This experiment was conducted to study the long-term effect of soil management treatments on crop yields, sustainability yield indices (SYI) and soil fertility in rainfed semi-arid tropical Alfisol at Hayathnagar Research Farm, Hyderabad, India, during the period 1995 to 2009. The experiment was conducted in a split–split plot design with conventional tillage (CT) and minimum tillage (MT) as main factors, surface application of sorghum stover @ 2 t ha−1 (SS), fresh Gliricidia loppings @ 2 t ha−1 (GL) and ‘no’ residue (NR) as sub-factors and levels of N viz. 0 (N0), 30 (N30), 60 (N60) and 90 (N90) kg N ha−1 as sub–sub factors in a castor–sorghum two-year rotation. On an average, CT maintained 30.4 and 57.0% higher grain yields of sorghum and castor, respectively, over MT. Between two residues, GL performed well in both the crops. The highest yields of sorghum (1425 kg ha−1) and castor (876 kg ha−1) were recorded at 90 kg N ha−1. CT maintained higher SYI of 0.44 compared to MT (0.38) and higher agronomic efficiency (AE) of 13.5 and 6.76 kg grain kg−1 N for sorghum and castor crop, respectively. Use of crop residue as mulch had an advantage in increasing the yield of both the crops with increase in rainfall under CT even without N application (control), probably by making the soil more receptive to water infiltration, better moisture storage and by reducing the evaporative losses. Using response functions, the optimum fertilizer N requirement was also computed for a given set of tillage and residue combinations. The revised optimum fertilizer N doses for sorghum and castor varied from 45 to 56 kg ha−1 and 46 to 74 kg ha−1, respectively, under different tillage and residue combinations and could be recommended depending upon the soil management practices.