feat: fabric and gateway health probe

Author: cheina97

cmd/fabric/main.go

@@ -31,6 +31,7 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/cache"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
+	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	"sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/metrics/server"
 
@@ -142,6 +143,14 @@ func run(cmd *cobra.Command, _ []string) error {
 		return fmt.Errorf("unable to create manager: %w", err)
 	}
 
+	// Register the healthiness probes.
+	if err := mgr.AddHealthzCheck("healthz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up healthz probe: %w", err)
+	}
+	if err := mgr.AddReadyzCheck("readyz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up readyz probe: %w", err)
+	}
+
 	gwr, err := sourcedetector.NewGatewayReconciler(
 		mgr.GetClient(),
 		mgr.GetScheme(),

cmd/gateway/geneve/main.go

@@ -25,6 +25,7 @@ import (
 	"k8s.io/klog/v2"
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
+	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	"sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/metrics/server"
 
@@ -95,6 +96,14 @@ func run(cmd *cobra.Command, _ []string) error {
 		return fmt.Errorf("unable to create manager: %w", err)
 	}
 
+	// Register the healthiness probes.
+	if err := mgr.AddHealthzCheck("healthz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up healthz probe: %w", err)
+	}
+	if err := mgr.AddReadyzCheck("readyz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up readyz probe: %w", err)
+	}
+
 	inr, err := geneve.NewInternalNodeReconciler(
 		mgr.GetClient(),
 		mgr.GetScheme(),

cmd/gateway/main.go

@@ -29,6 +29,7 @@ import (
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
+	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	"sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/metrics/server"
 
@@ -160,6 +161,14 @@ func run(cmd *cobra.Command, _ []string) error {
 		return fmt.Errorf("unable to create manager: %w", err)
 	}
 
+	// Register the healthiness probes.
+	if err := mgr.AddHealthzCheck("healthz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up healthz probe: %w", err)
+	}
+	if err := mgr.AddReadyzCheck("readyz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up readyz probe: %w", err)
+	}
+
 	if connoptions.EnableConnectionController {
 		// Setup the connection controller.
 		connr, err := connection.NewConnectionsReconciler(

cmd/gateway/wireguard/main.go

@@ -29,6 +29,7 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/cache"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
 	"sigs.k8s.io/controller-runtime/pkg/event"
+	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	"sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/metrics"
 	"sigs.k8s.io/controller-runtime/pkg/metrics/server"
@@ -107,6 +108,14 @@ func run(cmd *cobra.Command, _ []string) error {
 		return fmt.Errorf("unable to create manager: %w", err)
 	}
 
+	// Register the healthiness probes.
+	if err := mgr.AddHealthzCheck("healthz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up healthz probe: %w", err)
+	}
+	if err := mgr.AddReadyzCheck("readyz", healthz.Ping); err != nil {
+		return fmt.Errorf("unable to set up readyz probe: %w", err)
+	}
+
 	// Setup the controller.
 	pkr, err := wireguard.NewPublicKeysReconciler(
 		mgr.GetClient(),

deployments/liqo/README.md

@@ -86,6 +86,8 @@
 | networking.enabled | bool | `true` | Use the default Liqo networking module. |
 | networking.fabric.config.fullMasquerade | bool | `false` | Enabe/Disable the full masquerade mode for the fabric pod. It means that all traffic will be masquerade using the first external cidr IP, instead of using the pod IP. Full masquerade is useful when the cluster nodeports uses a PodCIDR IP to masqerade the incoming traffic. IMPORTANT: Please consider that enabling this feature will masquerade the source IP of traffic towards a remote cluster, making impossible for a pod that receives the traffic to know the original source IP. |
 | networking.fabric.config.gatewayMasqueradeBypass | bool | `false` | Enable/Disable the masquerade bypass for the gateway pods. It means that the packets from gateway pods will not be masqueraded from the host where the pod is scheduled. This is useful in scenarios where CNIs masquerade the traffic from pod to nodes. For example this is required when using the Azure CNI or Kindnet. |
+| networking.fabric.config.healthProbeBindAddressPort | string | `"8081"` | Set the port where the fabric pod will expose the health probe. To disable the health probe, set the port to 0. |
+| networking.fabric.config.metricsAddressPort | string | `"8082"` | Set the port where the fabric pod will expose the metrics. To disable the metrics, set the port to 0. |
 | networking.fabric.config.nftablesMonitor | bool | `true` | Enable/Disable the nftables monitor for the fabric pod. It means that the fabric pod will monitor the nftables rules and will restore them in case of changes. In some cases (like K3S), this monitor can cause a huge amount of CPU usage. If you are experiencing high CPU usage, you can disable this feature. |
 | networking.fabric.image.name | string | `"ghcr.io/liqotech/fabric"` | Image repository for the fabric pod. |
 | networking.fabric.image.version | string | `""` | Custom version for the fabric image. If not specified, the global tag is used. |

deployments/liqo/templates/liqo-fabric-daemonset.yaml

@@ -45,6 +45,8 @@ spec:
           - --podname=$(POD_NAME)
           - --nodename=$(NODE_NAME)
           - --geneve-port={{ .Values.networking.genevePort }}
+          - --health-probe-bind-address=:{{ .Values.networking.fabric.config.healthProbeBindAddressPort}}
+          - --metrics-address=:{{ .Values.networking.fabric.config.metricsAddressPort}}
           {{- if not .Values.requirements.kernel.enabled }}
           - --disable-kernel-version-check
           {{- end }}
@@ -79,6 +81,16 @@ spec:
               valueFrom:
                 fieldRef:
                   fieldPath: metadata.name
+          {{- if and .Values.networking.fabric.config.healthProbeBindAddressPort (ne .Values.networking.fabric.config.healthProbeBindAddressPort "0") }}
+          ports:
+          - name: healthz
+            containerPort: {{ .Values.networking.fabric.config.healthProbeBindAddressPort }}
+            protocol: TCP
+          readinessProbe:
+            httpGet:
+              path: /readyz
+              port: healthz
+          {{- end }}
       hostNetwork: true
       {{- if .Values.networking.fabric.pod.priorityClassName }}
       priorityClassName: {{ .Values.networking.fabric.pod.priorityClassName }}

deployments/liqo/templates/liqo-wireguard-gateway-client-template.yaml

@@ -79,6 +79,14 @@ spec:
                 - containerPort: 8082
                   name: gw-metrics
                 {{- end }}
+                - containerPort: 8083
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 env:
                 - name: NODE_NAME
                   valueFrom:
@@ -117,6 +125,14 @@ spec:
                 - containerPort: 8084
                   name: wg-metrics
                 {{- end }}
+                - containerPort: 8085
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 securityContext:
                   capabilities:
                     add:
@@ -155,6 +171,14 @@ spec:
                 - containerPort: 8086
                   name: gv-metrics
                 {{- end }} 
+                - containerPort: 8087
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 env:
                 - name: NODE_NAME
                   valueFrom:

deployments/liqo/templates/liqo-wireguard-gateway-server-template-eks.yaml

@@ -106,6 +106,15 @@ spec:
                 - containerPort: 8082
                   name: gw-metrics
                 {{- end }}
+                ports:
+                - containerPort: 8083
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 env:
                 - name: NODE_NAME
                   valueFrom:
@@ -143,6 +152,15 @@ spec:
                 - containerPort: 8084
                   name: wg-metrics
                 {{- end }}
+                ports:
+                - containerPort: 8085
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 securityContext:
                   capabilities:
                     add:
@@ -172,7 +190,7 @@ spec:
                 {{- if .Values.metrics.enabled }}
                 - --metrics-address=:8086
                 {{- end }}
-                - --health-probe-bind-address=:8086
+                - --health-probe-bind-address=:8087
                 volumeMounts: 
                 - name: ipc 
                   mountPath: /ipc
@@ -181,6 +199,15 @@ spec:
                 - containerPort: 8086
                   name: gv-metrics
                 {{- end }}
+                ports:
+                - containerPort: 8087
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 env:
                 - name: NODE_NAME
                   valueFrom:

deployments/liqo/templates/liqo-wireguard-gateway-server-template.yaml

@@ -80,9 +80,9 @@ spec:
                 {{- include "liqo.concatenateMap" $d | nindent 16 }}
                 {{- end }}
                 {{- if .Values.metrics.enabled }}
-                - --metrics-address=:8084
+                - --metrics-address=:8082
                 {{- end }}
-                - --health-probe-bind-address=:8085
+                - --health-probe-bind-address=:8083
                 - --ping-enabled=true
                 - --ping-loss-threshold={{ .Values.networking.gatewayTemplates.ping.lossThreshold }}
                 - --ping-interval={{ .Values.networking.gatewayTemplates.ping.interval }}
@@ -96,9 +96,17 @@ spec:
                   mountPath: /ipc
                 {{- if .Values.metrics.enabled }}
                 ports:
-                - containerPort: 8084
+                - containerPort: 8082
                   name: gw-metrics
                 {{- end }}
+                - containerPort: 8083
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 env:
                 - name: NODE_NAME
                   valueFrom:
@@ -127,15 +135,23 @@ spec:
                 - --mtu={{"{{ .Spec.MTU }}"}}
                 - --listen-port={{"{{ .Spec.Endpoint.Port }}"}}
                 {{- if .Values.metrics.enabled }}
-                - --metrics-address=:8082
+                - --metrics-address=:8084
                 {{- end }}
-                - --health-probe-bind-address=:8083
+                - --health-probe-bind-address=:8085
                 - --implementation={{ .Values.networking.gatewayTemplates.wireguard.implementation }}
                 {{- if .Values.metrics.enabled }}
                 ports:
-                - containerPort: 8082
+                - containerPort: 8084
                   name: wg-metrics
                 {{- end }}
+                - containerPort: 8085
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 securityContext:
                   capabilities:
                     add:
@@ -174,6 +190,14 @@ spec:
                 - containerPort: 8086
                   name: gv-metrics
                 {{- end }}
+                - containerPort: 8087
+                  name: healthz
+                # ATTENTION: uncomment the readinessProbe section if you are aware of the consequences.
+                # If you have more replicas of the same gateway, the passive ones will not reach the ready state.
+                #readinessProbe:
+                #  httpGet:
+                #    path: /readyz
+                #    port: healthz
                 env:
                 - name: NODE_NAME
                   valueFrom:

deployments/liqo/values.yaml

@@ -124,6 +124,12 @@ networking:
       # In some cases (like K3S), this monitor can cause a huge amount of CPU usage.
       # If you are experiencing high CPU usage, you can disable this feature.
       nftablesMonitor: true
+      # -- Set the port where the fabric pod will expose the health probe.
+      # To disable the health probe, set the port to 0.
+      healthProbeBindAddressPort: "8081"
+      # -- Set the port where the fabric pod will expose the metrics.
+      # To disable the metrics, set the port to 0.
+      metricsAddressPort: "8082"
 
 authentication:
   # -- Enable/Disable the authentication module.