GPU VulnDB

Database/Container, Kubernetes & orchestration

RHACM GitOpsCluster: tenant can redirect spoke cluster bearer tokens into a namespace they control

CVE-2026-70398Container, Kubernetes & orchestrationcurated

Impact

An authenticated tenant can manipulate the GitOpsCluster controller so that bearer tokens for managed spoke clusters are written into a namespace the tenant owns. Those tokens are the hub's credentials for driving the spokes, so recovering one means acting against a managed cluster directly and stepping around the ArgoCD AppProject boundaries that were supposed to limit which clusters a tenant can deploy to. In a GPU fleet managed by ACM, a spoke token is effectively a key to whatever the ACM agent can do on that cluster's nodes and workloads. The tokens are long-lived, so the exposure persists until they are rotated - patching alone does not undo a leak that already happened.

Who can reach it

Any authenticated ACM hub user (tenant) able to influence GitOpsCluster resources and read secrets in a namespace they control. Network-reachable via the hub API; no node access needed.

What to do

Apply the Red Hat errata for your ACM stream (RHSA-2026:60386 through 60390, covering 2.11 and 2.13 through 2.17). The fix is a hub-side operator update that rolls the multicloud-integrations controller pods - no spoke or GPU node reboot. Because this is a credential-disclosure flaw, also rotate spoke cluster bearer tokens and audit tenant namespaces for copied secrets after patching.

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.