DNS is the service everybody depends on and nobody thinks about, until it breaks. Then nothing works. Running your own BIND servers gives you control over what resolves, how fast, and who can change it. It also puts you on the hook when something goes wrong.
This series is how I run BIND: a recursive resolver for the network, an authoritative server for my own zones, and everything around them that makes the pair safe to depend on. Each part is a self-contained guide with working configuration. Read in order, they build one setup step by step.
What you’ll build #
- A caching resolver for your clients and a separate authoritative server for your zones. They are two roles that shouldn’t share a server.
- Secure management: remote control with
rndc, TSIG keys, views and ACLs that decide who may query, transfer or update what. - Automation: dynamic updates from DHCP and scripts, and catalog zones so secondaries pick up new zones by themselves.
- Trust: DNSSEC validation on the resolver, signed zones on the authoritative side, and Response Policy Zones to block malicious domains.
- Privacy: DNS over TLS and DNS over HTTPS, so queries don’t cross the network in plain text.
- Integration and visibility: Kubernetes records managed by External-DNS, and query metrics in Prometheus and Grafana.
How the series fits together #
- Foundations (parts 1–2). Install BIND as a resolver, then as an authoritative-only server. Everything after this assumes one or both.
- Running your zones (parts 3–6). Part 3 creates the TSIG keys that parts 5 and 12 reuse, so don’t skip it. Part 4 locks down who can query and transfer. Parts 5 and 6 take the manual work out of record and zone changes.
- Trust and protection (parts 7–9). Validate first (part 7), then sign your own zones (part 8). Part 9 uses RPZ to turn the resolver into a security filter.
- Encrypted DNS (parts 10–11). Part 10 covers native DoT in BIND 9.17 and later, plus stunnel for older versions. Part 11 adds DoH through nginx, with my honest opinion on when it is worth it.
- Integrate and observe (parts 12–13). Let Kubernetes manage its own records over RFC 2136, and monitor the lot.
Pick a path #
Not everyone needs all thirteen parts:
- A resolver for your home lab or office: parts 1, 7, 9 and 10.
- Hosting your own domains: parts 2, 3, 4 and 8, then 6 when the number of zones grows.
- DNS for Kubernetes: parts 2, 3, 5 and 12.
Before you start #
You need a Linux server with root access. The install steps cover RHEL and Rocky Linux as well as Debian and Ubuntu. You should know what A, NS and SOA records are. The rest is explained along the way. Where a feature needs a specific BIND version, the part says so.
Running two resolvers? Put them behind a load balancer with HAProxy DNS load balancing, part 6 of the HAProxy series.