There have been a number of threads on this list about needing to set spark.akka.frameSize to something higher than the default. The issue seems to come up most when one key in a groupByKey has particularly large amounts of data.
What is the downside to setting this configuration parameter to the maximum value by default? Andrew
