How to do staging data in data migration? do we need it really?

I am trying to write a framework using spark-sql to migrate data from oracle to cassandra, using datastax-connector. How to handle the errors if there is any connection break-outs? In sqoop we put it in staging tables then we process the data , here in spark-sql do we need the staging tables/area ? if so how to handle/design it ? Please help me ~thank you.