@IsmailKeskin , --conf spark.hadoop.fs.hdfs.impl.disable.cache=true should help, if not, please also try to add --conf mapreduce.job.complete.cancel.delegation.tokens=false See if it helps. Cheers Eric ... View more