I have enabled data loading in my cluster using INSERT OVERWRITE, fe pods are behind a AWS NLB. How to debug why the memory of only 1 fe pod is so high. All 3 fe pods are on a single EC2 only 
the highest load FE must be the leader one. it takes care of write/update meta operations. FE nodes are running in single-writer-multiple-readers mode. Vertically scale it by setting -Xmx is set correctly.