You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What part(s) of the page would you like to see updated?
The "Incremental table models" section currently documents insert_overwrite as always falling back to append when no partitions are defined.
dbt-labs/dbt-adapters#2082 proposes distinguishing an omitted strategy from an explicitly configured insert_overwrite for unpartitioned Hive models:
When the strategy is omitted, preserve the existing append fallback.
When insert_overwrite is explicit and dbt creates a new S3 location per run, which is the default, replace the complete table.
When the model reuses an S3 path through external_location or s3_data_naming, preserve append and emit an informational message.
Preserve existing behavior for partitioned Hive models and all other strategies.
The Athena configuration page should document this distinction when the adapter change is released.
Suggested replacement copy
Replace the current insert_overwrite bullet with:
insert_overwrite (default for Hive): For partitioned models, dbt deletes overlapping partitions from the destination table and inserts the new records. For unpartitioned models, an explicitly configured incremental_strategy='insert_overwrite' replaces the complete table when dbt creates a new S3 location per run (s3_data_naming contains unique, which is the default, and external_location is not configured). If the strategy is omitted, dbt preserves the historical append fallback. Models that explicitly request insert_overwrite but reuse the same S3 path also preserve append and emit an informational message.
Contributions
Link to the page on docs.getdbt.com requiring updates
https://docs.getdbt.com/reference/resource-configs/athena-configs
What part(s) of the page would you like to see updated?
The "Incremental table models" section currently documents
insert_overwriteas always falling back toappendwhen no partitions are defined.dbt-labs/dbt-adapters#2082 proposes distinguishing an omitted strategy from an explicitly configured
insert_overwritefor unpartitioned Hive models:appendfallback.insert_overwriteis explicit and dbt creates a new S3 location per run, which is the default, replace the complete table.external_locationors3_data_naming, preserveappendand emit an informational message.The Athena configuration page should document this distinction when the adapter change is released.
Suggested replacement copy
Replace the current
insert_overwritebullet with:insert_overwrite(default for Hive): For partitioned models, dbt deletes overlapping partitions from the destination table and inserts the new records. For unpartitioned models, an explicitly configuredincremental_strategy='insert_overwrite'replaces the complete table when dbt creates a new S3 location per run (s3_data_namingcontainsunique, which is the default, andexternal_locationis not configured). If the strategy is omitted, dbt preserves the historicalappendfallback. Models that explicitly requestinsert_overwritebut reuse the same S3 path also preserveappendand emit an informational message.Additional information
Feature issue: dbt-labs/dbt-adapters#2081
Implementation PR: dbt-labs/dbt-adapters#2082
The dbt-athena README is also updated in the implementation PR.