I discuss one exactly solvable model of disorder - the model of (2+1)-
dimensional Dirac fermions in a random non-Abelian gauge potential
where the exact solution can be obtained by three different
approaches (replicas, supersymmetry and the direct averaging over
disorder). All these approaches give identical results. The solution
containts typical features of disordered critcial models such as the presence of logarithms
etc.