How does one go about deploying a Spark streaming job on a CDH cluster? Does the job need to be deployed as a CSD parcel?
CSD and parcel are two different things. CSDs manage third party processes, parcels is a mechanism to distrubte the software. You could use a parcel to deploy your Spark code to all workers, but this would be overkill. Instead you can package a jar if using java or spark or submit a python file from the command line.