Support Argo Rollout controller object as a first class citizen

Makefile

@@ -2,7 +2,7 @@
 
 NAME ?= adobe/k8s-shredder
 K8S_SHREDDER_VERSION ?= "dev"
-KINDNODE_VERSION ?= "v1.28.0"
+KINDNODE_VERSION ?= "v1.30.4"
 COMMIT ?= $(shell git rev-parse --short HEAD)
 TEST_CLUSTERNAME ?= "k8s-shredder-test-cluster"
 

charts/k8s-shredder/Chart.yaml

@@ -12,5 +12,5 @@ maintainers:
   email: [email protected]
   url: https://adobe.com
 
-version: 0.1.0
+version: 0.1.1
 appVersion: v0.2.1

charts/k8s-shredder/templates/cluster-role.yaml

@@ -15,4 +15,7 @@ rules:
 - apiGroups: [apps, extensions]
   resources: [statefulsets, deployments, replicasets]
   verbs: [get, list, watch, update, patch]
+- apiGroups: [ "argoproj.io" ]
+  resources: [ rollouts ]
+  verbs: [ get, list, watch, update, patch ]
 {{ end }}

cmd/root.go

@@ -114,6 +114,7 @@ func discoverConfig() {
 	viper.SetDefault("RestartedAtAnnotation", "shredder.ethos.adobe.net/restartedAt")
 	viper.SetDefault("AllowEvictionLabel", "shredder.ethos.adobe.net/allow-eviction")
 	viper.SetDefault("ToBeDeletedTaint", "ToBeDeletedByClusterAutoscaler")
+	viper.SetDefault("ArgoRolloutsAPIVersion", "v1alpha1")
 
 	err := viper.ReadInConfig()
 	if err != nil {
@@ -144,6 +145,7 @@ func parseConfig() {
 		"RestartedAtAnnotation":              cfg.RestartedAtAnnotation,
 		"AllowEvictionLabel":                 cfg.AllowEvictionLabel,
 		"ToBeDeletedTaint":                   cfg.ToBeDeletedTaint,
+		"ArgoRolloutsAPIVersion":             cfg.ArgoRolloutsAPIVersion,
 	}).Info("Loaded configuration")
 }
 

internal/testing/e2e_test.go

@@ -23,6 +23,8 @@ import (
 	log "github.com/sirupsen/logrus"
 	"golang.org/x/exp/slices"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
+	"k8s.io/apimachinery/pkg/runtime/schema"
 	"os"
 	"strings"
 	"testing"
@@ -47,6 +49,56 @@ var (
 	}
 )
 
+func grabMetrics(shredderMetrics []string, t *testing.T) []string {
+	results := make([]string, 0)
+	warnings := make([]string, 0)
+
+	for _, shredderMetric := range shredderMetrics {
+		result, warning, err := prometheusQuery(shredderMetric)
+		if err != nil {
+			t.Errorf("Error querying Prometheus: %v\n", err)
+		}
+		warnings = append(warnings, warning...)
+		results = append(results, result.String())
+	}
+
+	if len(warnings) > 0 {
+		t.Logf("Warnings: %v\n", strings.Join(warnings, "\n"))
+	}
+
+	t.Logf("Results: \n%v\n", strings.Join(results, "\n"))
+
+	return results
+}
+
+func prometheusQuery(query string) (model.Value, v1.Warnings, error) {
+
+	client, err := api.NewClient(api.Config{
+		Address: "http://localhost:30007",
+	})
+	if err != nil {
+		fmt.Printf("Error creating client: %v\n", err)
+		os.Exit(1)
+	}
+
+	v1api := v1.NewAPI(client)
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+	return v1api.Query(ctx, query, time.Now(), v1.WithTimeout(5*time.Second))
+}
+
+func compareTime(expirationTime time.Time, t *testing.T, ch chan time.Time) {
+	currentTime := time.Now().UTC()
+
+	for !currentTime.After(expirationTime.UTC()) {
+		t.Logf("Node TTL haven't expired yet: current time(UTC): %s, expire time(UTC): %s", currentTime, expirationTime.UTC())
+		time.Sleep(10 * time.Second)
+		currentTime = time.Now().UTC()
+
+	}
+	ch <- currentTime
+}
+
 // Validates that k8s-shredder cleanup a parked node after its TTL expires
 func TestNodeIsCleanedUp(t *testing.T) {
 	var err error
@@ -110,18 +162,7 @@ func TestNodeIsCleanedUp(t *testing.T) {
 	}
 }
 
-func compareTime(expirationTime time.Time, t *testing.T, ch chan time.Time) {
-	currentTime := time.Now().UTC()
-
-	for !currentTime.After(expirationTime.UTC()) {
-		t.Logf("Node TTL haven't expired yet: current time(UTC): %s, expire time(UTC): %s", currentTime, expirationTime.UTC())
-		time.Sleep(10 * time.Second)
-		currentTime = time.Now().UTC()
-
-	}
-	ch <- currentTime
-}
-
+// Validates shredder metrics
 func TestShredderMetrics(t *testing.T) {
 
 	// Intentionally skipped the gauge metrics as they are going to be wiped out before every eviction loop
@@ -140,40 +181,43 @@ func TestShredderMetrics(t *testing.T) {
 	}
 }
 
-func grabMetrics(shredderMetrics []string, t *testing.T) []string {
-	results := make([]string, 0)
-	warnings := make([]string, 0)
+func TestArgoRolloutRestartAt(t *testing.T) {
+	var err error
 
-	for _, shredderMetric := range shredderMetrics {
-		result, warning, err := prometheusQuery(shredderMetric)
-		if err != nil {
-			t.Errorf("Error querying Prometheus: %v\n", err)
-		}
-		warnings = append(warnings, warning...)
-		results = append(results, result.String())
-	}
+	appContext, err := utils.NewAppContext(config.Config{
+		ParkedNodeTTL:                      30 * time.Second,
+		EvictionLoopInterval:               10 * time.Second,
+		RollingRestartThreshold:            0.1,
+		UpgradeStatusLabel:                 "shredder.ethos.adobe.net/upgrade-status",
+		ExpiresOnLabel:                     "shredder.ethos.adobe.net/parked-node-expires-on",
+		NamespacePrefixSkipInitialEviction: "",
+		RestartedAtAnnotation:              "shredder.ethos.adobe.net/restartedAt",
+		AllowEvictionLabel:                 "shredder.ethos.adobe.net/allow-eviction",
+		ToBeDeletedTaint:                   "ToBeDeletedByClusterAutoscaler",
+		ArgoRolloutsAPIVersion:             "v1alpha1",
+	}, false)
 
-	if len(warnings) > 0 {
-		t.Logf("Warnings: %v\n", strings.Join(warnings, "\n"))
+	if err != nil {
+		log.Fatalf("Failed to setup application context: %s", err)
 	}
 
-	t.Logf("Results: \n%v\n", strings.Join(results, "\n"))
-
-	return results
-}
+	gvr := schema.GroupVersionResource{
+		Group:    "argoproj.io",
+		Version:  appContext.Config.ArgoRolloutsAPIVersion,
+		Resource: "rollouts",
+	}
 
-func prometheusQuery(query string) (model.Value, v1.Warnings, error) {
+	rollout, err := appContext.DynamicK8SClient.Resource(gvr).Namespace("ns-team-k8s-shredder-test").Get(appContext.Context, "test-app-argo-rollout", metav1.GetOptions{})
+	if err != nil {
+		log.Fatalf("Failed to get the Argo Rollout object: %s", err)
+	}
+	_, found, err := unstructured.NestedString(rollout.Object, "spec", "restartAt")
 
-	client, err := api.NewClient(api.Config{
-		Address: "http://localhost:30007",
-	})
 	if err != nil {
-		fmt.Printf("Error creating client: %v\n", err)
-		os.Exit(1)
+		log.Fatalf("Failed to get the Argo Rollout spec.restartAt field: %s", err)
 	}
 
-	v1api := v1.NewAPI(client)
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-	defer cancel()
-	return v1api.Query(ctx, query, time.Now(), v1.WithTimeout(5*time.Second))
+	if !found {
+		t.Fatalf("Argo Rollout object does not have the spec.restartAt field set")
+	}
 }

internal/testing/local_env_prep.sh

@@ -41,9 +41,17 @@ kubectl apply -f "${test_dir}/k8s-shredder.yaml"
 echo "KIND: deploying prometheus..."
 kubectl apply -f "${test_dir}/prometheus_stuffs.yaml"
 
+echo "KIND: deploying Argo Rollouts CRD..."
+kubectl apply -f https://raw.githubusercontent.com/argoproj/argo-rollouts/v1.7.2/manifests/crds/rollout-crd.yaml
+
 echo "KIND: deploying test applications..."
 kubectl apply -f "${test_dir}/test_apps.yaml"
 
+# Adjust the correct UID for the test-app-argo-rollout ownerReference
+rollout_uid=$(kubectl -n ns-team-k8s-shredder-test get rollout test-app-argo-rollout -ojsonpath='{.metadata.uid}')
+cat "${test_dir}/test_apps.yaml" | sed "s/REPLACE_WITH_ROLLOUT_UID/${rollout_uid}/" | kubectl apply -f -
+
+
 echo "K8S_SHREDDER: waiting for k8s-shredder deployment to become ready!"
 retry_count=0
 i=1

internal/testing/prometheus_stuffs.yaml

@@ -18,7 +18,7 @@ spec:
     spec:
       containers:
         - name: prometheus
-          image: prom/prometheus:v2.42.0
+          image: prom/prometheus:v2.54.1
           args:
             - "--storage.tsdb.retention.time=1h"
             - "--config.file=/etc/prometheus/prometheus.yml"

internal/testing/rbac.yaml

@@ -31,4 +31,7 @@ rules:
   - apiGroups: [apps, extensions]
     resources: [statefulsets, deployments, replicasets]
     verbs: [get, list, watch, update, patch]
+  - apiGroups: [ "argoproj.io" ]
+    resources: [ rollouts ]
+    verbs: [ get, list, watch, update, patch ]
 

internal/testing/test_apps.yaml

@@ -239,4 +239,69 @@ spec:
   minAvailable: 1
   selector:
     matchLabels:
-      app: test-app-statefulset
+      app: test-app-statefulset
+#### FLEX ####
+# 1. Good citizen Argo Rollout in Flex world
+---
+apiVersion: apps/v1
+kind: ReplicaSet
+metadata:
+  name: test-app-argo-rollout
+  namespace: ns-team-k8s-shredder-test
+  ownerReferences:
+  - apiVersion: argoproj.io/v1alpha1
+    kind: Rollout
+    blockOwnerDeletion: true
+    name: test-app-argo-rollout
+    uid: REPLACE_WITH_ROLLOUT_UID
+spec:
+  replicas: 2
+  selector:
+    matchLabels:
+      app: test-app-argo-rollout
+  template:
+    metadata:
+      labels:
+        app: test-app-argo-rollout
+    spec:
+      affinity:
+        podAntiAffinity:
+          preferredDuringSchedulingIgnoredDuringExecution:
+          - weight: 100
+            podAffinityTerm:
+              labelSelector:
+                matchExpressions:
+                - key: app
+                  operator: In
+                  values:
+                  - test-app-argo-rollout
+              topologyKey: kubernetes.io/hostname
+      containers:
+      - name: test-app-argo-rollout
+        image: aaneci/canary
+        ports:
+        - containerPort: 8080
+          name: web
+---
+apiVersion: argoproj.io/v1alpha1
+kind: Rollout
+metadata:
+  name: test-app-argo-rollout
+  namespace: ns-team-k8s-shredder-test
+spec:
+  replicas: 2
+  workloadRef:
+    apiVersion: apps/v1
+    kind: ReplicaSet
+    name: test-app-argo-rollout
+---
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: test-app-argo-rollout
+  namespace: ns-team-k8s-shredder-test
+spec:
+  minAvailable: 10
+  selector:
+    matchLabels:
+      app: test-app-argo-rollout

pkg/config/config.go

@@ -33,4 +33,6 @@ type Config struct {
 	AllowEvictionLabel string
 	// ToBeDeletedTaint is used for skipping a subset of parked nodes
 	ToBeDeletedTaint string
+	// ArgoRolloutsAPIVersion is used for specifying the API version from `argoproj.io` apigroup to be used while handling Argo Rollouts objects
+	ArgoRolloutsAPIVersion string
 }