Hi @DavidOldelius ,

If the initial snapshot has been fully ingested, then in case of ingestion gateway restart the connector will start from where it left and not from the beginning.

Do you observe the behaviour you described during or after the snapshot being processed? You can check this in the gateway event logs. Look for origin.flow_name: {catalog}.{schema}.{table}_snapshot_flow for initial snapshot flows.

If ingestion gateway restart happens during the initial snapshot processing, a full snapshot refresh could be required if the table could not be chunked. For example, if a SQL Server table has no Primary Key, Unique Key, or index, there is little to no chance it could be split into chunks. Also, if the primary key values are skewed, this can also lead to a full refresh of a large part of the table.

I recommend to check the structure of the SQL Server tables to improve chunking. Otherwise, please contact Databricks support to work with you to improve performance.

Hope it helps.

Best regards,