Quick Question! Does project_id have to be hard-coded here?
JoshZastrow opened this issue · 0 comments
JoshZastrow commented
Love the example here! It almost perfectly fits our use case, and super clean!
We deploy our images and pipelines to a _staging
GCP project and then a _prod
GCP project if things look good. An easy way to do this is to have Project ID and Bucket as global variables in the deploy process.
It doesn't seem that the image of the component spec implementation is configurable, but we push images to a different container registry depending on the environment. Is the <project_id> in your example configurable, or this is a "one project-multiple environments" sort of approach?
i.e
implementation:
container:
image: gcr.io/${PROJECT_ID}/kubeflow/components/predict:latest
```