Landscape management affects species interactions and can have notable effects on food web structure. Local parasitoid populations in greenhouses usually migrate from outside crops; biological control of greenhouse aphids may be thus highly dependent on the composition of surrounding landscape. However, it is less clear how surrounding landscape composition affects primary-hyperparasitoid food webs and pest control services in greenhouses. We investigated the food web of parasitoids on melon-cotton aphid (Aphis gossypii Glover) in watermelon greenhouses in two suburban Beijing counties over two years. We used the quantitative food web metrics (generality, vulnerability, link density, and interaction evenness) to assess the effects of landscape composition on primary-hyperparasitoid food web structure. We found that landscape with more cropland within 1–3 km tended to have more primary parasitoids per hyperparasitoid species (generality). Higher proportions of woodland at the 0.5 km scale were negatively correlated with the mean numbers of hyperparasitoid per primary parasitoid species (vulnerability), as well as with hyperparasitism rate and hyperparasitoid richness. Link density, interaction evenness and aphid mortality caused by parasitoids (parasitism rate) were not affected by landscape factors. However, active primary parasitism (biocontrol potential) increased with the proportion of woodland. This suggested that the bottom-up effect induced by primary parasitoids might benefit hyperparasitoids, thus exerting little influence of primary parasitoids on pest control. The top-down effect of hyperparasitoids may reduce with increasing woodland proportion. To enhance the effects of primary parasitoids, landscape management programs should also target, and thus limit the impact of hyperparasitoids.