I would do both a cantilever bending beam analysis and a direct shear analysis. I think the real stress mechanism will be some combination of both. It will be a result of the geometry of the actual contact area.
Have you experimented? Do you know the failure mode?