I'm encountering something similar. Immediately upon starting a cluster and triggering a job run, my memory usage jumps from 0 to about 20GB used and 15GB cached (see the attached screenshot). The data I am working with should be very small (less tha...
