I'm trying to convert a spark batch application to a streaming application and wondering what function (or design pattern) I should use to execute a series of operations inside the driver upon arrival of each message (a text file inside an HDFS folder) before starting computation inside executors.
Thanks, Mohammad
