Going from 5s to 5ms: Benefits of a Node-Local DNSCache
summary
Kubernetes 1.13 introduces a node-local-dns-cache that runs a caching agent as a DaemonSet on each node, listening on a link‑local IP so pods use the local resolver instead of cluster DNS. It combines CoreDNS cache and forward plugins, upgrades upstream queries to TCP, exposes per‑node metrics, and uses a CoreDNS prefetch plugin to reduce misses. Real deployments report drop from 5s timeouts to sub‑millisecond medians, and a Postmates cluster handles ~50,000 QPS with 64‑core nodes.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:18Agenda
- 2:09What is Local DNSCache
- 2:33Why do we need this
- 3:00Why are DNS latency issues showing up
- 4:39Other DNS latency issues
- 6:19DNS workarounds
- 8:16How DNS fits in
- 11:33Visibility metrics
- 18:24How to run it
- 19:59Future features
- 24:05Autopath
- 25:27New workflow
- 26:10Search path expansion
- 28:08QA
