[client] fix flaky test around event aggregation (#6710)

* fix flaky test around event aggregation: control time.Now() from the test

Signed-off-by: Dmitri Dolguikh <dmitri.external@netbird.io>

* actually use passed in func to generate time

Signed-off-by: Dmitri Dolguikh <dmitri.external@netbird.io>

---------

Signed-off-by: Dmitri Dolguikh <dmitri.external@netbird.io>

client/internal/conn_mgr.go

@@ -109,7 +109,7 @@ func (e *ConnMgr) UpdatedRemoteFeatureFlag(ctx context.Context, enabled bool) er
 			return nil
 		}
 
-		log.Warnf("lazy connection manager is enabled by management feature flag")
+		log.Infof("lazy connection manager is enabled by the management feature flag")
 		e.initLazyManager(ctx)
 		e.statusRecorder.UpdateLazyConnection(true)
 		return e.addPeersToLazyConnManager()

client/internal/netflow/store/event_aggregation_test.go

@@ -175,7 +175,9 @@ func TestFlowAggregationOfUnknownProtocols(t *testing.T) {
 }
 
 func TestResetAggregationWindow(t *testing.T) {
-	store := NewAggregatingMemoryStore()
+	now := time.Now()
+	nowFunc := func() time.Time { return now }
+	store := NewAggregatingMemoryStoreWithTimeFunc(nowFunc)
 	store.StoreEvent(&types.Event{
 		ID:        uuid.New(),
 		Timestamp: time.Now(),
@@ -198,6 +200,7 @@ func TestResetAggregationWindow(t *testing.T) {
 		},
 	})
 
+	now = now.Add(1 * time.Second)
 	reset := store.ResetAggregationWindow()
 	previousEvents, ok := reset.(*AggregatingMemory)
 	assert.True(t, ok)

client/internal/netflow/store/memory.go

@@ -29,6 +29,7 @@ type AggregatingMemory struct {
 	WindowStart time.Time
 	WindowEnd   time.Time
 	rnd         *v2.PCG
+	nowFunc     func() time.Time
 }
 
 func (m *Memory) StoreEvent(event *types.Event) {
@@ -62,14 +63,19 @@ func (m *Memory) DeleteEvents(ids []uuid.UUID) {
 }
 
 func NewAggregatingMemoryStore() *AggregatingMemory {
-	return &AggregatingMemory{WindowStart: time.Now(), Memory: Memory{events: make(map[uuid.UUID]*types.Event)}, rnd: v2.NewPCG(rand.Uint64(), rand.Uint64())}
+	return NewAggregatingMemoryStoreWithTimeFunc(defaultNowFunc)
+}
+
+// used in tests when deterministic (less random) time intervals are required
+func NewAggregatingMemoryStoreWithTimeFunc(nowFunc func() time.Time) *AggregatingMemory {
+	return &AggregatingMemory{WindowStart: nowFunc(), Memory: Memory{events: make(map[uuid.UUID]*types.Event)}, nowFunc: nowFunc, rnd: v2.NewPCG(rand.Uint64(), rand.Uint64())}
 }
 
 func (am *AggregatingMemory) ResetAggregationWindow() types.FlowEventAggregator {
 	am.mux.Lock()
 	defer am.mux.Unlock()
 
-	now := time.Now()
+	now := am.nowFunc()
 	toret := AggregatingMemory{WindowStart: am.WindowStart, WindowEnd: now, Memory: Memory{events: am.events}, rnd: v2.NewPCG(rand.Uint64(), rand.Uint64())}
 
 	am.events = make(map[uuid.UUID]*types.Event)
@@ -152,3 +158,7 @@ func (am *AggregatingMemory) GetAggregatedEvents() []*types.Event {
 
 	return slices.Collect(maps.Values(aggregated)) // could return an iterator instead here
 }
+
+func defaultNowFunc() time.Time {
+	return time.Now()
+}

client/internal/profilemanager/service.go

@@ -11,6 +11,7 @@ import (
 	"runtime"
 	"sort"
 	"strings"
+	"syscall"
 
 	log "github.com/sirupsen/logrus"
 
@@ -439,7 +440,11 @@ func (s *ServiceManager) GetStatePath() string {
 
 	activeProf, err := s.GetActiveProfileState()
 	if err != nil {
-		log.Warnf("failed to get active profile state: %v", err)
+		if errors.Is(err, syscall.ENOSYS) {
+			log.Debugf("active profile state unavailable on this platform: %v", err)
+		} else {
+			log.Warnf("failed to get active profile state: %v", err)
+		}
 		return defaultStatePath
 	}
 

client/internal/routemanager/manager.go

@@ -12,6 +12,7 @@ import (
 	"strings"
 	"sync"
 	"sync/atomic"
+	"syscall"
 	"time"
 
 	"github.com/google/uuid"
@@ -264,7 +265,11 @@ func (m *DefaultManager) initSelector() *routeselector.RouteSelector {
 
 	// restore selector state if it exists
 	if err := m.stateManager.LoadState(state); err != nil {
-		log.Warnf("failed to load state: %v", err)
+		if errors.Is(err, syscall.ENOSYS) {
+			log.Debugf("route selector state unavailable on this platform: %v", err)
+		} else {
+			log.Warnf("failed to load state: %v", err)
+		}
 		return routeselector.NewRouteSelector()
 	}
 

client/wasm/internal/rdp/rdcleanpath.go

@@ -23,7 +23,7 @@ const (
 	RDCleanPathProxyHost   = "rdcleanpath.proxy.local"
 	RDCleanPathProxyScheme = "ws"
 
-	rdpDialTimeout = 15 * time.Second
+	rdpDialTimeout = 30 * time.Second
 
 	GeneralErrorCode = 1
 	WSAETimedOut     = 10060

shared/management/http/api/openapi.yml

@@ -9327,6 +9327,18 @@ paths:
           required: false
           schema:
             type: string
+        - name: source_id
+          in: query
+          description: Filter by source endpoint ID
+          required: false
+          schema:
+            type: string
+        - name: destination_id
+          in: query
+          description: Filter by destination endpoint ID
+          required: false
+          schema:
+            type: string
         - name: protocol
           in: query
           description: Filter by protocol

shared/management/http/api/types.gen.go

@@ -5857,6 +5857,12 @@ type GetApiEventsNetworkTrafficParams struct {
 	// ReporterId Filter by reporter ID
 	ReporterId *string `form:"reporter_id,omitempty" json:"reporter_id,omitempty"`
 
+	// SourceId Filter by source endpoint ID
+	SourceId *string `form:"source_id,omitempty" json:"source_id,omitempty"`
+
+	// DestinationId Filter by destination endpoint ID
+	DestinationId *string `form:"destination_id,omitempty" json:"destination_id,omitempty"`
+
 	// Protocol Filter by protocol
 	Protocol *int `form:"protocol,omitempty" json:"protocol,omitempty"`
 

shared/relay/client/dialer/ws/close_generic.go

@@ -0,0 +1,9 @@
+//go:build !js
+
+package ws
+
+// closeConn closes the underlying WebSocket immediately, skipping the close
+// handshake.
+func (c *Conn) closeConn() error {
+	return c.Conn.CloseNow()
+}

shared/relay/client/dialer/ws/close_js.go

@@ -0,0 +1,25 @@
+//go:build js
+
+package ws
+
+import (
+	"github.com/coder/websocket"
+	log "github.com/sirupsen/logrus"
+)
+
+// closeConn closes the browser WebSocket without blocking the caller.
+//
+// The browser close API only accepts codes 1000 and 3000-4999, so CloseNow's
+// 1001 (going away) throws an InvalidAccessError. Close with a valid code
+// waits for the browser close event before returning, which can park the
+// calling goroutine (the relay teardown path holds its mutexes while closing)
+// until the close handshake finishes. Run the close in the background and
+// report success; a teardown close error is not actionable.
+func (c *Conn) closeConn() error {
+	go func() {
+		if err := c.Conn.Close(websocket.StatusNormalClosure, ""); err != nil {
+			log.Debugf("failed to close relay websocket: %v", err)
+		}
+	}()
+	return nil
+}

shared/relay/client/dialer/ws/conn.go

@@ -77,5 +77,5 @@ func (c *Conn) SetDeadline(t time.Time) error {
 }
 
 func (c *Conn) Close() error {
-	return c.Conn.CloseNow()
+	return c.closeConn()
 }