I change the default *kylin.job.mapreduce.default.reduce.input.mb = 500 --> 5* then the build have successed! But the mr log shows it have failed two time in convert the cuboid to hfile(19 step), so how many time will the mr job try before the task failed, is it configured?
But I still can't understand why i decrease the volume of input to each reduce then the build step pass, does it mean the previous reduce was killed by yarn by out of 'memory'? -- View this message in context: http://apache-kylin.74782.x6.nabble.com/Failed-with-Build-Cube-in-18-step-tp6700p6704.html Sent from the Apache Kylin mailing list archive at Nabble.com.
