There are a few situations in which a cluster admin might want to trigger a rebuild of their OS image in addition to situations where cluster state may dictate that we should perform a rebuild. For example, if the custom Dockerfile changes or the machine-config-osimageurl changes, it would be desirable to perform a rebuild in that case. To that end, this particular story covers adding the foundation for a rebuild mechanism in the form of an annotation that can be applied to the target MachineConfigPool. What is out of scope for this story is applying this annotation in response to a change in cluster state (e.g., custom Dockerfile change).

Done When:

• BuildController is aware of and recognizes a special annotation on layered MachineConfigPools (e.g., machineconfiguration.openshift.io/rebuildImage).
• Upon recognizing that a MachineConfigPool has this annotation, BuildController will clear any failed build attempts, delete any failed builds and their related ephemeral objects (e.g., rendered Dockerfile / MachineConfig ConfigMaps), and schedule a new build to be performed.
• This annotation should be removed when the build process completes, regardless of outcome. In other words, should the build success or fail, the annotation should be removed.
•  [optional] BuildController keeps track of the number of retries for a given MachineConfigPool. This can occur via another annotation, e.g., machineconfiguration.openshift.io/buildRetries=1 . For now, this can be a hard-coded value (e.g., 5), but in the future, this could be wired up to an end-user facing knob. This annotation should be cleared upon a successful rebuild. If the rebuild is reached, then we should degrade.