test: Fix ConsolidationState() test failure in state testing (#1957) 1.1.x (#2039)

Co-authored-by: Jonathan Innis <[email protected]>

Author: rschalo

.github/actions/install-deps/action.yaml

@@ -16,7 +16,7 @@ runs:
     # Root path permission workaround for caching https://github.com/actions/cache/issues/845#issuecomment-1252594999
     - run: sudo chown "$USER" /usr/local
       shell: bash
-    - uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
+    - uses: actions/cache@d4323d4df104b026a6aa633fdb11d772146be0bf # v4.2.2
       id: cache-toolchain
       with:
         path: |

pkg/controllers/state/cluster.go

@@ -426,12 +426,17 @@ func (c *Cluster) ConsolidationState() time.Time {
 func (c *Cluster) Reset() {
 	c.mu.Lock()
 	defer c.mu.Unlock()
+	c.clusterState = time.Time{}
+	c.unsyncedStartTime = time.Time{}
 	c.nodes = map[string]*StateNode{}
 	c.nodeNameToProviderID = map[string]string{}
 	c.nodeClaimNameToProviderID = map[string]string{}
 	c.bindings = map[types.NamespacedName]string{}
 	c.antiAffinityPods = sync.Map{}
 	c.daemonSetPods = sync.Map{}
+	c.podAcks = sync.Map{}
+	c.podsSchedulingAttempted = sync.Map{}
+	c.podsSchedulableTimes = sync.Map{}
 }
 
 func (c *Cluster) GetDaemonSetPod(daemonset *appsv1.DaemonSet) *corev1.Pod {

pkg/controllers/state/suite_test.go

@@ -21,6 +21,7 @@ import (
 	"fmt"
 	"math/rand"
 	"net"
+	"sync"
 	"testing"
 	"time"
 
@@ -1474,7 +1475,7 @@ var _ = Describe("Consolidated State", func() {
 		cluster.MarkUnconsolidated()
 		Expect(cluster.ConsolidationState()).ToNot(Equal(state))
 	})
-	It("should update the consolidated value when consolidation timeout (5m) has passed and state hasn't changed", func() {
+	It("should update the consolidated value when state timeout (5m) has passed and state hasn't changed", func() {
 		state := cluster.ConsolidationState()
 
 		fakeClock.Step(time.Minute)
@@ -1497,14 +1498,21 @@ var _ = Describe("Consolidated State", func() {
 })
 
 var _ = Describe("Data Races", func() {
+	var wg sync.WaitGroup
+	var cancelCtx context.Context
+	var cancel context.CancelFunc
+	BeforeEach(func() {
+		cancelCtx, cancel = context.WithCancel(ctx)
+	})
+	AfterEach(func() {
+		cancel()
+		wg.Wait()
+	})
 	It("should ensure that calling Synced() is valid while making updates to Nodes", func() {
-		cancelCtx, cancel := context.WithCancel(ctx)
-		DeferCleanup(func() {
-			cancel()
-		})
-
 		// Keep calling Synced for the entirety of this test
+		wg.Add(1)
 		go func() {
+			defer wg.Done()
 			for {
 				_ = cluster.Synced(ctx)
 				if cancelCtx.Err() != nil {
@@ -1523,13 +1531,10 @@ var _ = Describe("Data Races", func() {
 		}
 	})
 	It("should ensure that calling Synced() is valid while making updates to NodeClaims", func() {
-		cancelCtx, cancel := context.WithCancel(ctx)
-		DeferCleanup(func() {
-			cancel()
-		})
-
 		// Keep calling Synced for the entirety of this test
+		wg.Add(1)
 		go func() {
+			defer wg.Done()
 			for {
 				_ = cluster.Synced(ctx)
 				if cancelCtx.Err() != nil {