It’s easy. Lift and shift, then fuck it up by not quite completely migrate everything to numerous badly managed kubernetes clusters. That’s what we did.