diff --git a/docs/feature.md b/docs/feature.md
index 650b18e310adda45c53f6ad6ed817eaffd89c2a2..63218319108501d22f01fcc2aa98de91a5f951a1 100644
--- a/docs/feature.md
+++ b/docs/feature.md
@@ -37,6 +37,11 @@ annotations:
   insmod /lib/modules/bwm/bwm.ko
   ```
 
+#### 约束限制
+  - 网络绝对抢占不支持统计主机网络命名空间内的网络带宽。包括：
+    -  忽略主机进程所使用的网络带宽；
+    -  忽略与主机共享网络命名空间的容器内部进程所使用的网络带宽（包括kube-proxy容器等）；
+
 ## dynCache 内存带宽和LLC限制
 rubik支持业务的Pod内存带宽(memory bandwidth)和LLC(Last Level Cache)限制，通过限制离线业务的内存带宽/LLC使用，减少其对在线业务的干扰。
 
diff --git a/pkg/core/typedef/podinfo.go b/pkg/core/typedef/podinfo.go
index bc63962d927fe9b55bbfc9e96b1f36b401601bfd..4af68639468964b8d0beaea7f64567678eba31f4 100644
--- a/pkg/core/typedef/podinfo.go
+++ b/pkg/core/typedef/podinfo.go
@@ -32,6 +32,7 @@ type PodInfo struct {
 	Labels              map[string]string         `json:"labels,omitempty"`
 	ID                  string                    `json:"id,omitempty"` // id of the sandbox container in pod
 	StartTime           *metav1.Time              `json:"startTime,omitempty"`
+	HostNetwork         bool                      `json:"hostNetwork,omitempty"`
 	nriContainerRequest map[string]ResourceMap
 	nriContainerLimit   map[string]ResourceMap
 }
@@ -47,6 +48,7 @@ func NewPodInfo(pod *RawPod) *PodInfo {
 		Annotations:     pod.DeepCopy().Annotations,
 		Labels:          pod.DeepCopy().Labels,
 		StartTime:       pod.Status.StartTime,
+		HostNetwork:     pod.Spec.HostNetwork,
 	}
 }
 
diff --git a/pkg/services/preemption/net.go b/pkg/services/preemption/net.go
index 8e4e361a66939c11ec7cd182dc79e5262cf26225..fecd8fb55e43c285f10f3fd5b81ae2f23ace756e 100644
--- a/pkg/services/preemption/net.go
+++ b/pkg/services/preemption/net.go
@@ -65,20 +65,24 @@ func validateNetResConf(conf *PreemptionConfig) error {
 	return nil
 }
 
-func enableNetRes(pod *typedef.PodInfo) error {
+func enableNetRes(pod *typedef.PodInfo) (bool, error) {
 	var err error
 	var pid string
 
 	if pid, err = getPodProcID(pod); err != nil {
-		return fmt.Errorf("failed to get Pod procID %s: %v", pid, err)
+		return true, fmt.Errorf("failed to get Pod procID %s: %v", pid, err)
+	}
+
+	if pod.HostNetwork {
+		return false, nil
 	}
 
 	if err = enablePodNetqos(pid); err != nil {
 		disablePodNetqos(pid)
-		return err
+		return true, err
 	}
 
-	return nil
+	return true, nil
 }
 
 func initNetRes(conf *PreemptionConfig) error {
diff --git a/pkg/services/preemption/preemption.go b/pkg/services/preemption/preemption.go
index 1dc9190292feaa3935ad30b2ad1f383b658ff559..d40b050587c4e06026a25b7ab59d3537181a5ca3 100644
--- a/pkg/services/preemption/preemption.go
+++ b/pkg/services/preemption/preemption.go
@@ -32,7 +32,7 @@ type resOpt struct {
 	getQosStr       func(int) string
 	validateResConf func(*PreemptionConfig) error
 	initRes         func(*PreemptionConfig) error
-	enableRes       func(*typedef.PodInfo) error
+	enableRes       func(*typedef.PodInfo) (bool, error) // Return true to indicate that the res needs to enable qos
 }
 
 var supportCgroupTypes = map[string]resOpt{
@@ -178,7 +178,12 @@ func (q *Preemption) SetQoSLevel(pod *typedef.PodInfo) error {
 	for _, r := range q.config.Resource {
 		resOpt := supportCgroupTypes[r]
 		if resOpt.enableRes != nil {
-			if err := resOpt.enableRes(pod); err != nil {
+			needQos, err := resOpt.enableRes(pod)
+			if !needQos {
+				log.Infof("ignore pod %s(%s) %s-qos", pod.Name, pod.UID, r)
+				continue
+			}
+			if err != nil {
 				return err
 			}
 		}