[management, proxy] Add require_subdomain capability for proxy clusters (#5628)

management/internals/modules/reverseproxy/domain/domain.go

@@ -17,6 +17,9 @@ type Domain struct {
 	// SupportsCustomPorts is populated at query time for free domains from the
 	// proxy cluster capabilities. Not persisted.
 	SupportsCustomPorts *bool `gorm:"-"`
+	// RequireSubdomain is populated at query time. When true, the domain
+	// cannot be used bare and a subdomain label must be prepended. Not persisted.
+	RequireSubdomain *bool `gorm:"-"`
 }
 
 // EventMeta returns activity event metadata for a domain

management/internals/modules/reverseproxy/domain/manager/api.go

@@ -47,6 +47,7 @@ func domainToApi(d *domain.Domain) api.ReverseProxyDomain {
 		Type:                domainTypeToApi(d.Type),
 		Validated:           d.Validated,
 		SupportsCustomPorts: d.SupportsCustomPorts,
+		RequireSubdomain:    d.RequireSubdomain,
 	}
 	if d.TargetCluster != "" {
 		resp.TargetCluster = &d.TargetCluster

management/internals/modules/reverseproxy/domain/manager/domain_test.go

@@ -0,0 +1,172 @@
+package manager
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+
+	"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/domain"
+)
+
+func TestExtractClusterFromFreeDomain(t *testing.T) {
+	clusters := []string{"eu1.proxy.netbird.io", "us1.proxy.netbird.io"}
+
+	tests := []struct {
+		name    string
+		domain  string
+		wantOK  bool
+		wantVal string
+	}{
+		{
+			name:    "subdomain of cluster matches",
+			domain:  "myapp.eu1.proxy.netbird.io",
+			wantOK:  true,
+			wantVal: "eu1.proxy.netbird.io",
+		},
+		{
+			name:    "deep subdomain of cluster matches",
+			domain:  "foo.bar.eu1.proxy.netbird.io",
+			wantOK:  true,
+			wantVal: "eu1.proxy.netbird.io",
+		},
+		{
+			name:    "bare cluster domain matches",
+			domain:  "eu1.proxy.netbird.io",
+			wantOK:  true,
+			wantVal: "eu1.proxy.netbird.io",
+		},
+		{
+			name:   "unrelated domain does not match",
+			domain: "example.com",
+			wantOK: false,
+		},
+		{
+			name:   "partial suffix does not match",
+			domain: "fakeu1.proxy.netbird.io",
+			wantOK: false,
+		},
+		{
+			name:    "second cluster matches",
+			domain:  "app.us1.proxy.netbird.io",
+			wantOK:  true,
+			wantVal: "us1.proxy.netbird.io",
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			cluster, ok := ExtractClusterFromFreeDomain(tc.domain, clusters)
+			assert.Equal(t, tc.wantOK, ok)
+			if ok {
+				assert.Equal(t, tc.wantVal, cluster)
+			}
+		})
+	}
+}
+
+func TestExtractClusterFromCustomDomains(t *testing.T) {
+	customDomains := []*domain.Domain{
+		{Domain: "example.com", TargetCluster: "eu1.proxy.netbird.io"},
+		{Domain: "proxy.corp.io", TargetCluster: "us1.proxy.netbird.io"},
+	}
+
+	tests := []struct {
+		name    string
+		domain  string
+		wantOK  bool
+		wantVal string
+	}{
+		{
+			name:    "subdomain of custom domain matches",
+			domain:  "app.example.com",
+			wantOK:  true,
+			wantVal: "eu1.proxy.netbird.io",
+		},
+		{
+			name:    "bare custom domain matches",
+			domain:  "example.com",
+			wantOK:  true,
+			wantVal: "eu1.proxy.netbird.io",
+		},
+		{
+			name:    "deep subdomain of custom domain matches",
+			domain:  "a.b.example.com",
+			wantOK:  true,
+			wantVal: "eu1.proxy.netbird.io",
+		},
+		{
+			name:    "subdomain of multi-level custom domain matches",
+			domain:  "app.proxy.corp.io",
+			wantOK:  true,
+			wantVal: "us1.proxy.netbird.io",
+		},
+		{
+			name:    "bare multi-level custom domain matches",
+			domain:  "proxy.corp.io",
+			wantOK:  true,
+			wantVal: "us1.proxy.netbird.io",
+		},
+		{
+			name:   "unrelated domain does not match",
+			domain: "other.com",
+			wantOK: false,
+		},
+		{
+			name:   "partial suffix does not match custom domain",
+			domain: "fakeexample.com",
+			wantOK: false,
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			cluster, ok := extractClusterFromCustomDomains(tc.domain, customDomains)
+			assert.Equal(t, tc.wantOK, ok)
+			if ok {
+				assert.Equal(t, tc.wantVal, cluster)
+			}
+		})
+	}
+}
+
+func TestExtractClusterFromCustomDomains_OverlappingDomains(t *testing.T) {
+	customDomains := []*domain.Domain{
+		{Domain: "example.com", TargetCluster: "cluster-generic"},
+		{Domain: "app.example.com", TargetCluster: "cluster-app"},
+	}
+
+	tests := []struct {
+		name    string
+		domain  string
+		wantVal string
+	}{
+		{
+			name:    "exact match on more specific domain",
+			domain:  "app.example.com",
+			wantVal: "cluster-app",
+		},
+		{
+			name:    "subdomain of more specific domain",
+			domain:  "api.app.example.com",
+			wantVal: "cluster-app",
+		},
+		{
+			name:    "subdomain of generic domain",
+			domain:  "other.example.com",
+			wantVal: "cluster-generic",
+		},
+		{
+			name:    "bare generic domain",
+			domain:  "example.com",
+			wantVal: "cluster-generic",
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			cluster, ok := extractClusterFromCustomDomains(tc.domain, customDomains)
+			assert.True(t, ok)
+			assert.Equal(t, tc.wantVal, cluster)
+		})
+	}
+}

management/internals/modules/reverseproxy/domain/manager/manager.go

@@ -35,6 +35,7 @@ type proxyManager interface {
 
 type clusterCapabilities interface {
 	ClusterSupportsCustomPorts(clusterAddr string) *bool
+	ClusterRequireSubdomain(clusterAddr string) *bool
 }
 
 type Manager struct {
@@ -98,6 +99,7 @@ func (m Manager) GetDomains(ctx context.Context, accountID, userID string) ([]*d
 		}
 		if m.clusterCapabilities != nil {
 			d.SupportsCustomPorts = m.clusterCapabilities.ClusterSupportsCustomPorts(cluster)
+			d.RequireSubdomain = m.clusterCapabilities.ClusterRequireSubdomain(cluster)
 		}
 		ret = append(ret, d)
 	}
@@ -115,6 +117,8 @@ func (m Manager) GetDomains(ctx context.Context, accountID, userID string) ([]*d
 		if m.clusterCapabilities != nil && d.TargetCluster != "" {
 			cd.SupportsCustomPorts = m.clusterCapabilities.ClusterSupportsCustomPorts(d.TargetCluster)
 		}
+		// Custom domains never require a subdomain by default since
+		// the account owns them and should be able to use the bare domain.
 		ret = append(ret, cd)
 	}
 
@@ -302,13 +306,19 @@ func (m Manager) DeriveClusterFromDomain(ctx context.Context, accountID, domain
 	return "", fmt.Errorf("domain %s does not match any available proxy cluster", domain)
 }
 
-func extractClusterFromCustomDomains(domain string, customDomains []*domain.Domain) (string, bool) {
-	for _, customDomain := range customDomains {
-		if strings.HasSuffix(domain, "."+customDomain.Domain) {
-			return customDomain.TargetCluster, true
+func extractClusterFromCustomDomains(serviceDomain string, customDomains []*domain.Domain) (string, bool) {
+	bestCluster := ""
+	bestLen := -1
+	for _, cd := range customDomains {
+		if serviceDomain != cd.Domain && !strings.HasSuffix(serviceDomain, "."+cd.Domain) {
+			continue
+		}
+		if l := len(cd.Domain); l > bestLen {
+			bestLen = l
+			bestCluster = cd.TargetCluster
 		}
 	}
-	return "", false
+	return bestCluster, bestLen >= 0
 }
 
 // ExtractClusterFromFreeDomain extracts the cluster address from a free domain.

management/internals/modules/reverseproxy/proxy/manager.go

@@ -35,4 +35,5 @@ type Controller interface {
 	UnregisterProxyFromCluster(ctx context.Context, clusterAddr, proxyID string) error
 	GetProxiesForCluster(clusterAddr string) []string
 	ClusterSupportsCustomPorts(clusterAddr string) *bool
+	ClusterRequireSubdomain(clusterAddr string) *bool
 }

management/internals/modules/reverseproxy/proxy/manager/controller.go

@@ -77,6 +77,12 @@ func (c *GRPCController) ClusterSupportsCustomPorts(clusterAddr string) *bool {
 	return c.proxyGRPCServer.ClusterSupportsCustomPorts(clusterAddr)
 }
 
+// ClusterRequireSubdomain returns whether the cluster requires a subdomain label.
+// Returns nil when no proxy has reported the capability (defaults to false).
+func (c *GRPCController) ClusterRequireSubdomain(clusterAddr string) *bool {
+	return c.proxyGRPCServer.ClusterRequireSubdomain(clusterAddr)
+}
+
 // GetProxiesForCluster returns all proxy IDs registered for a specific cluster.
 func (c *GRPCController) GetProxiesForCluster(clusterAddr string) []string {
 	proxySet, ok := c.clusterProxies.Load(clusterAddr)

management/internals/modules/reverseproxy/proxy/manager_mock.go

@@ -159,6 +159,20 @@ func (mr *MockControllerMockRecorder) ClusterSupportsCustomPorts(clusterAddr int
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "ClusterSupportsCustomPorts", reflect.TypeOf((*MockController)(nil).ClusterSupportsCustomPorts), clusterAddr)
 }
 
+// ClusterRequireSubdomain mocks base method.
+func (m *MockController) ClusterRequireSubdomain(clusterAddr string) *bool {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "ClusterRequireSubdomain", clusterAddr)
+	ret0, _ := ret[0].(*bool)
+	return ret0
+}
+
+// ClusterRequireSubdomain indicates an expected call of ClusterRequireSubdomain.
+func (mr *MockControllerMockRecorder) ClusterRequireSubdomain(clusterAddr interface{}) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "ClusterRequireSubdomain", reflect.TypeOf((*MockController)(nil).ClusterRequireSubdomain), clusterAddr)
+}
+
 // GetOIDCValidationConfig mocks base method.
 func (m *MockController) GetOIDCValidationConfig() OIDCValidationConfig {
 	m.ctrl.T.Helper()

management/internals/modules/reverseproxy/service/manager/l4_port_test.go

@@ -76,6 +76,7 @@ func setupL4Test(t *testing.T, customPortsSupported *bool) (*Manager, store.Stor
 
 	mockCtrl := proxy.NewMockController(ctrl)
 	mockCtrl.EXPECT().ClusterSupportsCustomPorts(gomock.Any()).Return(customPortsSupported).AnyTimes()
+	mockCtrl.EXPECT().ClusterRequireSubdomain(gomock.Any()).Return((*bool)(nil)).AnyTimes()
 	mockCtrl.EXPECT().SendServiceUpdateToCluster(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).AnyTimes()
 	mockCtrl.EXPECT().GetOIDCValidationConfig().Return(proxy.OIDCValidationConfig{}).AnyTimes()
 

management/internals/modules/reverseproxy/service/manager/manager.go

@@ -236,6 +236,10 @@ func (m *Manager) initializeServiceForCreate(ctx context.Context, accountID stri
 			return status.Errorf(status.PreconditionFailed, "could not derive cluster from domain %s: %v", service.Domain, err)
 		}
 		service.ProxyCluster = proxyCluster
+
+		if err := m.validateSubdomainRequirement(service.Domain, proxyCluster); err != nil {
+			return err
+		}
 	}
 
 	service.AccountID = accountID
@@ -261,6 +265,20 @@ func (m *Manager) initializeServiceForCreate(ctx context.Context, accountID stri
 	return nil
 }
 
+// validateSubdomainRequirement checks whether the domain can be used bare
+// (without a subdomain label) on the given cluster. If the cluster reports
+// require_subdomain=true and the domain equals the cluster domain, it rejects.
+func (m *Manager) validateSubdomainRequirement(domain, cluster string) error {
+	if domain != cluster {
+		return nil
+	}
+	requireSub := m.proxyController.ClusterRequireSubdomain(cluster)
+	if requireSub != nil && *requireSub {
+		return status.Errorf(status.InvalidArgument, "domain %s requires a subdomain label", domain)
+	}
+	return nil
+}
+
 func (m *Manager) persistNewService(ctx context.Context, accountID string, svc *service.Service) error {
 	return m.store.ExecuteInTransaction(ctx, func(transaction store.Store) error {
 		if svc.Domain != "" {
@@ -489,53 +507,61 @@ func (m *Manager) persistServiceUpdate(ctx context.Context, accountID string, se
 	var updateInfo serviceUpdateInfo
 
 	err := m.store.ExecuteInTransaction(ctx, func(transaction store.Store) error {
-		existingService, err := transaction.GetServiceByID(ctx, store.LockingStrengthUpdate, accountID, service.ID)
-		if err != nil {
-			return err
-		}
-
-		if err := validateProtocolChange(existingService.Mode, service.Mode); err != nil {
-			return err
-		}
-
-		updateInfo.oldCluster = existingService.ProxyCluster
-		updateInfo.domainChanged = existingService.Domain != service.Domain
-
-		if updateInfo.domainChanged {
-			if err := m.handleDomainChange(ctx, transaction, accountID, service); err != nil {
-				return err
-			}
-		} else {
-			service.ProxyCluster = existingService.ProxyCluster
-		}
-
-		m.preserveExistingAuthSecrets(service, existingService)
-		if err := validateHeaderAuthValues(service.Auth.HeaderAuths); err != nil {
-			return err
-		}
-		m.preserveServiceMetadata(service, existingService)
-		m.preserveListenPort(service, existingService)
-		updateInfo.serviceEnabledChanged = existingService.Enabled != service.Enabled
-
-		if err := m.ensureL4Port(ctx, transaction, service); err != nil {
-			return err
-		}
-		if err := m.checkPortConflict(ctx, transaction, service); err != nil {
-			return err
-		}
-		if err := validateTargetReferences(ctx, transaction, accountID, service.Targets); err != nil {
-			return err
-		}
-		if err := transaction.UpdateService(ctx, service); err != nil {
-			return fmt.Errorf("update service: %w", err)
-		}
-
-		return nil
+		return m.executeServiceUpdate(ctx, transaction, accountID, service, &updateInfo)
 	})
 
 	return &updateInfo, err
 }
 
+func (m *Manager) executeServiceUpdate(ctx context.Context, transaction store.Store, accountID string, service *service.Service, updateInfo *serviceUpdateInfo) error {
+	existingService, err := transaction.GetServiceByID(ctx, store.LockingStrengthUpdate, accountID, service.ID)
+	if err != nil {
+		return err
+	}
+
+	if err := validateProtocolChange(existingService.Mode, service.Mode); err != nil {
+		return err
+	}
+
+	updateInfo.oldCluster = existingService.ProxyCluster
+	updateInfo.domainChanged = existingService.Domain != service.Domain
+
+	if updateInfo.domainChanged {
+		if err := m.handleDomainChange(ctx, transaction, accountID, service); err != nil {
+			return err
+		}
+	} else {
+		service.ProxyCluster = existingService.ProxyCluster
+	}
+
+	if err := m.validateSubdomainRequirement(service.Domain, service.ProxyCluster); err != nil {
+		return err
+	}
+
+	m.preserveExistingAuthSecrets(service, existingService)
+	if err := validateHeaderAuthValues(service.Auth.HeaderAuths); err != nil {
+		return err
+	}
+	m.preserveServiceMetadata(service, existingService)
+	m.preserveListenPort(service, existingService)
+	updateInfo.serviceEnabledChanged = existingService.Enabled != service.Enabled
+
+	if err := m.ensureL4Port(ctx, transaction, service); err != nil {
+		return err
+	}
+	if err := m.checkPortConflict(ctx, transaction, service); err != nil {
+		return err
+	}
+	if err := validateTargetReferences(ctx, transaction, accountID, service.Targets); err != nil {
+		return err
+	}
+	if err := transaction.UpdateService(ctx, service); err != nil {
+		return fmt.Errorf("update service: %w", err)
+	}
+
+	return nil
+}
+
 func (m *Manager) handleDomainChange(ctx context.Context, transaction store.Store, accountID string, svc *service.Service) error {
 	if err := m.checkDomainAvailable(ctx, transaction, svc.Domain, svc.ID); err != nil {
 		return err

management/internals/modules/reverseproxy/service/manager/manager_test.go

@@ -1272,3 +1272,69 @@ func TestValidateTargetReferences_PeerValid(t *testing.T) {
 	}
 	require.NoError(t, validateTargetReferences(ctx, mockStore, accountID, targets))
 }
+
+func TestValidateSubdomainRequirement(t *testing.T) {
+	ptrBool := func(b bool) *bool { return &b }
+
+	tests := []struct {
+		name             string
+		domain           string
+		cluster          string
+		requireSubdomain *bool
+		wantErr          bool
+	}{
+		{
+			name:             "subdomain present, require_subdomain true",
+			domain:           "app.eu1.proxy.netbird.io",
+			cluster:          "eu1.proxy.netbird.io",
+			requireSubdomain: ptrBool(true),
+			wantErr:          false,
+		},
+		{
+			name:             "bare cluster domain, require_subdomain true",
+			domain:           "eu1.proxy.netbird.io",
+			cluster:          "eu1.proxy.netbird.io",
+			requireSubdomain: ptrBool(true),
+			wantErr:          true,
+		},
+		{
+			name:             "bare cluster domain, require_subdomain false",
+			domain:           "eu1.proxy.netbird.io",
+			cluster:          "eu1.proxy.netbird.io",
+			requireSubdomain: ptrBool(false),
+			wantErr:          false,
+		},
+		{
+			name:             "bare cluster domain, require_subdomain nil (default)",
+			domain:           "eu1.proxy.netbird.io",
+			cluster:          "eu1.proxy.netbird.io",
+			requireSubdomain: nil,
+			wantErr:          false,
+		},
+		{
+			name:             "custom domain apex is not the cluster",
+			domain:           "example.com",
+			cluster:          "eu1.proxy.netbird.io",
+			requireSubdomain: ptrBool(true),
+			wantErr:          false,
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			ctrl := gomock.NewController(t)
+
+			mockCtrl := proxy.NewMockController(ctrl)
+			mockCtrl.EXPECT().ClusterRequireSubdomain(tc.cluster).Return(tc.requireSubdomain).AnyTimes()
+
+			mgr := &Manager{proxyController: mockCtrl}
+			err := mgr.validateSubdomainRequirement(tc.domain, tc.cluster)
+			if tc.wantErr {
+				require.Error(t, err)
+				assert.Contains(t, err.Error(), "requires a subdomain label")
+			} else {
+				require.NoError(t, err)
+			}
+		})
+	}
+}

management/internals/shared/grpc/proxy.go

@@ -537,6 +537,35 @@ func (s *ProxyServiceServer) ClusterSupportsCustomPorts(clusterAddr string) *boo
 	return nil
 }
 
+// ClusterRequireSubdomain returns whether any connected proxy in the given
+// cluster reports that a subdomain is required. Returns nil if no proxy has
+// reported the capability (defaults to not required).
+func (s *ProxyServiceServer) ClusterRequireSubdomain(clusterAddr string) *bool {
+	if s.proxyController == nil {
+		return nil
+	}
+
+	var hasCapabilities bool
+	for _, pid := range s.proxyController.GetProxiesForCluster(clusterAddr) {
+		connVal, ok := s.connectedProxies.Load(pid)
+		if !ok {
+			continue
+		}
+		conn := connVal.(*proxyConnection)
+		if conn.capabilities == nil || conn.capabilities.RequireSubdomain == nil {
+			continue
+		}
+		if *conn.capabilities.RequireSubdomain {
+			return ptr(true)
+		}
+		hasCapabilities = true
+	}
+	if hasCapabilities {
+		return ptr(false)
+	}
+	return nil
+}
+
 func (s *ProxyServiceServer) Authenticate(ctx context.Context, req *proto.AuthenticateRequest) (*proto.AuthenticateResponse, error) {
 	service, err := s.serviceManager.GetServiceByID(ctx, req.GetAccountId(), req.GetId())
 	if err != nil {

management/internals/shared/grpc/proxy_test.go

@@ -57,6 +57,10 @@ func (c *testProxyController) ClusterSupportsCustomPorts(_ string) *bool {
 	return ptr(true)
 }
 
+func (c *testProxyController) ClusterRequireSubdomain(_ string) *bool {
+	return nil
+}
+
 func (c *testProxyController) GetProxiesForCluster(clusterAddr string) []string {
 	c.mu.Lock()
 	defer c.mu.Unlock()