Hi @mrTao    It is not good idea to use entire memory Cluster YARN memory.  You can tune the memory from spark side by adjusting the memory using following parameters: --conf spark.executor.instances=5 --conf spark.driver.memory=10g --conf spark.driver.memoryOverhead=1g --conf spark.executor.memory=10g --conf spark.executor.memoryOverhead=1g With the above memory configuration, YARN will allocate 66gb(Executor memory 11gb * Executor Instances(5) + Driver memory 11gb = 55g + 11g = 66g).   Better check your spark-submit once again and tune the above parameters according to your requirement.  ... View more