We want LibreDB Studio available in OpenShift's OperatorHub. Submissions go to redhat-openshift-ecosystem/community-operators-prod and have to be operators, so we will wrap the existing Helm chart with the operator-sdk helm plugin. No Go code is needed. The clusterpulse operator is a good reference since it does the same thing.
To do in this repo:
- Scaffold the operator:
operator-sdk init --plugins helm --domain libredb.org --group studio --version v1alpha1 --kind LibreDBStudio --helm-chart libredb-studio --helm-chart-repo https://libredb.org/libredb-studio/
- Add a release workflow that builds and pushes the controller image to
ghcr.io/libredb/libredb-studio-operator. It has to be public, because their CI installs it on a real cluster and has no credentials for a private registry.
The submission itself is a separate PR to community-operators-prod with just the bundle metadata. Their pipeline builds the bundle image, we do not push it.
Note on upkeep: a new bundle PR is needed for every chart release, and ci.yaml catalog_names has to be updated when a new OpenShift version ships, otherwise the checks fail. So this is something we keep maintaining, not set-and-forget.
We want LibreDB Studio available in OpenShift's OperatorHub. Submissions go to
redhat-openshift-ecosystem/community-operators-prodand have to be operators, so we will wrap the existing Helm chart with the operator-sdk helm plugin. No Go code is needed. Theclusterpulseoperator is a good reference since it does the same thing.To do in this repo:
operator-sdk init --plugins helm --domain libredb.org --group studio --version v1alpha1 --kind LibreDBStudio --helm-chart libredb-studio --helm-chart-repo https://libredb.org/libredb-studio/ghcr.io/libredb/libredb-studio-operator. It has to be public, because their CI installs it on a real cluster and has no credentials for a private registry.The submission itself is a separate PR to community-operators-prod with just the bundle metadata. Their pipeline builds the bundle image, we do not push it.
Note on upkeep: a new bundle PR is needed for every chart release, and
ci.yamlcatalog_names has to be updated when a new OpenShift version ships, otherwise the checks fail. So this is something we keep maintaining, not set-and-forget.