I have an H2O k-means model object trained on data with numeric variables and missing values. As documented here, the algorithm handles missing values automatically imputed