Currently, JVM (openjdk/Oracle) only exposes overall CPU or memory metrics for a process.As of now, it's not possible at jvm level to pull the individual pipeline memory details.
To know more about the the memory consumption we can analyse heap dumps ,thread dumps and gc.log when the issue is happening.
heap dumps:
sudo -u sdc jcmd <pid> GC.heap_dump <filename>.hprof
Thread dumps:
for i in {0..5} ; do echo $i ; jstack {pid} > /tmp/sdc-$(date +%Y-%m-%d.%H:%M:%S).jstack ; sleep 5 ; done