I've launched my HDFS balancer with the nohup command :
$ nohup hdfs balancer &
It is taking forever and I need to work on my cluster. Do you know how can I stop the process ?
It's a distributed process so it's quite difficult to stop by juste doing "kill PID" ...
Thank you
Actually, the HDFS balancer is not a distributed process. It is a single process that initiates block movements on the cluster; the balancing process is carried out in a distributed manner cooperatively by DataNodes, but the commands about which blocks to balance originate from a single process (the one created by hdfs balancer
). Doing a kill $PID
will be sufficient to stop any further balancing.
Source: HDFS Balancer documentation, personal experience.