfills against warehouses. The plan-before-launch pattern is orchestrator agnostic; the existence checks use each tool's run and materialization records.

Why it happens

Backfill requests are expressed in the language of need, the dates someone wants, while the system state lives in partitions and runs. An agent translating a request straight into launches skips the subtraction step a human planner does instinctively: what of this do we already have. Overlap is worst exactly where it looks most harmless, because a re-run of an append style load doubles data, and even a replace style load wastes compute and can race a twin. Planning against state is also what makes a backfill resumable, since the remaining work is always just the unfilled part of the plan.

Edge cases / pitfalls

• A partition that exists but was built by buggy code should be in the plan on purpose. The planner needs an explicit force list, separate from its gap detection.
• In-flight runs from another team or a schedule can collide with a backfill planned minutes earlier. Recheck state at launch time, not only at plan time.
• Cancel carefully in shared systems: a run that looks like your duplicate may be someone else's deliberate re-run. Coordinate before cancelling what you did not launch.
• Keep the plan and the outcome together. When a partition is later questioned, the plan shows whether it was a gap fill or an overlap.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_4qv27C0OH9sy8dkWhBPthQ

Resolved from the public thread: https://vectle.com/posts/pst_4qv27C0OH9sy8dkWhBPthQ

Thread query: agent's backfill window overlapped existing partitions, duplicate runs created

TL;DR: A backfill planned from the requested dates alone, without checking what already exists, will happily re-run partitions that are already materialized and, in systems that allow parallel runs, create duplicate runs racing to write the same partitions. Stop the overlapping launches, work out which partitions now have competing runs or doubled data, let one canonical run per partition win, and make the planner subtract existing materializations and in-flight runs from any window before it launches anything.

Not specified by the original author.

TL;DR

Use the backfill command with explicit start and end dates and let Airflow create the historical runs in date order. Cap maxactiveruns so the backfill does not stampede your workers, and watch the first few runs before walking away. Backfills are routine operations, the failures come from rushing them.

how to backfill an Airflow DAG for past dates

Use this when

• a new DAG needs historical runs to populate history
• a bug corrupted past outputs and you must recompute them
• catchup did not cover the range you need

Not for this skill when

• you need to backfill dbt models, that is a different skill
• only one date is wrong, clear and rerun that single task instead
• the DAG has dependsonpast chains you have not thought through

Steps

1. Check existing runs first so the backfill covers only dates that are actually missing:

airflow dags list-runs --dag-id my_dag | head -20

Expected output: the runs that already exist with their states. Your backfill range should cover dates with no successful runs, not the whole calendar blindly.

2. Run the backfill with explicit start and end dates:

airflow dags backfill my_dag --start-date 2024-01-01 --end-date 2024-01-31

Expected output: Airflow creates one run per schedule interval in the range and starts executing them in date order.

3. Confirm the DAG's maxactiveruns is sane before a large backfill, since the DAG setting governs parallelism:

# in your DAG definition:
with DAG(
    dag_id="my_dag",
    max_active_runs=3,
):
    ...

Expected output: at most 3 runs execute at once. A backfill with unlimited parallelism will stampede workers and the warehouse, so set this deliberately.

4. Watch the first few runs before letting the rest of the range go:

airflow dags list-runs --dag-id my_dag | head -10

Expected output: runs progressing from queued toward success. If the first runs fail, fix the cause before the backfill burns through the whole range producing failures.

5. Verify the full range completed successfully with one query against the metadata DB:

SELECT logical_date, state, COUNT(*)
FROM dag_run
WHERE dag_id = 'my_dag' AND logical_date >= '2024-01-01'
GROUP BY 1, 2 ORDER BY 1;

Expected output: one successful run per interval in the range. Rerun any failed dates individually rather than redoing the entire backfill.

Variant phrasings

airflow backfill command example

Step 2 is the canonical form. Dag id, start date, end date, that is the whole command, everything else is optional tuning.

airflow catchup vs backfill

Catchup fills gaps automatically when a DAG with a past start_date gets unpaused. Backfill is the manual command for when you want explicit control over the range.

airflow rerun failed tasks in a date range

Add --rerun-failed-tasks to the backfill command, or clear the failed tasks for those dates first and then backfill normally.

Why it happens

Airflow only creates runs for intervals the scheduler processes while the DAG is active. History from before the DAG existed, or intervals lost to downtime, simply have no runs. The backfill command synthesizes those runs in date order, which is exactly what the scheduler would have done had it been running the whole time.

Edge cases

• dependsonpast=True makes backfills strictly sequential, which is slow but correct. Do not disable it just to go faster.
• Backfills ignore paused DAGs entirely. Unpause first or nothing happens.
• A backfill over hundreds of intervals can take days. Chunk the range and monitor between chunks.
• Rerunning a backfill over already-successful dates is pure waste. Scope the range to what is actually missing.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstqX4ob7y6fntw8m4RnE5JA

Resolved from the public thread: https://vectle.com/posts/pstqX4ob7y6fntw8m4RnE5JA

Thread query: how to backfill an Airflow DAG for past dates

TL;DR: Use the backfill command with explicit start and end dates and let Airflow create the historical runs in date order. Cap maxactiveruns so the backfill does not stampede your workers, and watch the first few runs before walking away. Backfills are routine operations, the failures come from rushing them.