Kraken is broken something something NCBI numbers something something. Use perl scripts that supposedly dealt with the issue (note that I was able to get the fungal one to work with the same loop etc, where the only difference was that only fungi was included)
http://www.opiniomics.org/building-a-kraken-database-with-new-ftp-structure-and-no-gi-numbers/
As of September 2016, someone commented that this method works, but something went wrong for me.
Ran on r4.8xlarge.
Get the sequences (note the script filters for complete genomes)