New Methodology for images.txt, package-images.yaml, oci_package_list.txt, and Othersπ
BLUF (Bottom Line Up Front)π
- Big Bang image metadata now comes from explicit declarations (no more cluster scraping)
- New
images-v2-*artifacts show the dependency graph so you can pre-pull or allowlist exactly what you needβno Istio/Kyverno/Flux noise unless they are actual dependencies. images.txtis now a copy ofimages-v2-with-dependencies.txt(which uses the explicit declarations)- The
smoke testsstage and its jobs (clean install all-packagesandclean install oci all-packages) have been eliminated from the Big Bang release pipeline since each package is tested individually as part of the individual package pipeline. - The Big Bang release pipeline now completes in around 20 minutes rather than multiple hours, saving the release engineers considerable time on every release.
Terms / Glossaryπ
Throughout this article, the terms core, umbrella, and Big Bang chart are all used interchangeably.
Breaking Changesπ
None
What changed and when?π
The changes were introduced in Big Bang 3.4.0 on August 7, 2025.
The methodology used to compute the contents of the following files changed:
images.txtpackage-images.yamloci_package_list.txt
Three (3) new files are being generated:
images-v2-dependencies.yamlimages-v2-with-dependencies.txtimages-v2-no-dependencies.txt
The old methodology spun up the package on a cluster and then pulled a list of all images running in the cluster, whether they were for that package or not. Therefore, the list of images contained references to Istio, Kyverno, Flux CD, the package of concern, and sometimes more. This was an implicit approach and led to an imprecise list of images.
The new methodology uses an explicit approach where all images are defined purposefully in the following locations:
| Source / Location | Key Path | Scope | Example |
|---|---|---|---|
Package’s chart/Chart.yaml |
annotations."helm.sh/images" |
package, umbrella |
Example |
Subcharts listed in chart/Chart.yaml that contain their own Chart.yaml |
annotations."helm.sh/images" |
package, umbrella |
|
Flux kustomization in base/flux/kustomization.yaml |
images |
umbrella |
Example |
Test images in tests/images.txt |
N/A | umbrella |
Example |
If you are looking at the images.txt from a Big Bang release and you do not see the image you expect, it is likely not
explicitly defined in one of these locations.
If it is defined and you still do not see it, please
create an issue.
What are the images-v2-* files and should I care about them?π
images-v2-dependencies.yamlimages-v2-with-dependencies.txtimages-v2-no-dependencies.txt
These new files were created from the ground up to take a true dependency graph approach that will be fully leveraged in the future.
They are available in both the package repos and the umbrella release level.
See an example of the images-v2-dependencies.yaml from the Big Bang 3.6.0 release.
For more technical details, please review ADR 8: Generate Images Metadata from Explicit References
Where does images.txt come from now?π
- The new dependency graph model is used to generate
images-v2-dependencies.yaml. images-v2-with-dependencies.txtis created.- A copy of
images-v2-with-dependencies.txtis saved asimages.txt
Therefore, images.txt is an exact replica of images-v2-with-dependencies.txt.
Is there anything I need to do differently?π
No. If you were using images.txt before, keep on using it, but know that the images listed in it only show up because
they are explicitly defined in one of the locations detailed above.
If you want to begin leveraging images metadata and are not already using images.txt, the recommendation is to use
images-v2-dependencies.yaml directly to get as close as possible to the source of truth.
Deprecation Planπ
images.txt will remain for the foreseeable future. If you are already using it, keep doing so.
Questions?π
If you have any questions, concerns, or ideas about this shift, please reach out to the Pipelines channel on MatterMost IL4.