-
Notifications
You must be signed in to change notification settings - Fork 14
Description
For any given change, we want to be able to have CI build the base CoreOS image, then build a node image on to of that, and then take that node image and run a cluster test.
This is not possible right now because building that first base CoreOS image requires more privileges than available in the OpenShift builder. openshift/builder#410 was an attempt to work around that.
An idea I had for a crude way to work around this is to just build it in a different CI environment and push it somewhere, and then have the actual Containerfile for OpenShift CI "wait" for that CI build. Conveniently, Konflux pushes to a temporary repo built images, even on PRs. So this could be done on top of the Konflux work.
The Containerfile would look something like (pseudocode):
ARG COMMIT
FROM quay.io/fedora/fedora-coreos:latest
RUN # wait until quay.io/quay.io/redhat-user-workloads/coreos-tenant/coreos-assembler:on-pr-$COMMIT exists
FROM quay.io/konflux...:on-pr-$COMMIT exists
<build node image>
We would need OpenShift CI to auto-pass the git commit of the PR under test as a build arg, which may already be a thing it does.