Aims: To explore the mechanisms of continuous cropping obstacles of tobacco using co-occurrence network analyses to identify interactions between rhizosphere soil microbiota and metabolites. Methods: Using pot experiments, tobacco biomass, soil chemical properties were routinely determined over three continuous growth seasons. Rhizosphere microbiota and metabolites were respectively determined using phospholipid fatty acids (PLFAs) and gas chromatography-mass spectrometry, and then analysed using co-occurrence network analyses to explore growth obstacle mechanisms of tobacco. Results: Tobacco biomass was significantly lower in the 2nd- and 3rd-season soils when compared with soils from the 1st-season – indicating growth obstacles. Three PLFA biomarkers (a16:0, 17:1ω8c, and 20:0) and five (i14:0, i15:1G, 17:0, 11Me18:1ω7c, and 16:1ω5c) were distinct to the 2nd- and 3rd-season soils, respectively. In the 2nd-season, 33 metabolites (phenol, cyclopropanebutanoic acid, 16-octadecenoic acid, n-hexadecanoic acid, and [z]-13-docosenamide, etc.) were up-regulated, and 10 metabolites (d-(−)-ribofuranose, d-(+)-cellobiose, and myo-inositol, etc.) down-regulated. Co-occurrence network analyses indicated that 16-octadecenoic acid, n-hexadecanoic acid, oleic acid and [z]-13-docosenamide might act as “hubs” to alter the secondary metabolism, d-(−)-ribofuranose and d-(+)-cellobiose as key metabolites to induce the changes in microbial compositions, while myo-inositol as a “trigger” metabolite in negative feedback signaling between plants and microbes. Conclusion: We found that a combination of positive feedback involving allelochemicals (i.e. phenolic acids) and negative feedback involving metabolites (i.e. myo-inositol, D-(−)-ribofuranose and D-(+)-cellobiose) could result in changes to soil microbial composition associated with plant growth obstacles.