By chance do you know where in the Spark job this was happening for you? This really should be initialized (by Spark) and I suspect that we have process that is not initializing the pool before use. Reviewing the code, I note that we have a reduceByKeythat didn't have the initialization. Do you know if this is being spit out during the "message pass" phase?
--- I believe this is seen during the phase prior to message pass since we see it occur starting roughly 30 seconds into the read from the Datastore. Job starts at 10:24~ WARN 2016-03-11 10:24:12,658 Cluster.java:2105 - com.datastax.driver.core.Cluster: You listed /10.200.241.65:9042 in your contact points, but it wasn't found in the control host's system.peers at startup WARN 2016-03-11 10:25:32,278 HadoopPools.java:55 - org.apache.tinkerpop.gremlin.hadoop.structure.HadoopGraph: The HadoopPools has not been initialized, using the default pool Then the message then repeats for a significant period of time (perhaps right until the message passing phase occurs)
