I am hosting my own private conda channel on S3, but I don't understand how to avoid mirroring all the packages on my local hard drive. The source of the problem is the repodata.json, channeldata.json, etc. files.
If I build a single package and copy just the .tar.bz2 file to S3, conda-install does not see it in the channel. In order for conda-build to see the package, I have to copy the package and all of the repodata.json, etc. metadata files created by conda-build to S3.
My conundrum is: these JSON metadata files do not appear to be updated properly unless all of the packages I've ever built appear in my local package directory. I now have more than 4GB of conda packages on my local hard drive so that I get the correct metadata in these JSON files.
Is there a way to continue building conda packages with my local machine while avoiding mirroring my private S3 conda channel on my local hard drive? For what its worth, I realize that conda-index will build the JSON files, but it doesn't seem to work when you attempt to index a location on S3.