feature(serving): Add --autoscale-window (#614)

* feature(serving): Add --autoscale-window

Fixes #613

* chore: Update help-text for --autoscale-window

* chore: Add missing generated files

docs/cmd/kn_service_create.md

@@ -44,6 +44,7 @@ kn service create NAME --image IMAGE [flags]
 ```
       --annotation stringArray    Service annotation to set. name=value; you may provide this flag any number of times to set multiple annotations. To unset, specify the annotation name followed by a "-" (e.g., name-).
       --async                     Create service and don't wait for it to become ready.
+      --autoscale-window string   Duration to look back for making auto-scaling decisions. The service is scaled to zero if no request was received in during that time. (eg: 10s)
       --concurrency-limit int     Hard Limit of concurrent requests to be processed by a single replica.
       --concurrency-target int    Recommendation for when to scale up based on the concurrent number of incoming request. Defaults to --concurrency-limit when given.
   -e, --env stringArray           Environment variable to set. NAME=value; you may provide this flag any number of times to set multiple environment variables. To unset, specify the environment variable name followed by a "-" (e.g., NAME-).

docs/cmd/kn_service_update.md

@@ -40,6 +40,7 @@ kn service update NAME [flags]
 ```
       --annotation stringArray    Service annotation to set. name=value; you may provide this flag any number of times to set multiple annotations. To unset, specify the annotation name followed by a "-" (e.g., name-).
       --async                     Update service and don't wait for it to become ready.
+      --autoscale-window string   Duration to look back for making auto-scaling decisions. The service is scaled to zero if no request was received in during that time. (eg: 10s)
       --concurrency-limit int     Hard Limit of concurrent requests to be processed by a single replica.
       --concurrency-target int    Recommendation for when to scale up based on the concurrent number of incoming request. Defaults to --concurrency-limit when given.
   -e, --env stringArray           Environment variable to set. NAME=value; you may provide this flag any number of times to set multiple environment variables. To unset, specify the environment variable name followed by a "-" (e.g., NAME-).

pkg/kn/commands/revision/describe.go

@@ -129,7 +129,8 @@ func describe(w io.Writer, revision *v1alpha1.Revision, service *v1alpha1.Servic
 func WriteConcurrencyOptions(dw printers.PrefixWriter, revision *v1alpha1.Revision) {
 	target := clientserving.ConcurrencyTarget(&revision.ObjectMeta)
 	limit := revision.Spec.ContainerConcurrency
-	if target != nil || limit != nil && *limit != 0 {
+	autoscaleWindow := clientserving.AutoscaleWindow(&revision.ObjectMeta)
+	if target != nil || limit != nil && *limit != 0 || autoscaleWindow != "" {
 		section := dw.WriteAttribute("Concurrency", "")
 		if limit != nil && *limit != 0 {
 			section.WriteAttribute("Limit", strconv.FormatInt(int64(*limit), 10))
@@ -137,7 +138,11 @@ func WriteConcurrencyOptions(dw printers.PrefixWriter, revision *v1alpha1.Revisi
 		if target != nil {
 			section.WriteAttribute("Target", strconv.Itoa(*target))
 		}
+		if autoscaleWindow != "" {
+			section.WriteAttribute("Window", autoscaleWindow)
 		}
+	}
+
 }
 
 // Write the image attribute (with

pkg/kn/commands/service/configuration_edit_flags.go

@@ -41,6 +41,7 @@ type ConfigurationEditFlags struct {
 	MaxScale                   int
 	ConcurrencyTarget          int
 	ConcurrencyLimit           int
+	AutoscaleWindow            string
 	Port                       int32
 	Labels                     []string
 	NamePrefix                 string
@@ -113,6 +114,8 @@ func (p *ConfigurationEditFlags) addSharedFlags(command *cobra.Command) {
 	p.markFlagMakesRevision("min-scale")
 	command.Flags().IntVar(&p.MaxScale, "max-scale", 0, "Maximal number of replicas.")
 	p.markFlagMakesRevision("max-scale")
+	command.Flags().StringVar(&p.AutoscaleWindow, "autoscale-window", "", "Duration to look back for making auto-scaling decisions. The service is scaled to zero if no request was received in during that time. (eg: 10s)")
+	p.markFlagMakesRevision("autoscale-window")
 	command.Flags().IntVar(&p.ConcurrencyTarget, "concurrency-target", 0,
 		"Recommendation for when to scale up based on the concurrent number of incoming request. "+
 			"Defaults to --concurrency-limit when given.")
@@ -290,6 +293,13 @@ func (p *ConfigurationEditFlags) Apply(
 		}
 	}
 
+	if cmd.Flags().Changed("autoscale-window") {
+		err = servinglib.UpdateAutoscaleWindow(template, p.AutoscaleWindow)
+		if err != nil {
+			return err
+		}
+	}
+
 	if cmd.Flags().Changed("concurrency-target") {
 		err = servinglib.UpdateConcurrencyTarget(template, p.ConcurrencyTarget)
 		if err != nil {

pkg/serving/config_changes.go

@@ -21,6 +21,7 @@ import (
 	"sort"
 	"strconv"
 	"strings"
+	"time"
 	"unicode"
 
 	corev1 "k8s.io/api/core/v1"
@@ -183,6 +184,15 @@ func UpdateMaxScale(template *servingv1alpha1.RevisionTemplateSpec, max int) err
 	return UpdateRevisionTemplateAnnotation(template, autoscaling.MaxScaleAnnotationKey, strconv.Itoa(max))
 }
 
+// UpdateAutoscaleWindow updates the autoscale window annotation
+func UpdateAutoscaleWindow(template *servingv1alpha1.RevisionTemplateSpec, window string) error {
+	_, err := time.ParseDuration(window)
+	if err != nil {
+		return fmt.Errorf("invalid duration for 'autoscale-window': %v", err)
+	}
+	return UpdateRevisionTemplateAnnotation(template, autoscaling.WindowAnnotationKey, window)
+}
+
 // UpdateConcurrencyTarget updates container concurrency annotation
 func UpdateConcurrencyTarget(template *servingv1alpha1.RevisionTemplateSpec, target int) error {
 	return UpdateRevisionTemplateAnnotation(template, autoscaling.TargetAnnotationKey, strconv.Itoa(target))

pkg/serving/config_changes_test.go

@@ -225,7 +225,7 @@ func TestUpdateMinScale(t *testing.T) {
 	err := UpdateMinScale(template, 10)
 	assert.NilError(t, err)
 	// Verify update is successful or not
-	checkAnnotationValue(t, template, autoscaling.MinScaleAnnotationKey, 10)
+	checkAnnotationValueInt(t, template, autoscaling.MinScaleAnnotationKey, 10)
 	// Update with invalid value
 	err = UpdateMinScale(template, -1)
 	assert.ErrorContains(t, err, "minScale")
@@ -236,18 +236,29 @@ func TestUpdateMaxScale(t *testing.T) {
 	err := UpdateMaxScale(template, 10)
 	assert.NilError(t, err)
 	// Verify update is successful or not
-	checkAnnotationValue(t, template, autoscaling.MaxScaleAnnotationKey, 10)
+	checkAnnotationValueInt(t, template, autoscaling.MaxScaleAnnotationKey, 10)
 	// Update with invalid value
 	err = UpdateMaxScale(template, -1)
 	assert.ErrorContains(t, err, "maxScale")
 }
 
+func TestAutoscaleWindow(t *testing.T) {
+	template, _ := getV1alpha1RevisionTemplateWithOldFields()
+	err := UpdateAutoscaleWindow(template, "10s")
+	assert.NilError(t, err)
+	// Verify update is successful or not
+	checkAnnotationValue(t, template, autoscaling.WindowAnnotationKey, "10s")
+	// Update with invalid value
+	err = UpdateAutoscaleWindow(template, "blub")
+	assert.Check(t, util.ContainsAll(err.Error(), "invalid duration", "autoscale-window"))
+}
+
 func TestUpdateConcurrencyTarget(t *testing.T) {
 	template, _ := getV1alpha1RevisionTemplateWithOldFields()
 	err := UpdateConcurrencyTarget(template, 10)
 	assert.NilError(t, err)
 	// Verify update is successful or not
-	checkAnnotationValue(t, template, autoscaling.TargetAnnotationKey, 10)
+	checkAnnotationValueInt(t, template, autoscaling.TargetAnnotationKey, 10)
 	// Update with invalid value
 	err = UpdateConcurrencyTarget(template, -1)
 	assert.ErrorContains(t, err, "invalid")
@@ -708,13 +719,20 @@ func assertNoV1alpha1(t *testing.T, template *servingv1alpha1.RevisionTemplateSp
 	}
 }
 
-func checkAnnotationValue(t *testing.T, template *servingv1alpha1.RevisionTemplateSpec, key string, value int) {
+func checkAnnotationValueInt(t *testing.T, template *servingv1alpha1.RevisionTemplateSpec, key string, value int) {
 	anno := template.GetAnnotations()
 	if v, ok := anno[key]; !ok && v != strconv.Itoa(value) {
 		t.Errorf("Failed to update %s annotation key: got=%s, want=%d", key, v, value)
 	}
 }
 
+func checkAnnotationValue(t *testing.T, template *servingv1alpha1.RevisionTemplateSpec, key string, value string) {
+	anno := template.GetAnnotations()
+	if v, ok := anno[key]; !ok && v != value {
+		t.Errorf("Failed to update %s annotation key: got=%s, want=%s", key, v, value)
+	}
+}
+
 func checkContainerConcurrency(t *testing.T, template *servingv1alpha1.RevisionTemplateSpec, value *int64) {
 	if got, want := *template.Spec.ContainerConcurrency, *value; got != want {
 		t.Errorf("Failed to update containerConcurrency value: got=%d, want=%d", got, want)

pkg/serving/revision_template.go

@@ -67,6 +67,10 @@ func ConcurrencyTarget(m *metav1.ObjectMeta) *int {
 	return ret
 }
 
+func AutoscaleWindow(m *metav1.ObjectMeta) string {
+	return m.Annotations[autoscaling.WindowAnnotationKey]
+}
+
 func Port(revisionSpec *servingv1alpha1.RevisionSpec) *int32 {
 	c, err := ContainerOfRevisionSpec(revisionSpec)
 	if err != nil {

test/e2e/service_options_test.go

@@ -79,6 +79,14 @@ func TestServiceOptions(t *testing.T) {
 		test.validateServiceAnnotations(t, "svc3", map[string]string{"alpha": "direwolf", "brave": ""})
 		test.serviceDelete(t, "svc3")
 	})
+
+	t.Run("create, update and validate service with autoscale window option", func(t *testing.T) {
+		test.serviceCreateWithOptions(t, "svc4", []string{"--autoscale-window", "1m"})
+		test.validateAutoscaleWindow(t, "svc4", "1m")
+		test.serviceUpdate(t, "svc4", []string{"--autoscale-window", "15s"})
+		test.validateAutoscaleWindow(t, "svc4", "15s")
+		test.serviceDelete(t, "svc4")
+	})
 }
 
 func (test *e2eTest) serviceCreateWithOptions(t *testing.T, serviceName string, options []string) {
@@ -121,6 +129,13 @@ func (test *e2eTest) validateServiceConcurrencyTarget(t *testing.T, serviceName,
 	}
 }
 
+func (test *e2eTest) validateAutoscaleWindow(t *testing.T, serviceName, window string) {
+	jsonpath := "jsonpath={.items[0].spec.template.metadata.annotations.autoscaling\\.knative\\.dev/window}"
+	out, err := test.kn.RunWithOpts([]string{"service", "list", serviceName, "-o", jsonpath}, runOpts{})
+	assert.NilError(t, err)
+	assert.Equal(t, out, window)
+}
+
 func (test *e2eTest) validateServiceMinScale(t *testing.T, serviceName, minScale string) {
 	jsonpath := "jsonpath={.items[0].spec.template.metadata.annotations.autoscaling\\.knative\\.dev/minScale}"
 	out, err := test.kn.RunWithOpts([]string{"service", "list", serviceName, "-o", jsonpath}, runOpts{})