Breaking down OpenAI's outage: How to avoid a hidden DNS dependency in Kubernetes - in this article, folks at Render analyze the recent OpenAI outage and provide some insights on the possible cause of the issue.

This article also serves as a reminder that you'd better run DNS and the Control Plane on separate nodes and also that the data in ETCD is organized by namespace, thus listing "all" resources is very costly, especially in large clusters, since you need to scan everything.

#kubernetes #postmortem