Although there is an extensive tradition of research into the microbes that underlie the winemaking process, much remains to be learnt. We combined the high-throughput sequencing (HTS) tools of metabarcoding and metagenomics, to characterize how microbial communities of Riesling musts sampled at four different vineyards, and their subsequent spontaneously fermented derivatives, vary. We specifically explored community variation relating to three points: (i) how microbial communities vary by vineyard; (ii) how community biodiversity changes during alcoholic fermentation; and (iii) how microbial community varies between musts that successfully complete alcoholic fermentation and those that become ‘stuck’ in the process. Our metabarcoding data showed a general influence of microbial composition at the vineyard level. Two of the vineyards (4 and 5) had strikingly a change in the differential abundance of
Metschnikowia
. We therefore additionally performed shotgun metagenomic sequencing on a subset of the samples to provide preliminary insights into the potential relevance of this observation, and used the data to both investigate functional potential and reconstruct draft genomes (bins). At these two vineyards, we also observed an increase in non-Saccharomycetaceae fungal functions, and a decrease in bacterial functions during the early fermentation stage. The binning results yielded 11 coherent bins, with both vineyards sharing the yeast bins
Hanseniaspora
and
Saccharomyces
. Read recruitment and functional analysis of this data revealed that during fermentation, a high abundance of
Metschnikowia
might serve as a biocontrol agent against bacteria, via a putative iron depletion pathway, and this in turn could help
Saccharomyces
dominate the fermentation. During alcoholic fermentation, we observed a general decrease in biodiversity in both the metabarcoding and metagenomic data. Unexpected
Micrococcus
behavior was observed in vineyard 4 according to metagenomic analyses based on reference-based read mapping. Analysis of open reading frames using these data showed an increase of functions assigned to class Actinobacteria in the end of fermentation. Therefore, we hypothesize that bacteria might sit-and-wait until
Saccharomyces
activity slows down. Complementary approaches to annotation instead of relying a single database provide more coherent information true species. Lastly, our metabarcoding data enabled us to identify a relationship between stuck fermentations and
Starmerella
abundance. Given that robust chemical analysis indicated that although the stuck samples contained residual glucose, all fructose had been consumed, we hypothesize that this was because fructophilic
Starmerella
, rather than
Saccharomyces
, dominated these fermentations. Overall, our results showcase the differen...