Merge pull request #59 from sustainable-computing-io/issue30-sampling…

…rate

enable deployment customization of sampling rate

cmd/main.go

@@ -19,6 +19,8 @@ package main
 import (
 	"flag"
 	"os"
+	"time"
+	"strconv"
 
 	// Import all Kubernetes client auth plugins (e.g. Azure, GCP, OIDC, etc.)
 	// to ensure that exec-entrypoint and run can make use of them.
@@ -55,13 +57,15 @@ func main() {
 	var keplerPrometheusUrl string
 	var susqlPrometheusMetricsUrl string
 	var susqlPrometheusDatabaseUrl string
+	var samplingRate string
 
 	flag.StringVar(&metricsAddr, "metrics-bind-address", ":8080", "The address the metric endpoint binds to.")
 	flag.StringVar(&probeAddr, "health-probe-bind-address", ":8081", "The address the probe endpoint binds to.")
 	flag.BoolVar(&enableLeaderElection, "leader-elect", false, "Enable leader election for controller manager. Enabling this will ensure there is only one active controller manager.")
 	flag.StringVar(&keplerPrometheusUrl, "kepler-prometheus-url", "", "The URL for the Prometheus server where Kepler stores the energy data")
 	flag.StringVar(&susqlPrometheusDatabaseUrl, "susql-prometheus-database-url", "", "The URL for the Prometheus database where SusQL stores the energy data")
 	flag.StringVar(&susqlPrometheusMetricsUrl, "susql-prometheus-metrics-url", "", "The URL for the Prometheus metrics where SusQL exposes the energy data")
+	flag.StringVar(&samplingRate, "sampling-rate", "", "Sampling rate in seconds")
 
 	opts := zap.Options{
 		Development: true,
@@ -95,12 +99,18 @@ func main() {
 		os.Exit(1)
 	}
 
+	samplingRateInteger, err := strconv.Atoi(samplingRate)
+	if err != nil {
+		samplingRateInteger = 2
+	}
+
 	if err = (&controller.LabelGroupReconciler{
 		Client:                     mgr.GetClient(),
 		Scheme:                     mgr.GetScheme(),
 		KeplerPrometheusUrl:        keplerPrometheusUrl,
 		SusQLPrometheusDatabaseUrl: susqlPrometheusDatabaseUrl,
 		SusQLPrometheusMetricsUrl:  susqlPrometheusMetricsUrl,
+		SamplingRate:               time.Duration(samplingRateInteger) * time.Second,
 	}).SetupWithManager(mgr); err != nil {
 		setupLog.Error(err, "unable to create controller", "controller", "LabelGroup")
 		os.Exit(1)

deployment/deploy.sh

@@ -67,6 +67,11 @@ if [[ -z ${SUSQL_PROMETHEUS_URL} ]]; then
     fi
 fi
 
+if [[ -z ${SUSQL_SAMPLING_RATE} ]]; then
+    SUSQL_SAMPLING_RATE="2"
+fi
+
+
 # Check if namespace exists
 if [[ -z $(kubectl get namespaces --no-headers -o custom-columns=':{.metadata.name}' | grep ${SUSQL_NAMESPACE}) ]]; then
     echo "Namespace '${SUSQL_NAMESPACE}' doesn't exist. Creating it."
@@ -98,6 +103,7 @@ echo "PROMETHEUS_DOMAIN - '${PROMETHEUS_DOMAIN}'"
 echo "PROMETHEUS_PORT - '${PROMETHEUS_PORT}'"
 echo "KEPLER_PROMETHEUS_URL - '${KEPLER_PROMETHEUS_URL}'"
 echo "SUSQL_PROMETHEUS_URL - '${SUSQL_PROMETHEUS_URL}'"
+echo "SUSQL_SAMPLING_RATE - '${SUSQL_SAMPLING_RATE}'"
 echo "SUSQL_ENHANCED - '${SUSQL_ENHANCED}'"
 echo "SUSQL_REGISTRY - '${SUSQL_REGISTRY}'"
 echo "SUSQL_IMAGE_NAME - '${SUSQL_IMAGE_NAME}'"
@@ -125,6 +131,7 @@ echo "export PROMETHEUS_DOMAIN=${PROMETHEUS_DOMAIN}" >> ${LOGFILE}
 echo "export PROMETHEUS_PORT=${PROMETHEUS_PORT}" >> ${LOGFILE}
 echo "export KEPLER_PROMETHEUS_URL=${KEPLER_PROMETHEUS_URL}" >> ${LOGFILE}
 echo "export SUSQL_PROMETHEUS_URL=${SUSQL_PROMETHEUS_URL}" >> ${LOGFILE}
+echo "export SUSQL_SAMPLING_RATE=${SUSQL_SAMPLING_RATE}" >> ${LOGFILE}
 echo "export SUSQL_ENHANCED=${SUSQL_ENHANCED}" >> ${LOGFILE}
 echo "export SUSQL_REGISTRY=${SUSQL_REGISTRY}" >> ${LOGFILE}
 echo "export SUSQL_IMAGE_NAME=${SUSQL_IMAGE_NAME}" >> ${LOGFILE}
@@ -214,6 +221,7 @@ do
             --set keplerPrometheusUrl="${KEPLER_PROMETHEUS_URL}" \
             --set susqlPrometheusDatabaseUrl="${SUSQL_PROMETHEUS_URL}" \
             --set susqlPrometheusMetricsUrl="http://0.0.0.0:8082" \
+            --set samplingRate="${SUSQL_SAMPLING_RATE}" \
             --set imagePullPolicy="Always" \
             --set containerImage="${SUSQL_REGISTRY}/${SUSQL_IMAGE_NAME}:${SUSQL_IMAGE_TAG}"
         if [[ ! -z ${SUSQL_ENHANCED} ]]; then

deployment/susql-controller/templates/deployment.yaml

@@ -27,6 +27,7 @@ spec:
                       - "--kepler-prometheus-url={{ .Values.keplerPrometheusUrl }}"
                       - "--susql-prometheus-database-url={{ .Values.susqlPrometheusDatabaseUrl }}"
                       - "--susql-prometheus-metrics-url={{ .Values.susqlPrometheusMetricsUrl }}"
+                      - "--sampling-rate={{ .Values.samplingRate }}"
                   ports:
                       - name: metrics
                         containerPort: 8082

deployment/susql-controller/values.yaml

@@ -22,3 +22,4 @@ requests:
 keplerPrometheusUrl: "http://prometheus-k8s.monitoring.svc.cluster.local:9090"
 susqlPrometheusDatabaseUrl: "http://prometheus-susql.openshift-kepler-operator.svc.cluster.local:9090"
 susqlPrometheusMetricsUrl: "http://0.0.0.0:8082"
+samplingRate: "2"

internal/controller/labelgroup_controller.go

@@ -37,12 +37,12 @@ type LabelGroupReconciler struct {
 	KeplerPrometheusUrl        string
 	SusQLPrometheusDatabaseUrl string
 	SusQLPrometheusMetricsUrl  string
+	SamplingRate               time.Duration               // Sampling rate for all the label groups
 }
 
 const (
 	keplerMetricName = "kepler_container_joules_total" // Kepler metric to query
 	susqlMetricName  = "susql_total_energy_joules"     // SusQL metric to query
-	samplingRate     = 2 * time.Second                 // Sampling rate for all the label groups
 	fixingDelay      = 15 * time.Second                // Time to wait in the even the label group was badly constructed
 	errorDelay       = 1 * time.Second                 // Time to wait when an error happens due to network connectivity issues
 )
@@ -227,7 +227,7 @@ func (r *LabelGroupReconciler) Reconcile(ctx context.Context, req ctrl.Request)
 		r.SetAggregatedEnergyForLabels(totalEnergy, labelGroup.Status.PrometheusLabels)
 
 		// Requeue
-		return ctrl.Result{RequeueAfter: samplingRate}, nil
+		return ctrl.Result{RequeueAfter: r.SamplingRate}, nil
 
 	default:
 		// First time seeing this object