Updated docs and fixed WG bugs and ip's bugs

2026-03-17 04:53:00 +00:00 · 2026-01-31 07:09:09 +02:00 · 2026-01-31 07:09:09 +02:00 · 810094771d
commit 810094771d
parent 4acea72467
8 changed files with 299 additions and 73 deletions
--- a/docs/CLEAN_NODE.md
+++ b/docs/CLEAN_NODE.md
@ -0,0 +1,141 @@
+# Clean Node — Full Reset Guide
+
+How to completely remove all Orama Network state from a VPS so it can be reinstalled fresh.
+
+## Quick Clean (Copy-Paste)
+
+Run this as root or with sudo on the target VPS:
+
+```bash
+# 1. Stop and disable all services
+sudo systemctl stop debros-node debros-ipfs debros-ipfs-cluster debros-olric coredns caddy 2>/dev/null
+sudo systemctl disable debros-node debros-ipfs debros-ipfs-cluster debros-olric coredns caddy 2>/dev/null
+
+# 2. Remove systemd service files
+sudo rm -f /etc/systemd/system/debros-*.service
+sudo rm -f /etc/systemd/system/coredns.service
+sudo rm -f /etc/systemd/system/caddy.service
+sudo systemctl daemon-reload
+
+# 3. Tear down WireGuard
+# Must stop the systemd unit first — wg-quick@wg0 is a oneshot with
+# RemainAfterExit=yes, so it stays "active (exited)" even after the
+# interface is removed. Without "stop", a future "systemctl start" is a no-op.
+sudo systemctl stop wg-quick@wg0 2>/dev/null
+sudo wg-quick down wg0 2>/dev/null
+sudo systemctl disable wg-quick@wg0 2>/dev/null
+sudo rm -f /etc/wireguard/wg0.conf
+
+# 4. Reset UFW firewall
+sudo ufw --force reset
+sudo ufw allow 22/tcp
+sudo ufw --force enable
+
+# 5. Remove debros user and home directory
+sudo userdel -r debros 2>/dev/null
+sudo rm -rf /home/debros
+
+# 6. Remove sudoers files
+sudo rm -f /etc/sudoers.d/debros-access
+sudo rm -f /etc/sudoers.d/debros-deployments
+sudo rm -f /etc/sudoers.d/debros-wireguard
+
+# 7. Remove CoreDNS config
+sudo rm -rf /etc/coredns
+
+# 8. Remove Caddy config and data
+sudo rm -rf /etc/caddy
+sudo rm -rf /var/lib/caddy
+
+# 9. Remove deployment systemd services (dynamic)
+sudo rm -f /etc/systemd/system/orama-deploy-*.service
+sudo systemctl daemon-reload
+
+# 10. Clean temp files
+sudo rm -f /tmp/orama /tmp/network-source.tar.gz /tmp/network-source.zip
+sudo rm -rf /tmp/network-extract /tmp/coredns-build /tmp/caddy-build
+
+echo "Node cleaned. Ready for fresh install."
+```
+
+## What This Removes
+
+| Category | Paths |
+|----------|-------|
+| **User** | `debros` system user and `/home/debros/` |
+| **App data** | `/home/debros/.orama/` (configs, secrets, logs, IPFS, RQLite, Olric) |
+| **Source code** | `/home/debros/src/` |
+| **Binaries** | `/home/debros/bin/orama-node`, `/home/debros/bin/gateway` |
+| **Systemd** | `debros-*.service`, `coredns.service`, `caddy.service`, `orama-deploy-*.service` |
+| **WireGuard** | `/etc/wireguard/wg0.conf`, `wg-quick@wg0` systemd unit |
+| **Firewall** | All UFW rules (reset to default + SSH only) |
+| **Sudoers** | `/etc/sudoers.d/debros-*` |
+| **CoreDNS** | `/etc/coredns/Corefile` |
+| **Caddy** | `/etc/caddy/Caddyfile`, `/var/lib/caddy/` (TLS certs) |
+| **Temp files** | `/tmp/orama`, `/tmp/network-source.*`, build dirs |
+
+## What This Does NOT Remove
+
+These are shared system tools that may be used by other software. Remove manually if desired:
+
+| Binary | Path | Remove Command |
+|--------|------|----------------|
+| RQLite | `/usr/local/bin/rqlited` | `sudo rm /usr/local/bin/rqlited` |
+| IPFS | `/usr/local/bin/ipfs` | `sudo rm /usr/local/bin/ipfs` |
+| IPFS Cluster | `/usr/local/bin/ipfs-cluster-service` | `sudo rm /usr/local/bin/ipfs-cluster-service` |
+| Olric | `/usr/local/bin/olric-server` | `sudo rm /usr/local/bin/olric-server` |
+| CoreDNS | `/usr/local/bin/coredns` | `sudo rm /usr/local/bin/coredns` |
+| Caddy | `/usr/bin/caddy` | `sudo rm /usr/bin/caddy` |
+| xcaddy | `/usr/local/bin/xcaddy` | `sudo rm /usr/local/bin/xcaddy` |
+| Go | `/usr/local/go/` | `sudo rm -rf /usr/local/go` |
+| Orama CLI | `/usr/local/bin/orama` | `sudo rm /usr/local/bin/orama` |
+
+## Nuclear Clean (Remove Everything Including Binaries)
+
+```bash
+# Run quick clean above first, then:
+sudo rm -f /usr/local/bin/rqlited
+sudo rm -f /usr/local/bin/ipfs
+sudo rm -f /usr/local/bin/ipfs-cluster-service
+sudo rm -f /usr/local/bin/olric-server
+sudo rm -f /usr/local/bin/coredns
+sudo rm -f /usr/local/bin/xcaddy
+sudo rm -f /usr/bin/caddy
+sudo rm -f /usr/local/bin/orama
+```
+
+## Multi-Node Clean
+
+To clean all nodes at once from your local machine:
+
+```bash
+# Define your nodes
+NODES=(
+  "ubuntu@141.227.165.168:password1"
+  "ubuntu@141.227.165.154:password2"
+  "ubuntu@141.227.156.51:password3"
+)
+
+for entry in "${NODES[@]}"; do
+  IFS=: read -r userhost pass <<< "$entry"
+  echo "Cleaning $userhost..."
+  sshpass -p "$pass" ssh -o StrictHostKeyChecking=no "$userhost" 'bash -s' << 'CLEAN'
+sudo systemctl stop debros-node debros-ipfs debros-ipfs-cluster debros-olric coredns caddy 2>/dev/null
+sudo systemctl disable debros-node debros-ipfs debros-ipfs-cluster debros-olric coredns caddy 2>/dev/null
+sudo rm -f /etc/systemd/system/debros-*.service /etc/systemd/system/coredns.service /etc/systemd/system/caddy.service /etc/systemd/system/orama-deploy-*.service
+sudo systemctl daemon-reload
+sudo systemctl stop wg-quick@wg0 2>/dev/null
+sudo wg-quick down wg0 2>/dev/null
+sudo systemctl disable wg-quick@wg0 2>/dev/null
+sudo rm -f /etc/wireguard/wg0.conf
+sudo ufw --force reset && sudo ufw allow 22/tcp && sudo ufw --force enable
+sudo userdel -r debros 2>/dev/null
+sudo rm -rf /home/debros
+sudo rm -f /etc/sudoers.d/debros-access /etc/sudoers.d/debros-deployments /etc/sudoers.d/debros-wireguard
+sudo rm -rf /etc/coredns /etc/caddy /var/lib/caddy
+sudo rm -f /tmp/orama /tmp/network-source.tar.gz
+sudo rm -rf /tmp/network-extract /tmp/coredns-build /tmp/caddy-build
+echo "Done"
+CLEAN
+done
+```
--- a/pkg/cli/production/install/orchestrator.go
+++ b/pkg/cli/production/install/orchestrator.go
@ -199,15 +199,24 @@ func (o *Orchestrator) executeGenesisFlow() error {
 		return fmt.Errorf("service creation failed: %w", err)
 	}

-	// Phase 7: Seed DNS records
+	// Phase 7: Seed DNS records (with retry — migrations may still be running)
 	if o.flags.Nameserver && o.flags.BaseDomain != "" {
 		fmt.Printf("\n🌐 Phase 7: Seeding DNS records...\n")
-		fmt.Printf("  Waiting for RQLite to start (10s)...\n")
-		time.Sleep(10 * time.Second)
-		if err := o.setup.SeedDNSRecords(o.flags.BaseDomain, o.flags.VpsIP, o.peers); err != nil {
-			fmt.Fprintf(os.Stderr, "  ⚠️  Warning: Failed to seed DNS records: %v\n", err)
-		} else {
-			fmt.Printf("  ✓ DNS records seeded\n")
+		var seedErr error
+		for attempt := 1; attempt <= 6; attempt++ {
+			waitSec := 5 * attempt
+			fmt.Printf("  Waiting for RQLite + migrations (%ds, attempt %d/6)...\n", waitSec, attempt)
+			time.Sleep(time.Duration(waitSec) * time.Second)
+			seedErr = o.setup.SeedDNSRecords(o.flags.BaseDomain, o.flags.VpsIP, o.peers)
+			if seedErr == nil {
+				fmt.Printf("  ✓ DNS records seeded\n")
+				break
+			}
+			fmt.Fprintf(os.Stderr, "  ⚠️  Attempt %d failed: %v\n", attempt, seedErr)
+		}
+		if seedErr != nil {
+			fmt.Fprintf(os.Stderr, "  ⚠️  Warning: DNS seeding failed after all attempts.\n")
+			fmt.Fprintf(os.Stderr, "     Records will self-heal via node heartbeat once running.\n")
 		}
 	}

--- a/pkg/environments/production/installers.go
+++ b/pkg/environments/production/installers.go
@ -138,8 +138,8 @@ func (bi *BinaryInstaller) InstallCaddy() error {
 }

 // ConfigureCaddy creates Caddy configuration files
-func (bi *BinaryInstaller) ConfigureCaddy(domain string, email string, acmeEndpoint string) error {
-	return bi.caddy.Configure(domain, email, acmeEndpoint)
+func (bi *BinaryInstaller) ConfigureCaddy(domain string, email string, acmeEndpoint string, baseDomain string) error {
+	return bi.caddy.Configure(domain, email, acmeEndpoint, baseDomain)
 }

 // Mock system commands for testing (if needed)
--- a/pkg/environments/production/installers/caddy.go
+++ b/pkg/environments/production/installers/caddy.go
@ -6,6 +6,7 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
+	"strings"
 )

 const (
@ -158,15 +159,17 @@ func (ci *CaddyInstaller) Install() error {
 	return nil
 }

-// Configure creates Caddy configuration files
-func (ci *CaddyInstaller) Configure(domain string, email string, acmeEndpoint string) error {
+// Configure creates Caddy configuration files.
+// baseDomain is optional — if provided (and different from domain), Caddy will also
+// serve traffic for the base domain and its wildcard (e.g., *.dbrs.space).
+func (ci *CaddyInstaller) Configure(domain string, email string, acmeEndpoint string, baseDomain string) error {
 	configDir := "/etc/caddy"
 	if err := os.MkdirAll(configDir, 0755); err != nil {
 		return fmt.Errorf("failed to create config directory: %w", err)
 	}

 	// Create Caddyfile
-	caddyfile := ci.generateCaddyfile(domain, email, acmeEndpoint)
+	caddyfile := ci.generateCaddyfile(domain, email, acmeEndpoint, baseDomain)
 	if err := os.WriteFile(filepath.Join(configDir, "Caddyfile"), []byte(caddyfile), 0644); err != nil {
 		return fmt.Errorf("failed to write Caddyfile: %w", err)
 	}
@ -364,32 +367,31 @@ require (
 `
 }

-// generateCaddyfile creates the Caddyfile configuration
-func (ci *CaddyInstaller) generateCaddyfile(domain, email, acmeEndpoint string) string {
-	return fmt.Sprintf(`{
-    email %s
-}
-
-*.%s {
-    tls {
+// generateCaddyfile creates the Caddyfile configuration.
+// If baseDomain is provided and different from domain, Caddy also serves
+// the base domain and its wildcard (e.g., *.dbrs.space alongside *.node1.dbrs.space).
+func (ci *CaddyInstaller) generateCaddyfile(domain, email, acmeEndpoint, baseDomain string) string {
+	tlsBlock := fmt.Sprintf(`    tls {
        dns orama {
            endpoint %s
        }
-    }
-    reverse_proxy localhost:6001
-}
+    }`, acmeEndpoint)

-%s {
-    tls {
-        dns orama {
-            endpoint %s
-        }
-    }
-    reverse_proxy localhost:6001
-}
+	var sb strings.Builder
+	sb.WriteString(fmt.Sprintf("{\n    email %s\n}\n", email))

-:80 {
-    reverse_proxy localhost:6001
-}
-`, email, domain, acmeEndpoint, domain, acmeEndpoint)
+	// Node domain blocks (e.g., node1.dbrs.space, *.node1.dbrs.space)
+	sb.WriteString(fmt.Sprintf("\n*.%s {\n%s\n    reverse_proxy localhost:6001\n}\n", domain, tlsBlock))
+	sb.WriteString(fmt.Sprintf("\n%s {\n%s\n    reverse_proxy localhost:6001\n}\n", domain, tlsBlock))
+
+	// Base domain blocks (e.g., dbrs.space, *.dbrs.space) — for app routing
+	if baseDomain != "" && baseDomain != domain {
+		sb.WriteString(fmt.Sprintf("\n*.%s {\n%s\n    reverse_proxy localhost:6001\n}\n", baseDomain, tlsBlock))
+		sb.WriteString(fmt.Sprintf("\n%s {\n%s\n    reverse_proxy localhost:6001\n}\n", baseDomain, tlsBlock))
+	}
+
+	// HTTP fallback (handles plain HTTP and ACME challenges)
+	sb.WriteString("\n:80 {\n    reverse_proxy localhost:6001\n}\n")
+
+	return sb.String()
 }
--- a/pkg/environments/production/orchestrator.go
+++ b/pkg/environments/production/orchestrator.go
@ -557,7 +557,7 @@ func (ps *ProductionSetup) Phase4GenerateConfigs(peerAddresses []string, vpsIP s
 		}
 		email := "admin@" + caddyDomain
 		acmeEndpoint := "http://localhost:6001/v1/internal/acme"
-		if err := ps.binaryInstaller.ConfigureCaddy(caddyDomain, email, acmeEndpoint); err != nil {
+		if err := ps.binaryInstaller.ConfigureCaddy(caddyDomain, email, acmeEndpoint, baseDomain); err != nil {
 			ps.logf("  ⚠️  Caddy config warning: %v", err)
 		} else {
 			ps.logf("  ✓ Caddy config generated")
@ -686,7 +686,8 @@ func (ps *ProductionSetup) Phase5CreateSystemdServices(enableHTTPS bool) error {
 		}
 	}

-	// Start services in dependency order
+	// Restart services in dependency order (restart instead of start ensures
+	// services pick up new configs even if already running from a previous install)
 	ps.logf("  Starting services...")

 	// Start infrastructure first (IPFS, Olric, Anyone) - RQLite is managed internally by each node
@ -705,9 +706,9 @@ func (ps *ProductionSetup) Phase5CreateSystemdServices(enableHTTPS bool) error {
 			infraServices = append(infraServices, "debros-anyone-relay.service")
 		}
 	}
-	
+
 	for _, svc := range infraServices {
-		if err := ps.serviceController.StartService(svc); err != nil {
+		if err := ps.serviceController.RestartService(svc); err != nil {
 			ps.logf("  ⚠️  Failed to start %s: %v", svc, err)
 		} else {
 			ps.logf("    - %s started", svc)
@ -718,14 +719,14 @@ func (ps *ProductionSetup) Phase5CreateSystemdServices(enableHTTPS bool) error {
 	time.Sleep(2 * time.Second)

 	// Start IPFS Cluster
-	if err := ps.serviceController.StartService("debros-ipfs-cluster.service"); err != nil {
+	if err := ps.serviceController.RestartService("debros-ipfs-cluster.service"); err != nil {
 		ps.logf("  ⚠️  Failed to start debros-ipfs-cluster.service: %v", err)
 	} else {
 		ps.logf("    - debros-ipfs-cluster.service started")
 	}

 	// Start node service (gateway is embedded in node, no separate service needed)
-	if err := ps.serviceController.StartService("debros-node.service"); err != nil {
+	if err := ps.serviceController.RestartService("debros-node.service"); err != nil {
 		ps.logf("  ⚠️  Failed to start debros-node.service: %v", err)
 	} else {
 		ps.logf("    - debros-node.service started (with embedded gateway)")
@ -735,14 +736,14 @@ func (ps *ProductionSetup) Phase5CreateSystemdServices(enableHTTPS bool) error {
 	// Caddy depends on debros-node.service (gateway on :6001), so start after node
 	if ps.isNameserver {
 		if _, err := os.Stat("/usr/local/bin/coredns"); err == nil {
-			if err := ps.serviceController.StartService("coredns.service"); err != nil {
+			if err := ps.serviceController.RestartService("coredns.service"); err != nil {
 				ps.logf("  ⚠️  Failed to start coredns.service: %v", err)
 			} else {
 				ps.logf("    - coredns.service started")
 			}
 		}
 		if _, err := os.Stat("/usr/bin/caddy"); err == nil {
-			if err := ps.serviceController.StartService("caddy.service"); err != nil {
+			if err := ps.serviceController.RestartService("caddy.service"); err != nil {
 				ps.logf("  ⚠️  Failed to start caddy.service: %v", err)
 			} else {
 				ps.logf("    - caddy.service started")
--- a/pkg/environments/production/services.go
+++ b/pkg/environments/production/services.go
@ -216,8 +216,9 @@ func (ssg *SystemdServiceGenerator) GenerateNodeService() string {

 	return fmt.Sprintf(`[Unit]
 Description=DeBros Network Node
-After=debros-ipfs-cluster.service debros-olric.service
+After=debros-ipfs-cluster.service debros-olric.service wg-quick@wg0.service
 Wants=debros-ipfs-cluster.service debros-olric.service
+Requires=wg-quick@wg0.service

 [Service]
 Type=simple
--- a/pkg/environments/production/wireguard.go
+++ b/pkg/environments/production/wireguard.go
@ -157,8 +157,11 @@ func (wp *WireGuardProvisioner) Enable() error {
 		return fmt.Errorf("failed to enable wg-quick@wg0: %w\n%s", err, string(output))
 	}

-	// Start now
-	cmd = exec.Command("systemctl", "start", "wg-quick@wg0")
+	// Use restart instead of start. wg-quick@wg0 is a oneshot service with
+	// RemainAfterExit=yes, so "systemctl start" is a no-op if the service is
+	// already in "active (exited)" state (e.g. from a previous install that
+	// wasn't fully cleaned). "restart" always re-runs the ExecStart command.
+	cmd = exec.Command("systemctl", "restart", "wg-quick@wg0")
 	if output, err := cmd.CombinedOutput(); err != nil {
 		return fmt.Errorf("failed to start wg-quick@wg0: %w\n%s", err, string(output))
 	}
--- a/pkg/node/dns_registration.go
+++ b/pkg/node/dns_registration.go
@ -118,12 +118,15 @@ func (n *Node) updateDNSHeartbeat(ctx context.Context) error {
 // ensureBaseDNSRecords ensures this node's IP is present in the base DNS records.
 // This provides self-healing: if records are missing (fresh install, DB reset),
 // the node recreates them on startup. Each node only manages its own IP entries.
+//
+// Records are created for BOTH the base domain (dbrs.space) and the node domain
+// (node1.dbrs.space). The base domain records enable round-robin load balancing
+// across all nodes. The node domain records enable direct node access.
 func (n *Node) ensureBaseDNSRecords(ctx context.Context) error {
-	domain := n.config.Node.Domain
-	if domain == "" {
-		domain = n.config.HTTPGateway.BaseDomain
-	}
-	if domain == "" {
+	baseDomain := n.config.HTTPGateway.BaseDomain
+	nodeDomain := n.config.Node.Domain
+
+	if baseDomain == "" && nodeDomain == "" {
 		return nil // No domain configured, skip
 	}

@ -132,22 +135,32 @@ func (n *Node) ensureBaseDNSRecords(ctx context.Context) error {
 		return fmt.Errorf("failed to determine node IP: %w", err)
 	}

-	// Ensure trailing dot for FQDN format (as CoreDNS expects)
-	fqdn := domain + "."
-	wildcardFQDN := "*." + domain + "."
-
 	db := n.rqliteAdapter.GetSQLDB()

+	// Build list of A records to ensure
+	var records []struct {
+		fqdn  string
+		value string
+	}
+
+	// Base domain records (e.g., dbrs.space, *.dbrs.space) — for round-robin across all nodes
+	if baseDomain != "" {
+		records = append(records,
+			struct{ fqdn, value string }{baseDomain + ".", ipAddress},
+			struct{ fqdn, value string }{"*." + baseDomain + ".", ipAddress},
+		)
+	}
+
+	// Node-specific records (e.g., node1.dbrs.space, *.node1.dbrs.space) — for direct node access
+	if nodeDomain != "" && nodeDomain != baseDomain {
+		records = append(records,
+			struct{ fqdn, value string }{nodeDomain + ".", ipAddress},
+			struct{ fqdn, value string }{"*." + nodeDomain + ".", ipAddress},
+		)
+	}
+
 	// Insert root A record and wildcard A record for this node's IP
 	// ON CONFLICT DO NOTHING avoids duplicates (UNIQUE on fqdn, record_type, value)
-	records := []struct {
-		fqdn  string
-		value string
-	}{
-		{fqdn, ipAddress},
-		{wildcardFQDN, ipAddress},
-	}
-
 	for _, r := range records {
 		query := `INSERT INTO dns_records (fqdn, record_type, value, ttl, namespace, created_by, is_active, created_at, updated_at)
 			VALUES (?, 'A', ?, 300, 'system', 'system', TRUE, datetime('now'), datetime('now'))
@ -158,12 +171,64 @@ func (n *Node) ensureBaseDNSRecords(ctx context.Context) error {
 		}
 	}

-	// Claim an NS slot if available (ns1, ns2, or ns3)
-	n.claimNameserverSlot(ctx, domain, ipAddress)
+	// Ensure SOA and NS records exist for the base domain (self-healing)
+	if baseDomain != "" {
+		n.ensureSOAAndNSRecords(ctx, baseDomain)
+	}
+
+	// Claim an NS slot for the base domain (ns1/ns2/ns3)
+	if baseDomain != "" {
+		n.claimNameserverSlot(ctx, baseDomain, ipAddress)
+	}

 	return nil
 }

+// ensureSOAAndNSRecords creates SOA and NS records for the base domain if they don't exist.
+// These are normally seeded during install Phase 7, but if that fails (e.g. migrations
+// not yet run), the heartbeat self-heals them here.
+func (n *Node) ensureSOAAndNSRecords(ctx context.Context, baseDomain string) {
+	db := n.rqliteAdapter.GetSQLDB()
+	fqdn := baseDomain + "."
+
+	// Check if SOA exists
+	var count int
+	err := db.QueryRowContext(ctx,
+		`SELECT COUNT(*) FROM dns_records WHERE fqdn = ? AND record_type = 'SOA'`, fqdn,
+	).Scan(&count)
+	if err != nil || count > 0 {
+		return // SOA exists or query failed, skip
+	}
+
+	n.logger.ComponentInfo(logging.ComponentNode, "SOA/NS records missing, self-healing",
+		zap.String("domain", baseDomain))
+
+	// Create SOA record
+	soaValue := fmt.Sprintf("ns1.%s. admin.%s. %d 3600 1800 604800 300",
+		baseDomain, baseDomain, time.Now().Unix())
+	if _, err := db.ExecContext(ctx,
+		`INSERT INTO dns_records (fqdn, record_type, value, ttl, namespace, created_by, is_active, created_at, updated_at)
+		VALUES (?, 'SOA', ?, 300, 'system', 'system', TRUE, datetime('now'), datetime('now'))
+		ON CONFLICT(fqdn, record_type, value) DO NOTHING`,
+		fqdn, soaValue,
+	); err != nil {
+		n.logger.ComponentWarn(logging.ComponentNode, "Failed to create SOA record", zap.Error(err))
+	}
+
+	// Create NS records (ns1, ns2, ns3)
+	for i := 1; i <= 3; i++ {
+		nsValue := fmt.Sprintf("ns%d.%s.", i, baseDomain)
+		if _, err := db.ExecContext(ctx,
+			`INSERT INTO dns_records (fqdn, record_type, value, ttl, namespace, created_by, is_active, created_at, updated_at)
+			VALUES (?, 'NS', ?, 300, 'system', 'system', TRUE, datetime('now'), datetime('now'))
+			ON CONFLICT(fqdn, record_type, value) DO NOTHING`,
+			fqdn, nsValue,
+		); err != nil {
+			n.logger.ComponentWarn(logging.ComponentNode, "Failed to create NS record", zap.Error(err))
+		}
+	}
+}
+
 // claimNameserverSlot attempts to claim an available NS hostname (ns1/ns2/ns3) for this node.
 // If the node already has a slot, it updates the IP. If no slot is available, it does nothing.
 func (n *Node) claimNameserverSlot(ctx context.Context, domain, ipAddress string) {
@ -236,11 +301,11 @@ func (n *Node) cleanupStaleNodeRecords(ctx context.Context) {
 		return
 	}

-	domain := n.config.Node.Domain
-	if domain == "" {
-		domain = n.config.HTTPGateway.BaseDomain
+	baseDomain := n.config.HTTPGateway.BaseDomain
+	if baseDomain == "" {
+		baseDomain = n.config.Node.Domain
 	}
-	if domain == "" {
+	if baseDomain == "" {
 		return
 	}

@ -255,8 +320,12 @@ func (n *Node) cleanupStaleNodeRecords(ctx context.Context) {
 	}
 	defer rows.Close()

-	fqdn := domain + "."
-	wildcardFQDN := "*." + domain + "."
+	// Build all FQDNs to clean: base domain + node domain
+	var fqdnsToClean []string
+	fqdnsToClean = append(fqdnsToClean, baseDomain+".", "*."+baseDomain+".")
+	if n.config.Node.Domain != "" && n.config.Node.Domain != baseDomain {
+		fqdnsToClean = append(fqdnsToClean, n.config.Node.Domain+".", "*."+n.config.Node.Domain+".")
+	}

 	for rows.Next() {
 		var nodeID, ip string
@ -270,7 +339,7 @@ func (n *Node) cleanupStaleNodeRecords(ctx context.Context) {
 		}

 		// Remove the dead node's A records from round-robin
-		for _, f := range []string{fqdn, wildcardFQDN} {
+		for _, f := range fqdnsToClean {
 			if _, err := db.ExecContext(ctx, `DELETE FROM dns_records WHERE fqdn = ? AND record_type = 'A' AND value = ? AND namespace = 'system'`, f, ip); err != nil {
 				n.logger.ComponentWarn(logging.ComponentNode, "Failed to remove stale DNS record",
 					zap.String("fqdn", f), zap.String("ip", ip), zap.Error(err))
@ -284,7 +353,7 @@ func (n *Node) cleanupStaleNodeRecords(ctx context.Context) {

 		// Remove glue records for this node's IP (ns1.domain., ns2.domain., ns3.domain.)
 		for _, ns := range []string{"ns1", "ns2", "ns3"} {
-			nsFQDN := ns + "." + domain + "."
+			nsFQDN := ns + "." + baseDomain + "."
 			if _, err := db.ExecContext(ctx,
 				`DELETE FROM dns_records WHERE fqdn = ? AND record_type = 'A' AND value = ? AND namespace = 'system'`,
 				nsFQDN, ip,