bonded global contacts only apply to faces that are touching at the beginning of the analysis, otherwise you need to add contact sets. in your case where the bearing is, you'd need no penetration contact it sounds like because they slide with respect to each other and aren't glued together?
the other thing i'd be concerned about here is symmetry. that may overrestrain your model from a buckling perspective.
nonlinear buckling is ideal for all buckling problems. it is a truely nonlinear phenomena. you might want to double check how it is executed in simulation (linear buckling) and really outline what you want to learn in this analysis and what you expect before moving too much further.