I just went through figuring this out and I thought I would add my process, including the FASTA component, using Vibrio phage VP882 as my example and utilizing the gff strategy you mentioned in a comment to the other answer. Here is everything I did using an established snpEff installation. It worked when I ran my analysis using it, so this strategy is confirmed in my case:
#How to create a snpEff database using a gff3 and genomic DNA fasta file... (note, the chromosome names must match in the 2 files)
#NOTE: This uses /bin/tcsh...
setenv DBNAME Vibrio_phage_VP882
setenv GFF3 ~/Downloads/VibriophageVP882.gff3
setenv FASTA ~/Downloads/VibriophageVP882.fasta
#Go into the snpEff directory and create a directory for your files
#Copy the files into snpEff's directory structure
cp $GFF3 data/$DBNAME/genes.gff
cp $FASTA data/$DBNAME/sequences.fa
#Edit snpEff.config and insert your specific database information:
echo "$DBNAME.genome : $DBNAME" >> snpEff.config
#Build the database
java -jar snpEff.jar build -gff3 -v $DBNAME
I did not have any errors or warnings, so if you see anything untoward, you'll have to figure those things out.
You can set the 3 variable values at the top of this script and run the rest without changing it (unless your snpEff installation is in a different place.
And just for completeness, I downloaded the gff3 and fasta files directly from this page:
Using the complete record/file download option at the top right of the record, selecting gff and fasta in 2 separate downloads.