Liew Tze Hao Timothy
12/04/2025, 12:21 AMkapa.ai
12/04/2025, 12:22 AMairbyte_raw table and expecting Airbyte’s cursor to follow automatically. ### 1. What Airbyte supports Airbyte provides Refresh Syncs that explicitly reset the cursor for incremental streams, effectively starting them over “from the beginning.”[[Refreshing data](https://docs.airbyte.com/platform/operator-guides/refreshes)] - A Refresh Sync erases the current cursor for the selected incremental stream(s).[[Refreshing data](https://docs.airbyte.com/platform/operator-guides/refreshes)] - It will then re-import all records again for those streams (as long as they are still present in the source).[[Refreshing data](https://docs.airbyte.com/platform/operator-guides/refreshes)] This is not limited to the last two days; it replays the full history of that stream from the source. Separately, Airbyte may also drop and re-create final tables (“soft reset”) as part of typing & deduping, but that is driven by Airbyte itself for schema/refresh reasons, not by you manually deleting raw data.[[Final table re-creation](https://docs.airbyte.com/platform/using-airbyte/core-concepts/typing-deduping#on-final-table-re-creation)] There is a manual procedure documented to force a full re-build of a final table from raw data (drop final table, set _airbyte_loaded_at to NULL, then run a sync), but that assumes you still have the raw rows; it doesn’t support selectively reprocessing just a date range.[[Final table re-creation](https://docs.airbyte.com/platform/using-airbyte/core-concepts/typing-deduping#on-final-table-re-creation)] ### 2. About “partial refresh” for the last 2 days The documentation and forum content do not describe a supported way to: - Delete only the last N days of rows from airbyte_raw, and - Have Airbyte automatically adjust its cursor to that point and re-sync just that range. In fact, one forum answer explicitly notes that partial initial loads (starting from a later point instead of from the beginning) are not supported – the first sync is always full for incremental streams, and the suggested workaround is to adjust the source (e.g., sync from a view with filtered data).[[Partial data fetching](https://discuss.airbyte.io/t/3381)] So: - If you manually delete rows in airbyte_raw, Airbyte’s stored cursor/state will not automatically be reset to two days ago. - The documented “refresh” functionality is all-or-nothing per stream (rewind the cursor entirely), not “only the last X days.”[[Refreshing data](https://docs.airbyte.com/platform/operator-guides/refreshes)] ### 3. Practical options Based on the docs: 1. If you can afford a full re-read of the stream: Use a Refresh Sync for that stream (“refresh and retain” or “refresh and remove” depending on your sync mode and needs). This will reset the cursor and re-import all data from the source, fixing missing rows if they’re still present there.[[Refreshing data](https://docs.airbyte.com/platform/operator-guides/refreshes)] 2. If you truly need partial backfill only (last 2 days): The knowledge sources do not document an official, cursor-aware way to do this on a running connection. The recommended patterns in similar situations are: - Use a separate connection that does a one-time full re-ingestion of a defined period (e.g., last 12 months / last month) and then stop it, while your main incremental connection continues.[[One-time partial resync](https://discuss.airbyte.io/t/1992)] - Or adjust the source-side data (e.g., a view that only exposes the date range to re-ingest) and use full refresh on that auxiliary connection.[[Partial data fetching](https://discuss.airbyte.io/t/3381)] Given what’s in the docs, there is no documented, supported mechanism where simply deleting two days of data from airbyte_raw will cause Airbyte to reset its cursor to that point and backfill exactly those days.