Yarn job crashes with exit code 143 even though node has available memory

hadoop memory-management mapreduce hadoop-yarn

your mapper is performing some task which is taking more than 600 seconds, the nodemanager thinks its dead and hence killing it.
Please ack in between the performing the task, that the task is alive.
In case you are updating the counters the ack will be done automatically.
Increase the timeout more than 600 seconds

CodeHunter