triggering once is enough

etcd-io · Jun 9, 2023 · a030901 · a030901
1 parent aeaf52b
commit a030901
Show file tree

Hide file tree

Showing 2 changed files with 32 additions and 44 deletions.
diff --git a/tests/robustness/failpoints.go b/tests/robustness/failpoints.go
@@ -531,58 +531,50 @@ type noSpaceFailpoint struct {
 }
 
 func (f noSpaceFailpoint) Inject(ctx context.Context, t *testing.T, lg *zap.Logger, clus *e2e.EtcdProcessCluster) error {
-	quorumSize := (len(clus.Procs) / 2) + 1
-	for i := 0; i < quorumSize; i++ {
-		member := clus.Procs[i]
-		if member.IsRunning() {
-			lg.Info("Setting up gofailpoint", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
-			err := member.Failpoints().Setup(ctx, f.Name(), "return")
-			if err != nil {
-				lg.Info("goFailpoint setup failed", zap.String("failpoint", f.Name()), zap.Error(err))
+	memberIdx := rand.Int() % len(clus.Procs)
+	member := clus.Procs[memberIdx]
+	if member.IsRunning() {
+		lg.Info("Setting up gofailpoint", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
+		err := member.Failpoints().Setup(ctx, f.Name(), "return")
+		if err != nil {
+			lg.Info("goFailpoint setup failed", zap.String("failpoint", f.Name()), zap.Error(err))
+			return err
+		}
+
+		numRetries := 0
+		for {
+			if numRetries > 5 {
+				lg.Info("goFailpoint giving up after 5 retries", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
 				return err
 			}
 
-			numRetries := 0
-			for {
-				if numRetries > 5 {
-					lg.Info("goFailpoint giving up after 5 retries", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
-					return err
+			err = member.Etcdctl().Put(context.Background(), "failpoint_"+member.Config().Name, "", config.PutOptions{})
+			if err != nil {
+				if strings.Contains(err.Error(), rpctypes.ErrGRPCNoSpace.Error()) {
+					lg.Info("goFailpoint triggered successfully", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
+					break
 				}
 
-				err = member.Etcdctl().Put(context.Background(), "failpoint_"+member.Config().Name, "", config.PutOptions{})
-				if err != nil {
-					if strings.Contains(err.Error(), rpctypes.ErrGRPCNoSpace.Error()) {
-						lg.Info("goFailpoint triggered successfully", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
-						break
-					}
-
-					lg.Info("goFailpoint received unrelated error, retrying", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
-					time.Sleep(1 * time.Second)
-					numRetries++
-					continue
-				}
+				lg.Info("goFailpoint received unrelated error, retrying", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
+				time.Sleep(1 * time.Second)
+				numRetries++
+				continue
 			}
 		}
 	}
 
 	time.Sleep(f.alarmDuration)
 
-	for i := 0; i < quorumSize; i++ {
-		member := clus.Procs[i]
-		lg.Info("goFailpoint starting to disable", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
-
-		err := member.Failpoints().Disable(ctx, f.Name())
-		if err != nil {
-			lg.Info("goFailpoint disable failed", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
-			return err
-		}
-
-		lg.Info("goFailpoint disabled successfully", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
+	lg.Info("goFailpoint starting to disable", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
+	err := member.Failpoints().Disable(ctx, f.Name())
+	if err != nil {
+		lg.Info("goFailpoint disable failed", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
+		return err
 	}
 
-	// disarming can be done once at the end, clearing all alarms
-	member := clus.Procs[0]
-	_, err := member.Etcdctl().AlarmDisarm(ctx, nil)
+	lg.Info("goFailpoint disabled successfully", zap.String("failpoint", f.Name()), zap.String("target", member.Config().Name))
+
+	_, err = member.Etcdctl().AlarmDisarm(ctx, nil)
 	if err != nil {
 		lg.Info("goFailpoint alarm disarm failed", zap.String("failpoint", f.Name()), zap.Error(err), zap.String("target", member.Config().Name))
 		return err

diff --git a/tests/robustness/linearizability_test.go b/tests/robustness/linearizability_test.go
@@ -104,14 +104,10 @@ func TestRobustness(t *testing.T) {
 	scenarios = append(scenarios, testScenario{
 		name:      "IssueXYZ",
 		failpoint: PutReturnErrNoSpace,
-		traffic:   traffic.KubernetesTraffic,
+		traffic:   traffic.LowTraffic,
 		clusterOpts: []e2e.EPClusterOption{
 			e2e.WithIsPeerTLS(true),
-			e2e.WithSnapshotCount(100),
-			e2e.WithPeerProxy(true),
 			e2e.WithGoFailEnabled(true),
-			e2e.WithWatchProcessNotifyInterval(100 * time.Millisecond),
-			e2e.WithLogLevel("debug"),
 		},
 	})
 	if v.Compare(version.V3_5) >= 0 {