GPU VulnDB

Database/Control plane, storage & DevOps

AWX bulk job launch: read-level permission on an instance group is enough to run jobs on it

CVE-2026-84470Control plane, storage & DevOpscurated

Impact

The bulk job launch endpoint (POST /api/v2/bulk/job_launch/) checks only read permission on the requested instance groups, while the ordinary single-job launch path requires use permission on the same field. A principal holding execute on a job template plus read on an instance group - the built-in read-only System Auditor role is sufficient - can therefore place jobs onto execution nodes they were never authorized to use. Where instance groups are the placement boundary between environments, that boundary is gone for the bulk path: jobs land on execution nodes with those nodes' network reach and credentials, and consume their capacity. For fleets that carve out a dedicated execution group for GPU or otherwise sensitive nodes, this is the control that was supposed to keep unprivileged automation off them.

Who can reach it

Any authenticated automation-controller user who holds execute on a job template and read-level visibility of the target instance group, over the controller's REST API. No administrative role is needed.

What to do

No fixed version is stated in the record - it links the Red Hat CVE page and Bugzilla only. Apply the automation-controller erratum when Red Hat ships it and restart the controller services; the fix is application-side, so no node drain or reboot is involved. Until then, tighten who holds job-template execute permission, and review which roles have read visibility of instance groups that back sensitive execution nodes.

References

Related entries

All Control plane, storage & DevOps 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.