As a ETL process, we always do data deleting before incremental load to avoid data duplication. e.g. The previous loading contains error records or incomplete. if we load data for 2014-01-16, we may first delete from table where date='2014-01-16', then load the data again.
If the partition can be create as demand, we can partition data by date and drop special partition. But InfiniDB partition column automatically and can't be controlled. Data from same day may divide to 2-3 partition or some section in one partition and others in another partition. So it can't be deleted by dropping partitions.
Is there any efficient way to solve such problem. Thank you!