nfd-gc: rename -metrics to -port

Align with other daemons, -metrics has been deprecated. This patch also
refactors the code to make it easier to add healthz endpoint.

Author: marquiz

cmd/nfd-gc/main.go

@@ -83,7 +83,7 @@ func initFlags(flagset *flag.FlagSet) *nfdgarbagecollector.Args {
 		"interval between cleanup of obsolete api objects")
 	flagset.StringVar(&args.Kubeconfig, "kubeconfig", "",
 		"Kubeconfig to use")
-	flagset.IntVar(&args.MetricsPort, "metrics", 8081,
+	flagset.IntVar(&args.Port, "port", 8080,
 		"Port on which to expose metrics.")
 
 	klog.InitFlags(flagset)

deployment/base/gc/gc.yaml

@@ -29,5 +29,5 @@ spec:
           command:
             - "nfd-gc"
           ports:
-            - name: metrics
-              containerPort: 8081
+            - name: http
+              containerPort: 8080

deployment/helm/node-feature-discovery/templates/nfd-gc.yaml

@@ -70,8 +70,8 @@ spec:
           readOnlyRootFilesystem: true
           runAsNonRoot: true
         ports:
-          - name: metrics
-            containerPort: {{ .Values.gc.metricsPort | default "8081"}}
+          - name: http
+            containerPort: {{ .Values.gc.port | default "8080"}}
 
     {{- with .Values.gc.nodeSelector }}
       nodeSelector:

deployment/helm/node-feature-discovery/values.yaml

@@ -572,7 +572,7 @@ gc:
       cpu: 10m
       memory: 128Mi
 
-  metricsPort: 8081
+  port: 8080
 
   nodeSelector: {}
   tolerations: []

docs/deployment/helm.md

@@ -330,7 +330,7 @@ API's you need to install the prometheus operator in your cluster.
 | `gc.podDisruptionBudget.minAvailable`               | integer | 1            | Specifies minAvailable for the podDisruptionBudget configuration for nfd-gc                                                                                                                    |
 | `gc.podDisruptionBudget.maxUnavailable`             | integer | NULL         | Specifies maxUnavailable for the podDisruptionBudget configuration for nfd-gc                                                                                                                  |
 | `gc.podDisruptionBudget.unhealthyPodEvictionPolicy` | string | AlwaysAllow   | Specifies unhealthyPodEvictionPolicy for the podDisruptionBudget configuration for nfd-gc                                                                                                      |
-| `gc.metricsPort`                | integer | 8081                      | Port on which to serve Prometheus metrics. **DEPRECATED**: will be replaced by `gc.port` in NFD v0.18.                                                                                                |
+| `gc.port`                       | integer | 8080                      | Port on which to serve Prometheus metrics.                                                                                                                                                            |
 | `gc.nodeSelector`               | dict    | {}                        | Garbage collector pod [node selector](https://kubernetes.io/docs/concepts/scheduling-eviction/assign-pod-node/#nodeselector)                                                                          |
 | `gc.tolerations`                | dict    | {}                        | Garbage collector pod [node tolerations](https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/)                                                                               |
 | `gc.annotations`                | dict    | {}                        | Garbage collector pod [annotations](https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/)                                                                                   |

docs/reference/gc-commandline-reference.md

@@ -42,3 +42,15 @@ Example:
 ```bash
 nfd-gc -gc-interval=1h
 ```
+
+### -port
+
+The `-port` flag specifies the port on which metrics are served on.
+
+Default: 8080
+
+Example:
+
+```bash
+nfd-gc -port=12345
+```

pkg/nfd-gc/nfd-gc.go

@@ -19,9 +19,12 @@ package nfdgarbagecollector
 import (
 	"context"
 	"fmt"
+	"net/http"
 	"time"
 
 	topologyv1alpha2 "github.com/k8stopologyawareschedwg/noderesourcetopology-api/pkg/apis/topology/v1alpha2"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -46,9 +49,9 @@ var (
 
 // Args are the command line arguments
 type Args struct {
-	GCPeriod    time.Duration
-	Kubeconfig  string
-	MetricsPort int
+	GCPeriod   time.Duration
+	Kubeconfig string
+	Port       int
 }
 
 type NfdGarbageCollector interface {
@@ -240,15 +243,28 @@ func (n *nfdGarbageCollector) startNodeInformer() error {
 
 // Run is a blocking function that removes stale NRT objects when Node is deleted and runs periodic GC to make sure any obsolete objects are removed
 func (n *nfdGarbageCollector) Run() error {
-	if n.args.MetricsPort > 0 {
-		m := utils.CreateMetricsServer(n.args.MetricsPort,
-			buildInfo,
-			objectsDeleted,
-			objectDeleteErrors)
-		go m.Run()
-		registerVersion(version.Get())
-		defer m.Stop()
-	}
+	httpMux := http.NewServeMux()
+	promRegistry := prometheus.NewRegistry()
+	promRegistry.MustRegister(
+		buildInfo,
+		objectsDeleted,
+		objectDeleteErrors)
+	httpMux.Handle("/metrics", promhttp.HandlerFor(promRegistry, promhttp.HandlerOpts{}))
+	registerVersion(version.Get())
+
+	// TODO: health probe endpoint could be added here
+
+	// Start HTTP server
+	httpServer := http.Server{Addr: fmt.Sprintf(":%d", n.args.Port), Handler: httpMux}
+	go func() {
+		klog.InfoS("starting HTTP server", "port", n.args.Port)
+		if err := httpServer.ListenAndServe(); err != nil && err != http.ErrServerClosed {
+			klog.ErrorS(err, "failed to start HTTP server")
+		} else {
+			klog.InfoS("HTTP server stopped")
+		}
+	}()
+	defer httpServer.Close()
 
 	if err := n.startNodeInformer(); err != nil {
 		return err