> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mage.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Backfilling pipelines

> Run a pipeline multiple times.

A backfill creates 1 or more pipeline runs for a pipeline. There are 2 types of backfills:

1. Date and time window
2. Custom code

## Date and time window

Create 1 or more pipeline runs between 2 datetime values.

The datetime of each instance is used as the `execution_date` for the pipeline run.

For example, if the backfill has the following attributes:

| Attribute      | Value                 |
| -------------- | --------------------- |
| Start datetime | `2023-01-01T03:00:00` |
| End datetime   | `2023-01-05T03:00:00` |
| Interval type  | `day`                 |
| Interval units | `2`                   |

Then the following pipeline runs will be created:

| `id` | `execution_date`      | `ds`         | `hr` |
| ---- | --------------------- | ------------ | ---- |
| 1    | `2023-01-01T03:00:00` | `2023-01-01` | `03` |
| 2    | `2023-01-03T03:00:00` | `2023-01-03` | `03` |

***

## Custom code

The output of a backfill code will be used to generate the pipeline runs.

For example, if the backfill code has the following content:

```python theme={"system"}
backfill_data = []

for index, _ in range(3):
    backfill_data.append(dict(
        partition=index,
        power=5,
    ))

backfill_data
```

Then the following pipeline runs will be created:

| `id` | `execution_date`      | `ds`         | `hr` | `partition` | `power` |
| ---- | --------------------- | ------------ | ---- | ----------- | ------- |
| 1    | `2023-01-01T00:00:00` | `2023-01-01` | `00` | `0`         | `5`     |
| 2    | `2023-01-01T00:00:00` | `2023-01-01` | `00` | `1`         | `5`     |
| 3    | `2023-01-01T00:00:00` | `2023-01-01` | `00` | `2`         | `5`     |

***

## Extra runtime variables from pipeline run

If your pipeline run belongs to a trigger that is scheduled, then the following extra
variables are available in your Python block’s keyword arguments (e.g. `kwargs`).

| Key                                | Description                                                                       | Example                                      |
| ---------------------------------- | --------------------------------------------------------------------------------- | -------------------------------------------- |
| `interval_start_datetime`          | The `datetime` when the pipeline run is scheduled for.                            | `datetime.datetime(2023, 7, 23, 7, 0, 0, 0)` |
| `interval_end_datetime`            | The `datetime` when the next pipeline run is scheduled for.                       | `datetime.datetime(2023, 7, 24, 7, 0, 0, 0)` |
| `interval_seconds`                 | The number of seconds between the current pipeline run and the next pipeline run. | `86400`                                      |
| `interval_start_datetime_previous` | The `datetime` when the previous pipeline run was scheduled for.                  | `datetime.datetime(2023, 7, 22, 7, 0, 0, 0)` |

### SQL block

If you’re using a SQL block, here is an example of how you can access these variables:

```sql theme={"system"}
SELECT
    '{{ interval_start_datetime }}' AS interval_start_datetime
    , '{{ interval_end_datetime }}' AS interval_end_datetime
    , '{{ interval_seconds }}' AS interval_seconds
    , '{{ interval_start_datetime_previous }}' AS interval_start_datetime_previous
```