Merge pull request #968 from spidernet-io/fix/sync-node-status-to-egt

sync node status to egt
spidernet-io · Nov 16, 2023 · 3575cba · 3575cba · weizhoublue · Nov 16, 2023
2 parents 6daaf07 + b35428f
commit 3575cba
Show file tree

Hide file tree

Showing 6 changed files with 71 additions and 37 deletions.
diff --git a/charts/crds/egressgateway.spidernet.io_egresstunnels.yaml b/charts/crds/egressgateway.spidernet.io_egresstunnels.yaml
@@ -72,6 +72,7 @@ spec:
                 - Failed
                 - Ready
                 - HeartbeatTimeout
+                - NodeNotReady
                 type: string
               tunnel:
                 properties:

diff --git a/docs/reference/EgressTunnel.en.md b/docs/reference/EgressTunnel.en.md
@@ -29,4 +29,6 @@ status:
     - `Init`: successful tunnel IP allocation
     - `Ready`: the tunnel IP is allocated and tunnel is established
     - `Failed`: tunnel IP allocation fails
+    - `HeartbeatTimeout` heartbeat Timeout for Agent
+    - `NodeNotReady` Node Status is NotReady
 8. Packet mark value, one for each node. For example, if node A has egress traffic that needs to be forwarded to gateway node B, the traffic of node A will be marked with a mark.Each node is assigned a unique packet mark value. For instance, if Node A needs to forward Egress traffic to the gateway node B, it applies a specific mark to the packets originating from Node A.
diff --git a/docs/reference/EgressTunnel.zh.md b/docs/reference/EgressTunnel.zh.md
@@ -29,4 +29,6 @@ status:
     - `Init`：分配隧道 IP 成功
     - `Ready`：隧道 IP 已分配，且隧道已建成
     - `Failed`：隧道 IP 分配失败
+    - `HeartbeatTimeout` Agent 心跳超时
+    - `NodeNotReady` Node 状态处于 NotReady
 8. 数据包 mark 值，每个节点对应一个。例如节点 A 有 Egress 流量需要转发到网关节点 B，会对 A 节点的流量打 mark 进行标记。
diff --git a/pkg/agent/vxlan.go b/pkg/agent/vxlan.go
@@ -529,16 +529,16 @@ func (r *vxlanReconciler) ensureEgressTunnelStatus(node *egressv1.EgressTunnel)
 	return nil
 }
 
-func (r *vxlanReconciler) updateEgressTunnelStatus(node *egressv1.EgressTunnel, version int) error {
+func (r *vxlanReconciler) updateEgressTunnelStatus(tunnel *egressv1.EgressTunnel, version int) error {
 	parent, err := r.getParent(version)
 	if err != nil {
 		return err
 	}
 
-	if node == nil {
-		node = new(egressv1.EgressTunnel)
+	if tunnel == nil {
+		tunnel = new(egressv1.EgressTunnel)
 		ctx := context.Background()
-		err = r.client.Get(ctx, types.NamespacedName{Name: r.cfg.NodeName}, node)
+		err = r.client.Get(ctx, types.NamespacedName{Name: r.cfg.NodeName}, tunnel)
 		if err != nil {
 			if k8sErr.IsNotFound(err) {
 				return nil
@@ -548,43 +548,45 @@ func (r *vxlanReconciler) updateEgressTunnelStatus(node *egressv1.EgressTunnel,
 	}
 
 	needUpdate := false
-	if node.Status.Tunnel.Parent.Name != parent.Name {
+	if tunnel.Status.Tunnel.Parent.Name != parent.Name {
 		needUpdate = true
-		node.Status.Tunnel.Parent.Name = parent.Name
+		tunnel.Status.Tunnel.Parent.Name = parent.Name
 	}
 
 	if version == 4 {
-		if node.Status.Tunnel.Parent.IPv4 != parent.IP.String() {
+		if tunnel.Status.Tunnel.Parent.IPv4 != parent.IP.String() {
 			needUpdate = true
-			node.Status.Tunnel.Parent.IPv4 = parent.IP.String()
+			tunnel.Status.Tunnel.Parent.IPv4 = parent.IP.String()
 		}
-		if node.Status.Tunnel.Parent.IPv6 != "" {
+		if tunnel.Status.Tunnel.Parent.IPv6 != "" {
 			needUpdate = true
-			node.Status.Tunnel.Parent.IPv6 = ""
+			tunnel.Status.Tunnel.Parent.IPv6 = ""
 		}
 	} else {
-		if node.Status.Tunnel.Parent.IPv6 != parent.IP.String() {
+		if tunnel.Status.Tunnel.Parent.IPv6 != parent.IP.String() {
 			needUpdate = true
-			node.Status.Tunnel.Parent.IPv6 = parent.IP.String()
+			tunnel.Status.Tunnel.Parent.IPv6 = parent.IP.String()
 		}
-		if node.Status.Tunnel.Parent.IPv4 != "" {
+		if tunnel.Status.Tunnel.Parent.IPv4 != "" {
 			needUpdate = true
-			node.Status.Tunnel.Parent.IPv4 = ""
+			tunnel.Status.Tunnel.Parent.IPv4 = ""
 		}
 	}
 
 	// calculate whether the state has changed, update if the status changes.
-	vtep := r.parseVTEP(node.Status)
+	vtep := r.parseVTEP(tunnel.Status)
 	if vtep != nil {
 		phase := egressv1.EgressTunnelReady
-		if node.Status.Phase != phase {
+		// We should not overwrite the updated state of the controller.
+		if tunnel.Status.Phase != phase &&
+			tunnel.Status.Phase != egressv1.EgressTunnelNodeNotReady {
 			needUpdate = true
-			node.Status.Phase = phase
+			tunnel.Status.Phase = phase
 		}
 	}
 
 	if needUpdate {
-		err := r.updateTunnelStatus(node)
+		err := r.updateTunnelStatus(tunnel)
 		if err != nil {
 			return err
 		}
@@ -763,20 +765,20 @@ func (r *vxlanReconciler) keepVXLAN() {
 	}
 }
 
-func (r *vxlanReconciler) updateTunnelStatus(node *egressv1.EgressTunnel) error {
+func (r *vxlanReconciler) updateTunnelStatus(tunnel *egressv1.EgressTunnel) error {
 	ctx, cancel := context.WithTimeout(context.Background(), time.Duration(r.cfg.FileConfig.GatewayFailover.EipEvictionTimeout)*time.Second)
 	defer cancel()
 
-	node.Status.LastHeartbeatTime = metav1.Now()
+	tunnel.Status.LastHeartbeatTime = metav1.Now()
 	r.log.Info("update tunnel status",
-		"phase", node.Status.Phase,
-		"tunnelIPv4", node.Status.Tunnel.IPv4,
-		"tunnelIPv6", node.Status.Tunnel.IPv6,
-		"parentName", node.Status.Tunnel.Parent.Name,
-		"parentIPv4", node.Status.Tunnel.Parent.IPv4,
-		"parentIPv6", node.Status.Tunnel.Parent.IPv6,
+		"phase", tunnel.Status.Phase,
+		"tunnelIPv4", tunnel.Status.Tunnel.IPv4,
+		"tunnelIPv6", tunnel.Status.Tunnel.IPv6,
+		"parentName", tunnel.Status.Tunnel.Parent.Name,
+		"parentIPv4", tunnel.Status.Tunnel.Parent.IPv4,
+		"parentIPv6", tunnel.Status.Tunnel.Parent.IPv6,
 	)
-	err := r.client.Status().Update(ctx, node)
+	err := r.client.Status().Update(ctx, tunnel)
 	if err != nil {
 		return err
 	}

diff --git a/pkg/controller/egress_tunnel.go b/pkg/controller/egress_tunnel.go
@@ -196,8 +196,8 @@ func (r *egReconciler) reconcileNode(ctx context.Context, req reconcile.Request,
 		return reconcile.Result{Requeue: false}, nil
 	}
 
-	en := new(egressv1.EgressTunnel)
-	err = r.client.Get(ctx, req.NamespacedName, en)
+	egressTunnel := new(egressv1.EgressTunnel)
+	err = r.client.Get(ctx, req.NamespacedName, egressTunnel)
 	if err != nil {
 		log.Info("create egress tunnel")
 		if !k8serr.IsNotFound(err) {
@@ -210,6 +210,30 @@ func (r *egReconciler) reconcileNode(ctx context.Context, req reconcile.Request,
 		return reconcile.Result{}, nil
 	}
 
+	// If GatewayFailover is not enabled, we watch the health status of the node,
+	// and switch the active node of the egress IP based on the status of the node.
+	if !r.config.FileConfig.GatewayFailover.Enable {
+		phase := egressv1.EgressTunnelReady
+		for _, condition := range node.Status.Conditions {
+			if condition.Type == corev1.NodeReady {
+				if condition.Status != corev1.ConditionTrue {
+					phase = egressv1.EgressTunnelNodeNotReady
+				}
+			}
+		}
+		// don't overwrite the EgressTunnelHeartbeatTimeout status
+		if egressTunnel.Status.Phase != egressv1.EgressTunnelHeartbeatTimeout {
+			if egressTunnel.Status.Phase != phase {
+				log.Info("update egress tunnel", "status", egressTunnel.Status)
+				egressTunnel.Status.Phase = phase
+				err := r.updateEgressTunnel(*egressTunnel)
+				if err != nil {
+					return reconcile.Result{Requeue: true}, err
+				}
+			}
+		}
+	}
+
 	return reconcile.Result{Requeue: false}, nil
 }
 
@@ -479,23 +503,24 @@ func (r *egReconciler) keepEgressTunnel(node egressv1.EgressTunnel, log logr.Log
 }
 
 func (r *egReconciler) updateEgressTunnel(node egressv1.EgressTunnel) error {
-	phase := egressv1.EgressTunnelInit
+	if node.Status.Phase == "" {
+		node.Status.Phase = egressv1.EgressTunnelInit
+	}
 	if node.Status.Tunnel.Parent.Name == "" {
-		phase = egressv1.EgressTunnelInit
+		node.Status.Phase = egressv1.EgressTunnelInit
 	}
 	if node.Status.Mark == "" {
-		phase = egressv1.EgressTunnelPending
+		node.Status.Phase = egressv1.EgressTunnelPending
 	}
 	if node.Status.Tunnel.IPv4 == "" && r.allocatorV4 != nil {
-		phase = egressv1.EgressTunnelPending
+		node.Status.Phase = egressv1.EgressTunnelPending
 	}
 	if node.Status.Tunnel.IPv6 == "" && r.allocatorV6 != nil {
-		phase = egressv1.EgressTunnelPending
+		node.Status.Phase = egressv1.EgressTunnelPending
 	}
 	if node.Status.Tunnel.MAC == "" {
-		phase = egressv1.EgressTunnelPending
+		node.Status.Phase = egressv1.EgressTunnelPending
 	}
-	node.Status.Phase = phase
 
 	err := r.client.Status().Update(context.Background(), &node)
 	if err != nil {

diff --git a/pkg/k8s/apis/v1beta1/egresstunnel_types.go b/pkg/k8s/apis/v1beta1/egresstunnel_types.go
@@ -38,7 +38,7 @@ type EgressTunnelSpec struct{}
 type EgressTunnelStatus struct {
 	// +kubebuilder:validation:Optional
 	Tunnel Tunnel `json:"tunnel,omitempty"`
-	// +kubebuilder:validation:Enum=Pending;Init;Failed;Ready;HeartbeatTimeout
+	// +kubebuilder:validation:Enum=Pending;Init;Failed;Ready;HeartbeatTimeout;NodeNotReady
 	Phase EgressTunnelPhase `json:"phase,omitempty"`
 	// +kubebuilder:validation:Optional
 	Mark string `json:"mark,omitempty"`
@@ -77,6 +77,8 @@ const (
 	EgressTunnelFailed EgressTunnelPhase = "Failed"
 	// EgressTunnelHeartbeatTimeout tunnel heartbeat timeout
 	EgressTunnelHeartbeatTimeout EgressTunnelPhase = "HeartbeatTimeout"
+	// EgressTunnelNodeNotReady node not ready
+	EgressTunnelNodeNotReady EgressTunnelPhase = "NodeNotReady"
 	// EgressTunnelReady tunnel is available
 	EgressTunnelReady EgressTunnelPhase = "Ready"
 )