From ec1fabb19f6af216912bbad46f5df2314967df54 Mon Sep 17 00:00:00 2001
From: Geoffrey Israel <israelgeoffrey13@gmail.com>
Date: Tue, 3 Oct 2023 14:39:09 +0100
Subject: [PATCH] feat(lifecycle-operator): introduce metric showing readiness
 of operator (#2152)

Signed-off-by: Geoffrey Israel <israelgeoffrey13@gmail.com>
Co-authored-by: Giovanni Liva <giovanni.liva@dynatrace.com>
---
 lifecycle-operator/main.go                       | 16 ++++++++++++++++
 .../04-teststep-check-active-metrics.yaml        |  8 ++++++++
 test/testmetrics/metrics/check-active-metrics.sh | 12 ++++++++++++
 3 files changed, 36 insertions(+)
 create mode 100644 test/testmetrics/metrics/04-teststep-check-active-metrics.yaml
 create mode 100755 test/testmetrics/metrics/check-active-metrics.sh

diff --git a/lifecycle-operator/main.go b/lifecycle-operator/main.go
index 82a8298edd..42961ab3db 100644
--- a/lifecycle-operator/main.go
+++ b/lifecycle-operator/main.go
@@ -17,6 +17,7 @@ limitations under the License.
 package main
 
 import (
+	"context"
 	"flag"
 	"fmt"
 	"log"
@@ -49,6 +50,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus/promhttp"
 	"go.opentelemetry.io/otel"
 	otelprom "go.opentelemetry.io/otel/exporters/prometheus"
+	metricsapi "go.opentelemetry.io/otel/metric"
 	"go.opentelemetry.io/otel/sdk/metric"
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/runtime"
@@ -97,6 +99,8 @@ type envConfig struct {
 	KeptnOptionsCollectorURL string `envconfig:"OTEL_COLLECTOR_URL" default:""`
 }
 
+const KeptnLifecycleActiveMetric = "keptn_lifecycle_active"
+
 //nolint:funlen,gocognit,gocyclo
 func main() {
 	var env envConfig
@@ -121,6 +125,14 @@ func main() {
 	}
 	provider := metric.NewMeterProvider(metric.WithReader(exporter))
 	meter := provider.Meter("keptn/task")
+
+	keptnLifecycleActive, err := meter.Int64Counter(KeptnLifecycleActiveMetric, metricsapi.WithDescription("signals that Keptn Lifecycle Operator is installed correctly and ready"))
+
+	if err != nil {
+		setupLog.Error(err, "unable to create metric "+KeptnLifecycleActiveMetric)
+		os.Exit(1)
+	}
+
 	keptnMeters := telemetry.SetUpKeptnTaskMeters(meter)
 
 	// Start the prometheus HTTP server and pass the exporter Collector to it
@@ -350,6 +362,10 @@ func main() {
 		setupLog.Error(err, "unable to set up ready check")
 		os.Exit(1)
 	}
+
+	// Set the metric value as soon as the operator starts
+	setupLog.Info("Keptn lifecycle-operator is alive")
+	keptnLifecycleActive.Add(context.Background(), 1)
 	if !disableWebhook {
 		webhookBuilder = webhookBuilder.SetCertificateWatcher(
 			certificates.NewCertificateWatcher(
diff --git a/test/testmetrics/metrics/04-teststep-check-active-metrics.yaml b/test/testmetrics/metrics/04-teststep-check-active-metrics.yaml
new file mode 100644
index 0000000000..3e8904738a
--- /dev/null
+++ b/test/testmetrics/metrics/04-teststep-check-active-metrics.yaml
@@ -0,0 +1,8 @@
+apiVersion: kuttl.dev/v1
+kind: TestStep
+commands:
+  - script: ./check-active-metrics.sh "keptn_lifecycle_active"
+    validate:
+      assert:
+        - contains:
+            - "keptn_lifecycle_active"
diff --git a/test/testmetrics/metrics/check-active-metrics.sh b/test/testmetrics/metrics/check-active-metrics.sh
new file mode 100755
index 0000000000..7ccf661faf
--- /dev/null
+++ b/test/testmetrics/metrics/check-active-metrics.sh
@@ -0,0 +1,12 @@
+#!/bin/bash
+
+# Usage: ./retrieve-metrics.sh <component_name>
+
+component_name=$1
+
+metrics_url="http://lifecycle-operator-metrics-service.keptn-lifecycle-toolkit-system.svc.cluster.local:2222/metrics"
+
+# Fetch keptn_lifecycle_active metrics
+metrics=$(curl -s $metrics_url | grep "${component_name}")
+
+echo "$metrics"