Image validation workflow #297
Conversation
Signed-off-by: ruokun-niu <[email protected]>
Signed-off-by: ruokun-niu <[email protected]>
Signed-off-by: ruokun-niu <[email protected]>
Signed-off-by: ruokun-niu <[email protected]>
ruokun-niu
requested review from
agentofreality,
amansinghoriginal and
danielgerlag
| echo "" | ||
| echo "✅ SUCCESS: All images meet their architecture requirements for version $IMAGE_TAG" | ||
|
|
||
| - name: Validate azure-linux variant architecture support |
There was a problem hiding this comment.
Would a matrix strategy be better to reduce the workflow file size?
Something like:
multi-arch-validation:
...
strategy:
matrix:
variant: ["", "-azure-linux"]
steps:
- name: Validate ${{ matrix.variant || 'default' }} variant architecture support
run: |
IMAGE_TAG="${{ needs.get-version.outputs.image_tag }}${{ matrix.variant }}"
There was a problem hiding this comment.
Good point. Will update this
There was a problem hiding this comment.
Can it make use of the existing matrix?
|
|
||
| for image in $DRASI_IMAGES; do | ||
| # Skip reaction-dataverse unless we're on x86-64 runner | ||
| if echo "$AMD64_ONLY_IMAGES" | grep -q "$image" && [[ "${{ matrix.runner }}" != "oracle-vm-8cpu-32gb-x86-64" ]]; then |
There was a problem hiding this comment.
Should we check like [[ "${{ matrix.arch }}" == "arm64" ]] instead of the runner name?
amansinghoriginal
force-pushed
the
image-validation-workflow
branch
from
a0af0e5 to
28e49a5
Compare
| echo "📋 Using image tag: ${{ needs.get-version.outputs.image_tag }}-azure-linux" | ||
|
|
||
| # Images that only need amd64 support | ||
| AMD64_ONLY_IMAGES="ghcr.io/drasi-project/reaction-dataverse" |
There was a problem hiding this comment.
This could come from the matrix too
Signed-off-by: ruokun-niu <[email protected]>
ruokun-niu
force-pushed
the
image-validation-workflow
branch
from
28e49a5 to
8892617
Compare
| - arch: amd64 | ||
| runner: oracle-vm-8cpu-32gb-x86-64 | ||
| variant: "" | ||
| amd64_only_images: "ghcr.io/drasi-project/reaction-dataverse" |
There was a problem hiding this comment.
Why are we only doing an amd64 build for dataverse?
There was a problem hiding this comment.
The Dataverse reaction is only built with amd64 architecture in the release workflow. IIRC, we had a dependency that would only work when we build the image in amd architecture.
There was a problem hiding this comment.
Looking at the dependencies, there is nothing that stands out as being an issue. Can we confirm this?
There was a problem hiding this comment.
Yep let me check again
There was a problem hiding this comment.
I've added the arm64 variant of the dataverse reaction to both the release workflow and the image validation workflow
Signed-off-by: ruokun-niu <[email protected]>
Signed-off-by: ruokun-niu <[email protected]>
3 participants
This pull request introduces a new image validation workflow into the release process to ensure that all Drasi container images meet architecture requirements before a release is finalized. The workflow checks for the presence of required architectures (amd64 and arm64) for each image, tests pulling images on real hardware runners, and integrates these checks into the existing release pipeline.
Release process improvements:
validate-imagesjob to the.github/workflows/draft-release.ymlworkflow, which runs after manifest creation and before the final release, ensuring image validation is a required step for releases. [1] [2]Image validation workflow:
.github/workflows/image-validation.yml, which:reaction-dataverseonly requires amd64).Fixes: #issue_number