09-11-2018 02:16 PM
Some of the Hadoop processes like YarnChild, DataNode, NodeManager are slightly using swap memory (10MB), there is enough memory available (usage is only 20GB of 500GB, buff/cache 450GB), enough heap available for these processes.
CentOS 7, CDH enterprise 5.14.0,
When we run
echo 3 > /proc/sys/vm/drop_caches
its not swapping any more, again started swapping when buff/cache reached around 450GB
any suggestion? thanks