Understanding how fluid flow interacts with micro-textured surfaces is crucial for a broad range of key biological processes and engineering applications including particle dispersion, pathogenic infections, and drag manipulation by surface topology. We use high-speed digital holographic microscopy (DHM) in combination with a correlation based de-noising algorithm to overcome the optical interference generated by surface roughness and to capture a large number of 3D particle trajectories in a microfluidic channel with one surface patterned with micropillars. It allows us to obtain a 3D ensembled velocity field with an uncertainty of 0.06% and 2D wall shear stress distribution at the resolution of ~65 μPa. Contrary to laminar flow in most microfluidics, we find that the flow is three-dimensional and complex for the textured microchannel. While the micropillars affect the velocity flow field locally, their presence is felt globally in terms of wall shear stresses at the channel walls. These findings imply that micro-scale mixing and wall stress sensing/manipulation can be achieved through hydro-dynamically smooth but topologically rough micropillars.