Writing a CNI plugin from scratch

CNI (Container Network Interface) plugins are the cornerstone of Kubernetes networking.
CNI is the standardized way used by Kubernetes to expose network devices to pods,
responsible for pod to pod communication across physical nodes in your cluster.

During this talk we will:
- Explore the details of the CNI plugin interface
- Understand how it is used with Kubernetes
- Provide a detailed walkthrough of a simple CNI plugin from scratch

Attendees in this talk will gain insight into the process of creating a CNI plugin and get familiar with networking decisions required for having their pods connected and reachable from within the cluster and the internet.

All the examples in the slides are also available here: https://github.com/eranyanay/cni-from-scratch

Transcript

1. Writing a CNI plugin from
   scratch
   Eran Yanay
   

   View Slide

2. Eran Yanay
   Twistlock
   R&D Team Lead
   eran@twistlock.com
   Who am I
   

   View Slide

3. kubectl apply -f
   "https://cloud.weave.works/k8s/net?k8s-version
   =$(kubectl version | base64 | tr -d '\n')"
   

   View Slide

4. Writing a CNI (Container Network Interface) plugin from scratch, using only bash
   ● What is CNI?
   ● How CNI plugin works?
   ● What a CNI plugin is made of?
   ● How a CNI plugin is being used in K8s?
   ● How a CNI plugin is executed?
   ● Anatomy of pod networking
   ● Live demo
   Agenda
   

   View Slide

5. What is CNI?
   ● CNI stands for Container Networking Interface
   ● An interface between container runtime and the network implementation
   ● Configures the network interfaces and routes
   ● Concerns itself only with network connectivity
   ● https://github.com/containernetworking/cni/blob/spec-v0.4.0/SPEC.md
   

   View Slide

6. How CNI plugin works (in k8s)
   ● A CNI binary
   Handles connectivity - configures the network interface of the pod
   ● A daemon
   Handles reachability - manages routings across the cluster
   

   View Slide

7. # cat /etc/cni/net.d/10-my-cni-demo.conf
   {
   "cniVersion": "0.3.1",
   "name": "my-cni-demo",
   "type": "my-cni-demo",
   "podcidr": "10.240.0.0/24",
   }
   What a CNI plugin is made of?
   # cat /opt/cni/bin/my-cni-demo
   case $CNI_COMMAND in
   ADD)
   # Configure networking for a new container
   ;;
   DEL)
   # Cleanup when container is stopped
   ;;
   GET)
   ;;
   VERSION)
   # Get the plugin version
   ;;
   esac
   

   View Slide

8. The weave example..
   kind: DaemonSet
   spec:
   containers:
   - name: weave
   command:
   - /home/weave/launch.sh
   image: 'docker.io/weaveworks/weave-kube:2.5.1'
   volumeMounts:
   - name: weavedb
   mountPath: /weavedb
   - name: cni-bin
   mountPath: /host/opt
   - name: cni-bin2
   mountPath: /host/home
   - name: cni-conf
   mountPath: /host/etc
   hostNetwork: true
   

   View Slide

9. The weave example..
   $ cat /home/weave/launch.sh
   # ... previous non related code ...
   # Install CNI plugin binary to typical CNI bin location
   # with fall-back to CNI directory used by kube-up on GCI OS
   if ! mkdir -p $HOST_ROOT/opt/cni/bin ; then
   if mkdir -p $HOST_ROOT/home/kubernetes/bin ; then
   export WEAVE_CNI_PLUGIN_DIR=$HOST_ROOT/home/kubernetes/bin
   else
   echo "Failed to install the Weave CNI plugin" >&2
   exit 1
   fi
   fi
   mkdir -p $HOST_ROOT/etc/cni/net.d
   export HOST_ROOT
   /home/weave/weave --local setup-cni
   

   View Slide

10. How a CNI plugin is being used in K8s?
    API Server Kubelet
    Schedule a pod
    Host network
    namespace
    br
    

    View Slide

11. How a CNI plugin is being used in K8s?
    Host network
    namespace
    API Server Kubelet
    Create pod network ns
    Pod network
    namespace br
    lo
    

    View Slide

12. How a CNI plugin is being used in K8s?
    Host network
    namespace
    API Server Kubelet
    CNI_COMMAND: ADD
    Pod network
    namespace
    CNI
    br
    lo
    

    View Slide

13. How a CNI plugin is being used in K8s?
    Host network
    namespace
    API Server Kubelet
    Pod network
    namespace
    CNI
    eth0
    ● Create eth0
    ● Allocate IP
    ● Define routes
    br
    lo
    

    View Slide

14. How a CNI plugin is executed?
    Container
    info
    CNI
    config
    CNI
    env vars
    stdin
    

    View Slide

15. How a CNI plugin is executed?
    Container
    info
    CNI
    config
    CNI
    env vars
    stdin
    CNI_COMMAND=ADD
    CNI_CONTAINERID=b784318..
    CNI_NETNS=/proc/12345/ns/net
    CNI_IFNAME=eth0
    

    View Slide

16. How a CNI plugin is executed?
    Container
    info
    CNI
    config
    CNI
    env vars
    stdin
    {
    "cniVersion" : "0.3.1",
    "name": "my-cni-demo" ,
    "type": "my-cni-demo" ,
    "podcidr" : "10.240.0.0/24"
    }
    CNI_COMMAND=ADD
    CNI_CONTAINERID=b784318..
    CNI_NETNS=/proc/12345/ns/net
    CNI_IFNAME=eth0
    

    View Slide

17. {
    "cniVersion" : "0.3.1",
    "interfaces" : [
    {
    "name": "eth0",
    "mac": "ce:60:4c:b9:3a:06" ,
    "sandbox" : "/proc/15116/ns/net"
    }
    ],
    "ips": [
    {
    "version" : "4",
    "address" : "10.240.0.6/24" ,
    "gateway" : "10.240.0.1" ,
    "interface" : 0
    }
    ]
    }
    How a CNI plugin is executed?
    Container
    info
    CNI
    config
    CNI
    env vars
    stdin
    

    View Slide

18. Anatomy of pod networking
    Linux host
    pod1
    

    View Slide

19. Anatomy of pod networking
    Linux host
    bridge
    pod1
    

    View Slide

20. Anatomy of pod networking
    Linux host
    pod1
    veth
    bridge
    veth
    

    View Slide

21. Anatomy of pod networking
    Linux host
    bridge
    pod1
    eth0
    veth
    

    View Slide

22. Anatomy of pod networking
    Linux host
    pod1
    eth0
    eth0
    veth
    pod2 pod3 pod4
    veth
    veth
    veth
    bridge
    eth0 eth0 eth0
    

    View Slide

23. enp0s9
    enp0s9
    node1
    node2
    # cat /etc/cni/net.d/10-my-cni-demo.conf
    {
    "cniVersion": "0.3.1",
    "name": "my-cni-demo",
    "type": "my-cni-demo",
    "podcidr": "10.240.0.0/24"
    ,
    }
    # cat /opt/cni/bin/my-cni-demo
    case $CNI_COMMAND in
    ADD)
    ;; # Configure networking
    DEL)
    ;; # Cleanup
    GET)
    ;;
    VERSION)
    ;; # Print plugin version
    esac
    10.10.10.10/24
    10.240.0.0/24
    10.10.10.11/24
    10.240.1.0/24
    pod
    

    View Slide

24. CNI_CONTAINERID=b552f9…
    CNI_IFNAME=eth0
    CNI_COMMAND=ADD
    CNI_NETNS=/proc/6137/ns/net
    case $CNI_COMMAND in
    ADD)
    podcidr=$(cat /dev/stdin | jq -r ".podcidr") # 10.240.0.0/24
    podcidr_gw=$(echo $podcidr | sed "s:0/24:1:g") # 10.240.0.1
    ;;
    enp0s9
    node1
    10.10.10.10/24
    10.240.0.0/24
    pod
    bridge
    Container info
    

    View Slide

25. CNI_CONTAINERID=b552f9…
    CNI_IFNAME=eth0
    CNI_COMMAND=ADD
    CNI_NETNS=/proc/6137/ns/net
    case $CNI_COMMAND in
    ADD)
    podcidr=$(cat /dev/stdin | jq -r ".podcidr") # 10.240.0.0/24
    podcidr_gw=$(echo $podcidr | sed "s:0/24:1:g") # 10.240.0.1
    brctl addbr cni0 # create a new bridge (if doesnt exist), cni0
    ip link set cni0 up
    ip addr add "${podcidr_gw}/24" dev cni0 # assign 10.240.0.1/24 to cni0
    ;; enp0s9
    node1
    10.10.10.10/24
    10.240.0.0/24
    pod
    bridge
    Container info
    

    View Slide

26. CNI_CONTAINERID=b552f9…
    CNI_IFNAME=eth0
    CNI_COMMAND=ADD
    CNI_NETNS=/proc/6137/ns/net
    enp0s9
    node1
    10.10.10.10/24
    10.240.0.0/24
    pod
    bridge
    eth0
    veth1
    case $CNI_COMMAND in
    ADD)
    podcidr=$(cat /dev/stdin | jq -r ".podcidr") # 10.240.0.0/24
    podcidr_gw=$(echo $podcidr | sed "s:0/24:1:g") # 10.240.0.1
    brctl addbr cni0 # create a new bridge (if doesnt exist), cni0
    ip link set cni0 up
    ip addr add "${podcidr_gw}/24" dev cni0 # assign 10.240.0.1/24 to cni0
    host_ifname="veth$n" # n=1,2,3...
    ip link add $CNI_IFNAME type veth peer name $host_ifname
    ip link set $host_ifname up
    ;;
    Container info
    

    View Slide

27. CNI_CONTAINERID=b552f9…
    CNI_IFNAME=eth0
    CNI_COMMAND=ADD
    CNI_NETNS=/proc/6137/ns/net
    enp0s9
    node1
    10.10.10.10/24
    10.240.0.0/24
    pod
    bridge
    eth0
    case $CNI_COMMAND in
    ADD)
    podcidr=$(cat /dev/stdin | jq -r ".podcidr") # 10.240.0.0/24
    podcidr_gw=$(echo $podcidr | sed "s:0/24:1:g") # 10.240.0.1
    brctl addbr cni0 # create a new bridge (if doesnt exist), cni0
    ip link set cni0 up
    ip addr add "${podcidr_gw}/24" dev cni0 # assign 10.240.0.1/24 to cni0
    host_ifname="veth$n" # n=1,2,3...
    ip link add $CNI_IFNAME type veth peer name $host_ifname
    ip link set $host_ifname up
    ip link set $host_ifname master cni0 # connect veth1 to bridge
    ln -sfT $CNI_NETNS /var/run/netns/$CNI_CONTAINERID
    ip link set $CNI_IFNAME netns $CNI_CONTAINERID # move eth0 to pod ns
    ;;
    Container info
    veth1
    

    View Slide

28. CNI_CONTAINERID=b552f9…
    CNI_IFNAME=eth0
    CNI_COMMAND=ADD
    CNI_NETNS=/proc/6137/ns/net
    enp0s9
    node1
    10.10.10.10/24
    10.240.0.0/24
    pod
    bridge
    10.240.0.2
    case $CNI_COMMAND in
    ADD)
    podcidr=$(cat /dev/stdin | jq -r ".podcidr") # 10.240.0.0/24
    podcidr_gw=$(echo $podcidr | sed "s:0/24:1:g") # 10.240.0.1
    brctl addbr cni0 # create a new bridge (if doesnt exist), cni0
    ip link set cni0 up
    ip addr add "${podcidr_gw}/24" dev cni0 # assign 10.240.0.1/24 to cni0
    host_ifname="veth$n" # n=1,2,3...
    ip link add $CNI_IFNAME type veth peer name $host_ifname
    ip link set $host_ifname up
    ip link set $host_ifname master cni0 # connect veth1 to bridge
    ln -sfT $CNI_NETNS /var/run/netns/$CNI_CONTAINERID
    ip link set $CNI_IFNAME netns $CNI_CONTAINERID # move eth0 to pod ns
    # calculate $ip
    ip netns exec $CNI_CONTAINERID ip link set $CNI_IFNAME up
    ip netns exec $CNI_CONTAINERID ip addr add $ip/24 dev $CNI_IFNAME
    ip netns exec $CNI_CONTAINERID ip route add default via $podcidr_gw
    dev $CNI_IFNAME
    Container info
    eth0
    veth1
    

    View Slide

29. CNI_CONTAINERID=b552f9…
    CNI_IFNAME=eth0
    CNI_COMMAND=ADD
    CNI_NETNS=/proc/6137/ns/net
    enp0s9
    node1
    10.10.10.10/24
    10.240.0.0/24
    pod
    bridge
    10.240.0.2
    case $CNI_COMMAND in
    ADD)
    podcidr=$(cat /dev/stdin | jq -r ".podcidr") # 10.240.0.0/24
    podcidr_gw=$(echo $podcidr | sed "s:0/24:1:g") # 10.240.0.1
    brctl addbr cni0 # create a new bridge (if doesnt exist), cni0
    ip link set cni0 up
    ip addr add "${podcidr_gw}/24" dev cni0 # assign 10.240.0.1/24 to cni0
    host_ifname="veth$n" # n=1,2,3...
    ip link add $CNI_IFNAME type veth peer name $host_ifname
    ip link set $host_ifname up
    ip link set $host_ifname master cni0 # connect veth1 to bridge
    ln -sfT $CNI_NETNS /var/run/netns/$CNI_CONTAINERID
    ip link set $CNI_IFNAME netns $CNI_CONTAINERID # move eth0 to pod ns
    # calculate $ip
    ip netns exec $CNI_CONTAINERID ip link set $CNI_IFNAME up
    ip netns exec $CNI_CONTAINERID ip addr add $ip/24 dev $CNI_IFNAME
    ip netns exec $CNI_CONTAINERID ip route add default via $podcidr_gw
    dev $CNI_IFNAME
    Container info
    eth0
    veth1
    if [ -f /tmp/last_allocated_ip ]; then
    n=`cat /tmp/last_allocated_ip`
    else
    n=1
    fi
    ip=$(echo $podcidr | sed "s:0/24:$((
    $n+1)):g")
    echo $(($n+1)) > /tmp/last_allocated_ip
    

    View Slide

30. Lets see a demo!
    

    View Slide

31. Thank you!
    Full implementation is available at https://github.com/eranyanay/cni-from-scratch
    

    View Slide