In aquatic ecosystems, the health of fish depends greatly on the dynamics of microbial community structure in the background environment. Nonetheless, finding microbes with profound impacts on fish performance out of thousands of candidate species remains a major challenge. We here show that time-series analyses of microbial population dynamics illuminate core components and structure of fish-associated microbiomes. By targeting eel aquaculture microbiomes as model systems, we reconstructed the population dynamics of 9,605 bacterial and 303 archaeal species/strains across 128 days. Due to the remarkable increase/decrease of constituent microbial populations, the taxonomic compositions of microbiomes changed drastically through time. We then found that some specific microbial taxa showed positive relationship with eel activity level even after excluding cofounding effects of environmental parameters (pH and dissolved oxygen level) on population dynamics. In particular, a vitamin B12-producing bacteria, Cetobacterium somerae, consistently showed strong positive associations with eel activity level across the replicate time-series of the five aquaculture tanks. Network theoretical and metabolic modeling analyses further suggested that the highlighted bacterium formed compartments of close microbe-to-microbe interactions with some other bacterial taxa, forming potential core microbiomes with positive impacts on eels. Overall, these results suggest that integration of microbiology, ecological theory, and network science allows us to explore core species and interactions embedded within complex dynamics of fish-associated microbiomes.