Given the open and distributed architecture of Renku, I am wondering if you have any plans to support other forges than GitLab? Forgejo / Gitea are well-established alternatives, don’t require a licensed to be used efficiently and almost don’t need any resources when being self-hosted.
Hi @pat-s thanks for raising that question! Indeed we already support both GitLab and GitHub - but any other provider that allows us to register an OAuth application should be fine (I believe Gitea at least falls into that category). We have not considered these other providers yet in more detail and there may need to be some additional adaptations needed, but it seems that it should very much be possible.
Specifically Forgejo (Gitea fork) as this one is a german-based forge being user by quite a few teams already (and rising). While GL is still dominant, I think the future is to go with other solutions than GH/GL for reasons of simplicity, costs and privacy.
Oauth2 isn’t a problem. I thought more about your storage integration and everything else you’ve built around GL etc to be much more of a “blocker” in terms of support for alternative forges.
I’ve just recently dived into renku and might have also a wrong perception for some parts, so please correct me if something sounds wrong to you.
PS: I am a contributor/dev of Forgejo/Gitea (incl Helm chart maintainer).
@pat-s you’re right, the “legacy” Renku is quite dependent on GitLab - I’m assuming that is what you are referring to. However, in our new version of the platform (Renku 2.0) there is a lot more flexibility. In the new incarnation of the platform, we offer easy ways to connect projects to various resources, including storage and code repositories so there is practically zero dependence on GitLab. That’s what I was referring to when I suggested that we could integrate with Gitea/Forgejo similar to how we integrate with various GitLab instances and GitHub.
Also, it can be noted that any git repository publicly accessible via HTTP/HTTPS can be used in Renku 2.0 with read-only operations (git clone, git pull, etc.). Write operations like git push would ask the user for their credentials.
OAuth2 is not an issue per se, neither are/should be git operations. In the end, any forge operates in the same way
I was more wondering about the “storage” part as I saw that a lot around it was tightly connected to GL when browsing the docs (for the first time).
I guess I need to get a better feeling for it first, so I’ll go ahead and deploy it on my cluster first and come back with more fine-grained questions afterwards
I was more wondering about the “storage” part as I saw that a lot around it was tightly connected to GL when browsing the docs (for the first time).
@pat-s this used to be the case with Renku v1. But with v2 we are decoupling from Gitlab and you can integrate with many different git providers. You can also connect more than 1 git repository to a single project. The only data we store in our database is just metadata about the project, access control info and what repos you have in your project. For storing datasets that you work with you can mount and connect anything that is compatible with Rclone. That is what data connectors are for in Renku v2. In addition you can also mount datasets from Zenodo or Dataverse through their doi.
One more thing, deploying Renku is not a trivial thing. And our documentation on this front is lacking. If you are doing this just to explore the features I strongly recommend that you just try v2 at renkulab.io.
We will be actively working on streamlining our Helm charts and improving the documentation over the summer. So it should get better.
I am a DevOps professional and operating with Helm every day, in fact I am specialized on providing/managing data-science environments I need to explore Renku in depth anyhow to get a feeling whether I want to add it to my stack of possible data science platform options, so I am happy to report any missing pieces along the way
I did so with renku v1 some time ago, but it wasn’t quite “ready” yet back then. Excited to give it a new try!
@pat-s since the docs are not up to date. Reach out to me in a direct message and I can give you a quick intro to v2 with a focus on deploying it and an overview of the architecture. So that if you need to deploy it and run it, it is easier.