I use CDH 6.3.2 。 hive 2.1  hadoop 3.0 hive on spark 。yarn cluster 。   hive.merge.sparkfiles=true ; hive.merge.orcfile.stripe.level=true ; This configuration makes the 1099 reduce file result  merge into one file when the result is small 。Then the merged file has about 1099 stripes in one file 。 Then the result is so slow when it is read.   I tried  hive.merge.orcfile.stripe.level=false ; The result is desirable 。One small file with one stripe and read fast 。   Can anyone tell the difference between  true and false ？ Why " hive.merge.orcfile.stripe.level=true " is the default one ?     ... View more

Thanks for the solution, but since i am not familiar with rest api, solution by Matt looks easy to me. Will surely try yours one too. ... View more