ufna
January 22, 2021, 9:01pm
5
Rawalanche:
Can you elaborate? I tried Git LFS and it turned into a complete hell. For example - I created a few large binary assets (imported a bunch of high resolutions assets to try in the project), made a commit, then removed them, and made another commit. It was about 4GB of data. Of course, once they were removed they were no longer needed. The way Git works is not an issue with source code, since file sizes are negligible. But in terms of binary assets, just a couple days of several artists working can easily generate couple of dozen gigabytes of data.
That’s where the big issue came. Git LFS has some prune command, which removes outdated LFS files, but they are based on a very convoluted set of confusing and borderline insane rules. So there was just no way to prune the old content. In less than a week we have reached limit of out paid Git repository storage with absolutely no way to prune old LFS assets to make a space for new ones.
The literally only way to solve that issue was to remove git completely, and reinitialize the new git repo from scratch. Of course that has lead to loss of all previous VC data, so it completely defeats the point of having version control system in the first place.
I just don’t understand LFS. To me, it seems like if you add any large binary file, at any point in time, you are stuck with it forever, even if it’s way outdated and you will never need it again. So if you are working on a long term game project, which is actually serious AAA production, and contains serious high end assets, not some gimpy lowpoly indie stuff, it’s easy to approach a repository of over a terabyte in size in just a couple of months, assuming that there will be lots of iteration on the art assets. And hosting for repos of that size is nowhere near affordable, nor makes any sense.
Use bitbucket or gitlab if you want to prune LFS files. Eternal LFS it’s just a GitHub issue, and it’s definitely not a problem for AAA projects
Because: 1. devops can prune files if necessary. 2. space is not the issue, you can use hosted bitbucket solution in case of terrabytes of data.