GPU VulnDB

Database/Control plane, storage & DevOps

Grafana: unbounded request body on the Prometheus rule conversion API crashes the server via out-of-memory

CVSS 6.5CVE-2026-33374Control plane, storage & DevOpscurated

Impact

The Prometheus rule conversion endpoints accept a request body with no size limit, so an authenticated user can force unbounded memory allocation and take the Grafana process down with an out-of-memory kill. Availability only - no data is exposed or altered. It matters because Grafana is usually the single pane an operations team watches: losing it does not stop the GPU fleet, but it does leave the people on call without dashboards or the alert UI until the process comes back, and a user who can send the request can send it again. Any authenticated account is enough, which on a multi-tenant or widely shared instance is a large set of people.

Who can reach it

Any authenticated Grafana user able to reach the rule conversion API over the network. No elevated role is stated as required.

What to do

Upgrade to the fixed Grafana versions in the vendor advisory and restart the server. Until then, put a request body size limit in front of Grafana at the reverse proxy and make sure the Grafana process has a memory limit and automatic restart configured, so an OOM is a blip rather than an outage someone has to notice first.

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.