Deploy SMARTER Demo using K3s

K3s is a lightweight Kubernetes distribution, built to work well in embedded environments. It is optimized for Arm.

K3s setup

You will use a single node K3s cluster setup. To download K3s and set it up, run the following commands:

    

        
        
            export INSTALL_K3S_EXEC="server --disable traefik --disable metrics-server --disable coredns --disable local-storage --flannel-backend=none --cluster-dns 169.254.0.2 --container-runtime-endpoint=unix://var/run/containerd/containerd.sock"
        
    

    

        
        
            curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC=$INSTALL_K3S_EXEC sh -s -
        
    

This can take a few minutes to complete.

Make sure K3s is running. A snippet of the expected output is shown below:

    

        
        
            systemctl status k3s
        
    

    

        
        * k3s.service - Lightweight Kubernetes
     Loaded: loaded (/etc/systemd/system/k3s.service; enabled; preset: disabled)
     Active: active (running) since <date and time>
       Docs: https://k3s.io
   Main PID: 2069 (k3s-server)
      Tasks: 13
     Memory: 448.0M
     CGroup: /system.slice/k3s.service
             `-2069 "/usr/local/bin/k3s agent"

        
    

For things to work properly, you need to make K3s aware of the hybrid-runtime by configuring this with containerd. You can find the config file with the K3s example YAML files in GitHub.

Download the K3s demo example YAML files:

    

        
        
            wget https://github.com/smarter-project/hybrid-runtime/releases/download/v1.5/example.tar.gz
        
    

Extract the files:

    

        
        
            tar -xvf example.tar.gz
        
    

Create a containerd directory under /etc and copy the config file to there:

    

        
        
            mkdir /etc/containerd
mv example/config.toml /etc/containerd/
        
    

Restart containerd, and make sure that it’s running:

    

        
        
            systemctl restart containerd
systemctl status containerd
        
    

If you run the kubectl command below, you will see that the node is not ready:

    

        
        
            kubectl get nodes
        
    

    

        
        NAME     STATUS     ROLES                  AGE   VERSION
narsil   NotReady   control-plane,master   18m   v1.29.6+k3s2

        
    

To fix this, you need to apply a Container Network Interface (CNI). Run the command to use the smarter_cni, and label the node as follows:

    

        
        
            kubectl apply -f example/smarter_cni.yaml
kubectl label node narsil smarter.cni=deploy
        
    

Re-run the kubectl command:

    

        
        
            kubectl get nodes
        
    

This time, you should be able to see that the node is ready.

    

        
        root@narsil:~#
NAME     STATUS   ROLES                  AGE   VERSION
narsil   Ready    control-plane,master   24m   v1.29.6+k3s2

        
    

K3s demo

Deploy SMARTER camera demo

The smarter camera demo is used as an example to show the capabilities of K3s.

First, you need to set a runtimeClass in K3s. It allows you to select the container runtime we want to use:

    

        
        
            kubectl apply -f example/runtime_class.yaml
        
    

Once this is done, we can run the smarter demo:

    

        
        
            kubectl apply -f example/test_hybrid.yaml
        
    

The test_hybrid.yaml file contains the following:

    

        
        kind: Pod
apiVersion: v1
metadata:
  name: example3
  labels:
    k3s-app: example3
spec:
    runtimeClassName: hybrid
    containers:
      - name: example-hybrid-pod3
        image: ghcr.io/smarter-project/smart-camera-hybrid-application/hybrid_app_imx8mp:latest
        imagePullPolicy: IfNotPresent

        
    

There are two ways to check that the firmware is running:

1. Run and observe the output (a pod with the name example3 should be running).

    

        
        
            kubectl get pods -A
        
    

    

        
        root@narsil:~# kubectl get pods -A
NAMESPACE     NAME                READY   STATUS    RESTARTS       AGE
default       example3            1/1     Running   0              6m57s
kube-system   smarter-cni-wplzn   1/1     Running   3 (141m ago)   4h29m

        
    

2. Check the Cortex-M Console. If there is any output here, the firmware is running.

Stop the demo

To stop the demo, run the command shown below (the termination process can take a few minutes):

    

        
        
            kubectl delete pod example3 --grace-period=0 --force
        
    

To make sure the pod was terminated, check the following:

1. Go to the Cortex-M Console and check that there are no new outputs:

Figure 1. Cortex-M output

2. Check that the firmware is offline:

    

        
        
            cat /sys/class/remoteproc/remoteproc0/state
        
    

The output should be as follows:

    

        
        offline

        
    

3. Make sure the created pod above was deleted:

    

        
        
            kubectl get pods -A
        
    

    

        
        NAMESPACE     NAME                READY   STATUS    RESTARTS       AGE
kube-system   smarter-cni-wplzn   1/1     Running   3 (143m ago)   4h31m

        
    

4. Make sure all the container resources were deleted. The command below should give no output:

    

        
        
            ls /var/lib/hybrid-runtime/
        
    

Summary

The hybrid-runtime can be used to improve the experience for systems with multiple IPs on a single SoC. You now know how to use cloud tools such as K3s and containerd to deploy and run workloads on hybrid systems using the hybrid-runtime.

If you have an Arm Linux host, you can run the hello world example by following the instructions in the next section.