Open proppy opened 2 years ago
@kgugala looks like there is already a bucket named hdl-conda
that we could reuse.
depending if we want to land this before or after #184, we could use either https://github.com/google-github-actions/upload-cloud-storage#authenticating-via-workload-identity-federation or https://github.com/google-github-actions/upload-cloud-storage#via-application-default-credentials for authorizing the action to access the bucket.
@umarcor as discussed with @mithro on https://github.com/hdl/conda-eda/issues/184#issuecomment-1052923605, do you think we can get a bucket set-up on github-hdl
with either https://cloud.google.com/iam/docs/workload-identity-federation or https://cloud.google.com/iam/docs/creating-managing-service-accounts authorization scheme set so that we can upload artefact from the Upload packages
workflow?
@PiotrZierhoffer @ajelinski @AdamOlech can we get a bucket created in the same project that's currently running the runners?
I can see the settings of github-hdl
(I think). There, in Compute Engine, I can see the gha-runner-coordinator-conda-eda VM, which I believe is the runner (coordinator) used in this repository; and the Disks attached to it.
However, Section Cloud Storage > Buckets is empty (I cannot see anything at least). I can neither see the container images from hdl/containers. If I browse gcr.io/hdl-containers
I can see the images (left), but the "selected project" is empty. If I set the project to github-hdl, it seems that the feature is disabled (right):
I don't know who can clarify whether the images from hdl/containers and the runner in this repo belong to the same project and whether that project is github-hdl
. I believe it should be the same (a single) project and we should be able to create a bucket there. Yet, I'm unsure.
Since conda has the capability of installing package from arbitrary static http host (as long as it can fetch the index), we could consider mirroring the packages under a google cloud storage bucket.
That would have the following advantages:
https://conda.hdl.tools
as the canonical channel name)