Also, is there anything interesting in the yarn scheduler logs? Something about scheduling being skipped?
On Wed, Jan 31, 2018 at 05:16 Billy Watson <williamrwat...@gmail.com> wrote: > Ok, and your container settings? > > On Wed, Jan 31, 2018 at 02:38 nishchay malhotra < > nishchay.malht...@gmail.com> wrote: > >> yes my job has about 160,000 maps and my cluster not getting fully >> utilized around 6000 maps ran for 2 hrs and then I killed the job. At any >> point of time only 40 containers are running thats just 11% of my cluster >> capacity. >> >> { >> "classification": "mapred-site", >> "properties": { >> "mapreduce.job.reduce.slowstart.completedmaps":"1", >> "mapreduce.reduce.memory.mb": "3072", >> "mapreduce.map.memory.mb": "2208", >> "mapreduce.map.java.opts":"-Xmx1800m", >> "mapreduce.map.cpu.vcores":"1" >> } >> }, >> { >> "classification": "yarn-site", >> "properties": { >> "yarn.scheduler.minimum-allocation-mb": "32”, >> “yarn.scheduler.maximum-allocation-mb”:”253952”, >> “yarn.scheduler.maximum-allocation-vcores: “128” >> >> "yarn.nodemanager.vmem-pmem-ratio":"3", >> "yarn.nodemanager.vmem-check-enabled":"true", >> yarn.nodemanager.resource.cpu-vcores" ; "16”, >> yarn.nodemanager.resource.memory-mb: “23040" >> } >> >> Each node: capacity >> Disk-space=100gb >> memory=28gb >> processors: 8 >> >> >> -- > William Watson > > -- William Watson