feat: refactor monitor loop.

Signed-off-by: IRONICBo <[email protected]>
antrea-io · Apr 22, 2024 · 6915393 · 6915393
1 parent 3c8bb56
commit 6915393
Show file tree

Hide file tree

Showing 31 changed files with 493 additions and 424 deletions.
diff --git a/build/charts/antrea/crds/nodelatencymonitor.yaml b/build/charts/antrea/crds/nodelatencymonitor.yaml
@@ -5,7 +5,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -17,12 +17,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -31,8 +33,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/build/yamls/antrea-aks.yml b/build/yamls/antrea-aks.yml
@@ -4316,7 +4316,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -4328,12 +4328,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -4342,8 +4344,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/build/yamls/antrea-crds.yml b/build/yamls/antrea-crds.yml
@@ -4293,7 +4293,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -4305,12 +4305,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -4319,8 +4321,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/build/yamls/antrea-eks.yml b/build/yamls/antrea-eks.yml
@@ -4316,7 +4316,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -4328,12 +4328,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -4342,8 +4344,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/build/yamls/antrea-gke.yml b/build/yamls/antrea-gke.yml
@@ -4316,7 +4316,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -4328,12 +4328,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -4342,8 +4344,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/build/yamls/antrea-ipsec.yml b/build/yamls/antrea-ipsec.yml
@@ -4316,7 +4316,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -4328,12 +4328,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -4342,8 +4344,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/build/yamls/antrea.yml b/build/yamls/antrea.yml
@@ -4316,7 +4316,7 @@ metadata:
 spec:
   group: crd.antrea.io
   versions:
-    - name: v1alpha2
+    - name: v1alpha1
       served: true
       storage: true
       schema:
@@ -4328,12 +4328,14 @@ spec:
             spec:
               type: object
               required:
-                - pingInterval
+                - pingIntervalSeconds
               properties:
-                pingInterval:
+                pingIntervalSeconds:
                   type: integer
+                  format: int32
                   minimum: 1
-                  # Explanation: At least one second ("1") or higher in seconds.
+                  description: "Ping interval in seconds, must be at least 1."
+                  default: 10
             metadata:
               type: object
               properties:
@@ -4342,8 +4344,8 @@ spec:
                   pattern: '^default$'
       additionalPrinterColumns:
         - description: Specifies the interval between pings.
-          jsonPath: .spec.pingInterval
-          name: PingInterval
+          jsonPath: .spec.pingIntervalSeconds
+          name: PingIntervalSeconds
           type: string
         - jsonPath: .metadata.creationTimestamp
           name: Age

diff --git a/cmd/antrea-agent/agent.go b/cmd/antrea-agent/agent.go
@@ -124,7 +124,7 @@ func run(o *Options) error {
 	endpointsInformer := informerFactory.Core().V1().Endpoints()
 	endpointSliceInformer := informerFactory.Discovery().V1().EndpointSlices()
 	namespaceInformer := informerFactory.Core().V1().Namespaces()
-	nodeLatencyMonitorInformer := crdInformerFactory.Crd().V1alpha2().NodeLatencyMonitors()
+	nodeLatencyMonitorInformer := crdInformerFactory.Crd().V1alpha1().NodeLatencyMonitors()
 
 	// Create Antrea Clientset for the given config.
 	antreaClientProvider := agent.NewAntreaClientProvider(o.config.AntreaClientConnection, k8sClient)

diff --git a/pkg/agent/monitortool/latency_store.go b/pkg/agent/monitortool/latency_store.go
@@ -62,13 +62,11 @@ func NewLatencyStore(isNetworkPolicyOnly bool) *LatencyStore {
 	return store
 }
 
-func (l *LatencyStore) GetNodeIPLatencyEntryByKey(key string) (*NodeIPLatencyEntry, bool) {
+func (l *LatencyStore) GetNodeIPLatencyEntryByKey(key string) *NodeIPLatencyEntry {
 	l.mutex.RLock()
 	defer l.mutex.RUnlock()
 
-	entry, found := l.nodeIPLatencyMap[key]
-
-	return entry, found
+	return l.nodeIPLatencyMap[key]
 }
 
 func (l *LatencyStore) DeleteNodeIPLatencyEntryByKey(key string) {
@@ -122,6 +120,7 @@ func (l *LatencyStore) updateNodeMap(node *corev1.Node) {
 	if l.isNetworkPolicyOnly {
 		transportIPs, err := getTransportIPs(node)
 		if err != nil {
+			klog.Errorf("Failed to get transport IPs for Node %s: %v", node.Name, err)
 			return
 		}
 
@@ -130,12 +129,15 @@ func (l *LatencyStore) updateNodeMap(node *corev1.Node) {
 	} else {
 		gw0IPs, err := getGWIPs(node)
 		if err != nil {
+			klog.Errorf("Failed to get gateway IPs for Node %s: %v", node.Name, err)
 			return
 		}
 
 		// Add the node to the node IP map
 		l.nodeGatewayMap[node.Name] = gw0IPs
 	}
+
+	klog.Infof("Node %s has gateway IPs %v", node.Name, l.nodeGatewayMap[node.Name])
 }
 
 func getTransportIPs(node *corev1.Node) ([]net.IP, error) {
@@ -161,21 +163,23 @@ func getGWIPs(node *corev1.Node) ([]net.IP, error) {
 	podCIDRStrs := getPodCIDRsOnNode(node)
 	if len(podCIDRStrs) == 0 {
 		// Skip the node if it does not have a PodCIDR.
-		klog.Warningf("Node %s does not have a PodCIDR", node.Name)
+		klog.Errorf("Node %s does not have a PodCIDR", node.Name)
 		return gwIPs, errors.New("node does not have a PodCIDR")
 	}
 
 	// the 0th entry must match the podCIDR field. It may contain at most 1 value for
 	// each of IPv4 and IPv6.
+	// Both podCIDRStrs need to be parsed to get the gateway IP.
 	for _, podCIDR := range podCIDRStrs {
 		if podCIDR == "" {
 			klog.Errorf("PodCIDR is empty for Node %s", node.Name)
+			return gwIPs, errors.New("PodCIDR is empty")
 		}
 
 		peerPodCIDRAddr, _, err := net.ParseCIDR(podCIDR)
 		if err != nil {
 			klog.Errorf("Failed to parse PodCIDR %s for Node %s", podCIDR, node.Name)
-			continue
+			return gwIPs, err
 		}
 
 		// Add first ip in CIDR to the map