Merge pull request 'Notify secondaries immediately on primary zone changes' (#13) from benvin/notify-secondaries into main

Reviewed-on: #13

README.md

@@ -42,7 +42,7 @@ script picks one based on the pod ordinal.
 | `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). |
+| `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. |

api/v1alpha1/bindtsigkey_types.go

@@ -41,6 +41,24 @@ type BindTSIGKeySpec struct {
 	// `secret` key and the operator will not generate new material.
 	// +optional
 	ImportExisting bool `json:"importExisting,omitempty"`
+
+	// SecretTemplate customizes metadata written onto the managed key Secret.
+	// Useful, for example, to let secret-reflection tooling mirror the key into
+	// another namespace. Operator-managed labels are always preserved.
+	// +optional
+	SecretTemplate *SecretMetadata `json:"secretTemplate,omitempty"`
+}
+
+// SecretMetadata carries extra labels and annotations to stamp onto a
+// Secret managed by the operator.
+type SecretMetadata struct {
+	// Annotations to set on the Secret.
+	// +optional
+	Annotations map[string]string `json:"annotations,omitempty"`
+
+	// Labels to set on the Secret.
+	// +optional
+	Labels map[string]string `json:"labels,omitempty"`
 }
 
 // BindTSIGKeyStatus reports observed TSIG key state.

api/v1alpha1/zz_generated.deepcopy.go

@@ -691,7 +691,7 @@ func (in *BindTSIGKey) DeepCopyInto(out *BindTSIGKey) {
 	*out = *in
 	out.TypeMeta = in.TypeMeta
 	in.ObjectMeta.DeepCopyInto(&out.ObjectMeta)
-	out.Spec = in.Spec
+	in.Spec.DeepCopyInto(&out.Spec)
 	in.Status.DeepCopyInto(&out.Status)
 }
 
@@ -748,6 +748,11 @@ func (in *BindTSIGKeyList) DeepCopyObject() runtime.Object {
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *BindTSIGKeySpec) DeepCopyInto(out *BindTSIGKeySpec) {
 	*out = *in
+	if in.SecretTemplate != nil {
+		in, out := &in.SecretTemplate, &out.SecretTemplate
+		*out = new(SecretMetadata)
+		(*in).DeepCopyInto(*out)
+	}
 }
 
 // DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new BindTSIGKeySpec.
@@ -1208,3 +1213,32 @@ func (in *Record) DeepCopy() *Record {
 	in.DeepCopyInto(out)
 	return out
 }
+
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *SecretMetadata) DeepCopyInto(out *SecretMetadata) {
+	*out = *in
+	if in.Annotations != nil {
+		in, out := &in.Annotations, &out.Annotations
+		*out = make(map[string]string, len(*in))
+		for key, val := range *in {
+			(*out)[key] = val
+		}
+	}
+	if in.Labels != nil {
+		in, out := &in.Labels, &out.Labels
+		*out = make(map[string]string, len(*in))
+		for key, val := range *in {
+			(*out)[key] = val
+		}
+	}
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new SecretMetadata.
+func (in *SecretMetadata) DeepCopy() *SecretMetadata {
+	if in == nil {
+		return nil
+	}
+	out := new(SecretMetadata)
+	in.DeepCopyInto(out)
+	return out
+}

config/crd/bases/bind.unkin.net_bindtsigkeys.yaml

@@ -87,6 +87,23 @@ spec:
                   SecretName is the Secret the key material is written to (or read from when
                   ImportExisting is set). Defaults to "<name>-tsig".
                 type: string
+              secretTemplate:
+                description: |-
+                  SecretTemplate customizes metadata written onto the managed key Secret.
+                  Useful, for example, to let secret-reflection tooling mirror the key into
+                  another namespace. Operator-managed labels are always preserved.
+                properties:
+                  annotations:
+                    additionalProperties:
+                      type: string
+                    description: Annotations to set on the Secret.
+                    type: object
+                  labels:
+                    additionalProperties:
+                      type: string
+                    description: Labels to set on the Secret.
+                    type: object
+                type: object
             type: object
           status:
             description: BindTSIGKeyStatus reports observed TSIG key state.

config/crd/install.yaml

@@ -2376,6 +2376,23 @@ spec:
                   SecretName is the Secret the key material is written to (or read from when
                   ImportExisting is set). Defaults to "<name>-tsig".
                 type: string
+              secretTemplate:
+                description: |-
+                  SecretTemplate customizes metadata written onto the managed key Secret.
+                  Useful, for example, to let secret-reflection tooling mirror the key into
+                  another namespace. Operator-managed labels are always preserved.
+                properties:
+                  annotations:
+                    additionalProperties:
+                      type: string
+                    description: Annotations to set on the Secret.
+                    type: object
+                  labels:
+                    additionalProperties:
+                      type: string
+                    description: Labels to set on the Secret.
+                    type: object
+                type: object
             type: object
           status:
             description: BindTSIGKeyStatus reports observed TSIG key state.

config/samples/00-tsigkeys.yaml

@@ -11,7 +11,9 @@ spec:
   algorithm: hmac-sha256
 ---
 # TSIG key permitting external-dns (and DNSRecord objects) to send RFC2136
-# dynamic updates to the dynamic cluster's primary.
+# dynamic updates to the dynamic cluster's primary. secretTemplate mirrors the
+# generated Secret into the external-dns namespace via emberstack reflector, so
+# external-dns presents exactly the key the primary's allow-update accepts.
 apiVersion: bind.unkin.net/v1alpha1
 kind: BindTSIGKey
 metadata:
@@ -19,3 +21,9 @@ metadata:
   namespace: bind-externaldns
 spec:
   algorithm: hmac-sha256
+  secretTemplate:
+    annotations:
+      reflector.v1.k8s.emberstack.com/reflection-allowed: "true"
+      reflector.v1.k8s.emberstack.com/reflection-allowed-namespaces: "externaldns"
+      reflector.v1.k8s.emberstack.com/reflection-auto-enabled: "true"
+      reflector.v1.k8s.emberstack.com/reflection-auto-namespaces: "externaldns"

go.mod

@@ -3,6 +3,7 @@ module git.unkin.net/unkin/bind-operator
 go 1.25
 
 require (
+	github.com/go-logr/logr v1.4.2
 	k8s.io/api v0.34.4
 	k8s.io/apimachinery v0.34.4
 	k8s.io/client-go v0.34.4
@@ -17,7 +18,6 @@ require (
 	github.com/evanphx/json-patch/v5 v5.9.11 // indirect
 	github.com/fsnotify/fsnotify v1.9.0 // indirect
 	github.com/fxamacker/cbor/v2 v2.9.0 // indirect
-	github.com/go-logr/logr v1.4.2 // indirect
 	github.com/go-logr/zapr v1.3.0 // indirect
 	github.com/go-openapi/jsonpointer v0.21.0 // indirect
 	github.com/go-openapi/jsonreference v0.20.2 // indirect

internal/bind/seed.go

@@ -35,13 +35,19 @@ func (e *Executor) ZoneExists(ctx context.Context, namespace, pod, zone, view st
 func (e *Executor) WriteSeedZone(ctx context.Context, namespace, pod, zone, path, primaryIP string, serial int64) error {
 	origin := dot(zone)
 	ns := "ns1." + origin
+	// Short refresh/retry so a secondary that misses a NOTIFY (e.g. its pod IP
+	// changed and the primary's also-notify was briefly stale) still converges
+	// in minutes, not the hour a 3600s refresh would impose. minimum is the
+	// negative-cache TTL: keep it low so a stale-secondary NXDOMAIN does not
+	// stick in downstream resolvers for long. NOTIFY (also-notify on the
+	// primary) remains the fast path; these are the fallback.
 	content := fmt.Sprintf(`$TTL 3600
 @ IN SOA %s hostmaster.%s (
     %d ; serial
-    3600       ; refresh
-    900        ; retry
+    300        ; refresh
+    60         ; retry
     1209600    ; expire
-    300 )      ; minimum
+    60 )       ; minimum
 @ IN NS %s
 ns1 IN A %s
 `, ns, origin, serial, ns, primaryIP)

internal/controller/bindtsigkey_controller.go

@@ -60,7 +60,7 @@ func (r *BindTSIGKeyReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 			return ctrl.Result{}, genErr
 		}
 		newSecret := &corev1.Secret{
-			ObjectMeta: metav1.ObjectMeta{Name: secretName, Namespace: key.Namespace, Labels: map[string]string{managedByLabel: managedByValue}},
+			ObjectMeta: metav1.ObjectMeta{Name: secretName, Namespace: key.Namespace},
 			Data: map[string][]byte{
 				"algorithm": []byte(algorithm),
 				"keyName":   []byte(keyName),
@@ -68,6 +68,7 @@ func (r *BindTSIGKeyReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 				"key.conf":  []byte(bind.KeyClause(keyName, algorithm, material)),
 			},
 		}
+		applySecretTemplate(&newSecret.ObjectMeta, key.Spec.SecretTemplate)
 		if err := ctrl.SetControllerReference(&key, newSecret, r.Scheme); err != nil {
 			return ctrl.Result{}, err
 		}
@@ -77,6 +78,21 @@ func (r *BindTSIGKeyReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 		logger.Info("generated TSIG key", "key", key.Name, "secret", secretName)
 	case err != nil:
 		return ctrl.Result{}, err
+	default:
+		// Secret already exists: reconcile the template-managed metadata so that
+		// annotation/label changes on the CR (e.g. reflection hints) propagate
+		// without regenerating key material. Skip imported secrets, which are
+		// owned by an external manager (Vault/VSO, reflector) that we must not
+		// fight over metadata.
+		if key.Spec.ImportExisting {
+			break
+		}
+		if updated := applySecretTemplate(&secret.ObjectMeta, key.Spec.SecretTemplate); updated {
+			if err := r.Update(ctx, &secret); err != nil {
+				return ctrl.Result{}, err
+			}
+			logger.Info("updated TSIG key secret metadata", "key", key.Name, "secret", secretName)
+		}
 	}
 
 	key.Status.SecretName = secretName
@@ -90,6 +106,42 @@ func (r *BindTSIGKeyReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 	return ctrl.Result{}, nil
 }
 
+// applySecretTemplate stamps the operator-managed label plus any
+// user-supplied labels/annotations onto the Secret's metadata. It returns true
+// if it mutated meta, so callers can decide whether an update is needed.
+func applySecretTemplate(meta *metav1.ObjectMeta, tmpl *bindv1alpha1.SecretMetadata) bool {
+	changed := false
+	setLabel := func(k, v string) {
+		if meta.Labels == nil {
+			meta.Labels = map[string]string{}
+		}
+		if meta.Labels[k] != v {
+			meta.Labels[k] = v
+			changed = true
+		}
+	}
+	setAnnotation := func(k, v string) {
+		if meta.Annotations == nil {
+			meta.Annotations = map[string]string{}
+		}
+		if meta.Annotations[k] != v {
+			meta.Annotations[k] = v
+			changed = true
+		}
+	}
+
+	setLabel(managedByLabel, managedByValue)
+	if tmpl != nil {
+		for k, v := range tmpl.Labels {
+			setLabel(k, v)
+		}
+		for k, v := range tmpl.Annotations {
+			setAnnotation(k, v)
+		}
+	}
+	return changed
+}
+
 func (r *BindTSIGKeyReconciler) fail(ctx context.Context, key *bindv1alpha1.BindTSIGKey, reason, msg string) (ctrl.Result, error) {
 	key.Status.Ready = false
 	key.Status.ObservedGeneration = key.Generation

internal/controller/bindtsigkey_test.go

@@ -0,0 +1,71 @@
+package controller
+
+import (
+	"testing"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+	bindv1alpha1 "git.unkin.net/unkin/bind-operator/api/v1alpha1"
+)
+
+func TestApplySecretTemplate(t *testing.T) {
+	t.Run("nil template still stamps managed-by label", func(t *testing.T) {
+		var meta metav1.ObjectMeta
+		if !applySecretTemplate(&meta, nil) {
+			t.Fatal("expected change on empty meta")
+		}
+		if meta.Labels[managedByLabel] != managedByValue {
+			t.Errorf("managed-by label = %q, want %q", meta.Labels[managedByLabel], managedByValue)
+		}
+		if meta.Annotations != nil {
+			t.Errorf("annotations = %v, want nil", meta.Annotations)
+		}
+	})
+
+	t.Run("applies labels and annotations", func(t *testing.T) {
+		meta := metav1.ObjectMeta{Labels: map[string]string{managedByLabel: managedByValue}}
+		tmpl := &bindv1alpha1.SecretMetadata{
+			Annotations: map[string]string{"reflector.v1.k8s.emberstack.com/reflection-allowed": "true"},
+			Labels:      map[string]string{"team": "dns"},
+		}
+		if !applySecretTemplate(&meta, tmpl) {
+			t.Fatal("expected change when adding template metadata")
+		}
+		if got := meta.Annotations["reflector.v1.k8s.emberstack.com/reflection-allowed"]; got != "true" {
+			t.Errorf("reflection annotation = %q, want true", got)
+		}
+		if meta.Labels["team"] != "dns" {
+			t.Errorf("team label = %q, want dns", meta.Labels["team"])
+		}
+		// managed-by must survive user-supplied labels.
+		if meta.Labels[managedByLabel] != managedByValue {
+			t.Errorf("managed-by label dropped: %v", meta.Labels)
+		}
+	})
+
+	t.Run("idempotent when already applied", func(t *testing.T) {
+		tmpl := &bindv1alpha1.SecretMetadata{
+			Annotations: map[string]string{"a": "1"},
+			Labels:      map[string]string{"b": "2"},
+		}
+		meta := metav1.ObjectMeta{}
+		applySecretTemplate(&meta, tmpl)
+		if applySecretTemplate(&meta, tmpl) {
+			t.Error("expected no change on second apply")
+		}
+	})
+
+	t.Run("updates drifted annotation value", func(t *testing.T) {
+		meta := metav1.ObjectMeta{
+			Labels:      map[string]string{managedByLabel: managedByValue},
+			Annotations: map[string]string{"a": "old"},
+		}
+		tmpl := &bindv1alpha1.SecretMetadata{Annotations: map[string]string{"a": "new"}}
+		if !applySecretTemplate(&meta, tmpl) {
+			t.Fatal("expected change when annotation value drifts")
+		}
+		if meta.Annotations["a"] != "new" {
+			t.Errorf("annotation a = %q, want new", meta.Annotations["a"])
+		}
+	})
+}

internal/controller/bindzone_controller.go

@@ -80,7 +80,15 @@ func (r *BindZoneReconciler) Reconcile(ctx context.Context, req ctrl.Request) (c
 		return r.setPhase(ctx, &zone, "Pending", "PrimaryNotReady", "waiting for cluster primary to be ready")
 	}
 
-	zoneConfig, err := r.buildZoneConfig(ctx, &zone, r.zoneTransferKeyRef(ctx, &zone, cluster))
+	// Primary zones replicated to secondaries (catalog members) get an
+	// also-notify pointing at the secondary pods, so a dynamic update NOTIFYs
+	// them immediately rather than waiting for the SOA refresh.
+	var notifyTargets []string
+	if isPrimaryType(zone.Spec.Type) && catalogEnabled(&zone) {
+		notifyTargets = secondaryPodIPs(ctx, r.Client, cluster)
+	}
+
+	zoneConfig, err := r.buildZoneConfig(ctx, &zone, r.zoneTransferKeyRef(ctx, &zone, cluster), notifyTargets)
 	if err != nil {
 		return r.setPhase(ctx, &zone, "Error", "ConfigError", err.Error())
 	}
@@ -135,7 +143,7 @@ func (r *BindZoneReconciler) Reconcile(ctx context.Context, req ctrl.Request) (c
 // buildZoneConfig renders the inner clause passed to rndc addzone/modzone.
 // transferKey, when set, is the catalog transfer TSIG key name; catalog member
 // primary zones must allow AXFR with it so secondaries can pull them.
-func (r *BindZoneReconciler) buildZoneConfig(ctx context.Context, zone *bindv1alpha1.BindZone, transferKey string) (string, error) {
+func (r *BindZoneReconciler) buildZoneConfig(ctx context.Context, zone *bindv1alpha1.BindZone, transferKey string, notifyTargets []string) (string, error) {
 	zType := zone.Spec.Type
 	if zType == "" {
 		zType = bindv1alpha1.ZonePrimary
@@ -154,6 +162,12 @@ func (r *BindZoneReconciler) buildZoneConfig(ctx context.Context, zone *bindv1al
 			// Catalog member: permit key-authenticated AXFR from secondaries.
 			parts = append(parts, fmt.Sprintf("allow-transfer { key \"%s\"; }", transferKey))
 		}
+		// NOTIFY only the secondaries we know about (their apex NS is the primary
+		// itself, so default `notify yes` would reach no one). `notify explicit`
+		// keeps NOTIFY off the query-serving VIP and scoped to the pod IPs.
+		if len(notifyTargets) > 0 {
+			parts = append(parts, "notify explicit", fmt.Sprintf("also-notify { %s }", terminateInline(notifyTargets)))
+		}
 		if zone.Spec.DNSSECPolicyRef != "" {
 			parts = append(parts, fmt.Sprintf("dnssec-policy \"%s\"", zone.Spec.DNSSECPolicyRef), "inline-signing yes")
 		}

internal/controller/bindzone_notify_test.go

@@ -0,0 +1,48 @@
+package controller
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	bindv1alpha1 "git.unkin.net/unkin/bind-operator/api/v1alpha1"
+)
+
+// A primary zone with known secondaries renders notify explicit + also-notify
+// so a dynamic update NOTIFYs the secondaries immediately.
+func TestBuildZoneConfigPrimaryAlsoNotify(t *testing.T) {
+	r := &BindZoneReconciler{}
+	zone := &bindv1alpha1.BindZone{
+		Spec: bindv1alpha1.BindZoneSpec{
+			ZoneName: "main.unkin.net",
+			Type:     bindv1alpha1.ZonePrimary,
+		},
+	}
+
+	cfg, err := r.buildZoneConfig(context.Background(), zone, "transfer-key", []string{"10.42.2.6", "10.42.1.5"})
+	if err != nil {
+		t.Fatalf("buildZoneConfig: %v", err)
+	}
+	if !strings.Contains(cfg, "notify explicit") {
+		t.Errorf("expected notify explicit in %q", cfg)
+	}
+	if !strings.Contains(cfg, "also-notify { 10.42.2.6; 10.42.1.5; }") {
+		t.Errorf("expected also-notify with the secondary IPs in %q", cfg)
+	}
+}
+
+// With no secondaries, no also-notify is emitted (single-replica cluster).
+func TestBuildZoneConfigPrimaryNoNotifyTargets(t *testing.T) {
+	r := &BindZoneReconciler{}
+	zone := &bindv1alpha1.BindZone{
+		Spec: bindv1alpha1.BindZoneSpec{ZoneName: "main.unkin.net", Type: bindv1alpha1.ZonePrimary},
+	}
+
+	cfg, err := r.buildZoneConfig(context.Background(), zone, "transfer-key", nil)
+	if err != nil {
+		t.Fatalf("buildZoneConfig: %v", err)
+	}
+	if strings.Contains(cfg, "also-notify") || strings.Contains(cfg, "notify explicit") {
+		t.Errorf("did not expect notify clauses with no targets: %q", cfg)
+	}
+}

internal/controller/helpers.go

@@ -3,6 +3,7 @@ package controller
 import (
 	"context"
 	"fmt"
+	"sort"
 	"time"
 
 	corev1 "k8s.io/api/core/v1"
@@ -126,6 +127,34 @@ func primaryTransferAddress(ctx context.Context, c client.Client, cluster *bindv
 	return primaryPodIP(ctx, c, cluster)
 }
 
+// secondaryPodIPs returns the pod IPs of a cluster's secondary pods (every pod
+// except the ordinal-0 primary) that currently have an address. The primary
+// uses this list as its zone `also-notify` set, so a change to a primary zone
+// (in particular a dynamic update) triggers an immediate NOTIFY -> IXFR to the
+// secondaries instead of leaving them stale until the next SOA refresh. The
+// list is sorted so the rendered zone config is stable and does not churn
+// modzone on every reconcile. Pod IPs change across restarts, so the caller
+// relies on the zone controller's periodic requeue to refresh the set (a
+// restarted secondary re-transfers the whole zone on load regardless). Returns
+// nil for a single-replica cluster.
+func secondaryPodIPs(ctx context.Context, c client.Client, cluster *bindv1alpha1.BindCluster) []string {
+	var pods corev1.PodList
+	if err := c.List(ctx, &pods, client.InNamespace(cluster.Namespace), client.MatchingLabels(commonLabels(cluster.Name))); err != nil {
+		return nil
+	}
+	primary := primaryPodName(cluster.Name)
+	var ips []string
+	for i := range pods.Items {
+		p := &pods.Items[i]
+		if p.Name == primary || p.Status.PodIP == "" {
+			continue
+		}
+		ips = append(ips, p.Status.PodIP)
+	}
+	sort.Strings(ips)
+	return ips
+}
+
 // resolveTSIG reads the material of a BindTSIGKey into TSIG credentials.
 func resolveTSIG(ctx context.Context, c client.Client, namespace, keyRef string) (bind.TSIGCreds, error) {
 	var creds bind.TSIGCreds