Annotations provide descriptive information about the various rows and columns of a dataset, while clusterings allow one to divide the dataset into various subsets. Typical annotation are things like unique names and longer descriptions for each row, such as accession IDs and gene names respectively.
The primary annotation provide label that is unique for each row of your data. It must be set if you want to look at a specific gene vectors. Good choices are things like ORF name, Accession Number, etc.
The secondary annotation is used to provide a additional information for a particular row, good choices are gene annotation or gene name.