Sanitization

Sanitization removes cluster-generated noise from non-excluded resources so the exported YAML is closer to something a human would commit.

Pipeline

1. Deep-clone the live object.
2. Remove server-assigned metadata.
3. Remove noisy annotations.
4. Remove system finalizers.
5. Remove top-level runtime fields and common defaults.
6. Remove pod-template defaults.
7. Apply kind-specific cleanup.
8. Apply the selected Secret handling mode.

The original object returned by the console SDK is not modified.

Metadata Removed

These fields are removed from metadata:

• uid
• resourceVersion
• generation
• creationTimestamp
• managedFields
• selfLink
• ownerReferences

Annotation Cleanup

Exact annotation keys removed:

Annotation	Why removed
kubectl.kubernetes.io/last-applied-configuration	Client-side apply bookkeeping
deployment.kubernetes.io/revision	Runtime rollout state
openshift.io/generated-by	Generated object marker
openshift.io/host.generated	Route host generation marker

Annotation prefixes removed:

• pv.kubernetes.io/
• operator.openshift.io/
• openshift.io/build.

If metadata.annotations becomes empty, it is removed.

Runtime and Default Fields

The plugin removes status, spec.nodeName, and spec.schedulerName: default-scheduler.

For pod templates, it removes common Kubernetes defaults such as dnsPolicy: ClusterFirst, restartPolicy: Always, empty securityContext, and default container termination message fields.

Kind-specific Cleanup

Kind	Cleanup
Deployment	Drops default progress deadline, revision history, minimum ready seconds, and default rolling update strategy values.
Secret	Redacts values by default, omits the resource when Secret handling is omit, or keeps values when Secret handling is include.
PersistentVolumeClaim	Drops spec.volumeName.
Service	Drops cluster-assigned IP fields and default session affinity.
Service type LoadBalancer	Also drops load balancer implementation annotations.
ImageStream	Drops spec.dockerImageRepository.

Sanitization is intentionally practical, not exhaustive. The goal is to remove common cluster-generated drift while leaving enough context for review.