Our automated schema version bump workflow releases a new cellxgene-schema version to PyPI then triggers a rebuild of the prod processing image, so that it picks up the latest version we just pushed to PyPI. However, currently there is a race condition such that the rebuild's request to PyPI sometimes still pulls the previous, now-outdated version of cellxgene-schema.
Account for this lag between a package version publish and it being available for downstream consumers of PyPI either in the schema-migration GHA workflow in single-cell-data-portal (where the image rebuild + pypi install is run) or the trigger-schema-migration workflow in single-cell-curation (where package is published)
In cases where the error occurs, it is easy to get around by simply commenting out the test release so that the second run only goes on prod pypi. Can be implemented at a later date.
Our automated schema version bump workflow releases a new cellxgene-schema version to PyPI then triggers a rebuild of the prod processing image, so that it picks up the latest version we just pushed to PyPI. However, currently there is a race condition such that the rebuild's request to PyPI sometimes still pulls the previous, now-outdated version of cellxgene-schema.
Account for this lag between a package version publish and it being available for downstream consumers of PyPI either in the schema-migration GHA workflow in single-cell-data-portal (where the image rebuild + pypi install is run) or the trigger-schema-migration workflow in single-cell-curation (where package is published)