Chapter 19. Kubernetes NMState

19.1. About the Kubernetes NMState Operator

The Kubernetes NMState Operator provides a Kubernetes API for performing state-driven network configuration across the OpenShift Container Platform cluster’s nodes with NMState. The Kubernetes NMState Operator provides users with functionality to configure various network interface types, DNS, and routing on cluster nodes. Additionally, the daemons on the cluster nodes periodically report on the state of each node’s network interfaces to the API server.

Important

Kubernetes NMState Operator is a Technology Preview feature only. Technology Preview features are not supported with Red Hat production service level agreements (SLAs) and might not be functionally complete. Red Hat does not recommend using them in production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process.

For more information about the support scope of Red Hat Technology Preview features, see https://access.redhat.com/support/offerings/techpreview/.

Before you can use NMState with OpenShift Container Platform, you must install the Kubernetes NMState Operator.

19.1.1. Installing the Kubernetes NMState Operator

You must install the Kubernetes NMState Operator from the web console while logged in with administrator privileges. After it is installed, the Operator can deploy the NMState State Controller as a daemon set across all of the cluster nodes.

Procedure

  1. Select OperatorsOperatorHub.
  2. In the search field below All Items, enter nmstate and click Enter to search for the Kubernetes NMState Operator.
  3. Click on the Kubernetes NMState Operator search result.
  4. Click on Install to open the Install Operator window.
  5. Under Installed Namespace, ensure the namespace is openshift-nmstate. If openshift-nmstate does not exist in the combo box, click on Create Namespace and enter openshift-nmstate in the Name field of the dialog box and press Create.
  6. Click Install to install the Operator.
  7. After the Operator finishes installing, click View Operator.
  8. Under Provided APIs, click Create Instance to open the dialog box for creating an instance of kubernetes-nmstate.
  9. In the Name field of the dialog box, ensure the name of the instance is nmstate.

    Note

    The name restriction is a known issue. The instance is a singleton for the entire cluster.

  10. Accept the default settings and click Create to create the instance.

Summary

Once complete, the Operator has deployed the NMState State Controller as a daemon set across all of the cluster nodes.

19.2. Observing node network state

Node network state is the network configuration for all nodes in the cluster.

19.2.1. About nmstate

OpenShift Container Platform uses nmstate to report on and configure the state of the node network. This makes it possible to modify network policy configuration, such as by creating a Linux bridge on all nodes, by applying a single configuration manifest to the cluster.

Node networking is monitored and updated by the following objects:

NodeNetworkState
Reports the state of the network on that node.
NodeNetworkConfigurationPolicy
Describes the requested network configuration on nodes. You update the node network configuration, including adding and removing interfaces, by applying a NodeNetworkConfigurationPolicy manifest to the cluster.
NodeNetworkConfigurationEnactment
Reports the network policies enacted upon each node.

OpenShift Container Platform supports the use of the following nmstate interface types:

  • Linux Bridge
  • VLAN
  • Bond
  • Ethernet
Note

If your OpenShift Container Platform cluster uses OVN-Kubernetes as the default Container Network Interface (CNI) provider, you cannot attach a Linux bridge or bonding to the default interface of a host because of a change in the host network topology of OVN-Kubernetes. As a workaround, you can use a secondary network interface connected to your host, or switch to the OpenShift SDN default CNI provider.

19.2.2. Viewing the network state of a node

A NodeNetworkState object exists on every node in the cluster. This object is periodically updated and captures the state of the network for that node.

Procedure

  1. List all the NodeNetworkState objects in the cluster:

    $ oc get nns
  2. Inspect a NodeNetworkState object to view the network on that node. The output in this example has been redacted for clarity:

    $ oc get nns node01 -o yaml

    Example output

    apiVersion: nmstate.io/v1beta1
    kind: NodeNetworkState
    metadata:
      name: node01 1
    status:
      currentState: 2
        dns-resolver:
    ...
        interfaces:
    ...
        route-rules:
    ...
        routes:
    ...
      lastSuccessfulUpdateTime: "2020-01-31T12:14:00Z" 3

    1
    The name of the NodeNetworkState object is taken from the node.
    2
    The currentState contains the complete network configuration for the node, including DNS, interfaces, and routes.
    3
    Timestamp of the last successful update. This is updated periodically as long as the node is reachable and can be used to evalute the freshness of the report.

19.3. Updating node network configuration

You can update the node network configuration, such as adding or removing interfaces from nodes, by applying NodeNetworkConfigurationPolicy manifests to the cluster.

19.3.1. About nmstate

OpenShift Container Platform uses nmstate to report on and configure the state of the node network. This makes it possible to modify network policy configuration, such as by creating a Linux bridge on all nodes, by applying a single configuration manifest to the cluster.

Node networking is monitored and updated by the following objects:

NodeNetworkState
Reports the state of the network on that node.
NodeNetworkConfigurationPolicy
Describes the requested network configuration on nodes. You update the node network configuration, including adding and removing interfaces, by applying a NodeNetworkConfigurationPolicy manifest to the cluster.
NodeNetworkConfigurationEnactment
Reports the network policies enacted upon each node.

OpenShift Container Platform supports the use of the following nmstate interface types:

  • Linux Bridge
  • VLAN
  • Bond
  • Ethernet
Note

If your OpenShift Container Platform cluster uses OVN-Kubernetes as the default Container Network Interface (CNI) provider, you cannot attach a Linux bridge or bonding to the default interface of a host because of a change in the host network topology of OVN-Kubernetes. As a workaround, you can use a secondary network interface connected to your host, or switch to the OpenShift SDN default CNI provider.

19.3.2. Creating an interface on nodes

Create an interface on nodes in the cluster by applying a NodeNetworkConfigurationPolicy manifest to the cluster. The manifest details the requested configuration for the interface.

By default, the manifest applies to all nodes in the cluster. To add the interface to specific nodes, add the spec: nodeSelector parameter and the appropriate <key>:<value> for your node selector.

Procedure

  1. Create the NodeNetworkConfigurationPolicy manifest. The following example configures a Linux bridge on all worker nodes:

    apiVersion: nmstate.io/v1beta1
    kind: NodeNetworkConfigurationPolicy
    metadata:
      name: <br1-eth1-policy> 1
    spec:
      nodeSelector: 2
        node-role.kubernetes.io/worker: "" 3
      desiredState:
        interfaces:
          - name: br1
            description: Linux bridge with eth1 as a port 4
            type: linux-bridge
            state: up
            ipv4:
              dhcp: true
              enabled: true
            bridge:
              options:
                stp:
                  enabled: false
              port:
                - name: eth1
    1
    Name of the policy.
    2
    Optional: If you do not include the nodeSelector parameter, the policy applies to all nodes in the cluster.
    3
    This example uses the node-role.kubernetes.io/worker: "" node selector to select all worker nodes in the cluster.
    4
    Optional: Human-readable description for the interface.
  2. Create the node network policy:

    $ oc apply -f <br1-eth1-policy.yaml> 1
    1
    File name of the node network configuration policy manifest.

Additional resources

19.3.3. Confirming node network policy updates on nodes

A NodeNetworkConfigurationPolicy manifest describes your requested network configuration for nodes in the cluster. The node network policy includes your requested network configuration and the status of execution of the policy on the cluster as a whole.

When you apply a node network policy, a NodeNetworkConfigurationEnactment object is created for every node in the cluster. The node network configuration enactment is a read-only object that represents the status of execution of the policy on that node. If the policy fails to be applied on the node, the enactment for that node includes a traceback for troubleshooting.

Procedure

  1. To confirm that a policy has been applied to the cluster, list the policies and their status:

    $ oc get nncp
  2. Optional: If a policy is taking longer than expected to successfully configure, you can inspect the requested state and status conditions of a particular policy:

    $ oc get nncp <policy> -o yaml
  3. Optional: If a policy is taking longer than expected to successfully configure on all nodes, you can list the status of the enactments on the cluster:

    $ oc get nnce
  4. Optional: To view the configuration of a particular enactment, including any error reporting for a failed configuration:

    $ oc get nnce <node>.<policy> -o yaml

19.3.4. Removing an interface from nodes

You can remove an interface from one or more nodes in the cluster by editing the NodeNetworkConfigurationPolicy object and setting the state of the interface to absent.

Removing an interface from a node does not automatically restore the node network configuration to a previous state. If you want to restore the previous state, you will need to define that node network configuration in the policy.

If you remove a bridge or bonding interface, any node NICs in the cluster that were previously attached or subordinate to that bridge or bonding interface are placed in a down state and become unreachable. To avoid losing connectivity, configure the node NIC in the same policy so that it has a status of up and either DHCP or a static IP address.

Note

Deleting the node network policy that added an interface does not change the configuration of the policy on the node. Although a NodeNetworkConfigurationPolicy is an object in the cluster, it only represents the requested configuration.
Similarly, removing an interface does not delete the policy.

Procedure

  1. Update the NodeNetworkConfigurationPolicy manifest used to create the interface. The following example removes a Linux bridge and configures the eth1 NIC with DHCP to avoid losing connectivity:

    apiVersion: nmstate.io/v1beta1
    kind: NodeNetworkConfigurationPolicy
    metadata:
      name: <br1-eth1-policy> 1
    spec:
      nodeSelector: 2
        node-role.kubernetes.io/worker: "" 3
      desiredState:
        interfaces:
        - name: br1
          type: linux-bridge
          state: absent 4
        - name: eth1 5
          type: ethernet 6
          state: up 7
          ipv4:
            dhcp: true 8
            enabled: true 9
    1
    Name of the policy.
    2
    Optional: If you do not include the nodeSelector parameter, the policy applies to all nodes in the cluster.
    3
    This example uses the node-role.kubernetes.io/worker: "" node selector to select all worker nodes in the cluster.
    4
    Changing the state to absent removes the interface.
    5
    The name of the interface that is to be unattached from the bridge interface.
    6
    The type of interface. This example creates an Ethernet networking interface.
    7
    The requested state for the interface.
    8
    Optional: If you do not use dhcp, you can either set a static IP or leave the interface without an IP address.
    9
    Enables ipv4 in this example.
  2. Update the policy on the node and remove the interface:

    $ oc apply -f <br1-eth1-policy.yaml> 1
    1
    File name of the policy manifest.

19.3.5. Example policy configurations for different interfaces

19.3.5.1. Example: Linux bridge interface node network configuration policy

Create a Linux bridge interface on nodes in the cluster by applying a NodeNetworkConfigurationPolicy manifest to the cluster.

The following YAML file is an example of a manifest for a Linux bridge interface. It includes samples values that you must replace with your own information.

apiVersion: nmstate.io/v1beta1
kind: NodeNetworkConfigurationPolicy
metadata:
  name: br1-eth1-policy 1
spec:
  nodeSelector: 2
    kubernetes.io/hostname: <node01> 3
  desiredState:
    interfaces:
      - name: br1 4
        description: Linux bridge with eth1 as a port 5
        type: linux-bridge 6
        state: up 7
        ipv4:
          dhcp: true 8
          enabled: true 9
        bridge:
          options:
            stp:
              enabled: false 10
          port:
            - name: eth1 11
1
Name of the policy.
2
Optional: If you do not include the nodeSelector parameter, the policy applies to all nodes in the cluster.
3
This example uses a hostname node selector.
4
Name of the interface.
5
Optional: Human-readable description of the interface.
6
The type of interface. This example creates a bridge.
7
The requested state for the interface after creation.
8
Optional: If you do not use dhcp, you can either set a static IP or leave the interface without an IP address.
9
Enables ipv4 in this example.
10
Disables stp in this example.
11
The node NIC to which the bridge attaches.

19.3.5.2. Example: VLAN interface node network configuration policy

Create a VLAN interface on nodes in the cluster by applying a NodeNetworkConfigurationPolicy manifest to the cluster.

The following YAML file is an example of a manifest for a VLAN interface. It includes samples values that you must replace with your own information.

apiVersion: nmstate.io/v1beta1
kind: NodeNetworkConfigurationPolicy
metadata:
  name: vlan-eth1-policy 1
spec:
  nodeSelector: 2
    kubernetes.io/hostname: <node01> 3
  desiredState:
    interfaces:
    - name: eth1.102 4
      description: VLAN using eth1 5
      type: vlan 6
      state: up 7
      vlan:
        base-iface: eth1 8
        id: 102 9
1
Name of the policy.
2
Optional: If you do not include the nodeSelector parameter, the policy applies to all nodes in the cluster.
3
This example uses a hostname node selector.
4
Name of the interface.
5
Optional: Human-readable description of the interface.
6
The type of interface. This example creates a VLAN.
7
The requested state for the interface after creation.
8
The node NIC to which the VLAN is attached.
9
The VLAN tag.

19.3.5.3. Example: Bond interface node network configuration policy

Create a bond interface on nodes in the cluster by applying a NodeNetworkConfigurationPolicy manifest to the cluster.

Note

OpenShift Container Platform only supports the following bond modes:

  • mode=1 active-backup
  • mode=2 balance-xor
  • mode=4 802.3ad
  • mode=5 balance-tlb
  • mode=6 balance-alb

The following YAML file is an example of a manifest for a bond interface. It includes samples values that you must replace with your own information.

apiVersion: nmstate.io/v1beta1
kind: NodeNetworkConfigurationPolicy
metadata:
  name: bond0-eth1-eth2-policy 1
spec:
  nodeSelector: 2
    kubernetes.io/hostname: <node01> 3
  desiredState:
    interfaces:
    - name: bond0 4
      description: Bond enslaving eth1 and eth2 5
      type: bond 6
      state: up 7
      ipv4:
        dhcp: true 8
        enabled: true 9
      link-aggregation:
        mode: active-backup 10
        options:
          miimon: '140' 11
        slaves: 12
        - eth1
        - eth2
      mtu: 1450 13
1
Name of the policy.
2
Optional: If you do not include the nodeSelector parameter, the policy applies to all nodes in the cluster.
3
This example uses a hostname node selector.
4
Name of the interface.
5
Optional: Human-readable description of the interface.
6
The type of interface. This example creates a bond.
7
The requested state for the interface after creation.
8
Optional: If you do not use dhcp, you can either set a static IP or leave the interface without an IP address.
9
Enables ipv4 in this example.
10
The driver mode for the bond. This example uses an active backup mode.
11
Optional: This example uses miimon to inspect the bond link every 140ms.
12
The subordinate node NICs in the bond.
13
Optional: The maximum transmission unit (MTU) for the bond. If not specified, this value is set to 1500 by default.

19.3.5.4. Example: Ethernet interface node network configuration policy

Configure an Ethernet interface on nodes in the cluster by applying a NodeNetworkConfigurationPolicy manifest to the cluster.

The following YAML file is an example of a manifest for an Ethernet interface. It includes sample values that you must replace with your own information.

apiVersion: nmstate.io/v1beta1
kind: NodeNetworkConfigurationPolicy
metadata:
  name: eth1-policy 1
spec:
  nodeSelector: 2
    kubernetes.io/hostname: <node01> 3
  desiredState:
    interfaces:
    - name: eth1 4
      description: Configuring eth1 on node01 5
      type: ethernet 6
      state: up 7
      ipv4:
        dhcp: true 8
        enabled: true 9
1
Name of the policy.
2
Optional: If you do not include the nodeSelector parameter, the policy applies to all nodes in the cluster.
3
This example uses a hostname node selector.
4
Name of the interface.
5
Optional: Human-readable description of the interface.
6
The type of interface. This example creates an Ethernet networking interface.
7
The requested state for the interface after creation.
8
Optional: If you do not use dhcp, you can either set a static IP or leave the interface without an IP address.
9
Enables ipv4 in this example.

19.3.5.5. Example: Multiple interfaces in the same node network configuration policy

You can create multiple interfaces in the same node network configuration policy. These interfaces can reference each other, allowing you to build and deploy a network configuration by using a single policy manifest.

The following example snippet creates a bond that is named bond10 across two NICs and a Linux bridge that is named br1 that connects to the bond.

...
    interfaces:
    - name: bond10
      description: Bonding eth2 and eth3 for Linux bridge
      type: bond
      state: up
      link-aggregation:
        slaves:
        - eth2
        - eth3
    - name: br1
      description: Linux bridge on bond
      type: linux-bridge
      state: up
      bridge:
        port:
        - name: bond10
...

19.3.6. Examples: IP management

The following example configuration snippets demonstrate different methods of IP management.

These examples use the ethernet interface type to simplify the example while showing the related context in the policy configuration. These IP management examples can be used with the other interface types.

19.3.6.1. Static

The following snippet statically configures an IP address on the Ethernet interface:

...
    interfaces:
    - name: eth1
      description: static IP on eth1
      type: ethernet
      state: up
      ipv4:
        dhcp: false
        address:
        - ip: 192.168.122.250 1
          prefix-length: 24
        enabled: true
...
1
Replace this value with the static IP address for the interface.

19.3.6.2. No IP address

The following snippet ensures that the interface has no IP address:

...
    interfaces:
    - name: eth1
      description: No IP on eth1
      type: ethernet
      state: up
      ipv4:
        enabled: false
...

19.3.6.3. Dynamic host configuration

The following snippet configures an Ethernet interface that uses a dynamic IP address, gateway address, and DNS:

...
    interfaces:
    - name: eth1
      description: DHCP on eth1
      type: ethernet
      state: up
      ipv4:
        dhcp: true
        enabled: true
...

The following snippet configures an Ethernet interface that uses a dynamic IP address but does not use a dynamic gateway address or DNS:

...
    interfaces:
    - name: eth1
      description: DHCP without gateway or DNS on eth1
      type: ethernet
      state: up
      ipv4:
        dhcp: true
        auto-gateway: false
        auto-dns: false
        enabled: true
...

19.3.6.4. DNS

The following snippet sets DNS configuration on the host.

...
    interfaces:
       ...
    dns-resolver:
      config:
        search:
        - example.com
        - example.org
        server:
        - 8.8.8.8
...

19.3.6.5. Static routing

The following snippet configures a static route and a static IP on interface eth1.

...
    interfaces:
    - name: eth1
      description: Static routing on eth1
      type: ethernet
      state: up
      ipv4:
        dhcp: false
        address:
        - ip: 192.0.2.251 1
          prefix-length: 24
        enabled: true
    routes:
      config:
      - destination: 198.51.100.0/24
        metric: 150
        next-hop-address: 192.0.2.1 2
        next-hop-interface: eth1
        table-id: 254
...
1
The static IP address for the Ethernet interface.
2
Next hop address for the node traffic. This must be in the same subnet as the IP address set for the Ethernet interface.

19.4. Troubleshooting node network configuration

If the node network configuration encounters an issue, the policy is automatically rolled back and the enactments report failure. This includes issues such as:

  • The configuration fails to be applied on the host.
  • The host loses connection to the default gateway.
  • The host loses connection to the API server.

19.4.1. Troubleshooting an incorrect node network configuration policy configuration

You can apply changes to the node network configuration across your entire cluster by applying a node network configuration policy. If you apply an incorrect configuration, you can use the following example to troubleshoot and correct the failed node network policy.

In this example, a Linux bridge policy is applied to an example cluster that has three control plane nodes (master) and three compute (worker) nodes. The policy fails to be applied because it references an incorrect interface. To find the error, investigate the available NMState resources. You can then update the policy with the correct configuration.

Procedure

  1. Create a policy and apply it to your cluster. The following example creates a simple bridge on the ens01 interface:

    apiVersion: nmstate.io/v1beta1
    kind: NodeNetworkConfigurationPolicy
    metadata:
      name: ens01-bridge-testfail
    spec:
      desiredState:
        interfaces:
          - name: br1
            description: Linux bridge with the wrong port
            type: linux-bridge
            state: up
            ipv4:
              dhcp: true
              enabled: true
            bridge:
              options:
                stp:
                  enabled: false
              port:
                - name: ens01
    $ oc apply -f ens01-bridge-testfail.yaml

    Example output

    nodenetworkconfigurationpolicy.nmstate.io/ens01-bridge-testfail created

  2. Verify the status of the policy by running the following command:

    $ oc get nncp

    The output shows that the policy failed:

    Example output

    NAME                    STATUS
    ens01-bridge-testfail   FailedToConfigure

    However, the policy status alone does not indicate if it failed on all nodes or a subset of nodes.

  3. List the node network configuration enactments to see if the policy was successful on any of the nodes. If the policy failed for only a subset of nodes, it suggests that the problem is with a specific node configuration. If the policy failed on all nodes, it suggests that the problem is with the policy.

    $ oc get nnce

    The output shows that the policy failed on all nodes:

    Example output

    NAME                                   STATUS
    control-plane-1.ens01-bridge-testfail        FailedToConfigure
    control-plane-2.ens01-bridge-testfail        FailedToConfigure
    control-plane-3.ens01-bridge-testfail        FailedToConfigure
    compute-1.ens01-bridge-testfail              FailedToConfigure
    compute-2.ens01-bridge-testfail              FailedToConfigure
    compute-3.ens01-bridge-testfail              FailedToConfigure

  4. View one of the failed enactments and look at the traceback. The following command uses the output tool jsonpath to filter the output:

    $ oc get nnce compute-1.ens01-bridge-testfail -o jsonpath='{.status.conditions[?(@.type=="Failing")].message}'

    This command returns a large traceback that has been edited for brevity:

    Example output

    error reconciling NodeNetworkConfigurationPolicy at desired state apply: , failed to execute nmstatectl set --no-commit --timeout 480: 'exit status 1' ''
    ...
    libnmstate.error.NmstateVerificationError:
    desired
    =======
    ---
    name: br1
    type: linux-bridge
    state: up
    bridge:
      options:
        group-forward-mask: 0
        mac-ageing-time: 300
        multicast-snooping: true
        stp:
          enabled: false
          forward-delay: 15
          hello-time: 2
          max-age: 20
          priority: 32768
      port:
      - name: ens01
    description: Linux bridge with the wrong port
    ipv4:
      address: []
      auto-dns: true
      auto-gateway: true
      auto-routes: true
      dhcp: true
      enabled: true
    ipv6:
      enabled: false
    mac-address: 01-23-45-67-89-AB
    mtu: 1500
    
    current
    =======
    ---
    name: br1
    type: linux-bridge
    state: up
    bridge:
      options:
        group-forward-mask: 0
        mac-ageing-time: 300
        multicast-snooping: true
        stp:
          enabled: false
          forward-delay: 15
          hello-time: 2
          max-age: 20
          priority: 32768
      port: []
    description: Linux bridge with the wrong port
    ipv4:
      address: []
      auto-dns: true
      auto-gateway: true
      auto-routes: true
      dhcp: true
      enabled: true
    ipv6:
      enabled: false
    mac-address: 01-23-45-67-89-AB
    mtu: 1500
    
    difference
    ==========
    --- desired
    +++ current
    @@ -13,8 +13,7 @@
           hello-time: 2
           max-age: 20
           priority: 32768
    -  port:
    -  - name: ens01
    +  port: []
     description: Linux bridge with the wrong port
     ipv4:
       address: []
      line 651, in _assert_interfaces_equal\n    current_state.interfaces[ifname],\nlibnmstate.error.NmstateVerificationError:

    The NmstateVerificationError lists the desired policy configuration, the current configuration of the policy on the node, and the difference highlighting the parameters that do not match. In this example, the port is included in the difference, which suggests that the problem is the port configuration in the policy.

  5. To ensure that the policy is configured properly, view the network configuration for one or all of the nodes by requesting the NodeNetworkState object. The following command returns the network configuration for the control-plane-1 node:

    $ oc get nns control-plane-1 -o yaml

    The output shows that the interface name on the nodes is ens1 but the failed policy incorrectly uses ens01:

    Example output

       - ipv4:
     ...
          name: ens1
          state: up
          type: ethernet

  6. Correct the error by editing the existing policy:

    $ oc edit nncp ens01-bridge-testfail
    ...
              port:
                - name: ens1

    Save the policy to apply the correction.

  7. Check the status of the policy to ensure it updated successfully:

    $ oc get nncp

    Example output

    NAME                    STATUS
    ens01-bridge-testfail   SuccessfullyConfigured

The updated policy is successfully configured on all nodes in the cluster.