0.1.0 windup

README.md

@@ -86,28 +86,31 @@ Crane is composed of the following components:
 #### Installing prometheus components with helm chart
 
 > Note:
-> If you already deployed prometheus, kube-state-metric, prometheus-node-exporter, then you can skip this step.
+> If you already deployed prometheus, prometheus-node-exporter, then you can skip this step.
 
-Crane use prometheus to be the default metric provider. Using following command to install prometheus with the release name `[RELEASE_NAME]`. It will also install dependent chart: kube-state-metric.
+Export the following env if you want to use default settings, or specify customized value if you want to customize the installation.
 
 ```console
-helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
-helm repo update
-helm install [RELEASE_NAME] -n [NAMESPACE] --create-namespace  prometheus-community/prometheus
+export NAMESPACE=monitoring
+export RELEASE_NAME=myprometheus
 ```
 
-Using following command to install prometheus-node-exporter with another release name `[NODE_EXPORTER_RELEASE_NAME]`.
+Crane use prometheus to be the default metric provider. Using following command to install prometheus components.
 
 ```console
-helm install [NODE_EXPORTER_RELEASE_NAME] -n [NAMESPACE] --create-namespace  prometheus-community/prometheus-node-exporter
+helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
+helm repo update
+
+helm install $RELEASE_NAME -n $NAMESPACE --set kubeStateMetrics.enabled=false --set pushgateway.enabled=false --set alertmanager.enabled=false --set server.persistentVolume.enabled=false --create-namespace  prometheus-community/prometheus
 ```
 
 #### Configure Prometheus Address
 
-The following command will configure your prometheus http address for crane, please change `YOUR_PROMETHEUS` to actual Prometheus address. If you're following above guide to install prometheus with helm chart then **YOUR_PROMETHEUS** would be: `http:\/\/[RELEASE_NAME]-prometheus-server.[NAMESPACE].svc.cluster.local`
+The following command will configure prometheus http address for crane. Specify `CUSTOMIZE_PROMETHEUS` if you have existing prometheus server.
 
 ```console
-PROMETHEUS_ADDRESS="YOUR_PROMETHEUS" && sed -i '' "s/PROMETHEUS_ADDRESS/${YOUR_ADDRESS}/" deploy/craned/deployment.yaml
+export CUSTOMIZE_PROMETHEUS=
+if [ ! $CUSTOMIZE_PROMETHEUS ]; then sed -i '' "s/PROMETHEUS_ADDRESS/http:\/\/${RELEASE_NAME}-server.${NAMESPACE}.svc.cluster.local/" deploy/craned/deployment.yaml ; else sed -i '' "s/PROMETHEUS_ADDRESS/${CUSTOMIZE_PROMETHEUS}/" deploy/craned/deployment.yaml ; fi
 ```
 
 #### Deploying Crane
@@ -119,11 +122,3 @@ kubectl apply -f deploy/manifests
 kubectl apply -f deploy/craned 
 kubectl apply -f deploy/metric-adapter
 ```
-
-#### Deploying Crane-agent
-
-If you want to try `QoS Ensurance`, then deploy `Crane-agent` by apply YAML declaration which will create a DaemonSet in your cluster.
-
-```console
-kubectl apply -f deploy/crane-agent 
-```

cmd/craned/app/manager.go

@@ -3,7 +3,6 @@ package app
 import (
 	"context"
 	"flag"
-	"fmt"
 	"os"
 	"strings"
 
@@ -33,7 +32,6 @@ import (
 	"github.com/gocrane/crane/pkg/controller/timeseriesprediction"
 	"github.com/gocrane/crane/pkg/features"
 	"github.com/gocrane/crane/pkg/known"
-	"github.com/gocrane/crane/pkg/log"
 	"github.com/gocrane/crane/pkg/metrics"
 	"github.com/gocrane/crane/pkg/prediction"
 	"github.com/gocrane/crane/pkg/prediction/dsp"
@@ -105,7 +103,7 @@ func Run(ctx context.Context, opts *options.Options) error {
 	}
 
 	if opts.WebhookConfig.Enabled {
-		initializationWebhooks(mgr, opts)
+		initializationWebhooks(mgr)
 	}
 	initializationControllers(ctx, mgr, opts)
 	klog.Info("Starting crane manager")
@@ -126,9 +124,7 @@ func initializationMetricCollector(mgr ctrl.Manager) {
 	metrics.CustomCollectorRegister(metrics.NewTspMetricCollector(mgr.GetClient()))
 }
 
-func initializationWebhooks(mgr ctrl.Manager, opts *options.Options) {
-	log.Logger().Info(fmt.Sprintf("opts %v", opts))
-
+func initializationWebhooks(mgr ctrl.Manager) {
 	if certDir := os.Getenv("WEBHOOK_CERT_DIR"); len(certDir) > 0 {
 		mgr.GetWebhookServer().CertDir = certDir
 	}
@@ -144,7 +140,6 @@ func initializationControllers(ctx context.Context, mgr ctrl.Manager, opts *opti
 	nodeResource := utilfeature.DefaultFeatureGate.Enabled(features.CraneNodeResource)
 	clusterNodePrediction := utilfeature.DefaultFeatureGate.Enabled(features.CraneClusterNodePrediction)
 	analysis := utilfeature.DefaultMutableFeatureGate.Enabled(features.CraneAnalysis)
-	// todo: add more features
 	timeseriespredict := utilfeature.DefaultFeatureGate.Enabled(features.CraneTimeSeriesPrediction)
 
 	discoveryClientSet, err := discovery.NewDiscoveryClientForConfig(mgr.GetConfig())
@@ -245,7 +240,6 @@ func initializationControllers(ctx context.Context, mgr ctrl.Manager, opts *opti
 	if analysis {
 		if err := (&analytics.Controller{
 			Client:     mgr.GetClient(),
-			Logger:     log.Logger().WithName("analytics-controller"),
 			Scheme:     mgr.GetScheme(),
 			RestMapper: mgr.GetRESTMapper(),
 			Recorder:   mgr.GetEventRecorderFor("analytics-controller"),

deploy/craned/deployment.yaml

@@ -35,21 +35,28 @@ spec:
           operator: Exists
       containers:
         - name: craned
-          image: docker.io/gocrane/craned
+          image: docker.io/gocrane/craned:v0.1.0
           imagePullPolicy: IfNotPresent
           command:
             - /craned
             - --prometheus-address=PROMETHEUS_ADDRESS
+            - --feature-gates=Analysis=true,TimeSeriesPrediction=true,Autoscaling=true
+            - --recommendation-config-file=/tmp/recommendation-config/config_set.yaml
             - -v=4
           volumeMounts:
             - mountPath: /tmp/k8s-webhook-server/serving-certs
               name: cert
               readOnly: true
+            - name: config
+              mountPath: /tmp/recommendation-config
       volumes:
         - name: cert
           secret:
             defaultMode: 420
             secretName: webhook-server-tls
+        - name: config
+          configMap:
+            name: recommendation-config
 ---
 apiVersion: v1
 data:
@@ -60,3 +67,19 @@ metadata:
   name: webhook-server-tls
   namespace: crane-system
 type: kubernetes.io/tls
+
+---
+
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: recommendation-config
+  namespace: crane-system
+data:
+  config_set.yaml: |-
+    apiVersion: analysis.crane.io/v1alpha1
+    kind: ConfigSet
+    configs:
+      - targets: []
+        properties:
+          cpu-request-percentile: "0.98"

deploy/metric-adapter/deployment.yaml

@@ -27,7 +27,7 @@ spec:
       serviceAccountName: metric-adapter
       containers:
         - name: metric-adapter
-          image: docker.io/gocrane/metric-adapter
+          image: docker.io/gocrane/metric-adapter:v0.1.0
           imagePullPolicy: IfNotPresent
           args:
             - /metric-adapter

docs/tutorials/using-effective-hpa-to-scaling-with-effectiveness.md

@@ -20,7 +20,7 @@ spec:
     name: php-apache
   minReplicas: 1        # MinReplicas is the lower limit replicas to the scale target which the autoscaler can scale down to.
   maxReplicas: 10       # MaxReplicas is the upper limit replicas to the scale target which the autoscaler can scale up to.
-  scaleStrategy: Auto   # ScaleStrategy indicate the strategy to scaling target, value can be "Auto" and "Manual".
+  scaleStrategy: Auto   # ScaleStrategy indicate the strategy to scaling target, value can be "Auto" and "Preview".
   # Metrics contains the specifications for which to use to calculate the desired replica count.
   metrics:
   - type: Resource
@@ -45,7 +45,7 @@ spec:
 * spec.scaleTargetRef defines the reference to the workload that should be scaled.
 * spec.minReplicas is the lower limit replicas to the scale target which the autoscaler can scale down to.
 * spec.maxReplicas is the upper limit replicas to the scale target which the autoscaler can scale up to.
-* spec.metrics indicate the strategy to scaling target, value can be "Auto" and "Manual".
+* spec.metrics indicate the strategy to scaling target, value can be "Auto" and "Preview".
 * spec.metrics contains the specifications for which to use to calculate the desired replica count. Please refer to the details:
 * spec.prediction defines configurations for predict resources.If unspecified, defaults don't enable prediction.
 
@@ -138,22 +138,24 @@ We can see significant improvement with EffectiveHorizontalPodAutoscaler:
 * Fewer replicas changes than HorizontalPodAutoscaler
 
 ### ScaleStrategy
-EffectiveHorizontalPodAutoscaler provides two strategies for scaling: `Auto` and `Manual`. User can change the strategy at runtime, and it will take effect on the fly.
+EffectiveHorizontalPodAutoscaler provides two strategies for scaling: `Auto` and `Preview`. User can change the strategy at runtime, and it will take effect on the fly.
 
 #### Auto
 Auto strategy achieves automatic scaling based on metrics. It is the default strategy. With this strategy, EffectiveHorizontalPodAutoscaler will create and control a HorizontalPodAutoscaler instance in backend. We don't recommend explicit configuration on the underlying HorizontalPodAutoscaler because it will be overridden by EffectiveHPAController. If user delete EffectiveHorizontalPodAutoscaler, HorizontalPodAutoscaler will be cleaned up too.
 
-#### Manual
-Manual strategy means user can specify replicas of target. User can switch from default strategy to this one by applying `spec.scaleStrategy` to `Manual`. It will take effect immediately, During the switch, EffectiveHPAController will disable HorizontalPodAutoscaler if exists and scale the target to the value `spec.specificReplicas`, if user not set `spec.specificReplicas`, when ScaleStrategy is change to Manual, it will just stop scaling.
+#### Preview
+Preview strategy means EffectiveHorizontalPodAutoscaler won't change target's replicas automatically, so you can preview the calculated replicas and control target's replicas by themselves. User can switch from default strategy to this one by applying `spec.scaleStrategy` to `Preview`. It will take effect immediately, During the switch, EffectiveHPAController will disable HorizontalPodAutoscaler if exists and scale the target to the value `spec.specificReplicas`, if user not set `spec.specificReplicas`, when ScaleStrategy is change to Preview, it will just stop scaling.
 
-A sample manual configuration looks like following:
+A sample preview configuration looks like following:
 ```yaml
 apiVersion: autoscaling.crane.io/v1alpha1
 kind: EffectiveHorizontalPodAutoscaler
 spec:
-  scaleStrategy: Auto   # ScaleStrategy indicate the strategy to scaling target, value can be "Auto" and "Manual".
-  specificReplicas: 5   # SpecificReplicas specify the target replicas.
-
+  scaleStrategy: Preview   # ScaleStrategy indicate the strategy to scaling target, value can be "Auto" and "Preview".
+  specificReplicas: 5      # SpecificReplicas specify the target replicas.
+status:
+  expectReplicas: 4        # expectReplicas is the calculated replicas that based on prediction metrics or spec.specificReplicas.
+  currentReplicas: 4       # currentReplicas is actual replicas from target
 ```
 
 ### HorizontalPodAutoscaler compatible

examples/config_set.yaml

@@ -0,0 +1,6 @@
+apiVersion: analysis.crane.io/v1alpha1
+kind: ConfigSet
+configs:
+  - targets: []
+    properties:
+      cpu-request-percentile: "0.98"

go.mod

@@ -5,7 +5,6 @@ go 1.17
 require (
 	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/go-echarts/go-echarts/v2 v2.2.4
-	github.com/go-logr/logr v0.4.0
 	github.com/gocrane/api v0.0.0-20220101093604-2129a9ce85e3
 	github.com/google/go-cmp v0.5.6 // indirect
 	github.com/mjibson/go-dsp v0.0.0-20180508042940-11479a337f12
@@ -47,6 +46,7 @@ require (
 	github.com/evanphx/json-patch v4.11.0+incompatible // indirect
 	github.com/felixge/httpsnoop v1.0.1 // indirect
 	github.com/fsnotify/fsnotify v1.4.9 // indirect
+	github.com/go-logr/logr v0.4.0 // indirect
 	github.com/go-ole/go-ole v1.2.5 // indirect
 	github.com/go-openapi/jsonpointer v0.19.5 // indirect
 	github.com/go-openapi/jsonreference v0.19.5 // indirect