unkinben 6a07f91ea1
ci/woodpecker/pr/pre-commit Pipeline was successful
ci/woodpecker/pr/test Pipeline was successful
ci/woodpecker/pr/build Pipeline was successful
Notify secondaries immediately on primary zone changes
Dynamically-updated primary zones were only reaching the secondary pods on
the hardcoded 1h SOA refresh: the operator emitted no NOTIFY, and a zone's
only apex NS is the primary itself, so default 'notify yes' reached no one.
Queries load-balanced across the serve VIP hit stale secondaries and
returned NXDOMAIN (negatively cached downstream for the 300s SOA minimum),
so records flapped for up to an hour after every update.

Add 'notify explicit' + 'also-notify' with the secondary pod IPs to primary
zone stanzas so an update NOTIFYs the secondaries for an immediate IXFR.
Applied via modzone, so existing zones pick it up on the next reconcile.
Also shorten the seed SOA refresh/retry/minimum as a fallback for missed
NOTIFYs and to shrink stale-NXDOMAIN negative caching.
2026-07-21 00:15:43 +10:00

bind-operator

A Kubernetes operator that manages fleets of BIND9 servers declaratively: StatefulSet-backed clusters with primary/secondary replication, and zones, views, TSIG keys, ACLs, catalog zones, RPZ policies and DNSSEC policies as custom resources.

Architecture

Each BindCluster is a StatefulSet plus a headless Service (stable per-pod DNS) and a client Service. Ordinal-0 is the primary; the remaining pods are secondaries that replicate via AXFR/IXFR + NOTIFY. A per-pod PVC holds zone databases and journals.

Zone content is delivered dynamically: the operator execs rndc addzone and TSIG nsupdate against the primary pod (the same write path external-dns uses). Cluster-wide config — options, controls, ACLs, views, dnssec-policy blocks and response-policy clauses — is rendered into a ConfigMap-backed named.conf and reloaded with rndc reconfig. New zones land on the secondaries automatically through a catalog zone, so secondaries never need per-zone reconfiguration.

BindCluster ──> StatefulSet (pod-0 = primary, pod-N = secondaries)
                 ├─ headless Service   (pod-0.<cluster>-headless.<ns>.svc…)
                 ├─ client Service     (ClusterIP / LoadBalancer)
                 ├─ ConfigMap          (named.conf.primary / .secondary + entrypoint)
                 ├─ Secret <cluster>-keys   (TSIG key clauses, included by named.conf)
                 └─ Secret <cluster>-rndc   (rndc control key)

BindZone / DNSRecord ──rndc addzone + nsupdate──> primary ──catalog + AXFR──> secondaries

The named.conf is rendered in two variants (primary/secondary); an entrypoint script picks one based on the pod ordinal.

Custom Resources

Kind Purpose
BindCluster A set of BIND9 servers. spec.mode: authoritative or resolver.
BindZone A forward/reverse zone (primary/secondary/forward/stub), records inline, optional dynamic-update + DNSSEC + catalog membership.
DNSRecord A single record set applied via TSIG nsupdate — external-dns as a CRD.
BindView A split-horizon view (match-clients, ordering, per-view recursion).
BindTSIGKey A TSIG key; the operator generates material into a Secret (never stored in the CR). spec.secretTemplate stamps extra labels/annotations onto that Secret (e.g. reflection hints to mirror it into another namespace).
BindACL A reusable named address_match_list.
BindCatalogZone A BIND catalog zone so secondaries auto-provision member zones.
BindPolicy A Response Policy Zone (RPZ) / DNS firewall.
BindDNSSECPolicy A dnssec-policy for automated signing.

See config/samples/ for worked examples.

Migration mapping

The three Puppet-managed BIND roles map onto three BindClusters:

Puppet role BindCluster Mode
3× authoritative masters auth authoritative (pod-0 primary, 2 secondaries)
3× only-resolvers resolver resolver (3 identical recursive servers)
3× external-dns externaldns authoritative (zones set dynamicUpdate for RFC2136 TSIG updates)

Development

make generate   # regenerate deepcopy, CRDs and RBAC from kubebuilder markers
make build      # build the operator binary
make test       # go test -race
make lint fmt   # go vet / gofmt

Local (kind)

kind create cluster --name bind
docker build -t bind-operator:dev -f Dockerfile.operator .
kind load docker-image bind-operator:dev --name bind

kubectl apply -f config/crd/bases/
kubectl apply -f hack/kind/manifests/
kubectl apply -f config/samples/

CI

Woodpecker runs pre-commit (gofmt + vet), test, and a dry-run image build on pull requests; pushing a v* tag builds and pushes git.unkin.net/unkin/bind-operator to the Gitea registry.

Notes & caveats

  • The BIND container image (spec.image, default internetsystemsconsortium/bind9:9.20) must ship named, rndc and nsupdate. The operator projects its config at /etc/bind-operator (leaving the image's own /etc/bind, including bind.keys, intact) and runs named -g -c /run/named/named.conf.
  • Dynamic updates authenticate with nsupdate -y; the TSIG secret is passed on the argv of an exec'd process inside the pod.
  • RPZ IP-trigger encodings (ip, client-ip, nsip) are emitted verbatim; QNAME and NSDNAME triggers are fully supported.
S
Description
Kubernetes operator for managing BIND9 DNS clusters, zones, views, and TSIG keys
Readme 527 KiB
Languages
Go 99.1%
Makefile 0.9%