Thanks Sibani, I will try with .pyMaxAllowedFailRecords this approach. Our requirement is to process nearly 10 million records by batch. Any data issue occurs we can't run the 10 million records again. So we need to continue our processing except those failed records. At the same time if more number of records (above threshold) failed may be our input system sends wrong data we need to stop batch processing and run again entire 10 million records.