Feat: [WIP] Add Graphite metrics provider to address issue #1403

.github/workflows/release.yaml

@@ -17,6 +17,10 @@ jobs:
         with:
           ref: ${{ github.event.inputs.tag }}
 
+      - name: Get SHA
+        id: get-sha
+        run: echo "::set-output name=sha::$(git log -1 --format='%H')"
+
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v1
 
@@ -27,9 +31,7 @@ jobs:
         uses: actions/cache@v2
         with:
           path: /tmp/.buildx-cache
-          key: ${{ runner.os }}-buildx-${{ github.sha }}
-          restore-keys: |
-            ${{ runner.os }}-buildx-
+          key: ${{ runner.os }}-buildx-${{ steps.get-sha.outputs.sha }}
 
       - name: Print Disk Usage
         run: |
@@ -45,6 +47,8 @@ jobs:
             ghcr.io/argoproj/argo-rollouts
           tags: |
             type=semver,pattern={{version}},prefix=v,value=${{ github.event.inputs.tag }}
+          flavor: |
+            latest=false
 
       - name: Docker meta (plugin)
         id: plugin-meta
@@ -55,6 +59,8 @@ jobs:
             ghcr.io/argoproj/kubectl-argo-rollouts
           tags: |
             type=semver,pattern={{version}},prefix=v,value=${{ github.event.inputs.tag }}
+          flavor: |
+            latest=false
 
       - name: Login to GitHub Container Registry
         if: github.event_name != 'pull_request'
@@ -75,6 +81,7 @@ jobs:
       - name: Build and push (controller-image)
         uses: docker/build-push-action@v2
         with:
+          context: .
           platforms: linux/amd64,linux/arm64
           push: true
           tags: ${{ steps.controller-meta.outputs.tags }}
@@ -84,6 +91,7 @@ jobs:
       - name: Build and push (plugin-image)
         uses: docker/build-push-action@v2
         with:
+          context: .
           target: kubectl-argo-rollouts
           platforms: linux/amd64,linux/arm64
           push: true

cmd/rollouts-controller/main.go

@@ -99,7 +99,7 @@ func newCommand() *cobra.Command {
 
 			kubeClient, err := kubernetes.NewForConfig(config)
 			checkError(err)
-			rolloutClient, err := clientset.NewForConfig(config)
+			argoprojClient, err := clientset.NewForConfig(config)
 			checkError(err)
 			dynamicClient, err := dynamic.NewForConfig(config)
 			checkError(err)
@@ -150,7 +150,7 @@ func newCommand() *cobra.Command {
 			cm := controller.NewManager(
 				namespace,
 				kubeClient,
-				rolloutClient,
+				argoprojClient,
 				dynamicClient,
 				smiClient,
 				discoveryClient,

controller/controller.go

@@ -5,8 +5,8 @@ import (
 	"fmt"
 	"time"
 
-	"github.com/argoproj/notifications-engine/pkg/api"
-	"github.com/argoproj/notifications-engine/pkg/controller"
+	notificationapi "github.com/argoproj/notifications-engine/pkg/api"
+	notificationcontroller "github.com/argoproj/notifications-engine/pkg/controller"
 	"github.com/pkg/errors"
 	smiclientset "github.com/servicemeshinterface/smi-sdk-go/pkg/gen/client/split/clientset/versioned"
 	log "github.com/sirupsen/logrus"
@@ -72,7 +72,7 @@ type Manager struct {
 	analysisController      *analysis.Controller
 	serviceController       *service.Controller
 	ingressController       *ingress.Controller
-	notificationsController controller.NotificationController
+	notificationsController notificationcontroller.NotificationController
 
 	rolloutSynced                 cache.InformerSynced
 	experimentSynced              cache.InformerSynced
@@ -148,10 +148,10 @@ func NewManager(
 	ingressWorkqueue := workqueue.NewNamedRateLimitingQueue(queue.DefaultArgoRolloutsRateLimiter(), "Ingresses")
 
 	refResolver := rollout.NewInformerBasedWorkloadRefResolver(namespace, dynamicclientset, discoveryClient, argoprojclientset, rolloutsInformer.Informer())
-	apiFactory := api.NewFactory(record.NewAPIFactorySettings(), defaults.Namespace(), secretInformer.Informer(), configMapInformer.Informer())
+	apiFactory := notificationapi.NewFactory(record.NewAPIFactorySettings(), defaults.Namespace(), secretInformer.Informer(), configMapInformer.Informer())
 	recorder := record.NewEventRecorder(kubeclientset, metrics.MetricRolloutEventsTotal, apiFactory)
-	notificationsController := controller.NewController(dynamicclientset.Resource(v1alpha1.RolloutGVR), rolloutsInformer.Informer(), apiFactory,
-		controller.WithToUnstructured(func(obj metav1.Object) (*unstructured.Unstructured, error) {
+	notificationsController := notificationcontroller.NewController(dynamicclientset.Resource(v1alpha1.RolloutGVR), rolloutsInformer.Informer(), apiFactory,
+		notificationcontroller.WithToUnstructured(func(obj metav1.Object) (*unstructured.Unstructured, error) {
 			data, err := json.Marshal(obj)
 			if err != nil {
 				return nil, err

docs/features/bluegreen.md

@@ -68,6 +68,25 @@ spec:
       scaleDownDelayRevisionLimit: *int32
 ```
 
+## Sequence of Events
+
+The following describes the sequence of events that happen during a blue-green update.
+
+1. Beginning at a fully promoted, steady-state, a revision 1 ReplicaSet is pointed to by both the `activeService` and `previewService`.
+1. A user initiates an update by modifying the pod template (`spec.template.spec`).
+1. The revision 2 ReplicaSet is created with size 0.
+1. The preview service is modified to point to the revision 2 ReplicaSet. The `activeService` remains pointing to revision 1.
+1. The revision 2 ReplicaSet is scaled to either `spec.replicas` or `previewReplicaCount` if set.
+1. Once revision 2 ReplicaSet Pods are fully available, `prePromotionAnalysis` begins.
+1. Upon success of `prePromotionAnalysis`, the blue/green pauses if `autoPromotionEnabled` is false, or `autoPromotionSeconds` is non-zero.
+1. The rollout is resumed either manually by a user, or automatically by surpassing `autoPromotionSeconds`.
+1. The revision 2 ReplicaSet is scaled to the `spec.replicas`, if the `previewReplicaCount` feature was used.
+1. The rollout "promotes" the revision 2 ReplicaSet by updating the `activeService` to point to it. At this point, there are no services pointing to revision 1
+1. `postPromotionAnalysis` analysis begins
+1. Once `postPromotionAnalysis` completes successfully, the update is successful and the revision 2 ReplicaSet is marked as stable. The rollout is considered fully-promoted.
+1. After waiting `scaleDownDelaySeconds` (default 30 seconds), the revision 1 ReplicaSet is scaled down 
+
+
 ### autoPromotionEnabled
 The AutoPromotionEnabled will make the rollout automatically promote the new ReplicaSet to the active service once the new ReplicaSet is healthy. This field is defaulted to true if it is not specified.
 
@@ -111,15 +130,6 @@ This feature is used to provide an endpoint that can be used to test a new versi
 
 Defaults to an empty string
 
-Here is a timeline of how the active and preview services work (if you use a preview service):
-
-1. During the Initial deployment there is only one ReplicaSet. Both active and preview services point to it. This is the **old** version of the application.
-1. A change happens in the Rollout resource. A new ReplicaSet is created. This is the **new** version of the application. The preview service is modified to point to the new ReplicaSet. The active service still points to the old version.
-1. The blue/green deployment is "promoted". Both active and preview services are pointing to the new version. The old version is still there but no service is pointing at it.
-1. Once the the blue/green deployment is scaled down (see the `scaleDownDelaySeconds` field) the old ReplicaSet is has 0 replicas and we are back to the initial state. Both active and preview services point to the new version (which is the only one present anyway)
-
-
-
 ### previewReplicaCount
 The PreviewReplicaCount field will indicate the number of replicas that the new version of an application should run.  Once the application is ready to promote to the active service, the controller will scale the new ReplicaSet to the value of the `spec.replicas`. The rollout will not switch over the active service to the new ReplicaSet until it matches the `spec.replicas` count.
 
@@ -136,3 +146,4 @@ Defaults to 30
 The ScaleDownDelayRevisionLimit limits the number of old active ReplicaSets to keep scaled up while they wait for the scaleDownDelay to pass after being removed from the active service. 
 
 If omitted, all ReplicaSets will be retained for the specified scaleDownDelay
+

docs/features/specification.md

@@ -11,6 +11,14 @@ spec:
   # Number of desired pods.
   # Defaults to 1.
   replicas: 5
+  analysis:
+    # limits the number of successful analysis runs and experiments to be stored in a history
+    # Defaults to 5.
+    successfulRunHistoryLimit: 10
+    # limits the number of unsuccessful analysis runs and experiments to be stored in a history. 
+    # Stages for unsuccessful: "Error", "Failed", "Inconclusive"
+    # Defaults to 5.
+    unsuccessfulRunHistoryLimit: 10
 
   # Label selector for pods. Existing ReplicaSets whose pods are selected by
   # this will be the ones affected by this rollout. It must match the pod

docs/installation.md

@@ -9,6 +9,9 @@ kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/rele
 
 This will create a new namespace, `argo-rollouts`, where Argo Rollouts controller will run.
 
+!!! tip
+    If you are using another namspace name, please update `install.yall` clusterrolebinding's serviceaccount namespace name.
+
 !!! tip
     When installing Argo Rollouts on Kubernetes v1.14 or lower, the CRD manifests must be kubectl applied with the --validate=false option. This is caused by use of new CRD fields introduced in v1.15, which are rejected by default in lower API servers.
 

examples/analysis-templates.yaml

@@ -24,7 +24,7 @@ spec:
                 args: [exit 0]
               restartPolicy: Never
           backoffLimit: 0
-
+    count: 1
 ---
 # This AnalysisTemplate will run a Kubernetes Job every 5 seconds, with a 50% chance of failure.
 # When the number of accumulated failures exceeds failureLimit, it will cause the analysis run to
@@ -36,6 +36,7 @@ metadata:
 spec:
   metrics:
   - name: random-fail
+    count: 2
     interval: 5s
     failureLimit: 1
     provider:

manifests/crds/rollout-crd.yaml

@@ -47,6 +47,15 @@ spec:
             type: object
           spec:
             properties:
+              analysis:
+                properties:
+                  successfulRunHistoryLimit:
+                    format: int32
+                    type: integer
+                  unsuccessfulRunHistoryLimit:
+                    format: int32
+                    type: integer
+                type: object
               minReadySeconds:
                 format: int32
                 type: integer

manifests/install.yaml

@@ -9745,6 +9745,15 @@ spec:
             type: object
           spec:
             properties:
+              analysis:
+                properties:
+                  successfulRunHistoryLimit:
+                    format: int32
+                    type: integer
+                  unsuccessfulRunHistoryLimit:
+                    format: int32
+                    type: integer
+                type: object
               minReadySeconds:
                 format: int32
                 type: integer

manifests/namespace-install.yaml

@@ -9745,6 +9745,15 @@ spec:
             type: object
           spec:
             properties:
+              analysis:
+                properties:
+                  successfulRunHistoryLimit:
+                    format: int32
+                    type: integer
+                  unsuccessfulRunHistoryLimit:
+                    format: int32
+                    type: integer
+                type: object
               minReadySeconds:
                 format: int32
                 type: integer