Make release health-gate tests synchronize on observed state

Wait for the agent supervisor to check false readiness before asserting its commit gate, and give the race-instrumented fixture a non-expiring window. Hold the monitoring request at target resolution after its history snapshot so replacement is deterministic. Keep production paths and assertions unchanged.

Change-source: pulse-maintainer

cmd/pulse-agent/main_test.go

@@ -1901,11 +1901,13 @@ func (s *pendingUpdateSupervisorStub) Rollback(_ context.Context, activation age
 func testPendingUpdate(t *testing.T, stateDir string) *agentupdate.PendingPrivilegedUpdate {
 	t.Helper()
 	activation := agenthelper.UpdateResult{
-		Action:           "pending",
-		ActivationID:     "pulse-agent-0123456789abcdef0123456789abcdef:0123456789abcdef",
-		ActiveSHA256:     strings.Repeat("a", 64),
-		RollbackSHA256:   strings.Repeat("b", 64),
-		RollbackDeadline: time.Now().Add(2 * time.Second).UTC(),
+		Action:         "pending",
+		ActivationID:   "pulse-agent-0123456789abcdef0123456789abcdef:0123456789abcdef",
+		ActiveSHA256:   strings.Repeat("a", 64),
+		RollbackSHA256: strings.Repeat("b", 64),
+		// Leave enough time for the race-instrumented supervisor to be scheduled.
+		// The test below checks the health gates, not the expiry path.
+		RollbackDeadline: time.Now().Add(30 * time.Second).UTC(),
 	}
 	if err := agentupdate.PersistPendingPrivilegedUpdate(stateDir, "1.0.0", activation); err != nil {
 		t.Fatal(err)
@@ -1932,14 +1934,38 @@ func TestPendingPrivilegedUpdateCommitsOnlyAfterReadinessAndAcceptedReport(t *te
 	reportAccepted := make(chan struct{})
 	close(reportAccepted)
 	var ready atomic.Bool
+	notReadyChecks := make(chan struct{}, 2)
+	locallyReady := func() bool {
+		if ready.Load() {
+			return true
+		}
+		select {
+		case notReadyChecks <- struct{}{}:
+		default:
+		}
+		return false
+	}
 	result := make(chan error, 1)
 	go func() {
-		result <- supervisePendingPrivilegedUpdate(context.Background(), stub, pending, stateDir, pending.Activation.ActiveSHA256, ready.Load, reportAccepted, time.Millisecond, nil)
+		result <- supervisePendingPrivilegedUpdate(context.Background(), stub, pending, stateDir, pending.Activation.ActiveSHA256, locallyReady, reportAccepted, time.Millisecond, nil)
 	}()
+	// Observe the supervisor checking the false readiness state on two passes.
+	// A sleep cannot prove it started before we flip readiness under -race.
+	for range 2 {
+		select {
+		case <-notReadyChecks:
+		case <-stub.commitCalls:
+			t.Fatal("pending update committed before local readiness")
+		case err := <-result:
+			t.Fatalf("pending update supervisor stopped before local readiness: %v", err)
+		case <-time.After(10 * time.Second):
+			t.Fatal("pending update supervisor did not check local readiness")
+		}
+	}
 	select {
 	case <-stub.commitCalls:
 		t.Fatal("pending update committed before local readiness")
-	case <-time.After(20 * time.Millisecond):
+	default:
 	}
 	ready.Store(true)
 	select {
@@ -1947,7 +1973,9 @@ func TestPendingPrivilegedUpdateCommitsOnlyAfterReadinessAndAcceptedReport(t *te
 		if activation != pending.Activation {
 			t.Fatalf("commit activation = %#v", activation)
 		}
-	case <-time.After(time.Second):
+	case err := <-result:
+		t.Fatalf("pending update supervisor stopped before committing: %v", err)
+	case <-time.After(10 * time.Second):
 		t.Fatal("pending update was not committed after both health signals")
 	}
 	if err := <-result; err != nil {

internal/monitoring/metric_window_provider_test.go

@@ -2,12 +2,35 @@ package monitoring
 
 import (
 	"strings"
+	"sync"
 	"testing"
 	"time"
 
 	"github.com/rcourtman/pulse-go-rewrite/internal/alerts"
+	"github.com/rcourtman/pulse-go-rewrite/internal/models"
+	"github.com/rcourtman/pulse-go-rewrite/internal/unifiedresources"
 )
 
+// This resolver blocks only after metricWindowPoints has captured its history.
+// It lets the test replace that history at an exact point without scheduling
+// sleeps or exposing a production-only hook.
+type blockingMetricTargetStore struct {
+	entered chan struct{}
+	resume  <-chan struct{}
+}
+
+func (*blockingMetricTargetStore) ShouldSkipAPIPolling(string) bool { return false }
+func (*blockingMetricTargetStore) GetPollingRecommendations() map[string]float64 {
+	return nil
+}
+func (*blockingMetricTargetStore) GetAll() []unifiedresources.Resource       { return nil }
+func (*blockingMetricTargetStore) PopulateFromSnapshot(models.StateSnapshot) {}
+func (s *blockingMetricTargetStore) MetricsTargetForResource(string) *unifiedresources.MetricsTarget {
+	s.entered <- struct{}{}
+	<-s.resume
+	return nil
+}
+
 func TestMetricWindowPointsUsesInMemoryMetricAlias(t *testing.T) {
 	now := time.Now().UTC()
 	history := NewMetricsHistory(32, time.Hour)
@@ -77,33 +100,47 @@ func TestMetricWindowPointsKeepsHistorySnapshotDuringReplacement(t *testing.T) {
 		},
 	}
 
-	history.metricWindowMu.Lock()
-	result := make(chan []alerts.MetricWindowPoint, 1)
+	resume := make(chan struct{})
+	var resumeOnce sync.Once
+	resumeRequest := func() { resumeOnce.Do(func() { close(resume) }) }
+	defer resumeRequest()
+	entered := make(chan struct{}, 1)
+	monitor.resourceStore = &blockingMetricTargetStore{entered: entered, resume: resume}
+	type requestResult struct {
+		points []alerts.MetricWindowPoint
+		err    error
+	}
+	result := make(chan requestResult, 1)
 	go func() {
-		points, _ := monitor.metricWindowPoints(alerts.MetricWindowRequest{
+		points, err := monitor.metricWindowPoints(alerts.MetricWindowRequest{
 			ResourceID:   "vm-1",
 			ResourceType: "vm",
 			Metric:       "cpu",
 			Start:        now.Add(-5 * time.Minute),
 			End:          now,
 		})
-		result <- points
+		result <- requestResult{points: points, err: err}
 	}()
 
-	// Give the request time to snapshot history and block on its cache mutex,
-	// matching the mock seed replacement that exposed the startup panic.
-	time.Sleep(20 * time.Millisecond)
+	select {
+	case <-entered:
+	case <-time.After(10 * time.Second):
+		t.Fatal("metric request did not reach target resolution after snapshot")
+	}
 	monitor.mu.Lock()
 	monitor.metricsHistory = NewMetricsHistory(32, time.Hour)
 	monitor.mu.Unlock()
-	history.metricWindowMu.Unlock()
+	resumeRequest()
 
 	select {
-	case points := <-result:
-		if len(points) != 1 || points[0].Value != 42 {
-			t.Fatalf("metricWindowPoints = %+v, want cached point from the request snapshot", points)
+	case got := <-result:
+		if got.err != nil {
+			t.Fatalf("metricWindowPoints returned error: %v", got.err)
 		}
-	case <-time.After(time.Second):
+		if len(got.points) != 1 || got.points[0].Value != 42 {
+			t.Fatalf("metricWindowPoints = %+v, want cached point from the request snapshot", got.points)
+		}
+	case <-time.After(10 * time.Second):
 		t.Fatal("metricWindowPoints did not finish after history replacement")
 	}
 }