Rackspace told Spot customers that an attacker used leaked credentials on November 17 to act as legitimate users, create pods, and read Kubernetes Secrets and Spot API tokens in some customer clusters. Rackspace revoked tokens and rotated keys. Affected customers must reissue tokens, download a new kubeconfig and rotate every credential stored in their cluster Secrets.
If your team runs Kubernetes clusters on Rackspace Spot, treat every secret in them as exposed until you have rotated it. Rackspace told affected customers that an attacker used leaked credentials to get into parts of the Spot platform on November 17, between about 00:20 and 22:00 UTC, according to the notice obtained by BleepingComputer. Rackspace says it learned of the break-in on November 18 from a customer report.
What the attacker could do
- Act as your users. In some customer organizations, the attacker could operate as a legitimate user, Rackspace’s notice says.
- Run code in your cluster. That access let them create pods, list nodes and volumes, and read Kubernetes Secrets.
- Take your tokens. Rackspace warns the attacker could have read any Secret in an affected user’s clusters, plus any Spot API tokens or refresh tokens that user held.
Kubernetes Secrets usually hold database passwords, cloud keys and API tokens, so the exposure reaches past Spot to whatever those credentials open.
Rackspace says it revoked the affected tokens and sessions, removed the attacker’s pods, rotated keys, and tightened access controls and monitoring. It has not said how many customers were affected or whether data was taken; BleepingComputer asked and got no answer. To clean up, Rackspace shut off access to the Spot console, which left customers unable to manage clusters or rotate credentials for days. Sirish Raghuram, co-founder of Platform9, which built Spot with Rackspace, said on Reddit that shutting all access and sanitizing the environment was a deliberate choice. Rackspace said service was fully restored on November 21.
What to do now
Rackspace’s notice lists these steps for affected customers. They are worth doing even if you are not sure you were in scope:
- Reissue your Spot API tokens and replace any refresh tokens used in Terraform, CI or other automation.
- Download a fresh kubeconfig for each cluster and delete the old files.
- Rotate every credential stored in cluster Secrets. Start with cloud provider keys, database passwords, registry tokens and AI API keys.
- Check logs back to November 17 for your databases, container registries, Git hosts, cloud storage and third-party services, looking for logins or API calls you can’t explain.
This is our analysis. The pattern is the same one behind recent breaches at Northflank and Railway: a hosting platform gets in, and the keys your apps store there get out. Keys with spending limits and narrow scopes do less damage when that happens. Google’s research on stolen cloud keys running up $1 million AI bills shows how fast an exposed key gets used.
Related, November 21: Microsoft Entra Connect Health flaw exposed tenant admin emails. Expect targeted phishing.
