Cluster BareMetal Operator (CBO)

This is a work-in-progress Second Level Operator (SLO) for bare metal.

See OpenShift enhancement proposal #212 where the design is being discussed

Hacking

To hack on his operator, you should first deploy a bare metal cluster using openshift-metal3/dev-scripts.

Then build and push an image to your private quay.io repo:

$> make image push
Building image quay.io/markmc/cluster-baremetal-operator:v0.0.1...
...
Pushing images...
...

Finally, use the hack/run-image.sh script to scale down the Machine API Operator (MAO), and create a CBO deployment with your newly built image:

$> hack/run-image.sh

TODO

This work is in its very early stages, so this is just a rough TODO list for now.

Short term:

• Allow make image push and hack/run-image.sh use other quay repos than markmc/cluster-baremetal-operator
• Build a release image which includes CBO and test that it works as expected
• Update the enhancement with latest thinking and findings - e.g. continuing to use the openshift-machine-api namespace
• Implement correct ClusterOperator semantics
• Figure out an in-place upgrade plan
• Implement "disabled SLO" behavior on non-baremetal platforms
• Unit tests - what have we got, what are we missing?
• E2E tests

Exiting prototype phase:

• Move this repo into the openshift org
• Add https://github.com/openshift/release config for building and publishing images

MAO changes to pick up:

• openshift/machine-api-operator#498 - hardware inventory management DaemonSet
• openshift/machine-api-operator#547 - podman support

Longer-term or lower priority:

• Move in-memory metal3 deployment resource generation tracking to a CR with OperatorStatus
• Metrics, alerts, events - what have we got? what are we missing?
• RBAC audit - do our roles have the minimal set of permissions?