Updated on 2024-01-26 GMT+08:00

Network Policies

Network policies are designed by Kubernetes to restrict pod access. It is equivalent to a firewall at the application layer to enhance network security. The capabilities supported by network policies depend on the capabilities of the network add-ons of the cluster.

By default, if a namespace does not have any policy, pods in the namespace accept traffic from any source and send traffic to any destination.

Network policies are classified into the following types:

  • namespaceSelector: selects particular namespaces for which all pods should be allowed as ingress sources or egress destinations.
  • podSelector: selects particular pods in the same namespace as the network policy which should be allowed as ingress sources or egress destinations.
  • ipBlock: selects particular IP blocks to allow as ingress sources or egress destinations. (Only egress rules support IP blocks.)

Constraints

  • Only clusters that use the tunnel network model support network policies. Network policies are classified into the following types:
    • Ingress: All versions support this type.
    • Egress: Only the following OSs and cluster versions support egress rules.

      OS

      Cluster Version

      Verified Kernel Version

      CentOS

      v1.23 or later

      3.10.0-1062.18.1.el7.x86_64

      3.10.0-1127.19.1.el7.x86_64

      3.10.0-1160.25.1.el7.x86_64

      EulerOS 2.5

      v1.23 or later

      3.10.0-862.14.1.5.h591.eulerosv2r7.x86_64

      3.10.0-862.14.1.5.h687.eulerosv2r7.x86_64

      EulerOS 2.9

      v1.23 or later

      4.18.0-147.5.1.6.h541.eulerosv2r9.x86_64

      4.18.0-147.5.1.6.h766.eulerosv2r9.x86_64

  • Network isolation is not supported for IPv6 addresses.
  • If upgrade to a cluster version that supports egress rules is performed in in-place mode, you cannot use egress rules because the node OS is not upgraded. In this case, reset the node.

Using Ingress Rules

  • Using podSelector to specify the access scope
    apiVersion: networking.k8s.io/v1
    kind: NetworkPolicy
    metadata:
      name: test-network-policy
      namespace: default
    spec:
      podSelector:                  # The rule takes effect for pods with the role=db label.
        matchLabels:
          role: db
      ingress:                      # This is an ingress rule.
      - from:
        - podSelector:              # Only traffic from the pods with the "role=frontend" label is allowed.
            matchLabels:
              role: frontend
        ports:                      # Only TCP can be used to access port 6379.
        - protocol: TCP
          port: 6379

    The following figure shows how podSelector works.

    Figure 1 podSelector
  • Using namespaceSelector to specify the access scope
    apiVersion: networking.k8s.io/v1
    kind: NetworkPolicy
    metadata:
      name: test-network-policy
    spec:
      podSelector:                  # The rule takes effect for pods with the role=db label.
        matchLabels:
          role: db
      ingress:                      # This is an ingress rule.
      - from:
        - namespaceSelector:        # Only traffic from the pods in the namespace with the "project=myproject" label is allowed.
            matchLabels:
              project: myproject
        ports:                      # Only TCP can be used to access port 6379.
        - protocol: TCP
          port: 6379

    The following figure shows how namespaceSelector works.

    Figure 2 namespaceSelector

Using Egress Rules

Egress supports not only podSelector and namespaceSelector, but also ipBlock.

Only clusters of version 1.23 or later support egress rules. Currently, nodes running EulerOS 2.5, EulerOS 2.9, and CentOS 7.X are supported.

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: deny-client-a-via-except-cidr-egress-rule
  namespace: default
spec:
  policyTypes:                  # Must be specified for an egress rule.
    - Egress
  podSelector:                  # The rule takes effect for pods with the role=db label.
    matchLabels:
      role: db
  egress:                       # Egress rule
  - to:
    - ipBlock:
        cidr: 172.16.0.16/16    # Allow access to this CIDR block.
        except:
        - 172.16.0.40/32        # This CIDR block cannot be accessed. This value must fall within the range specified by cidr.

The following figure shows how ipBlock works.

Figure 3 ipBlock

You can define ingress and egress in the same rule.

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: test-network-policy
  namespace: default
spec:
  policyTypes:
  - Ingress
  - Egress
  podSelector:                  # The rule takes effect for pods with the role=db label.
    matchLabels:
      role: db
  ingress:                      # This is an ingress rule.
  - from:
    - podSelector:              # Only traffic from the pods with the "role=frontend" label is allowed.
        matchLabels:
          role: frontend
    ports:                      # Only TCP can be used to access port 6379.
    - protocol: TCP
      port: 6379
  egress:                       # Egress rule
  - to:
    - podSelector:              # Only pods with the role=web label can be accessed.
        matchLabels:
          role: web

The following figure shows how to use ingress and egress together.

Figure 4 Using both ingress and egress

Creating a Network Policy on the Console

  1. Log in to the CCE console and click the cluster name to access the cluster console.
  2. Choose Networking in the navigation pane, click the Network Policies tab, and click Create Network Policy in the upper right corner.

    • Policy Name: Specify a network policy name.
    • Namespace: Select a namespace in which the network policy is applied.
    • Selector: Enter a label, select the pod to be associated, and click Confirm. You can also click Reference Workload Label to reference the label of an existing workload.
    • Inbound Rule: Click to add an inbound rule. For details about parameter settings, see Table 1.

      Table 1 Adding an inbound rule

      Parameter

      Description

      Protocol & Port

      Select the protocol type and port. Currently, TCP and UDP are supported.

      Source Namespace

      Select a namespace whose objects can be accessed. If this parameter is not specified, the object belongs to the same namespace as the current policy.

      Source Pod Label

      Allow accessing the pods with this label. If this parameter is not specified, all pods in the namespace can be accessed.

  3. Click OK.