Grapevine downy mildew (Plasmopara viticola) is a major disease of European cultivated grapevine (Vitis vinifera L.) against which a large amount of synthetic pesticides are used. Developing microbial biocontrol of P. viticola could reduce the use of pesticides in viticulture and preserve human and environmental health. To achieve this goal, it is necessary to better understand the interactions between P. viticola and the vine foliar microbiome. Here we present metabarcoding datasets describing the bacterial and fungal communities from more than 500 leaf samples collected during powdery mildew epidemics in three major wine-producing regions in France. The microbiome of both healthy tissues and disease lesions was characterized. P. viticola abundance was quantified using qPCR. We provide the raw metabarcoding datasets, the Amplicon Sequence Variant (ASV) tables obtained after bioinformatic processing, the metadata describing sampling sites and tissue health condition and the code used for bioinformatic and descriptive statistical analysis.