It's not clear if you are seeking steady RANS solutions or whether you will be using the URANS (unsteady RANS) approach that helps capture large scale unsteady structures within the flow (e.g., vortex shedding behind the bluff body).  There is evidence (see Ong et al., Marine Structures 22 (2009) 142–153) that the URANS method may provide reasonable engineering estimates in the context of bluff body flows.  URANS computations would be more expensive than the RANS computations because of the need to perform a time accurate simulation and to accumulate a long enough time series to compute flow statistics.  However, keep in mind that the RANS computations may or may not converge to a stationary flow for bluff body flows. 

Back to your query, the grid resolution requirements would vary significantly from the front of the bluff body (where the flow is attached) from those in the rear (i.e., the "base" flow, where the flow is separated).  Even in the attached flow region, the flow in the vicinity of the stagnation point is bound to be laminar, although a quick transition to turbulent flow could ensue depending on the surface roughness, freestream turbulent intensity, and the flow Reynolds number. If the laminar region is small, and therefore, can be ignored for the purpose of your computation, then you may use a skin friction estimate based on van Driest (Van Driest, E. R., “On the Aerodynamic Heating of Blunt Bodies,” ZAMP, 17, Vol. 9, issue 5-6, pp. 233-248. 1958) to ensure that your near wall spacing (i.e., first cell height) is consistent with the requirements of the turbulence model.

However, assuming that you are interested in relatively high Reynolds numbers, I would suggest an iterative approach based on URANS that begins with a grid that is based on a preliminary knowledge of the flowfield.  After scrutinizing the computed flowfield on the preliminary mesh, you can refine the mesh to ensure that all "relevant" flow features are adequately resolved and that the flow metrics of interest have become insensitive (to the order of the desired accuracy) to further refinements and/or some degree of grid coarsening. This is also the most convenient approach for any new class of fluid flow simulations.

You generally could continue to consider the guideline for the flat plate, considering the normal direction to the body. However, the flow around a bluff body has a separated flow region of strong intensity. The grid must be refined also in this region.

Finally, be aware that RANS formulation can be not very suitable for such flow.

Try to define a region separately

It's not clear if you are seeking steady RANS solutions or whether you will be using the URANS (unsteady RANS) approach that helps capture large scale unsteady structures within the flow (e.g., vortex shedding behind the bluff body).  There is evidence (see Ong et al., Marine Structures 22 (2009) 142–153) that the URANS method may provide reasonable engineering estimates in the context of bluff body flows.  URANS computations would be more expensive than the RANS computations because of the need to perform a time accurate simulation and to accumulate a long enough time series to compute flow statistics.  However, keep in mind that the RANS computations may or may not converge to a stationary flow for bluff body flows. 

Back to your query, the grid resolution requirements would vary significantly from the front of the bluff body (where the flow is attached) from those in the rear (i.e., the "base" flow, where the flow is separated).  Even in the attached flow region, the flow in the vicinity of the stagnation point is bound to be laminar, although a quick transition to turbulent flow could ensue depending on the surface roughness, freestream turbulent intensity, and the flow Reynolds number. If the laminar region is small, and therefore, can be ignored for the purpose of your computation, then you may use a skin friction estimate based on van Driest (Van Driest, E. R., “On the Aerodynamic Heating of Blunt Bodies,” ZAMP, 17, Vol. 9, issue 5-6, pp. 233-248. 1958) to ensure that your near wall spacing (i.e., first cell height) is consistent with the requirements of the turbulence model.

However, assuming that you are interested in relatively high Reynolds numbers, I would suggest an iterative approach based on URANS that begins with a grid that is based on a preliminary knowledge of the flowfield.  After scrutinizing the computed flowfield on the preliminary mesh, you can refine the mesh to ensure that all "relevant" flow features are adequately resolved and that the flow metrics of interest have become insensitive (to the order of the desired accuracy) to further refinements and/or some degree of grid coarsening. This is also the most convenient approach for any new class of fluid flow simulations.

Meelan Choudhari , thanks a lot for your detailed answer. Very helpful indeed.

Yes I was talking about URANS. Precisely, I want to ask how well does URANS perform and to what extent is it reliable in case of detached boundary layer flow? Are any wall functions valid in detached area of boundary layer?