A ~6 minute read — just three concepts that, once you know them, change how you reason about DNS inside a cluster.
While chasing some DNS timeouts recently, I went down a rabbit hole and came out with three concepts I wish I'd known earlier. None of them is exotic, but together they explain a surprising amount of "why is DNS being weird" behaviour on Kubernetes-on-AWS.
ndots — why one hostname lookup can become many DNS queries
NodeLocal DNS — the per-node caching layer your queries actually hit first
The EC2 per-ENI DNS packet limit — a hard ceiling most people never hear about until they hit it
