I have just deleted two forks I had of the GitLab project in my gitlab.com account. I did this after receiving a warning that quotas would now start to be enforced. It didn't say that I was going over quota, so I actually had to go look in the usage quotas page, which stated I was using 5.6GB of storage. So far so good, I'm not going to get billed because I'm below the 10GB threshold.

But still, I found that number puzzling. That's a lot of data! Maybe wallabako? I build images there in CI... Or the ISOs in stressant?

Nope. The biggest disk users were... my forks of gitlab-ce and gitlab-ee (now respectively called gitlab-foss and gitlab-ee, but whatever). CE was taking up roughly 1GB and EE was taking up the rest.

So I deleted both repos, which means that the next time I want to contribute a fix to their documentation — which is as far as I managed to contribute to GitLab — I will need to re-fork those humongous repositories.

Maybe I'm reading this wrong. Maybe there's a bug in the quotas system. Or, if I'm reading this right, GitLab is actually double-billing people: once for the source repository, and once for the fork. Because surely repos are not duplicating all those blobs on disk... right? RIGHT?

Either case, that's rather a bad move on their part, I feel like. With GitHub charging 4$/user/month, it feels like GitLab is going to have to trouble to compete by charging 20$/user/month as a challenger...

(Update: as noted in the comments by Jim Sorenson, this is actually an issue with older versions of GitLab. Deleting and re-forking the repos will actually fix the issue so, in a way, I did exactly what I should. Another workaround is to run the housekeeping jobs on the repo, although I cannot confirm this works myself.)

But maybe it's just me: I'm not an economist, surely there's some brilliant plan here I'm missing...

In the meantime, free-ish alternatives include Quay.io (currently free for public repos) and sr.ht (2$/mth, but at least not open-core, unfortunately no plan for a container registry). And of course, you can painfully self-host GitLab, sr.ht, gitea, pagure, or whatever it is the current fancy gitweb.

Known issue with a work-around

Yes, this is an issue with forks that were created with previous versions of GitLab. You can see the two work-arounds in this closed issue:

This was eventually resolved with new forks with this epic:

However...you would think that they would have come up with some way to retroactively resolve existing forks. Alas, no. Use the work-arounds in the first link.

Comment by Jim Sorenson

Oh that makes sense, thanks, that's very useful.

I'll make sure to correct the article to clarify this has been fixed in later releases. In a way, then, me deleting the repos is exactly what I should have done...

Comment by anarcat
Comments on this page are closed.
Created . Edited .