1.2.0 changes how Enterprise image scanning works on Kubernetes. Update the image tag and restart; the database migrates on start.
The node pulls the image
A scan Job’s pod now runs the image itself, pinned to the digest your pods run, so the cluster pulls it exactly as it does for your workloads: through the nodes’ own registry access (for example an EKS node role reading ECR) or the image pull secrets your pods use. FleetOps copies those secrets into the fleetops-scan namespace for the scan and deletes them afterwards. FleetOps itself never needs registry credentials.
Nothing in the image runs. Its entrypoint is replaced by a Trivy binary that init containers copy in after refreshing the vulnerability and Java databases. The scan pod gets no service account token, and runs on a node of the same OS and architecture as your pods, so distroless and multi-platform images scan correctly.
When a scan can’t pull the image, the environment names the image and the reason, and a scan that can’t be scheduled fails fast instead of waiting out its deadline. A mirror Trivy image must be a copy of the official aquasec/trivy; turning scanning on checks this. Every setting that affects findings is pinned, so the same image gets the same result on Docker and on Kubernetes.
The permissions this needs are in the Kubernetes guide. The Cluster admin agent access level has all of them; with Standard, create the fleetops-scan namespace first.
Next
Port-forward, then API tokens, SIEM forwarding and long audit retention for Enterprise. See the roadmap.