Adult survival is perhaps the fitness parameter most important to population growth in long-lived species. Intrinsic and extrinsic covariates of survival are therefore likely to be important drivers of population dynamics. We used long-term mark-recapture data to identify genetic, individual and environmental covariates of local survival in a natural population of mountain brushtail possums (Trichosurus cunninghami). Rainfall and intra-individual diversity at microsatellite DNA markers were associated with increased local survival of adults and juveniles. We contrasted the performance of several microsatellite heterozygosity measures, including internal relatedness (IR), homozygosity by loci (HL) and the mean multilocus estimate of the squared difference in microsatellite allele sizes within an individual (mean d (2)). However, the strongest effect on survival was not associated with multilocus microsatellite diversity (which would indicate a genome-wide inbreeding effect), but a subset of two loci. This included a major histocompatibility complex (MHC)-linked marker and a putatively neutral microsatellite locus. For both loci, diversity measures incorporating allele size information had stronger associations with survival than measures based on heterozygosity, whether or not allele frequency information was included (such as IR). Increased survival was apparent among heterozygotes at the MHC-linked locus, but the benefits of heterozygosity to survival were reduced in heterozygotes with larger differences in allele size. The effect of heterozygosity on fitness-related traits was supported by data on endoparasites in a subset of the individuals studied in this population. There was no apparent density dependence in survival, nor an effect of sex, age or immigrant status. Our findings suggest that in the apparent absence of inbreeding, variation at specific loci can generate strong associations between fitness and diversity at linked markers.