feat(cluster): PG Logical Replication + VIP/Keepalived + config_hash sync (v1.2.1–1.2.2)

- PG Logical Replication: edgeguard_shared PUBLICATION auf Primary,
  edgeguard_sub SUBSCRIPTION auf Secondary. Nur geteilte Config-Tabellen
  werden repliziert; node-eigene Daten (network_interfaces, ip_addresses,
  static_routes, cluster_settings, dns_settings, ntp_settings) bleiben
  lokal — OPNsense-Muster.
- cluster-init-replication: Erstellt PUBLICATION, Rolle + pg_hba-Einträge
  (logical + replication), WAL-Level auf logical.
- cluster-setup-standby: Erstellt SUBSCRIPTION (copy_data=true), pollt
  pg_subscription_rel bis alle Tabellen sync = 'r', rendert dann Configs.
- promote: manueller Failover via pg_promote() + touch recovery.signal.
- VIP/Keepalived: cluster_settings-Tabelle (vip_address, vip_interface,
  vrrp_router_id), /cluster/vip-settings API, Keepalived-Config-Generator
  mit VRRP + check_script + notify-Skripten in /usr/lib/edgeguard/scripts/.
- config_hash sync: Secondary pusht alle 5 Min seinen Hash via mTLS an
  Primary (PushSelfToPrimary). Heartbeat schreibt nur LOCAL, daher ohne
  aktiven Push wäre Primary-Sicht des Secondary-Hash stale gewesen.
- runSecondaryConfigRender: Goroutine auf Secondary rendert HAProxy+nftables
  neu wenn config_hash sich ändert (Logical-Replication-Nachzügler).
- confighash: node-spezifische Tabellen aus hashSpec entfernt.
- postinst: Keepalived-Skripte installieren, sudoers für keepalived.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

internal/cluster/confighash.go

@@ -35,10 +35,12 @@ import (
 
 // hashTable beschreibt eine Tabelle die in den config-hash einfließt.
 type hashTable struct {
-	Name          string
-	Singleton     bool     // dns_settings, ntp_settings → eine row, id=1
-	ExtraExclude  []string // Spalten die zusätzlich aus to_jsonb gefiltert werden
-	SkipUpdatedAt bool     // setze true wenn updated_at semantisch relevant ist
+	Name             string
+	Singleton        bool     // dns_settings, ntp_settings → eine row, id=1
+	ExtraExclude     []string // Spalten die zusätzlich aus to_jsonb gefiltert werden
+	SkipUpdatedAt    bool     // setze true wenn updated_at semantisch relevant ist
+	MigrationDefault bool     // Tabelle hat migrations-erzeugte Default-Rows (firewall_zones, ntp_pools…)
+	                          // → zählt nicht als "user hat config" bei der Empty-DB-Erkennung
 }
 
 // hashSpec ist die Reihenfolge-stabile Liste. NEUE Tabellen hier
@@ -49,15 +51,13 @@ var hashSpec = []hashTable{
 	{Name: "backends"},
 	{Name: "backend_servers"},
 	{Name: "routing_rules"},
-	{Name: "network_interfaces"},
-	{Name: "ip_addresses"},
 	{Name: "tls_certs", ExtraExclude: []string{"last_renewed_at", "last_error"}},
 
-	{Name: "firewall_zones"},
+	{Name: "firewall_zones", MigrationDefault: true},
 	{Name: "firewall_address_objects"},
 	{Name: "firewall_address_groups"},
-	{Name: "firewall_services"},
-	{Name: "firewall_service_groups"},
+	{Name: "firewall_services", MigrationDefault: true},
+	{Name: "firewall_service_groups", MigrationDefault: true},
 	{Name: "firewall_rules"},
 	{Name: "firewall_nat_rules"},
 
@@ -67,12 +67,12 @@ var hashSpec = []hashTable{
 
 	{Name: "dns_zones"},
 	{Name: "dns_records"},
-	{Name: "dns_settings", Singleton: true},
 
-	{Name: "ntp_pools"},
-	{Name: "ntp_settings", Singleton: true},
+	{Name: "ntp_pools", MigrationDefault: true},
 
-	{Name: "static_routes"},
+	// network_interfaces, ip_addresses, static_routes, dns_settings, ntp_settings
+	// sind node-spezifisch (jeder Node hat eigene IPs/Routes/Listen-Adressen)
+	// und fließen NICHT in den Drift-Hash ein.
 }
 
 // hashSQL rendert die SHA-Input-SQL für eine Tabelle.
@@ -100,22 +100,35 @@ func hashSQL(t hashTable) string {
 // ComputeConfigHash gibt den 16-hex-char-Hash über alle Spec-Tabellen
 // zurück. Fehlende Tabellen (transienter schema-flux) werden als
 // leerer Per-Table-Hash behandelt — kein Abbruch.
+//
+// Gibt "" zurück wenn alle user-konfigurierbaren Tabellen leer sind
+// (Singleton- und MigrationDefault-Tabellen zählen nicht als User-Config).
+// Das verhindert False-Positive-Drift-Banner auf frisch gejointen Secondaries.
 func ComputeConfigHash(ctx context.Context, pool *pgxpool.Pool) (string, error) {
 	if pool == nil {
 		return "", fmt.Errorf("nil pool")
 	}
 	h := sha256.New()
+	hasUserConfig := false
 	for _, t := range hashSpec {
 		var s string
 		if err := pool.QueryRow(ctx, hashSQL(t)).Scan(&s); err != nil {
 			// Migration fehlt o.ä. → leeren string nehmen, weiter.
 			s = ""
 		}
+		if s != "" && !t.Singleton && !t.MigrationDefault {
+			hasUserConfig = true
+		}
 		h.Write([]byte(t.Name))
 		h.Write([]byte{0})
 		h.Write([]byte(s))
 		h.Write([]byte{0})
 	}
+	if !hasUserConfig {
+		// Frisch gejoincter Secondary oder komplett leere DB →
+		// leerer String signalisiert "kein Drift prüfen" im Status-Handler.
+		return "", nil
+	}
 	return hex.EncodeToString(h.Sum(nil))[:16], nil
 }
 

internal/database/migrations/0029_cluster_vip.sql

@@ -0,0 +1,36 @@
+-- +goose Up
+-- +goose StatementBegin
+
+-- pg_role: Rolle dieser Node in der PG-Replikation.
+-- "standalone" = kein Streaming-Replication-Setup
+-- "primary"    = WAL-Sender, repliziert an Standby(s)
+-- "standby"    = Hot-Standby, liest WAL vom Primary
+ALTER TABLE ha_nodes ADD COLUMN IF NOT EXISTS pg_role TEXT NOT NULL DEFAULT 'standalone';
+
+-- cluster_settings: VIP + VRRP-Konfiguration (Singleton, id=1).
+-- vip_address   = die virtuelle IP-Adresse (z.B. "89.163.205.10")
+-- vip_interface = Netzwerk-Interface (z.B. "eth0")
+-- vip_auth_pass = VRRP-Authentication-Passwort (max. 8 Zeichen, Keepalived-Limit)
+-- vrrp_router_id = VRRP Virtual Router ID (1–255, muss im Subnetz eindeutig sein)
+CREATE TABLE IF NOT EXISTS cluster_settings (
+    id              INTEGER PRIMARY KEY DEFAULT 1,
+    vip_address     TEXT,
+    vip_interface   TEXT,
+    vip_auth_pass   TEXT,
+    vrrp_router_id  INTEGER NOT NULL DEFAULT 51,
+    created_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated_at      TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    CONSTRAINT cluster_settings_singleton CHECK (id = 1)
+);
+
+INSERT INTO cluster_settings (id) VALUES (1) ON CONFLICT DO NOTHING;
+
+-- +goose StatementEnd
+
+-- +goose Down
+-- +goose StatementBegin
+
+DROP TABLE IF EXISTS cluster_settings;
+ALTER TABLE ha_nodes DROP COLUMN IF EXISTS pg_role;
+
+-- +goose StatementEnd

internal/keepalived/keepalived.conf.tpl

@@ -0,0 +1,40 @@
+global_defs {
+    router_id {{ .RouterID }}
+    script_user root
+    enable_script_security
+    vrrp_garp_interval 0
+    vrrp_gna_interval 0
+}
+
+vrrp_script chk_edgeguard {
+    script "/usr/lib/edgeguard/keepalived-check.sh"
+    interval 2
+    weight -50
+    fall 3
+    rise 2
+}
+
+vrrp_instance VI_1 {
+    state {{ .State }}
+    interface {{ .Interface }}
+    virtual_router_id {{ .RouterID }}
+    priority {{ .Priority }}
+    advert_int 1
+{{ if .SrcIP }}    unicast_src_ip {{ .SrcIP }}
+    unicast_peer {
+        {{ .PeerIP }}
+    }
+{{ end }}    authentication {
+        auth_type PASS
+        auth_pass {{ .AuthPass }}
+    }
+    virtual_ipaddress {
+        {{ .VIP }}
+    }
+    track_script {
+        chk_edgeguard
+    }
+    notify_master  "/usr/lib/edgeguard/keepalived-master.sh"
+    notify_backup  "/usr/lib/edgeguard/keepalived-backup.sh"
+    notify_fault   "/usr/lib/edgeguard/keepalived-backup.sh"
+}

internal/keepalived/keepalived.go

@@ -0,0 +1,154 @@
+// Package keepalived rendert /etc/keepalived/keepalived.conf aus
+// cluster_settings (VIP/VRRP-Config) und ha_nodes (local vs. peer).
+//
+// Split-Brain-Strategie: kein Auto-Promote. notify_master loggt nur
+// und sendet einen internen Alert. Promotion ist immer manuell via
+// "edgeguard-ctl promote" — das ist die einzig sichere Option ohne
+// externes Quorum in einem 2-Node-Cluster.
+package keepalived
+
+import (
+	"bytes"
+	"context"
+	_ "embed"
+	"fmt"
+	"os"
+	"os/exec"
+	"text/template"
+
+	"github.com/jackc/pgx/v5/pgxpool"
+
+	"git.netcell-it.de/projekte/edgeguard-native/internal/configgen"
+	"git.netcell-it.de/projekte/edgeguard-native/internal/models"
+)
+
+const ConfPath = "/etc/keepalived/keepalived.conf"
+
+//go:embed keepalived.conf.tpl
+var cfgTpl string
+
+var tpl = template.Must(template.New("keepalived").Parse(cfgTpl))
+
+// View ist der Template-Kontext.
+type View struct {
+	State    string // MASTER | BACKUP
+	Interface string
+	RouterID int
+	Priority int    // MASTER=200, BACKUP=100
+	SrcIP    string // eigene Public-IP (für unicast_src_ip)
+	PeerIP   string // Peer-Public-IP (für unicast_peer)
+	AuthPass string
+	VIP      string
+}
+
+type generator struct {
+	pool    *pgxpool.Pool
+	localID string
+}
+
+func New(pool *pgxpool.Pool, localID string) configgen.Generator {
+	return &generator{pool: pool, localID: localID}
+}
+
+func (g *generator) Name() string { return "keepalived" }
+
+func (g *generator) Render(ctx context.Context) error {
+	cs, local, peer, err := g.loadData(ctx)
+	if err != nil {
+		return fmt.Errorf("keepalived: load: %w", err)
+	}
+	if cs.VIPAddress == nil || *cs.VIPAddress == "" {
+		// Kein VIP konfiguriert → keepalived.conf nicht schreiben.
+		return nil
+	}
+	v := g.buildView(cs, local, peer)
+	var buf bytes.Buffer
+	if err := tpl.Execute(&buf, v); err != nil {
+		return fmt.Errorf("keepalived: template: %w", err)
+	}
+	if err := configgen.AtomicWrite(ConfPath, buf.Bytes(), 0o640); err != nil {
+		return fmt.Errorf("keepalived: write: %w", err)
+	}
+	if err := reloadKeepalived(); err != nil {
+		return fmt.Errorf("keepalived: reload: %w", err)
+	}
+	return nil
+}
+
+func (g *generator) loadData(ctx context.Context) (*models.ClusterSettings, *models.HANode, *models.HANode, error) {
+	var cs models.ClusterSettings
+	row := g.pool.QueryRow(ctx, `SELECT id, vip_address, vip_interface, vip_auth_pass, vrrp_router_id FROM cluster_settings WHERE id = 1`)
+	if err := row.Scan(&cs.ID, &cs.VIPAddress, &cs.VIPInterface, &cs.VIPAuthPass, &cs.VRRPRouterID); err != nil {
+		return nil, nil, nil, fmt.Errorf("cluster_settings: %w", err)
+	}
+
+	rows, err := g.pool.Query(ctx, `SELECT id, fqdn, role, pg_role, public_ip, status FROM ha_nodes ORDER BY joined_at`)
+	if err != nil {
+		return nil, nil, nil, fmt.Errorf("ha_nodes: %w", err)
+	}
+	defer rows.Close()
+
+	var local, peer *models.HANode
+	for rows.Next() {
+		n := &models.HANode{}
+		if err := rows.Scan(&n.ID, &n.FQDN, &n.Role, &n.PGRole, &n.PublicIP, &n.Status); err != nil {
+			continue
+		}
+		if n.ID == g.localID {
+			local = n
+		} else {
+			peer = n
+		}
+	}
+	if local == nil {
+		return nil, nil, nil, fmt.Errorf("local node %s not in ha_nodes", g.localID)
+	}
+	return &cs, local, peer, nil
+}
+
+func (g *generator) buildView(cs *models.ClusterSettings, local, peer *models.HANode) View {
+	v := View{
+		RouterID: cs.VRRPRouterID,
+		VIP:      deref(cs.VIPAddress),
+		Interface: deref(cs.VIPInterface),
+		AuthPass: deref(cs.VIPAuthPass),
+	}
+	if v.Interface == "" {
+		v.Interface = "eth0"
+	}
+	if v.AuthPass == "" {
+		v.AuthPass = "edgeguard"
+	}
+
+	// Primary-Node bekommt höhere Priorität und startet als MASTER.
+	if local.PGRole == "primary" || local.Role == "primary" {
+		v.State = "MASTER"
+		v.Priority = 200
+	} else {
+		v.State = "BACKUP"
+		v.Priority = 100
+	}
+
+	if local.PublicIP != nil {
+		v.SrcIP = *local.PublicIP
+	}
+	if peer != nil && peer.PublicIP != nil {
+		v.PeerIP = *peer.PublicIP
+	}
+	return v
+}
+
+func reloadKeepalived() error {
+	if _, err := os.Stat("/run/keepalived.pid"); os.IsNotExist(err) {
+		// keepalived läuft noch nicht — erster Render beim Start.
+		return nil
+	}
+	return exec.Command("systemctl", "reload-or-restart", "keepalived").Run()
+}
+
+func deref(s *string) string {
+	if s == nil {
+		return ""
+	}
+	return *s
+}

internal/models/cluster_settings.go

@@ -0,0 +1,17 @@
+package models
+
+import "time"
+
+// ClusterSettings ist die Singleton-Tabelle (id=1) für VIP/VRRP-
+// und Replikations-Konfiguration. Angelegt in Migration 0029.
+type ClusterSettings struct {
+	ID           int       `gorm:"column:id;primaryKey"   json:"id"`
+	VIPAddress   *string   `gorm:"column:vip_address"     json:"vip_address,omitempty"`
+	VIPInterface *string   `gorm:"column:vip_interface"   json:"vip_interface,omitempty"`
+	VIPAuthPass  *string   `gorm:"column:vip_auth_pass"   json:"vip_auth_pass,omitempty"`
+	VRRPRouterID int       `gorm:"column:vrrp_router_id"  json:"vrrp_router_id"`
+	CreatedAt    time.Time `gorm:"column:created_at"      json:"created_at"`
+	UpdatedAt    time.Time `gorm:"column:updated_at"      json:"updated_at"`
+}
+
+func (ClusterSettings) TableName() string { return "cluster_settings" }

internal/models/ha_node.go

@@ -4,7 +4,7 @@ import "time"
 
 // HANode mirrort eine Row der ha_nodes-Tabelle. Erweitert in Migration
 // 0020 um version/config_hash/mgmt_ip/status für Cluster-Phase-3-
-// Drift-Detection + Health-State.
+// Drift-Detection + Health-State. Migration 0029 fügt PGRole hinzu.
 type HANode struct {
 	ID         string     `gorm:"column:id;primaryKey"           json:"id"`
 	Name       string     `gorm:"column:name"                    json:"name"`
@@ -14,6 +14,7 @@ type HANode struct {
 	InternalIP *string    `gorm:"column:internal_ip;type:inet"   json:"internal_ip,omitempty"`
 	MgmtIP     *string    `gorm:"column:mgmt_ip;type:inet"       json:"mgmt_ip,omitempty"`
 	Role       string     `gorm:"column:role"                    json:"role"`
+	PGRole     string     `gorm:"column:pg_role"                 json:"pg_role"`
 	Version    *string    `gorm:"column:version"                 json:"version,omitempty"`
 	ConfigHash *string    `gorm:"column:config_hash"             json:"config_hash,omitempty"`
 	Status     string     `gorm:"column:status"                  json:"status"`

internal/services/clusterjoin/join.go

@@ -128,7 +128,7 @@ func Join(req Request) error {
 	// synchronous on the primary side.
 	var autoRegErr error
 	for i := 0; i < 3; i++ {
-		if err := autoRegister(primary, tlsDir, req.CommonName, req.Version, req.NodeID); err == nil {
+		if err := autoRegister(primary, tlsDir, req.CommonName, req.Version, req.NodeID, ""); err == nil {
 			autoRegErr = nil
 			break
 		} else {
@@ -217,7 +217,18 @@ func issueCert(primary, token, csr string, insecure bool) (caCert, peerCert stri
 	return env.Data.CACert, env.Data.PeerCert, nil
 }
 
-func autoRegister(primary, tlsDir, commonName, version, nodeID string) error {
+// PushSelfToPrimary sends this node's current identity + configHash to the
+// primary via mTLS. Exported for use by the API server's periodic push
+// goroutine so the primary's ha_nodes always reflects the secondary's actual
+// config_hash (not the stale join-time value).
+func PushSelfToPrimary(primaryURL, tlsDir, nodeID, fqdn, version, configHash string) error {
+	if tlsDir == "" {
+		tlsDir = clustertls.DefaultDir
+	}
+	return autoRegister(primaryURL, tlsDir, fqdn, version, nodeID, configHash)
+}
+
+func autoRegister(primary, tlsDir, commonName, version, nodeID, configHash string) error {
 	u, err := url.Parse(primary)
 	if err != nil {
 		return err
@@ -231,11 +242,12 @@ func autoRegister(primary, tlsDir, commonName, version, nodeID string) error {
 	}
 	hostname, _ := os.Hostname()
 	body, _ := json.Marshal(map[string]string{
-		"id":      nodeID,
-		"name":    hostname,
-		"fqdn":    commonName,
-		"api_url": "https://" + commonName + ":3443",
-		"version": version,
+		"id":          nodeID,
+		"name":        hostname,
+		"fqdn":        commonName,
+		"api_url":     "https://" + commonName + ":3443",
+		"version":     version,
+		"config_hash": configHash,
 	})
 
 	pair, err := tls.LoadX509KeyPair(tlsDir+"/peer.crt", tlsDir+"/peer.key")