Kaniko can work without needing to fetch the remote context itself. This would open the workflow to non-GCS remote contexts.
Alternatively, we could just target the S3 compatible API, which is supported by GCP and AWS. We'd need a more generic way of getting passing credentials (I think?)
https://github.com/graymeta/stow seems providing abstract interface for GCS, S3, Azure storage, Swift, and Oracle storage.
Hi @r2d4 I would be interested in this feature for using Minio or similar within a hybrid/on-premise solution.
Would fetching from a Git repo over SSH/HTTPs be within scope?
Alex
Hey @alexellis , I think that would be within scope since Docker allows for Github repos as build contexts.
Could you open a separate issue for this? It'll be easier to keep track of, and if you're interested in contributing this feature I can provide more details!
This is how I did it via a Gist + init container. https://gist.github.com/alexellis/87d732a4b5fe056f5bf903aa6e6437ed
It'd be far cleaner within kaniko - I'd also be interested in some details, but am unsure if I can commit to contribute it until I get CLA approved and can size the work.
Alex
Hi,
In our CI/CD environment, we are using gerrit and we can get the tar.gz of a particular commit from an https url.
kaniko does not allow using random https urls, as https:// scheme is checked against azure urls:
https://github.com/GoogleContainerTools/kaniko/blob/master/pkg/buildcontext/buildcontext.go#L61
What is the reason to limit https urls to Azure? Could the check for Azure url pattern be revmoed so any https url is accepted?
Most helpful comment
https://github.com/graymeta/stow seems providing abstract interface for GCS, S3, Azure storage, Swift, and Oracle storage.