Introduction:
We want to inform our customers about a known issue affecting Artifactory due to recent changes made by the Conda Forge upstream. This article explains the issue, its impact, and provides a recommended workaround to mitigate any disruptions.
Brief About the Upstream Changes:
On January 30 2025, it was identified that conda-forge added petsc-3.22.3-cuda11_complex_he184e62_101.conda 2 days ago that introduced a breaking change where the "track_features" property is holding a value of type List instead of the expected String. Conda Forge project uses the v1 format and the build rattler/rattler-build is used to build the project which had a bug with de/serializiation causing this issue of populating Lists instead of space delimited strings. This is potentially causing issues for anyone using rattler/rattler-build. Github pull request can be found on this link. The above change has introduced an issue when using Artifactory's virtual Conda repositories, which rely on data from the upstream repository.
Impact:
The above change is causing many integrations to break including Conda mutual customers using Conda and Artifactory users. Users with a Virtual Conda repository in Artifactory that includes a remote Conda Forge repository ((https://conda.anaconda.org/conda-forge) may experience issues when trying to resolve or fetch Conda packages.
Specifically, the failure occurs when trying to fetch the repodata.json file, which is crucial for resolving package metadata and dependencies.
As a result, users may encounter 404, 400, or 500 errors and be unable to use the Virtual Conda repository for package downloads. This failure prevents the Virtual Conda repository from functioning properly, making it impossible to retrieve package data through Artifactory.
Action items:
Feb 01: JFrog team is actively working to fix the issue with the relevant stakeholders of the Conda forge to get this rectified at the earliest. An issue has been raised on this github issue ticket.
Feb 02: JFrog has fixed the issue and the fix is implemented to all JFrog hosted SaaS servers. In case if any user is facing an issue, a user might need to Zap the cache of the repository as mentioned in the workaround section.
For On-prem installations, the JFrog Team has implemented the fix. More details are present on Artifactory Self Hosted Release Notes page.
We recommend you to follow this Article or the above github issues link for further update
Current Workaround:
Use the Conda Remote Repository Directly: Instead of relying on the Virtual Conda repository, users can directly use the Conda remote repository (repo.anaconda.com/conda-forge) to download packages.
Issues after upgrade:
If after upgrading you still experience the issue, it's due to the problematic artifact has not yet expired and still exists in the cache. To resolve this issue, you need to Zap the following folder to clear it from the Cache <repo-name>-cache/linux-64/.
