I have setup a remote git-annex repository (ssh into some machine in our local network) which I will call "common_repo". Multiple contributors will then be cloning this repo into their laptops (I will call "client_repos").

When these client_repos change and do a "git annex sync --content", all the files from their laptops are successfully synchronised. The problem arises when:

  1. CLIENT_A creates somefile.csv
  2. CLIENT_A commits somefile.csv
        git annex add somefile.csv
        git commit -a -m "uploaded somefile.csv"
  3. CLIENT_A modifies somefile.csv
  4. CLIENT_A commits modified somefile.csv
        git add somefile.csv
        git commit -a -m "updated somefile.csv"
  5. CLIENT_A synchronises with common_repo
    i.e.: git annex sync --content

If I check the .git/annex/objects of common_repo, I can't seem to find a copy of the unmodified somefile.csv. It only has a copy of the latest somefile.csv.

This is problematic if one client tries to checkout a revision of the project that uses the original somefile.csv.

I learned that I can change the preferred content of git-annex. So, the appropriate preset for common_repo seemed to be "backup". After running the commands in common_repo:

git annex wanted . standard
git annex group . backup

I've done another test of the scenario above, and common_repo is still missing the previous revision of the file!! The preferred file content of common_repo should be "include=* or unused". In my case, the previous version of somefile.csv will probably fall in the "unused" category. But I still cannot find it.

A workaround is using two commands from the client(s):

git annex copy --to --all
git annex sync --content

But I can imagine my users forgetting to run 'copy' and my repo will go to shit over time.

Any ideas why I can't synchronise 'unused' files?