fix: limit number of resources in appset status (argoproj#24690)

Signed-off-by: Alexander Matyushentsev <[email protected]>

Author: alexmt

applicationset/controllers/applicationset_controller.go

@@ -100,6 +100,7 @@ type ApplicationSetReconciler struct {
 	GlobalPreservedAnnotations []string
 	GlobalPreservedLabels      []string
 	Metrics                    *metrics.ApplicationsetMetrics
+	MaxResourcesStatusCount    int
 }
 
 // +kubebuilder:rbac:groups=argoproj.io,resources=applicationsets,verbs=get;list;watch;create;update;patch;delete
@@ -1398,6 +1399,11 @@ func (r *ApplicationSetReconciler) updateResourcesStatus(ctx context.Context, lo
 	sort.Slice(statuses, func(i, j int) bool {
 		return statuses[i].Name < statuses[j].Name
 	})
+
+	if r.MaxResourcesStatusCount > 0 && len(statuses) > r.MaxResourcesStatusCount {
+		logCtx.Warnf("Truncating ApplicationSet %s resource status from %d to max allowed %d entries", appset.Name, len(statuses), r.MaxResourcesStatusCount)
+		statuses = statuses[:r.MaxResourcesStatusCount]
+	}
 	appset.Status.Resources = statuses
 	// DefaultRetry will retry 5 times with a backoff factor of 1, jitter of 0.1 and a duration of 10ms
 	err := retry.RetryOnConflict(retry.DefaultRetry, func() error {

applicationset/controllers/applicationset_controller_test.go

@@ -6410,10 +6410,11 @@ func TestUpdateResourceStatus(t *testing.T) {
 	require.NoError(t, err)
 
 	for _, cc := range []struct {
-		name              string
-		appSet            v1alpha1.ApplicationSet
-		apps              []v1alpha1.Application
-		expectedResources []v1alpha1.ResourceStatus
+		name                    string
+		appSet                  v1alpha1.ApplicationSet
+		apps                    []v1alpha1.Application
+		expectedResources       []v1alpha1.ResourceStatus
+		maxResourcesStatusCount int
 	}{
 		{
 			name: "handles an empty application list",
@@ -6577,6 +6578,73 @@ func TestUpdateResourceStatus(t *testing.T) {
 			apps:              []v1alpha1.Application{},
 			expectedResources: nil,
 		},
+		{
+			name: "truncates resources status list to",
+			appSet: v1alpha1.ApplicationSet{
+				ObjectMeta: metav1.ObjectMeta{
+					Name:      "name",
+					Namespace: "argocd",
+				},
+				Status: v1alpha1.ApplicationSetStatus{
+					Resources: []v1alpha1.ResourceStatus{
+						{
+							Name:   "app1",
+							Status: v1alpha1.SyncStatusCodeOutOfSync,
+							Health: &v1alpha1.HealthStatus{
+								Status:  health.HealthStatusProgressing,
+								Message: "this is progressing",
+							},
+						},
+						{
+							Name:   "app2",
+							Status: v1alpha1.SyncStatusCodeOutOfSync,
+							Health: &v1alpha1.HealthStatus{
+								Status:  health.HealthStatusProgressing,
+								Message: "this is progressing",
+							},
+						},
+					},
+				},
+			},
+			apps: []v1alpha1.Application{
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: "app1",
+					},
+					Status: v1alpha1.ApplicationStatus{
+						Sync: v1alpha1.SyncStatus{
+							Status: v1alpha1.SyncStatusCodeSynced,
+						},
+						Health: v1alpha1.AppHealthStatus{
+							Status: health.HealthStatusHealthy,
+						},
+					},
+				},
+				{
+					ObjectMeta: metav1.ObjectMeta{
+						Name: "app2",
+					},
+					Status: v1alpha1.ApplicationStatus{
+						Sync: v1alpha1.SyncStatus{
+							Status: v1alpha1.SyncStatusCodeSynced,
+						},
+						Health: v1alpha1.AppHealthStatus{
+							Status: health.HealthStatusHealthy,
+						},
+					},
+				},
+			},
+			expectedResources: []v1alpha1.ResourceStatus{
+				{
+					Name:   "app1",
+					Status: v1alpha1.SyncStatusCodeSynced,
+					Health: &v1alpha1.HealthStatus{
+						Status: health.HealthStatusHealthy,
+					},
+				},
+			},
+			maxResourcesStatusCount: 1,
+		},
 	} {
 		t.Run(cc.name, func(t *testing.T) {
 			kubeclientset := kubefake.NewSimpleClientset([]runtime.Object{}...)
@@ -6587,13 +6655,14 @@ func TestUpdateResourceStatus(t *testing.T) {
 			argodb := db.NewDB("argocd", settings.NewSettingsManager(t.Context(), kubeclientset, "argocd"), kubeclientset)
 
 			r := ApplicationSetReconciler{
-				Client:        client,
-				Scheme:        scheme,
-				Recorder:      record.NewFakeRecorder(1),
-				Generators:    map[string]generators.Generator{},
-				ArgoDB:        argodb,
-				KubeClientset: kubeclientset,
-				Metrics:       metrics,
+				Client:                  client,
+				Scheme:                  scheme,
+				Recorder:                record.NewFakeRecorder(1),
+				Generators:              map[string]generators.Generator{},
+				ArgoDB:                  argodb,
+				KubeClientset:           kubeclientset,
+				Metrics:                 metrics,
+				MaxResourcesStatusCount: cc.maxResourcesStatusCount,
 			}
 
 			err := r.updateResourcesStatus(t.Context(), log.NewEntry(log.StandardLogger()), &cc.appSet, cc.apps)

cmd/argocd-applicationset-controller/commands/applicationset_controller.go

@@ -79,6 +79,7 @@ func NewCommand() *cobra.Command {
 		enableScmProviders           bool
 		webhookParallelism           int
 		tokenRefStrictMode           bool
+		maxResourcesStatusCount      int
 	)
 	scheme := runtime.NewScheme()
 	_ = clientgoscheme.AddToScheme(scheme)
@@ -231,6 +232,7 @@ func NewCommand() *cobra.Command {
 				GlobalPreservedAnnotations: globalPreservedAnnotations,
 				GlobalPreservedLabels:      globalPreservedLabels,
 				Metrics:                    &metrics,
+				MaxResourcesStatusCount:    maxResourcesStatusCount,
 			}).SetupWithManager(mgr, enableProgressiveSyncs, maxConcurrentReconciliations); err != nil {
 				log.Error(err, "unable to create controller", "controller", "ApplicationSet")
 				os.Exit(1)
@@ -275,6 +277,7 @@ func NewCommand() *cobra.Command {
 	command.Flags().IntVar(&webhookParallelism, "webhook-parallelism-limit", env.ParseNumFromEnv("ARGOCD_APPLICATIONSET_CONTROLLER_WEBHOOK_PARALLELISM_LIMIT", 50, 1, 1000), "Number of webhook requests processed concurrently")
 	command.Flags().StringSliceVar(&metricsAplicationsetLabels, "metrics-applicationset-labels", []string{}, "List of Application labels that will be added to the argocd_applicationset_labels metric")
 	command.Flags().BoolVar(&enableGitHubAPIMetrics, "enable-github-api-metrics", env.ParseBoolFromEnv("ARGOCD_APPLICATIONSET_CONTROLLER_ENABLE_GITHUB_API_METRICS", false), "Enable GitHub API metrics for generators that use the GitHub API")
+	command.Flags().IntVar(&maxResourcesStatusCount, "max-resources-status-count", env.ParseNumFromEnv("ARGOCD_APPLICATIONSET_CONTROLLER_MAX_RESOURCES_STATUS_COUNT", 0, 0, math.MaxInt), "Max number of resources stored in appset status.")
 
 	return &command
 }

docs/operator-manual/argocd-cmd-params-cm.yaml

@@ -292,6 +292,8 @@ data:
   applicationsetcontroller.global.preserved.labels: "acme.com/label1,acme.com/label2"
   # Enable GitHub API metrics for generators that use GitHub API
   applicationsetcontroller.enable.github.api.metrics: "false"
+  # The maximum number of resources stored in the status of an ApplicationSet. This is a safeguard to prevent the status from growing too large.
+  applicationsetcontroller.status.max.resources.count: "5000"
 
   ## Argo CD Notifications Controller Properties
   # Set the logging level. One of: debug|info|warn|error (default "info")

docs/operator-manual/server-commands/argocd-applicationset-controller.md

@@ -187,6 +187,12 @@ spec:
                   name: argocd-cmd-params-cm
                   key: applicationsetcontroller.requeue.after
                   optional: true
+            - name: ARGOCD_APPLICATIONSET_CONTROLLER_MAX_RESOURCES_STATUS_COUNT
+              valueFrom:
+                configMapKeyRef:
+                  name: argocd-cmd-params-cm
+                  key: applicationsetcontroller.status.max.resources.count
+                  optional: true
           volumeMounts:
             - mountPath: /app/config/ssh
               name: ssh-known-hosts