UpdateStateByKey and your batch data could be filling up your executor memory and hence it might be hitting the disk, you can verify it by looking at the memory footprint while your job is running. Looking at the executor logs will also give you a better understanding of whats going on.
Thanks Best Regards On Fri, Dec 4, 2015 at 8:24 AM, Steven Pearson <spearso...@gmail.com> wrote: > I'm running a Spark Streaming job on 1.3.1 which contains an > updateStateByKey. The job works perfectly fine, but at some point (after a > few runs), it starts shuffling to disk no matter how much memory I give the > executors. > > I have tried changing --executor-memory on > spark-submit, spark.shuffle.memoryFraction, spark.storage.memoryFraction, > and spark.storage.unrollFraction. But no matter how I configure these, it > always spills to disk around 2.5GB. > > What is the best way to avoid spilling shuffle to disk? > >
