Structure and heterojunction of substrates are widely regarded as critical factors to improve Raman detection and recyclibility, but the exact roles are rarely explored for the mixed contributions. Herein, the arrays of ZnO/MoS2 coaxial nanorods (NRs) and MoS2 nanotubes (NTs) were successfully constructed for the surface-enhanced Raman scattering (SERS). The influences of NR and NT length on the morphology, light absorption and emission, as well as SERS performance were systematically investigated. Both the arrays grown for 40 min displayed optimal SERS performance, in which MoS2 NT array owned stronger enhancement with a detection limit of 5 × 10−7 M than that of 1 × 10−6 M for ZnO/MoS2 NR array. The NTs also effectively enhanced photocatalysis, in which rhodamine 6G pollutant was completely degraded within 140 min under ultraviolet light, surpassing that by the NRs. The analysis revealed that band alignment and structure characteristic of the specimens played critical roles in the spectral enhancement and self-cleaning, while the nanostructure surpassed heterojunction as it supplied void interior for extending light propagation and more contact area for adsorbing pollutants. These results would contribute to a better understanding of the Raman enhancement and could be extended to the design of SERS substrates to harness their potential application.