Feature State: alpha as of version v1.123.1

Config Connector can be configured to enable experimental versions of direct controllers for reconciling specific resources. This will allow users to test the new direct controller code for bug fixes or features not available in legacy controllers.

You can enable direct controllers for all resources of a specific kind within a namespace using the experiments.controllerOverrides field in the ConfigConnectorContext resource. For more details and examples, see the Controller Implementation Overrides documentation.

To verify the direct controller code is in-use for a particular resource, check the logs for the phrase “Running reconcile” associated with the resource.

$ kubectl -n cnrm-system logs pod/cnrm-controller-manager-0 | grep -e "Running reconcile" -e "my-sqlinstance"

You should see output similar to the following:

{“severity”:”info”,”timestamp”:”2024-11-04T21:56:28.704Z”,”msg”:”Running reconcile”,”controller”:”sqlinstance-controller”,”controllerGroup”:”sql.cnrm.cloud.google.com”,”controllerKind”:”SQLInstance”,”SQLInstance”:{“name”:”my-sqlinstance”,”namespace”:”default”},”namespace”:”default”,”name”:”my-sqlinstance”,”reconcileID”:”577fe58d-14eb-4e5d-9642-c9ec4b1a3137″,”resource”:{“name”:”my-sqlinstance”,”namespace”:”default”}}

If the legacy controller is enabled (instead of the new direct controller), the logs will show “starting reconcile” for the resource instead of “Running reconcile”.

{“severity”:”info”,”timestamp”:”2024-10-30T23:47:49.726Z”,”logger”:”sqlinstance-controller”,”msg”:”starting reconcile”,”resource”:{“name”:”my-sqlinstance”,”namespace”:”default”}}

If you’re unsure if a particular resource has an experimental direct controller available and want to try out the new code wherever possible (or if you’re just curious to learn more about the internals of this feature in Config Connector), read on.

To find out if an experimental direct controller is available for a resource, check CRD for the resource.

If the CRD specifies either of the following legacy controller labels, then it is potentially possible to enable an experimental direct controller for resources of that kind.

• cnrm.cloud.google.com/tf2crd: true
• cnrm.cloud.google.com/dcl2crd: true

To check the CRD labels for a resource type, you can run the following command:

kubectl get crd sqlinstances.sql.cnrm.cloud.google.com -ojson | jq '.metadata.labels'

Alternatively, you can check in the source tree under config/crds/resources/ for the CRD YAML, but remember to double-check your source checkout matches your running Config Connector version.

However, not all resources with the legacy controller annotation on their CRDs will have an experimental direct controller available. This is because not all resources with the legacy controller annotation have a direct controller implemented yet; many of the direct controllers are not yet implemented. To confirm if there is an experimental direct controller available for a particular resource, check to make sure there is an implementation of the controller for that type in pkg/controller/direct. The controllers are found in *_controller.go [example]. Also, verify that the direct controller is registered in pkg/controller/direct/register/register.go [example].

At this point, if there is both a legacy controller label on the CRD and a direct controller implemented + registered, then you can be reasonably sure there is a direct controller available. To try it out, add the override in the ConfigConnectorContext for a resource of that kind, and use the steps above to verify the direct controller is enabled.