GPU VulnDB

Database/Container, Kubernetes & orchestration

RHACM multicloud-operators-channel: managed-cluster agent reads all Secrets in hub Channel namespaces

CVE-2026-73122Container, Kubernetes & orchestrationcurated

Impact

A single compromised managed cluster gets to read every Secret and ConfigMap in any Channel namespace on the RHACM hub, which is where the Git and Helm repository credentials for all the other managed clusters are kept. In a fleet where RHACM governs several GPU clusters for different tenants or business units, one foothold on the weakest cluster yields the deployment credentials for the rest, and those credentials typically allow pushing manifests that land as running workloads on GPU nodes. Red Hat scores it scope-changed (S:C) for exactly that reason: the disclosure crosses the boundary between the compromised cluster and the hub. Confidentiality only - the flaw does not by itself let the agent write to the hub.

Who can reach it

An attacker who already controls the RHACM agent on any single managed cluster, authenticated to the hub with that agent's own low-privilege credentials. No hub administrator access and no user interaction required.

What to do

Apply the RHSA errata matching your RHACM version (2.11 through 2.17 are all listed); this updates the hub multicloud-operators-channel controller and requires a rolling restart of the hub operator pods, not a node reboot. Because the exposure is credential disclosure, patching alone is not enough on a cluster you believe was already compromised: rotate the Git and Helm repository credentials stored in Channel namespaces afterwards.

References

Related entries

All Container, Kubernetes & orchestration entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.