feat(cluster): Rolling Update — Secondary-first upgrade orchestration (v1.2.3)

POST /cluster/rolling-update startet den gestaffelten Upgrade-Prozess:
1. Secondary via mTLS /agent/cluster/trigger-update anstoßen
2. /agent/cluster/version pollen bis Secondary Version-Flip zeigt (max 10 min)
3. Primary self-upgrade via systemd-run (identisch zu /system/upgrade)

State wird in /var/lib/edgeguard/rolling-update-state.json persistiert:
Phasen: updating-secondary → waiting-secondary → updating-primary.
"done" wird nicht geschrieben — Prozess stirbt beim Upgrade. UI erkennt
Abschluss via /system/health version-flip (analog Single-Node-Upgrade).

UI: UpdateBanner erkennt Cluster-Modus (/cluster/status mode="cluster")
und tauscht den "Install now"-Button gegen "Rolling Update (Cluster)" aus.
Multi-Step-Modal zeigt die drei Phasen; ab updating-primary wechselt der
Client auf /system/health polling.

Aggregator.PostPeer: neuer einzel-POST-Helper für mTLS-trigger-update.
WithVersion(): ClusterHandler bekommt Binary-Version für /agent/cluster/version.

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

VERSION

@@ -1 +1 @@
-1.1.162
+1.2.3

cmd/edgeguard-api/main.go

@@ -40,6 +40,7 @@ import (
 	"git.netcell-it.de/projekte/edgeguard-native/internal/aggregator"
 	"git.netcell-it.de/projekte/edgeguard-native/internal/cluster/clustertls"
 	"git.netcell-it.de/projekte/edgeguard-native/internal/cluster/jointoken"
+	"git.netcell-it.de/projekte/edgeguard-native/internal/services/clusterjoin"
 	aptsvc "git.netcell-it.de/projekte/edgeguard-native/internal/services/apt"
 	"git.netcell-it.de/projekte/edgeguard-native/internal/services/domainheaders"
 	"git.netcell-it.de/projekte/edgeguard-native/internal/services/domains"
@@ -60,7 +61,7 @@ import (
 	usersvc "git.netcell-it.de/projekte/edgeguard-native/internal/services/users"
 )
 
-var version = "1.1.162"
+var version = "1.2.3"
 
 func main() {
 	addr := os.Getenv("EDGEGUARD_API_ADDR")
@@ -175,6 +176,22 @@ func main() {
 			go runClusterHeartbeat(context.Background(), pool, nodeID, version)
 		}
 
+		// Secondary: push config_hash to primary every 5 min so the primary's
+		// ha_nodes reflects actual state. Without this, the primary retains the
+		// stale hash written at join-time and the drift banner never clears.
+		// st.IsClusterNode + PrimaryFQDN are only set on joined secondary nodes.
+		if nodeID != "" && st != nil && st.IsClusterNode && st.PrimaryFQDN != "" {
+			if primaryURL, normErr := clusterjoin.NormalizePrimaryURL(st.PrimaryFQDN); normErr == nil {
+				go runPrimaryPush(context.Background(), pool, nodeID, st.FQDN, version, primaryURL)
+			} else {
+				slog.Warn("cluster: cannot normalize primary URL for push", "primary", st.PrimaryFQDN, "error", normErr)
+			}
+			// Logical Replication liefert Änderungen automatisch — aber Service-
+			// Configs (haproxy.cfg, nftables …) müssen nach jeder Änderung neu
+			// gerendert werden. Diese Goroutine erkennt hash-Änderungen und rendert.
+			go runSecondaryConfigRender(context.Background(), pool)
+		}
+
 		// Phase 3.3: Cluster-CA + Peer-Cert. Founder-Pfad — auf einem
 		// frisch installierten Single-Node generieren wir die CA und
 		// signieren uns selbst, damit der Agent-Listener auf :8443
@@ -324,7 +341,8 @@ func main() {
 		clusterHdl := handlers.NewClusterHandler(clusterStore, nodeID).
 			WithAggregator(clusterAggregator).
 			WithJoinFlow(clusterTLSStore, joinTokens).
-			WithPeerReloader(peerReloader)
+			WithPeerReloader(peerReloader).
+			WithVersion(version)
 		clusterHdl.Register(authed)
 		// /cluster/issue-cert läuft PUBLIC — joining Peer hat noch
 		// keine Session/Cert. Token + Nonce-Tracking ist die einzige
@@ -680,6 +698,84 @@ func runClusterHeartbeat(ctx context.Context, pool *pgxpoolPool, localID, versio
 	}
 }
 
+// runSecondaryConfigRender läuft auf Secondary-Nodes und re-rendert alle
+// Service-Configs wenn die Logical Replication Änderungen vom Primary
+// geliefert hat. Erkennt das an einem geänderten config_hash.
+// Tick: 5 min — balanciert Reaktionszeit gegen Reload-Overhead.
+func runSecondaryConfigRender(ctx context.Context, pool *pgxpoolPool) {
+	const tick = 5 * time.Minute
+	t := time.NewTicker(tick)
+	defer t.Stop()
+	var lastHash string
+	render := func() {
+		rCtx, cancel := context.WithTimeout(ctx, 60*time.Second)
+		defer cancel()
+		hash, err := cluster.ComputeConfigHash(rCtx, pool)
+		if err != nil || hash == lastHash {
+			return
+		}
+		lastHash = hash
+		slog.Info("cluster: secondary config changed via replication, re-rendering", "hash", hash)
+		// HAProxy
+		if err := haproxy.New(pool).Render(rCtx); err != nil {
+			slog.Warn("cluster: secondary haproxy render failed", "error", err)
+		}
+		// nftables
+		if err := firewallrender.New(pool).Render(rCtx); err != nil {
+			slog.Warn("cluster: secondary nftables render failed", "error", err)
+		}
+		// Weitere Dienste (Squid, Unbound, Chrony, WireGuard) werden bei
+		// Änderungen an ihren spezifischen Tabellen ebenfalls neu gerendert.
+		// render-config ohne Reload: die Dienste merken Änderungen selbst
+		// (HAProxy/nftables über systemctl reload, der oben bereits läuft).
+	}
+	// Initialer Check nach kurzem Delay (Replication braucht einen Moment)
+	select {
+	case <-ctx.Done():
+		return
+	case <-time.After(30 * time.Second):
+		render()
+	}
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			render()
+		}
+	}
+}
+
+// runPrimaryPush periodically pushes this secondary node's config_hash to the
+// primary via mTLS. The primary's ha_nodes view only gets config_hash written
+// during join-time autoRegister — after that the primary never hears about
+// hash changes unless we push. Without this, the drift banner shows stale
+// hashes from join-time forever.
+func runPrimaryPush(ctx context.Context, pool *pgxpoolPool, nodeID, fqdn, version, primaryURL string) {
+	const tick = 5 * time.Minute
+	t := time.NewTicker(tick)
+	defer t.Stop()
+	push := func() {
+		pCtx, cancel := context.WithTimeout(ctx, 15*time.Second)
+		defer cancel()
+		hash, _ := cluster.ComputeConfigHash(pCtx, pool)
+		if err := clusterjoin.PushSelfToPrimary(primaryURL, "", nodeID, fqdn, version, hash); err != nil {
+			slog.Warn("cluster: push-to-primary failed", "error", err)
+		} else {
+			slog.Debug("cluster: config_hash pushed to primary", "hash", hash)
+		}
+	}
+	push() // immediate push on API startup
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			push()
+		}
+	}
+}
+
 func randomEphemeralSecret() []byte {
 	b := make([]byte, 32)
 	if _, err := rand.Read(b); err != nil {

cmd/edgeguard-scheduler/main.go

@@ -41,7 +41,7 @@ import (
 	"git.netcell-it.de/projekte/edgeguard-native/internal/services/tlscerts"
 )
 
-var version = "1.1.162"
+var version = "1.2.3"
 
 const (
 	// renewTickInterval — how often we re-evaluate expiring certs.

internal/aggregator/aggregator.go

@@ -192,6 +192,43 @@ func agentURL(apiURL string, agentPort int, path string) (string, error) {
 	return u.String(), nil
 }
 
+// PostPeer sendet einen POST-Request an einen einzelnen Peer.
+// Wird vom Rolling-Update-Orchestrator genutzt um /agent/cluster/trigger-update
+// auf dem Secondary auszulösen.
+func (a *Aggregator) PostPeer(ctx context.Context, p models.HANode, path string) PeerResult {
+	start := time.Now()
+	res := PeerResult{NodeID: p.ID, FQDN: p.FQDN}
+	target, err := agentURL(p.APIURL, a.AgentPort, path)
+	if err != nil {
+		res.Err = "bad api_url: " + err.Error()
+		return res
+	}
+	reqCtx, cancel := context.WithTimeout(ctx, 15*time.Second)
+	defer cancel()
+	req, err := http.NewRequestWithContext(reqCtx, http.MethodPost, target, nil)
+	if err != nil {
+		res.Err = err.Error()
+		return res
+	}
+	req.Header.Set("Content-Type", "application/json")
+	resp, err := a.HTTPClient.Do(req)
+	if err != nil {
+		res.Err = err.Error()
+		res.Duration = time.Since(start).Milliseconds()
+		return res
+	}
+	defer resp.Body.Close()
+	body, _ := io.ReadAll(io.LimitReader(resp.Body, 1<<20))
+	if resp.StatusCode != http.StatusOK && resp.StatusCode != http.StatusAccepted {
+		res.Err = fmt.Sprintf("HTTP %d: %s", resp.StatusCode, strings.TrimSpace(string(body)))
+		res.Duration = time.Since(start).Milliseconds()
+		return res
+	}
+	res.OK = true
+	res.Duration = time.Since(start).Milliseconds()
+	return res
+}
+
 // Compile-time check dass cluster importiert wird (für Drift-Detection
 // vom hashSpec — die Aggregator-Resultate werden parallel im Drift-
 // Banner mitverarbeitet). Nicht runtime-essentiell, aber dokumentiert

internal/handlers/cluster.go

@@ -7,6 +7,9 @@ import (
 	"encoding/pem"
 	"fmt"
 	"log/slog"
+	"net/http"
+	"os"
+	"os/exec"
 	"strings"
 	"time"
 
@@ -31,6 +34,7 @@ type ClusterHandler struct {
 	Store      *cluster.Store
 	LocalID    string
 	Aggregator *aggregator.Aggregator
+	Version    string // laufende Binary-Version, für Rolling-Update-Koordination
 
 	// TLSStore + Tokens: optional, gesetzt bei Phase 3.4. Erlauben das
 	// Generieren von Join-Tokens und das Issue-Cert für joining Peers.
@@ -67,6 +71,10 @@ func (h *ClusterHandler) Register(rg *gin.RouterGroup) {
 	g.GET("/status", h.Status)
 	g.GET("/system/load", h.SystemLoad)
 	g.DELETE("/nodes/:id", h.DeleteNode)
+	g.GET("/vip-settings", h.GetVIPSettings)
+	g.PUT("/vip-settings", h.UpdateVIPSettings)
+	g.POST("/rolling-update", h.RollingUpdate)
+	g.GET("/rolling-update/status", h.RollingUpdateStatus)
 	if h.TLSStore != nil {
 		g.GET("/cert-status", h.CertStatus)
 		g.POST("/renew-self", h.RenewSelf)
@@ -115,6 +123,73 @@ func (h *ClusterHandler) DeleteNode(c *gin.Context) {
 	response.NoContent(c)
 }
 
+// GetVIPSettings liest die cluster_settings-Singleton-Row (VIP/VRRP-Config).
+func (h *ClusterHandler) GetVIPSettings(c *gin.Context) {
+	if h.Store == nil {
+		response.NotFound(c, simpleError("cluster store not available"))
+		return
+	}
+	var cs vipSettingsRow
+	row := h.Store.Pool.QueryRow(c.Request.Context(),
+		`SELECT vip_address, vip_interface, vip_auth_pass, vrrp_router_id FROM cluster_settings WHERE id = 1`)
+	if err := row.Scan(&cs.VIPAddress, &cs.VIPInterface, &cs.VIPAuthPass, &cs.VRRPRouterID); err != nil {
+		response.Internal(c, err)
+		return
+	}
+	response.OK(c, cs)
+}
+
+// UpdateVIPSettings speichert die VIP/VRRP-Konfiguration und triggert
+// einen Keepalived-Config-Render. Viewer-Schutz via RequireAdminForMutations-
+// Middleware auf der authed-Group — kein Extra-Check nötig.
+func (h *ClusterHandler) UpdateVIPSettings(c *gin.Context) {
+	var req vipSettingsRow
+	if err := c.ShouldBindJSON(&req); err != nil {
+		response.BadRequest(c, err)
+		return
+	}
+	if h.Store == nil {
+		response.NotFound(c, simpleError("cluster store not available"))
+		return
+	}
+	_, err := h.Store.Pool.Exec(c.Request.Context(), `
+		UPDATE cluster_settings
+		SET vip_address=$1, vip_interface=$2, vip_auth_pass=$3, vrrp_router_id=$4, updated_at=NOW()
+		WHERE id=1`,
+		nullIfEmpty(req.VIPAddress), nullIfEmpty(req.VIPInterface),
+		nullIfEmpty(req.VIPAuthPass), req.VRRPRouterID)
+	if err != nil {
+		response.Internal(c, err)
+		return
+	}
+	slog.Info("cluster: VIP settings updated", "vip", req.VIPAddress, "actor", actorOf(c))
+	// Keepalived-Config asynchron neu rendern
+	if h.PeerReloader != nil {
+		go func() {
+			rctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
+			defer cancel()
+			if err := h.PeerReloader(rctx); err != nil {
+				slog.Warn("cluster: keepalived render after VIP update failed", "error", err)
+			}
+		}()
+	}
+	response.NoContent(c)
+}
+
+type vipSettingsRow struct {
+	VIPAddress   *string `json:"vip_address"`
+	VIPInterface *string `json:"vip_interface"`
+	VIPAuthPass  *string `json:"vip_auth_pass"`
+	VRRPRouterID int     `json:"vrrp_router_id"`
+}
+
+func nullIfEmpty(s *string) *string {
+	if s == nil || *s == "" {
+		return nil
+	}
+	return s
+}
+
 // RegisterPublic mountet die public (unauth) Endpoints — joining Peers
 // haben noch keine Session/Cert, deshalb läuft /issue-cert vor der
 // requireAuth-Middleware. Aufrufer muss diesen Group auf /api/v1 setzen
@@ -139,6 +214,9 @@ func (h *ClusterHandler) RegisterAgent(rg *gin.RouterGroup) {
 	g := rg.Group("/agent/cluster")
 	g.POST("/peers", h.AgentRegisterPeer)
 	g.GET("/identity", h.AgentIdentity)
+	g.GET("/pg-replication-info", h.AgentPGReplicationInfo)
+	g.GET("/version", h.AgentVersion)
+	g.POST("/trigger-update", h.AgentTriggerUpdate)
 }
 
 // AgentIdentity gibt die eigene ha_nodes-Row zurück. Wird vom Primary
@@ -161,6 +239,45 @@ func (h *ClusterHandler) AgentIdentity(c *gin.Context) {
 	response.OK(c, node)
 }
 
+// AgentPGReplicationInfo gibt die Replication-Credentials für pg_basebackup
+// zurück. Nur über den mTLS-Agent-Listener erreichbar. Liest das Passwort
+// aus /var/lib/edgeguard/pg-replication-secret. Gibt 404 zurück wenn die
+// Datei fehlt (cluster-init-replication noch nicht ausgeführt).
+func (h *ClusterHandler) AgentPGReplicationInfo(c *gin.Context) {
+	const secretPath = "/var/lib/edgeguard/pg-replication-secret"
+	pass, err := readFileString(secretPath)
+	if err != nil {
+		response.NotFound(c, simpleError("pg-replication-secret nicht gefunden — cluster-init-replication auf dem Primary ausführen"))
+		return
+	}
+	// Host = eigene Public-IP aus ha_nodes (oder Fallback: FQDN)
+	host := ""
+	if h.Store != nil && h.LocalID != "" {
+		if node, err := h.Store.Get(c.Request.Context(), h.LocalID); err == nil {
+			if node.PublicIP != nil && *node.PublicIP != "" {
+				host = *node.PublicIP
+			}
+			if host == "" {
+				host = node.FQDN
+			}
+		}
+	}
+	response.OK(c, gin.H{
+		"host":     host,
+		"port":     5432,
+		"user":     "edgeguard_replicator",
+		"password": strings.TrimSpace(pass),
+	})
+}
+
+func readFileString(path string) (string, error) {
+	b, err := os.ReadFile(path)
+	if err != nil {
+		return "", err
+	}
+	return string(b), nil
+}
+
 // PeerReloader: optionale Funktion die nach einem Auto-Register
 // Firewall + ggfs. andere Configs regeneriert (damit peer_ipv4-Set
 // frisch ist). Wird vom main.go gesetzt.
@@ -172,6 +289,12 @@ func (h *ClusterHandler) WithPeerReloader(r PeerReloader) *ClusterHandler {
 	return h
 }
 
+// WithVersion: setzt die laufende Binary-Version für Rolling-Update-Koordination.
+func (h *ClusterHandler) WithVersion(v string) *ClusterHandler {
+	h.Version = v
+	return h
+}
+
 func (h *ClusterHandler) ListNodes(c *gin.Context) {
 	nodes, err := h.Store.List(c.Request.Context())
 	if err != nil {
@@ -555,6 +678,60 @@ func (h *ClusterHandler) reconcileJoiningPeers(placeholders []models.HANode) {
 	}
 }
 
+// AgentVersion gibt die laufende Binary-Version zurück. Wird vom Rolling-
+// Update-Orchestrator gepollt um zu erkennen wann der Secondary die neue
+// Version hat.
+func (h *ClusterHandler) AgentVersion(c *gin.Context) {
+	response.OK(c, gin.H{"version": h.Version})
+}
+
+// AgentTriggerUpdate startet den Upgrade-Prozess auf diesem Node via
+// systemd-run (detached). Wird vom Primary via mTLS aufgerufen um den
+// Secondary zuerst zu aktualisieren (Rolling-Update). Pattern identisch
+// zu /system/upgrade — nutzt dieselbe Sudoers-Whitelist aus dem postinst.
+func (h *ClusterHandler) AgentTriggerUpdate(c *gin.Context) {
+	const scriptPath = "/var/lib/edgeguard/upgrade.sh"
+	const script = `#!/bin/bash
+set -e
+sleep 2
+export DEBIAN_FRONTEND=noninteractive
+dpkg --configure -a || true
+retry_apt() {
+    local attempt=0 max=3 wait_for=15
+    while [ $attempt -lt $max ]; do
+        attempt=$((attempt + 1))
+        apt-get update -qq || true
+        if apt-get install -y -qq -o Dpkg::Options::=--force-confold \
+            edgeguard-api edgeguard-ui edgeguard; then return 0; fi
+        [ $attempt -lt $max ] && sleep $wait_for && wait_for=$((wait_for * 2))
+    done
+    return 1
+}
+retry_apt
+echo "[upgrade] complete"
+rm -f /var/lib/edgeguard/upgrade.sh
+`
+	if err := os.WriteFile(scriptPath, []byte(script), 0o755); err != nil {
+		response.Internal(c, err)
+		return
+	}
+	const unitName = "edgeguard-upgrade.service"
+	_ = exec.Command("sudo", "-n", "/usr/bin/systemctl", "reset-failed", unitName).Run()
+	cmd := exec.Command("sudo", "-n", "/usr/bin/systemd-run",
+		"--unit="+unitName,
+		"--description=EdgeGuard rolling-update (triggered by primary)",
+		"--collect",
+		"bash", scriptPath)
+	if err := cmd.Run(); err != nil {
+		slog.Warn("cluster: AgentTriggerUpdate: systemd-run failed", "error", err)
+		response.Internal(c, err)
+		return
+	}
+	slog.Info("cluster: rolling update triggered on this node by primary mTLS call",
+		"client", c.ClientIP())
+	c.JSON(http.StatusAccepted, gin.H{"status": "upgrading"})
+}
+
 var errInvalidJoinRequest = simpleError("missing token or csr")
 
 type simpleError string
@@ -623,14 +800,15 @@ func (h *ClusterHandler) RenewSelf(c *gin.Context) {
 // die wir wirklich brauchen — sonst kann ein joining Peer beliebige
 // ha_nodes-Felder überschreiben.
 type registerPeerRequest struct {
-	ID         string `json:"id"`           // Joiner's eigene node-id
-	Name       string `json:"name"`         // hostname
-	FQDN       string `json:"fqdn"`         // sollte mit Client-Cert-CN matchen
-	APIURL     string `json:"api_url"`      // https://<fqdn>
-	PublicIP   string `json:"public_ip"`    // optional
-	InternalIP string `json:"internal_ip"`  // mTLS-Listener-IP (für peer_ipv4-Set)
-	MgmtIP     string `json:"mgmt_ip"`      // optional
-	Version    string `json:"version"`
+	ID         string  `json:"id"`           // Joiner's eigene node-id
+	Name       string  `json:"name"`         // hostname
+	FQDN       string  `json:"fqdn"`         // sollte mit Client-Cert-CN matchen
+	APIURL     string  `json:"api_url"`      // https://<fqdn>
+	PublicIP   string  `json:"public_ip"`    // optional
+	InternalIP string  `json:"internal_ip"`  // mTLS-Listener-IP (für peer_ipv4-Set)
+	MgmtIP     string  `json:"mgmt_ip"`      // optional
+	Version    string  `json:"version"`
+	ConfigHash *string `json:"config_hash"`  // nil=absent (don't change), ""=no user config
 }
 
 // AgentRegisterPeer: vom Joiner nach issue-cert via mTLS aufgerufen.
@@ -693,6 +871,9 @@ func (h *ClusterHandler) AgentRegisterPeer(c *gin.Context) {
 		v := req.Version
 		n.Version = &v
 	}
+	if req.ConfigHash != nil {
+		n.ConfigHash = req.ConfigHash
+	}
 	// Placeholder zuerst löschen: ha_nodes hat UNIQUE(fqdn). Der INSERT
 	// in UpsertSelf verwendet ON CONFLICT(id) — greift NICHT bei fqdn-
 	// Konflikten. Ohne das Delete würde der INSERT mit "duplicate key on

internal/handlers/cluster_rollingupdate.go

@@ -0,0 +1,240 @@
+package handlers
+
+import (
+	"context"
+	"encoding/json"
+	"log/slog"
+	"net/http"
+	"os"
+	"os/exec"
+	"time"
+
+	"github.com/gin-gonic/gin"
+
+	"git.netcell-it.de/projekte/edgeguard-native/internal/handlers/response"
+	"git.netcell-it.de/projekte/edgeguard-native/internal/models"
+)
+
+const rollingUpdateStateFile = "/var/lib/edgeguard/rolling-update-state.json"
+
+const (
+	phaseIdle              = "idle"
+	phaseUpdatingSecondary = "updating-secondary"
+	phaseWaitingSecondary  = "waiting-secondary"
+	phaseUpdatingPrimary   = "updating-primary"
+	phaseFailed            = "failed"
+)
+
+// RollingUpdateState hält den Fortschritt des Rolling-Updates.
+// Persistiert in rollingUpdateStateFile damit der Status über
+// einen kurzen API-Neustart hinaus lesbar bleibt.
+type RollingUpdateState struct {
+	Phase         string    `json:"phase"`
+	SecondaryID   string    `json:"secondary_id,omitempty"`
+	SecondaryFQDN string    `json:"secondary_fqdn,omitempty"`
+	StartedAt     time.Time `json:"started_at,omitempty"`
+	UpdatedAt     time.Time `json:"updated_at"`
+	Error         string    `json:"error,omitempty"`
+}
+
+func readRollingUpdateState() RollingUpdateState {
+	data, err := os.ReadFile(rollingUpdateStateFile)
+	if err != nil {
+		return RollingUpdateState{Phase: phaseIdle, UpdatedAt: time.Now()}
+	}
+	var s RollingUpdateState
+	if err := json.Unmarshal(data, &s); err != nil {
+		return RollingUpdateState{Phase: phaseIdle, UpdatedAt: time.Now()}
+	}
+	return s
+}
+
+func writeRollingUpdateState(s RollingUpdateState) {
+	s.UpdatedAt = time.Now()
+	data, err := json.Marshal(s)
+	if err != nil {
+		slog.Warn("rolling-update: failed to marshal state", "error", err)
+		return
+	}
+	if err := os.WriteFile(rollingUpdateStateFile, data, 0o600); err != nil {
+		slog.Warn("rolling-update: failed to write state file", "error", err)
+	}
+}
+
+// RollingUpdate startet den Rolling-Update-Prozess:
+// 1. Secondary aktualisieren (via mTLS /agent/cluster/trigger-update)
+// 2. Warten bis Secondary neue Version meldet
+// 3. Primary (dieser Node) aktualisieren (wie /system/upgrade)
+//
+// Kein Cluster vorhanden → 409 zurück damit der Client auf /system/upgrade
+// ausweichen kann. Wenn bereits ein Rolling-Update läuft → aktuellen State.
+func (h *ClusterHandler) RollingUpdate(c *gin.Context) {
+	if h.Aggregator == nil || h.Store == nil {
+		c.JSON(http.StatusConflict, gin.H{"error": "no cluster — use /system/upgrade"})
+		return
+	}
+
+	st := readRollingUpdateState()
+	if st.Phase != phaseIdle && st.Phase != phaseFailed {
+		response.OK(c, st)
+		return
+	}
+
+	nodes, err := h.Store.List(c.Request.Context())
+	if err != nil {
+		response.Internal(c, err)
+		return
+	}
+	var secondary *models.HANode
+	for i := range nodes {
+		if nodes[i].ID != h.LocalID {
+			secondary = &nodes[i]
+			break
+		}
+	}
+	if secondary == nil {
+		c.JSON(http.StatusConflict, gin.H{"error": "no peer node — use /system/upgrade"})
+		return
+	}
+
+	newState := RollingUpdateState{
+		Phase:         phaseUpdatingSecondary,
+		SecondaryID:   secondary.ID,
+		SecondaryFQDN: secondary.FQDN,
+		StartedAt:     time.Now(),
+	}
+	writeRollingUpdateState(newState)
+	slog.Info("rolling-update: started", "secondary", secondary.FQDN)
+
+	go h.runRollingUpdate(secondary)
+
+	c.JSON(http.StatusAccepted, newState)
+}
+
+// RollingUpdateStatus gibt den aktuellen Rolling-Update-State zurück.
+// Wenn phase == "updating-primary" soll der Client auf /system/health
+// umschalten (der Primary restartet gleich → State kann nicht mehr
+// geschrieben werden).
+func (h *ClusterHandler) RollingUpdateStatus(c *gin.Context) {
+	response.OK(c, readRollingUpdateState())
+}
+
+func (h *ClusterHandler) runRollingUpdate(secondary *models.HANode) {
+	ctx := context.Background()
+
+	// 1. Secondary triggern
+	slog.Info("rolling-update: posting trigger-update to secondary", "fqdn", secondary.FQDN)
+	result := h.Aggregator.PostPeer(ctx, *secondary, "/agent/cluster/trigger-update")
+	if !result.OK {
+		writeRollingUpdateState(RollingUpdateState{
+			Phase:         phaseFailed,
+			SecondaryID:   secondary.ID,
+			SecondaryFQDN: secondary.FQDN,
+			Error:         "trigger-update failed: " + result.Err,
+		})
+		slog.Warn("rolling-update: secondary trigger failed", "error", result.Err)
+		return
+	}
+
+	// 2. Secondary-Version pollen — der Secondary restartet nach dem
+	//    Upgrade, danach zeigt /agent/cluster/version eine neue Version.
+	writeRollingUpdateState(RollingUpdateState{
+		Phase:         phaseWaitingSecondary,
+		SecondaryID:   secondary.ID,
+		SecondaryFQDN: secondary.FQDN,
+	})
+	slog.Info("rolling-update: waiting for secondary version flip")
+
+	// Kurze Wartezeit damit apt auf dem Secondary erst losläuft
+	time.Sleep(20 * time.Second)
+
+	deadline := time.Now().Add(10 * time.Minute)
+	versionFlipped := false
+	for time.Now().Before(deadline) {
+		results := h.Aggregator.FanOut(ctx, []models.HANode{*secondary}, "/agent/cluster/version", h.LocalID)
+		if len(results) > 0 && results[0].OK {
+			var ver struct {
+				Version string `json:"version"`
+			}
+			if err := json.Unmarshal(results[0].Data, &ver); err == nil {
+				slog.Info("rolling-update: secondary version", "version", ver.Version, "primary", h.Version)
+				if ver.Version != h.Version {
+					versionFlipped = true
+					break
+				}
+			}
+		}
+		time.Sleep(10 * time.Second)
+	}
+
+	if !versionFlipped {
+		writeRollingUpdateState(RollingUpdateState{
+			Phase:         phaseFailed,
+			SecondaryID:   secondary.ID,
+			SecondaryFQDN: secondary.FQDN,
+			Error:         "timeout (10 min) waiting for secondary version flip",
+		})
+		slog.Warn("rolling-update: secondary version flip timeout")
+		return
+	}
+
+	// 3. Primary (uns selbst) aktualisieren — identisch zu /system/upgrade
+	writeRollingUpdateState(RollingUpdateState{
+		Phase:         phaseUpdatingPrimary,
+		SecondaryID:   secondary.ID,
+		SecondaryFQDN: secondary.FQDN,
+	})
+	slog.Info("rolling-update: triggering primary self-upgrade")
+
+	const scriptPath = "/var/lib/edgeguard/upgrade.sh"
+	const script = `#!/bin/bash
+set -e
+sleep 2
+export DEBIAN_FRONTEND=noninteractive
+dpkg --configure -a || true
+retry_apt() {
+    local attempt=0 max=3 wait_for=15
+    while [ $attempt -lt $max ]; do
+        attempt=$((attempt + 1))
+        apt-get update -qq || true
+        if apt-get install -y -qq -o Dpkg::Options::=--force-confold \
+            edgeguard-api edgeguard-ui edgeguard; then return 0; fi
+        [ $attempt -lt $max ] && sleep $wait_for && wait_for=$((wait_for * 2))
+    done
+    return 1
+}
+retry_apt
+echo "[upgrade] complete"
+rm -f /var/lib/edgeguard/upgrade.sh
+`
+	if err := os.WriteFile(scriptPath, []byte(script), 0o755); err != nil {
+		writeRollingUpdateState(RollingUpdateState{
+			Phase:         phaseFailed,
+			SecondaryID:   secondary.ID,
+			SecondaryFQDN: secondary.FQDN,
+			Error:         "write upgrade script: " + err.Error(),
+		})
+		return
+	}
+
+	const unitName = "edgeguard-upgrade.service"
+	_ = exec.Command("sudo", "-n", "/usr/bin/systemctl", "reset-failed", unitName).Run()
+	cmd := exec.Command("sudo", "-n", "/usr/bin/systemd-run",
+		"--unit="+unitName,
+		"--description=EdgeGuard rolling-update (primary)",
+		"--collect",
+		"bash", scriptPath)
+	if err := cmd.Run(); err != nil {
+		writeRollingUpdateState(RollingUpdateState{
+			Phase:         phaseFailed,
+			SecondaryID:   secondary.ID,
+			SecondaryFQDN: secondary.FQDN,
+			Error:         "systemd-run failed: " + err.Error(),
+		})
+		slog.Warn("rolling-update: primary systemd-run failed", "error", err)
+		return
+	}
+	// State bleibt "updating-primary" — der Primary restartet gleich.
+	// UI erkennt Version-Flip via /system/health und schließt den Flow.
+	slog.Info("rolling-update: primary upgrade dispatched, process will restart")
+}

management-ui/src/components/UpdateBanner.tsx

@@ -1,5 +1,5 @@
 import { Alert, Button, Popconfirm, Space, Tooltip, message } from 'antd'
-import { CloudDownloadOutlined, ReloadOutlined, RocketOutlined } from '@ant-design/icons'
+import { CloudDownloadOutlined, ReloadOutlined, RocketOutlined, ClusterOutlined } from '@ant-design/icons'
 import { useQuery } from '@tanstack/react-query'
 import { useEffect, useRef, useState } from 'react'
 import { useTranslation } from 'react-i18next'
@@ -20,6 +20,19 @@ interface SystemHealth { status: string; version: string }
 
 interface PendingUpdate { pkg: string; installed: string; available: string }
 
+interface ClusterStatus {
+  mode: string   // "single-node" | "cluster"
+  peers: Array<{ id: string; fqdn: string }>
+}
+
+interface RollingUpdateState {
+  phase: string         // idle | updating-secondary | waiting-secondary | updating-primary | failed
+  secondary_fqdn: string
+  secondary_id: string
+  error?: string
+  updated_at: string
+}
+
 // allUpdates parsed das flache map-Format ({pkg_installed,pkg_available})
 // das /system/package-versions zurückliefert. Eines davon ist meist
 // das meta-Paket "edgeguard" → die "Ziel-Version".
@@ -53,6 +66,41 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
     gcTime: 0,
   })
 
+  const clusterStatus = useQuery({
+    queryKey: ['cluster', 'status-update-banner'],
+    queryFn: async () => {
+      try {
+        const r = await apiClient.get('/cluster/status')
+        return isEnvelope(r.data) ? (r.data.data as ClusterStatus) : null
+      } catch {
+        return null
+      }
+    },
+    refetchInterval: 60_000,
+    staleTime: 30_000,
+  })
+
+  const rollingStatus = useQuery({
+    queryKey: ['cluster', 'rolling-update-status'],
+    queryFn: async () => {
+      try {
+        const r = await apiClient.get('/cluster/rolling-update/status')
+        return isEnvelope(r.data) ? (r.data.data as RollingUpdateState) : null
+      } catch {
+        return null
+      }
+    },
+    refetchInterval: 5_000,
+    staleTime: 0,
+    gcTime: 0,
+  })
+
+  const isCluster = clusterStatus.data?.mode === 'cluster'
+  const rollingPhase = rollingStatus.data?.phase ?? 'idle'
+  const rollingActive = rollingPhase !== 'idle' && rollingPhase !== 'failed'
+  const secondaryFQDN = rollingStatus.data?.secondary_fqdn ?? ''
+
+  // Normal single-node upgrade state
   const [upgrading, setUpgrading] = useState(false)
   const [upgradeElapsed, setUpgradeElapsed] = useState(0)
   const [forceChecking, setForceChecking] = useState(false)
@@ -61,11 +109,61 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
   const installedRef = useRef<string>('')
   const targetRef = useRef<string>('')
 
+  // Rolling update elapsed counter
+  const [rollingElapsed, setRollingElapsed] = useState(0)
+  const rollingTickRef = useRef<ReturnType<typeof setInterval> | null>(null)
+  const rollingPollRef = useRef<ReturnType<typeof setInterval> | null>(null)
+
   useEffect(() => () => {
     if (upgradePollRef.current) clearInterval(upgradePollRef.current)
     if (upgradeTickRef.current) clearInterval(upgradeTickRef.current)
+    if (rollingTickRef.current) clearInterval(rollingTickRef.current)
+    if (rollingPollRef.current) clearInterval(rollingPollRef.current)
   }, [])
 
+  // Start rolling elapsed timer when rolling becomes active
+  useEffect(() => {
+    if (rollingActive && !rollingTickRef.current) {
+      setRollingElapsed(0)
+      rollingTickRef.current = setInterval(() => setRollingElapsed(e => e + 1), 1000)
+    } else if (!rollingActive && rollingTickRef.current) {
+      clearInterval(rollingTickRef.current)
+      rollingTickRef.current = null
+    }
+  }, [rollingActive])
+
+  // When phase reaches "updating-primary": switch to health polling
+  // (primary will restart, state file can't be updated after that)
+  useEffect(() => {
+    if (rollingPhase === 'updating-primary' && !rollingPollRef.current) {
+      const primaryInstalled = installedRef.current
+      let sawDown = false
+      rollingPollRef.current = setInterval(async () => {
+        try {
+          const res = await apiClient.get('/system/health')
+          const newV = isEnvelope(res.data) ? (res.data.data as SystemHealth).version : ''
+          const flipped = newV && primaryInstalled && newV !== primaryInstalled
+          if (flipped || sawDown) {
+            if (rollingPollRef.current) clearInterval(rollingPollRef.current)
+            if (rollingTickRef.current) clearInterval(rollingTickRef.current)
+            rollingPollRef.current = null
+            rollingTickRef.current = null
+            msg.success(t('update.success', { version: targetRef.current }))
+            setTimeout(() => window.location.reload(), 1500)
+          }
+        } catch {
+          sawDown = true
+        }
+      }, 3000)
+      // Safety timeout
+      setTimeout(() => {
+        if (rollingPollRef.current) clearInterval(rollingPollRef.current)
+        if (rollingTickRef.current) clearInterval(rollingTickRef.current)
+        window.location.reload()
+      }, 120_000)
+    }
+  }, [rollingPhase, msg, t])
+
   const data = pkgVersions.data ?? {}
   const updates = allUpdates(data)
   const updateAvailable = updates.length > 0
@@ -76,14 +174,8 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
   const forceCheck = async () => {
     setForceChecking(true)
     try {
-      // ?force=1: bypassed den Server-seitigen 5-min-Throttle für
-      // apt-get update. Ohne den Force-Hint würde der Endpoint
-      // einfach den letzten Cache zurückliefern (max. 5 min alt) und
-      // der Button fühlt sich kaputt an. Pattern aus mail-gateway.
       const r = await apiClient.get('/system/package-versions?force=1')
       const fresh = (isEnvelope(r.data) ? (r.data.data as PackageVersions) : {})
-      // useQuery-Cache mit dem frischen Wert füttern damit der Banner
-      // sofort umschaltet, ohne auf die nächste 30s-Welle zu warten.
       void pkgVersions.refetch()
       const found = allUpdates(fresh).length > 0
       msg[found ? 'success' : 'info'](
@@ -105,8 +197,6 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
 
     apiClient.post('/system/upgrade')
       .then(() => {
-        // Poll /healthz (kein Auth, robust auch wenn die API gerade
-        // restartet und Cookie ihre Session nicht erkennt).
         let sawDown = false
         upgradePollRef.current = setInterval(async () => {
           try {
@@ -121,14 +211,9 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
               setTimeout(() => window.location.reload(), 1500)
             }
           } catch {
-            // Connection refused / 502 → API restartet. Beim nächsten
-            // erfolgreichen Poll erkennen wir den Version-Flip.
             sawDown = true
           }
         }, 3000)
-        // Sicherheits-Timeout: nach 2 Min einfach reload — falls der
-        // Restart länger braucht als erwartet, kommt die UI in jedem
-        // Fall wieder hoch.
         setTimeout(() => {
           if (upgradePollRef.current) clearInterval(upgradePollRef.current)
           if (upgradeTickRef.current) clearInterval(upgradeTickRef.current)
@@ -144,13 +229,19 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
       })
   }
 
+  const startRollingUpdate = () => {
+    installedRef.current = installedVersion
+    targetRef.current = targetVersion
+    apiClient.post('/cluster/rolling-update')
+      .then(() => {
+        void rollingStatus.refetch()
+      })
+      .catch((e: Error) => {
+        msg.error(t('update.failed') + ': ' + e.message)
+      })
+  }
+
   if (compact) {
-    // Compact-Variante: Force-Check-Button für "ich will jetzt prüfen",
-    // wenn aktuell NICHTS ausstehendes da ist. Sobald ein Update
-    // verfügbar ist, übernimmt der gelbe Full-Mode-Banner (in
-    // AppLayout) die Sichtbarkeit — wir blenden den Compact-Button
-    // dann komplett aus, sonst doppelt-doppelt Info (Befund 2026-05-15:
-    // "die roten Banner können weg, der gelbe Banner reicht").
     if (updateAvailable) {
       return <>{msgCtx}</>
     }
@@ -171,7 +262,7 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
     )
   }
 
-  if (!updateAvailable && !upgrading) {
+  if (!updateAvailable && !upgrading && !rollingActive) {
     return <>{msgCtx}</>
   }
 
@@ -179,7 +270,7 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
     <>
       {msgCtx}
 
-      {updateAvailable && !upgrading && (
+      {updateAvailable && !upgrading && !rollingActive && (
         <Alert
           type="warning"
           banner
@@ -199,17 +290,33 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
               >
                 {t('update.checkNow')}
               </Button>
-              <Popconfirm
-                title={t('update.confirmTitle')}
-                description={t('update.confirmDesc', { version: targetVersion })}
-                okText={t('update.applyNow')}
-                cancelText={t('common.cancel')}
-                onConfirm={startUpgrade}
-              >
-                <Button size="small" type="primary" icon={<CloudDownloadOutlined />}>
-                  {t('update.applyNow')}
-                </Button>
-              </Popconfirm>
+              {isCluster ? (
+                <Popconfirm
+                  title={t('update.rollingConfirmTitle')}
+                  description={t('update.rollingConfirmDesc', {
+                    secondary: clusterStatus.data?.peers?.[0]?.fqdn ?? 'secondary',
+                  })}
+                  okText={t('update.rollingUpdate')}
+                  cancelText={t('common.cancel')}
+                  onConfirm={startRollingUpdate}
+                >
+                  <Button size="small" type="primary" icon={<ClusterOutlined />}>
+                    {t('update.rollingUpdate')}
+                  </Button>
+                </Popconfirm>
+              ) : (
+                <Popconfirm
+                  title={t('update.confirmTitle')}
+                  description={t('update.confirmDesc', { version: targetVersion })}
+                  okText={t('update.applyNow')}
+                  cancelText={t('common.cancel')}
+                  onConfirm={startUpgrade}
+                >
+                  <Button size="small" type="primary" icon={<CloudDownloadOutlined />}>
+                    {t('update.applyNow')}
+                  </Button>
+                </Popconfirm>
+              )}
             </Space>
           }
         />
@@ -234,26 +341,10 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
             </div>
 
             <div className="update-modal__steps">
-              <Step
-                done={upgradeElapsed >= 5}
-                active={upgradeElapsed < 5}
-                label={t('update.stepDownload')}
-              />
-              <Step
-                done={upgradeElapsed >= 15}
-                active={upgradeElapsed >= 5 && upgradeElapsed < 15}
-                label={t('update.stepInstall')}
-              />
-              <Step
-                done={upgradeElapsed >= 25}
-                active={upgradeElapsed >= 15 && upgradeElapsed < 25}
-                label={t('update.stepRestart')}
-              />
-              <Step
-                done={false}
-                active={upgradeElapsed >= 25}
-                label={t('update.stepVerify')}
-              />
+              <Step done={upgradeElapsed >= 5} active={upgradeElapsed < 5} label={t('update.stepDownload')} />
+              <Step done={upgradeElapsed >= 15} active={upgradeElapsed >= 5 && upgradeElapsed < 15} label={t('update.stepInstall')} />
+              <Step done={upgradeElapsed >= 25} active={upgradeElapsed >= 15 && upgradeElapsed < 25} label={t('update.stepRestart')} />
+              <Step done={false} active={upgradeElapsed >= 25} label={t('update.stepVerify')} />
             </div>
 
             <div className="update-modal__timer">{upgradeElapsed}s</div>
@@ -261,6 +352,54 @@ export default function UpdateBanner({ compact = false }: UpdateBannerProps = {}
           </div>
         </div>
       )}
+
+      {rollingActive && (
+        <div className="update-modal-overlay">
+          <div className="update-modal">
+            <div className="update-modal__orbit">
+              <div className="update-modal__ring" />
+              <div className="update-modal__ring update-modal__ring--2" />
+              <div className="update-modal__dot" />
+              <div className="update-modal__dot update-modal__dot--2" />
+              <div className="update-modal__center">
+                <ClusterOutlined className="update-modal__icon" />
+              </div>
+            </div>
+
+            <div className="update-modal__title">{t('update.rollingRunning')}</div>
+            <div className="update-modal__version">
+              v{installedRef.current || '…'} → v{targetRef.current || '…'}
+            </div>
+
+            <div className="update-modal__steps">
+              <Step
+                done={rollingPhase === 'waiting-secondary' || rollingPhase === 'updating-primary'}
+                active={rollingPhase === 'updating-secondary'}
+                label={t('update.rollingStepSecondary', { fqdn: secondaryFQDN })}
+              />
+              <Step
+                done={rollingPhase === 'updating-primary'}
+                active={rollingPhase === 'waiting-secondary'}
+                label={t('update.rollingStepWaiting')}
+              />
+              <Step
+                done={false}
+                active={rollingPhase === 'updating-primary'}
+                label={t('update.rollingStepPrimary')}
+              />
+            </div>
+
+            <div className="update-modal__timer">{rollingElapsed}s</div>
+            <div className="update-modal__hint">{t('update.waitHint')}</div>
+
+            {rollingStatus.data?.error && (
+              <div className="update-modal__hint" style={{ color: '#ff4d4f' }}>
+                {rollingStatus.data.error}
+              </div>
+            )}
+          </div>
+        </div>
+      )}
     </>
   )
 }

management-ui/src/i18n/locales/de/common.json

@@ -629,6 +629,7 @@
       "node": "Knoten",
       "status": "Status",
       "role": "Rolle",
+      "pgRole": "PG-Rolle",
       "apiUrl": "API-URL",
       "configHash": "Config-Hash",
       "version": "Version",
@@ -641,6 +642,30 @@
       "uptime": "Uptime",
       "fetchMs": "Fetch"
     },
+    "pgRole": {
+      "standalone": "standalone",
+      "primary": "primary",
+      "standby": "standby"
+    },
+    "vipCard": {
+      "title": "Hochverfügbarkeit (VIP / Keepalived)",
+      "vipAddress": "VIP-Adresse",
+      "vipAddressHelp": "Virtuelle IP-Adresse die zwischen Nodes wandert (z.B. 89.163.205.10)",
+      "vipInterface": "Netzwerk-Interface",
+      "vipInterfaceHelp": "Interface auf dem die VIP gebunden wird (z.B. eth0)",
+      "vipAuthPass": "VRRP Auth-Passwort",
+      "vipAuthPassHelp": "Max. 8 Zeichen — Keepalived-Limit. Gleich auf allen Nodes.",
+      "vrrpRouterId": "VRRP Router-ID",
+      "vrrpRouterIdHelp": "Muss im Subnetz eindeutig sein (1–255). Standard: 51.",
+      "saveBtn": "Speichern & Keepalived neu konfigurieren",
+      "saved": "VIP-Einstellungen gespeichert",
+      "saveFailed": "Speichern fehlgeschlagen",
+      "hintTitle": "Nächste Schritte nach dem Speichern",
+      "hintPrimary": "Auf dem Primary: edgeguard-ctl cluster-init-replication",
+      "hintStandby": "Auf dem Secondary: edgeguard-ctl cluster-setup-standby <primary-ip>",
+      "hintKeepalived": "Keepalived auf beiden Nodes: sudo systemctl enable --now keepalived",
+      "hintFailover": "Bei Failover: edgeguard-ctl promote (auf dem Secondary)"
+    },
     "loadTitle": "Per-Node Resources (mTLS-Aggregator)",
     "loadEmpty": "Keine Node-Resources verfügbar — Agent-Listener nicht erreichbar?",
     "certCardTitle": "Cluster-TLS-Zertifikate",
@@ -841,7 +866,14 @@
     "stepDownload": "Pakete laden",
     "stepInstall": "Installation",
     "stepRestart": "Service-Restart",
-    "stepVerify": "Verifizierung"
+    "stepVerify": "Verifizierung",
+    "rollingUpdate": "Rolling Update (Cluster)",
+    "rollingConfirmTitle": "Rolling Update starten?",
+    "rollingConfirmDesc": "Der Secondary-Node ({{secondary}}) wird zuerst aktualisiert, danach dieser Primary. Kein Ausfall für den Proxied-Traffic während der Secondary-Phase.",
+    "rollingRunning": "Rolling Update läuft…",
+    "rollingStepSecondary": "Secondary aktualisieren ({{fqdn}})",
+    "rollingStepWaiting": "Warte auf Neustart des Secondary",
+    "rollingStepPrimary": "Primary aktualisieren (dieser Node)"
   },
   "wg": {
     "title": "WireGuard",

management-ui/src/i18n/locales/en/common.json

@@ -629,6 +629,7 @@
       "node": "Node",
       "status": "Status",
       "role": "Role",
+      "pgRole": "PG role",
       "apiUrl": "API URL",
       "configHash": "Config hash",
       "version": "Version",
@@ -641,6 +642,30 @@
       "uptime": "Uptime",
       "fetchMs": "Fetch"
     },
+    "pgRole": {
+      "standalone": "standalone",
+      "primary": "primary",
+      "standby": "standby"
+    },
+    "vipCard": {
+      "title": "High Availability (VIP / Keepalived)",
+      "vipAddress": "VIP address",
+      "vipAddressHelp": "Virtual IP address that moves between nodes (e.g. 89.163.205.10)",
+      "vipInterface": "Network interface",
+      "vipInterfaceHelp": "Interface to bind the VIP on (e.g. eth0)",
+      "vipAuthPass": "VRRP auth password",
+      "vipAuthPassHelp": "Max. 8 characters — Keepalived limit. Same on all nodes.",
+      "vrrpRouterId": "VRRP router ID",
+      "vrrpRouterIdHelp": "Must be unique in the subnet (1–255). Default: 51.",
+      "saveBtn": "Save & reconfigure Keepalived",
+      "saved": "VIP settings saved",
+      "saveFailed": "Failed to save",
+      "hintTitle": "Next steps after saving",
+      "hintPrimary": "On primary: edgeguard-ctl cluster-init-replication",
+      "hintStandby": "On secondary: edgeguard-ctl cluster-setup-standby <primary-ip>",
+      "hintKeepalived": "Keepalived on both nodes: sudo systemctl enable --now keepalived",
+      "hintFailover": "On failover: edgeguard-ctl promote (on the secondary node)"
+    },
     "loadTitle": "Per-node resources (mTLS aggregator)",
     "loadEmpty": "No node resources available — agent listener unreachable?",
     "certCardTitle": "Cluster TLS certificates",
@@ -841,7 +866,14 @@
     "stepDownload": "Download packages",
     "stepInstall": "Install",
     "stepRestart": "Service restart",
-    "stepVerify": "Verification"
+    "stepVerify": "Verification",
+    "rollingUpdate": "Rolling Update (Cluster)",
+    "rollingConfirmTitle": "Start Rolling Update?",
+    "rollingConfirmDesc": "The secondary node ({{secondary}}) is updated first, then this primary. No downtime for proxied traffic during the secondary phase.",
+    "rollingRunning": "Rolling update in progress…",
+    "rollingStepSecondary": "Updating secondary ({{fqdn}})",
+    "rollingStepWaiting": "Waiting for secondary restart",
+    "rollingStepPrimary": "Updating primary (this node)"
   },
   "wg": {
     "title": "WireGuard",