In this day's live online demonstration, Yang Xijie from Edgenesis demonstrated how to use Shifu Cloud to access three physical devices and develop applications based on them. Let's review the process together!

Creating the cluster & Installing shifu​

Before using Shifu Cloud, we need to make sure we have already installed Shifu on our computer.

# Start the k8s cluster locally using kind
$ sudo kind create cluster --image="kindest/node:v1.24.0"

# Clone the shifu repository and install it
$ git clone https://github.com/Edgenesis/shifu.git
$ cd shifu
$ sudo kubectl apply -f pkg/k8s/crd/install/shifu_install.yml

If you would like to know how to install Shifu and test it locally, you may want to check out download and install and local installing testing.

Connecting to thermometer and LED​

The devices we want to connect to are respectively an RS485 thermohygrometer and an RS485 LED display. The thermohygrometer is connected to the host PC (computer) via a serial server, and the LED display is connected to the host PC via an RS485 to USB chip. Here we don't want to bother you with any detail, and after the host computer opens the HTTP service:

• You may visit localhost:23330/temperature to get the temperature from the thermohygrometer
• You may visit localhost:23330/humidity to get the humidity from the thermohygrometer
• You may visit localhost:23331/setfloat\?value=123.4, fill value with the number you need to display on the LED

curl localhost:23330/temperature
curl localhost:23330/humidity
curl localhost:23331/setfloat\?value=123.4

Next we're going to integrate the two devices into Shifu, which means that the two pjysical devices (edgeDevices) are converted to digital twins (deviceShifus) in the k8s cluster.

Generating configuration file with one click​

Shifu Cloud can easily generate configuration files for deviceShifu.

After logging in, click "All projects" to add devices. Both devices use HTTP protocol, so choose Public Protocol > HTTP. The IP address of the device cannot be written as localhost, you need to find the local IP in the network settings of your computer. For the demonstration, we used 192.168.0.123:23330 and 192.168.0.123:23331.

Once the information is filled in, a command pops up on the website, click the button on the right to copy it to the terminal and execute it to deploy the device to the local k8s cluster. This saves the time of manually writing YAML configuration files and is more convenient.

Testing if the device has been successfully integrated​

We can go into the cluster to see if we could access the digital twin of both devices using the network address within the cluster:.

## Run an nginx container
$ sudo kubectl run --image=nginx:1.21 nginx
# Enter the nginx container
$ sudo kubectl exec -it nginx -- bash
# Interact with the device
$ curl http://deviceshifu-mythermometer-service.deviceshifu.svc.cluster.local/humidity
$ curl http://deviceshifu-mythermometer-service.deviceshifu.svc.cluster.local/temperature
$ curl http://deviceshifu-myled-service.deviceshifu.svc.cluster.local/setfloat?value=321

You can see that the thermometer reading and the LED display settings are working properly, which means that devices have been successfully converted to digital twins.

Packaging the application as an image​

We are going to develop the application based on the thermometer and LED, here's the Python program we wrote:

main.py

import time
import requests
import json

isLocal = False
localIp = "192.168.0.123"
flag = -1
while True:
    flag += 1

    # [get data]
    if flag % 2 == 0:
        # Get temperature
        url = f "http://{localIp}:23330/temperature" if isLocal else "http://deviceshifu-mythermometer-service.deviceshifu.svc.cluster.local/temperature"
    else:
        # Get humidity
        url = f "http://{localIp}:23330/humidity" if isLocal else "http://deviceshifu-mythermometer-service.deviceshifu.svc.cluster.local/humidity"
    res = requests.get(url)

    # [convert data]
    try:
        value = json.loads(res.text)['value']
        print("DEBUG", value)
        # [display data]
        led_url = f "http://{localIp}:23331/setfloat?value={value}" if isLocal else f "http://deviceshifu-myled-service.deviceshifu.svc.cluster.local/setfloat?value={value}"
        requests.get(led_url)
    except:
        print("DEBUG", res.text)

    time.sleep(2)

The program reads the temperature and humidity of the thermohygrometer alternately every 2 seconds and displays the readings on the LED display.

Next we want to package this program as an image, so that we can load the image in the cluster and run:

requirements.txt

requests

Dockerfile

FROM python:3.9-slim-bullseye

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY main.py .
CMD ["python3", "main.py"]

# Build the image
$ sudo docker build -t xxx/connection:v0.0.1 .
# Make sure the image has been built
$ sudo docker images | grep connection
xxx/connection v0.0.1 a9526147ddad 2 minutes ago 125MB
# Load the image into the cluster
$ sudo kind load docker-image xxx/connection:v0.0.1
# Run the container as an instance of the image
$ sudo kubectl run --image=xxx/connection:v0.0.1 connection-name

Accessing the camera​

Next we want to access the Hikvision camera. The camera is not nearby, it is connected wirelessly by Wi-Fi.

You can see that Shifu Cloud supports this module of Hikvision, click on it and configure the camera ip address, username and password to access it with one click. The device name in the demo is mycamera.

# Enter the nginx container
$ sudo kubectl exec -it nginx -- bash
# Check camera information
$ curl http://deviceshifu-mycamera-service.deviceshifu.svc.cluster.local/info

This means that the Hikvision camera is integrated into Shifu and has been converted to a digital twin.

Controlling the position of a camera​

Immediately after that, we want to adjust the camera's position, and the camera's orientation can be controlled using APIs like move/up move/down move/left move/right.

$ curl http://deviceshifu-mycamera-service.deviceshifu.svc.cluster.local/move/up

To check the outcome, we turn on the video player on our computer and open that streaming address: rtsp://<user_name>:<password>@<ip_address>. In macOS, we can open IINA.app, go to menu bar > Open Url... > paste and enter, and we can see the live surveillance video stream.

As you can see, the camera position has changed, and we managed to reorient the camera to the angle we need(the angle of the camera moved from the ceiling to the back the computer screenon the desk).

Summary​

In this demonstration, we used Shifu Cloud to access three devices, and if you compare it to our first Meetup, you will see that we have achieved faster integration and lowered the threshold of integrating a device.

Shifu Cloud includes an Aha Moment to help you get familiar with this website.

In the future, Shifu Cloud will offer:

• More protocol support (both Edgenesis and the open source community will continue to support more protocols to improve the coverage of fast access)
• Application development support (for now we still need to package applications we developed locally, but in the future, we can develope applications on Shifu Cloud
• App store support (which contains apps uploaded by developers or third-party plugins that users can install with one click)

Thank you for reading this, let's stay tuned to future progresses of Shifu Cloud!

Five devices were displayed in this activity, namely MQTT server, RS485 temperature and humidity meter and RS485 LED, Siemens S7 PLC, and Hikvision camera - all of which are relatively common IoT devices. Let's review the integration process below.

Create a Cluster and Install Shifu​

First we need to start Docker locally. Just open Docker Desktop using Windows or macOS Search and minimize it to the background.

After that we need to create a k8s cluster with kind. Later on Shifu and the digital twins of IoT devices will exist in this cluster as Pod.

# Create a Cluster
$ sudo kind create cluster --image="kindest/node:v1.24.0"

# Prepare the image in advance for import into the cluster
$ sudo docker pull bitnami/kube-rbac-proxy:0.13.1 
$ sudo docker pull edgehub/shifu-controller:v0.1.1
$ sudo docker pull nginx:1.21
$ sudo kind load docker-image bitnami/kube-rbac-proxy:0.13.1 edgehub/shifu-controller:v0.1.1 nginx:1.21

Shifu supports one-click installation, just clone Shifu repository first, and deploy it later with one single command:

# Install shifu
$ git clone https://github.com/Edgenesis/shifu.git
$ cd shifu
$ sudo kubectl apply -f pkg/k8s/crd/install/shifu_install.yml

# Run an application that will be used later
$ sudo kubectl run --image=nginx:1.21 nginx

You can also check out a more detailed tutorial on installing Shifu locally.

Device Integration​

MQTT​

Test MQTT Server​

We have deployed an MQTT server and can test it by first opening two shells.

# shellA
$ mosquitto_sub -h 82.157.170.202 -t topic0

# shellB
$ mosquitto_pub -h 82.157.170.202 -t topic0 -m "lol"

You can see that the message sent can be received correctly.

Integrate the Device​

Next we can modify the corresponding configuration, download the corresponding image, and then use the kubectl apply command to integrate the MQTT server into Shifu as a digital twin.

Modify spec.address to 82.157.170.202:1883 and spec.protocolSettings.MQTTSetting.MQTTTopic to topic0 in examples/my_mqtt/mqtt_deploy.

$ sudo docker pull edgehub/deviceshifu-http-mqtt:v0.1.1
$ sudo kind load docker-image edgehub/deviceshifu-http-mqtt:v0.1.1
$ sudo kubectl apply -f examples/my_mqtt/mqtt_deploy

Read the Data​

We can interact with the digital twin by starting an nginx application in the cluster.

$ sudo kubectl exec -it nginx -- bash

$ curl http://deviceshifu-mqtt.deviceshifu.svc.cluster.local/mqtt_data

Connect Thermometer and LED​

Connect Device to Computer​

• The thermometer is connected to the computer using a serial server via a network cable
• LEDs are connected to the computer using an RS485 to USB chip

Start HTTP Service Locally​

Since Shifu does not support Modbus protocol right now, we need to convert the data read by Modbus to HTTP data.

$ cd api_thermometer
$ uvicorn --host 0.0.0.0 --port 23330 main:app

$ cd api_led
$ uvicorn --host 0.0.0.0 --port 23331 main:app

contents in api_thermometer​

from fastapi import FastAPI
from typing import List
from pymodbus.client.sync import ModbusTcpClient

app = FastAPI()

def getHumidityAndTemperature() -> List[float]:
    """
    return temperature and humidity datas retrieved from TAS-LAN-460
    """
    client = ModbusTcpClient(host='192.168.0.80', port=10123) # the port of TAS-LAN-460
    client.connect()
    SLAVE = 0x01
    r = client.read_holding_registers(address=0x0000, count=2, unit=SLAVE)
    print("collected data", r.registers)
    client.close()

    result = [r.registers[0] / 10, r.registers[1] / 10]
    return result

@app.get("/")
def root():
    return { "message": "Hello World" }

@app.get("/temperature")
def getTemperature():
    temperature = getHumidityAndTemperature()[1]
    return { "value": f"{temperature}" }

@app.get("/humidity")
def getHumidity():
    humidity = getHumidityAndTemperature()[0]
    return { "value": f"{humidity}" }

fastapi
pymodbus

contents in api_led​

from fastapi import FastAPI
from pymodbus.client.sync import ModbusSerialClient
from typing import List, Dict

app = FastAPI()

class ZhongshengLed:
    """
    DEVICE_NAME = "ZhongshengLed"
    """

    def __init__(self, device_address: int = 0x01, port: str = '/dev/tty.usbserial-14120') -> None:
        self.device_address = device_address
        self.client = ModbusSerialClient(method='rtu', port=port,  stopbits=1, bytesize=8, parity='N', baudrate=9600, timeout=2.0)
    
    def setLedCharacter(self, position: int, character: str):
        self.setLedAscii(position=position, ascii_value=ZhongshengLed.character2ascii[character])

    def setLedAscii(self, position: int, ascii_value: int):
        self.client.connect()
        self.client.write_register(address=position, value=ascii_value, unit=self.device_address)
        self.client.close()

    def setFourLedsString(self, string: str):
        self.setFourLedsAsciis(ascii_values=[ZhongshengLed.character2ascii[string[0]], ZhongshengLed.character2ascii[string[1]], ZhongshengLed.character2ascii[string[2]], ZhongshengLed.character2ascii[string[3]]])

    def setFourLedsAsciis(self, ascii_values: List[int]):
        self.client.connect()
        self.client.write_registers(address=ZhongshengLed.LedPosition.one, values=ascii_values, unit=self.device_address)
        self.client.close()

    class LedPosition:
        one = 0
        two = 1
        three = 2
        four = 3
    
    character2ascii: Dict[str, int] = {
        "0": 0x30, "1": 0x31, "2": 0x32, "3": 0x33, "4": 0x34, 
        "5": 0x35, "6": 0x36, "7": 0x37, "8": 0x38, "9": 0x39,
        ".": 0x2e, "-": 0x2d, " ": 0x20
    }


    def setDot(self, count: int = 1):
        self.client.connect()
        self.client.write_register(address=16, value=count, unit=self.device_address)
        self.client.close()

    def setNegative(self, isNegative: bool = False):
        self.client.connect()
        self.client.write_register(address=17, value=1 if isNegative else 0, unit=self.device_address)
        self.client.close()
    
    def setFloat(self, value: float):
        """
        display one decimal place
        """
        self.setDot(count=1)
        if value < 0:
            self.setNegative(True)
        else:
            self.setNegative(False)

        data = int(abs(value) * 10)

        self.client.connect()
        self.client.write_register(address=7, value=data, unit=self.device_address)

        # self.client.write_register(address=16, value=value, unit=self.device_address)
        self.client.close()

    def setBrightness(self, brightness: int = 7):
        self.client.connect()
        self.client.write_register(address=14, value=brightness, unit=self.device_address)
        self.client.close()

device = ZhongshengLed()

@app.get("/")
def root():
    return { "message": "Hello World" }

@app.get("/setfloat/{value}")
def setTemperature(value: float):
    device.setFloat(value=value)
    return { "OK": "OK" }

@app.get("/setfloat/{value}")
def setTemperature(value: float):
    device.setFloat(value=value)
    return { "OK": "OK" }

@app.get("/setfloat")
def setTemperature(value: float = 0.0):
    device.setFloat(value=value)
    return { "OK": "OK" }

fastapi
pymodbus

Local Verification​

$ curl http://localhost:23330/temperature
$ curl http://localhost:23330/humidity
$ curl http://localhost:23331/setfloat\?value\=123.4

Device Integration​

• Modify the IP address inhttp_thermometer/deployment/http_edgedevice.yaml.
• Modify the IP address inhttp_led/deployment/http_edgedevice.yaml。

$ sudo docker pull edgehub/deviceshifu-http-http:v0.1.1
$ sudo kind load docker-image edgehub/deviceshifu-http-http:v0.1.1
$ sudo kubectl apply -f examples/my_http_led/deployment
$ sudo kubectl apply -f examples/my_http_thermometer/deployment

Interact with Devices​

Start nginx to interact with the thermohygrometer：

$ sudo kubectl exec -it nginx -- bash

$ curl http://my-thermometer.deviceshifu.svc.cluster.local/temperature
$ curl http://my-thermometer.deviceshifu.svc.cluster.local/humidity
$ curl http://my-led.deviceshifu.svc.cluster.local/setfloat?value=23.4

Application Development​

Read the temperature and humidity and then display it alternately on the LED.

$ sudo docker build -t yangxijie/connection:v0.0.1 .
$ sudo docker images | grep connection
yangxijie/connection  v0.0.1  a9526147ddad  2 minutes ago  125MB
$ sudo kind load docker-image yangxijie/connection:v0.0.1
$ sudo kubectl run --image=yangxijie/connection:v0.0.1 connection-name

The illustration of this application is as follows：

You can see the temperature and humidity are displayed alternely on the LED once the application is running.

files in current folder​

import time
import requests
import json

isLocal = False
localIp = "192.168.31.138"
flag = -1
while True:
    flag += 1

    # [get data]
    if flag % 2 == 0:
        # get temperature
        url = f"http://{localIp}:23330/temperature" if isLocal else "http://my-thermometer.deviceshifu.svc.cluster.local/temperature"
    else:
        # get humidity
        url = f"http://{localIp}:23330/humidity" if isLocal else "http://my-thermometer.deviceshifu.svc.cluster.local/humidity"
    res = requests.get(url)

    # [converse data]
    try:
        value = json.loads(res.text)['value']
        print("DEBUG", value)
        # [display data]
        led_url = f"http://{localIp}:23331/setfloat?value={value}" if isLocal else f"http://my-led.deviceshifu.svc.cluster.local/setfloat?value={value}"
        requests.get(led_url)
    except:
        print("DEBUG", res.text)

    time.sleep(2)

requests

FROM python:3.9-slim-bullseye

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY main.py .
CMD ["python3", "main.py"]

Simons PLC​

Device Integration​

First change the IP address to that of the PLC.

$ sudo docker pull edgehub/deviceshifu-http-http:v0.1.1
$ sudo docker pull edgehub/plc-device:v0.0.1
$ sudo kind load docker-image edgehub/deviceshifu-http-http:v0.1.1 edgehub/plc-device:v0.0.1
$ sudo kubectl apply -f examples/my_plc/plc-deployment

Interact with the Device​

Here we modify one bit on the PLC and you can see the indicator light on the PLC turns on. In actual scenarios, the PLC will control large equipments such as an robotic arm.

$ sudo kubectl run nginx --image=nginx:1.21 -n deviceshifu 
$ sudo kubectl exec -it nginx -n deviceshifu -- bash

$ curl "deviceshifu-plc/sendsinglebit?rootaddress=Q&address=0&start=0&digit=1&value=1"; echo

Hikivision Camera​

Device Integration​

Get the IP address of the camera and then replace the IP address in rtsp/camera-deployment/deviceshifu-camera-deployment.yaml.

$ sudo docker pull edgehub/deviceshifu-http-http:v0.1.1
$ sudo docker pull edgehub/camera-python:v0.0.1
$ sudo kind load docker-image edgehub/deviceshifu-http-http:v0.1.1 edgehub/camera-python:v0.0.1
$ sudo kubectl apply -f examples/my_rtsp/camera-deployment

Interact with Device​

Check device information in cluster using nginx:

# Interact through `curl` in cluster
$ sudo kubectl exec -it nginx -- bash

$ curl http://deviceshifu-camera.deviceshifu.svc.cluster.local/info

The command that the digital twin of the camera supports is:

To view the images captured by the camera, we need to forward the port to the local machine and access it on the browser.

# Local browser access
$ sudo kubectl port-forward svc/deviceshifu-camera -n deviceshifu 8080:80
# Enter`localhost:8080/info`check information
# Enter`localhost:8080/capture`access pictures
# Enter`localhost:8080/move/{up|down|left|right}`move the camera
# Enter`localhost:8080/stream?timeout=0`access real-time streamline

Summary​

The Shifu Meetup event was a huge success. As you can see Shifu enables developers to quickly access devices, unify various protocols to HTTP for easy management and subsequent application development. Shifu also has many advantages including running with no single point failure and good isolation.

If you are interested in Shifu, please visit Shifu official website to learn more. You are also welcome to give the project a star at Shifu's GitHub repository!

Background 🌇​

When we use Shifu to collect data, it usually happens that the data collected from the device is in a different format from the data we need. To solve this problem, we can use Shifu + WasmEdge to process the data collected by Shifu through WasmEdge and then return it to our application.

The following is the simple logic.

Prepare 🗂​

1. kubectl v1.24.2
2. docker 20.10.16
3. kind v0.14.0
4. git 2.36.1

Deployment 🔨​

To make this article faster for you, you can download the program from Github with the following command. 🚀

git clone https://github.com/Edgenesis/wasm-shifu-demo.git
cd wasm-shifu-demo

Create a K8s Cluster 🐝​

Use the following command to create a k8s cluster.

$ kind delete cluster && kind create cluster
Creating cluster "kind" ...
 ✓ Ensuring node image (kindest/node:v1.24.0) 🖼
 ✓ Preparing nodes 📦  
 ✓ Writing configuration 📜 
 ✓ Starting control-plane 🕹️ 
 ✓ Installing CNI 🔌 
 ✓ Installing StorageClass 💾 
Set kubectl context to "kind-kind"
You can now use your cluster with:
kubectl cluster-info --context kind-kind
Have a question, bug, or feature request? Let us know! https://kind.sigs.k8s.io/#community 🙂

Build Shifu image 🪞​

Use the following command to build a Shifu image.

$ make -f shifu/Makefile build-image-deviceshifu
$ kind load docker-image edgehub/deviceshifu-http-http:v0.0.6
$ docker images | grep edgehub/deviceshifu-http-http 
edgehub/deviceshifu-http-http v0.0.6 1d6b3544b8ad 54 minutes ago 36.1MB

Run Virtual Devices 🔌​

To make your experience easier, here we use a virtual appliance for simulation.

Install and run the virtual appliance with port number 8099.

$ docker build -f mockDevice/dockerfile -t mockdevice:v0.0.1 .
$ docker run -p 8099:8099 -itd mockdevice:v0.0.1 
bdfd2b1323be mockdevice:v0.0.1 ". /mockDevice" 19 seconds ago Up 18 seconds 0.0.0.0:8099->8099/tcp admiring_feistel

Write Rules & Compile Wasm​

You can write rules by using JavaScript. If you are not familiar with JavaScript, you can just use the default rules. 🥮

Rule file path: wasmEdge/js-func/src/js/run.js You can achieve different functions by modifying the rule.

$ docker build -t wasm:v0.0.1 -f wasmEdge/js.dockerfile .
$ kind load docker-image wasm:v0.0.1
$ kubectl apply -f wasmEdge/k8s

You can check the pod operation of WasmEdge with the following command.

$ kubectl get pod -n wasmedge
NAME READY STATUS RESTARTS AGE
wasm-deployment-fbc9564d8-td428 1/1 Running 0 1s

Install and Run Shifu​

Install Shifu.

$ kubectl apply -f shifuConfig/shifu_install.yml
$ kubectl get pod -n shifu-crd-system
NAME READY STATUS RESTARTS AGE
shifu-crd-controller-manager-5bbdb4d786-s6h4m 2/2 Running 0 1s

Install deviceShifu to connect with mockDeivce. Before doing so, please change the address in the shifuConfig/task3/task3.yaml file to the IP of your computer.

spec:
  sku: "E93"
  connection: Ethernet
  address: "192.168.14.163:8099"

Deploy and run deviceShifu with the following command. 🏖

$ kubectl apply -f shifuConfig/task3
$ kubectl get pod -n deviceshifu
NAME READY STATUS RESTARTS AGE
deviceshifu-demodevice-deployment-5589b55569-l5nb2 1/1 Running 0 4s

Experience 🕹​

You can start a nginx to communicate with deviceShifu.

$ kubectl run nginx --image=nginx:1.21
$ kubectl get pod 
NAME READY STATUS RESTARTS AGE
nginx 1/1 Running 0 3s

With the following command, you can interact with Shifu to clean the data collected from IoT devices. 🛁

$ kubectl exec -it nginx -- curl -v http://deviceshifu-demodevice-service.deviceshifu.svc.cluster.local/get_info;echo

[
   {
      "code":375287,
      "name": "atmospheric temperature",
      "val": "24.56",
      "unit":"°C",
      "exception": "Temperature is too high"
   },
   {
      "code":375287,
      "name": "Atmospheric Humidity",
      "val": "81.63",
      "unit":"%RH",
      "exception": "Humidity too high"
   }
]

Also we can use the following command to check the data generated by the IoT device.

$ curl localhost:8099/getInfo

{
   "statusCode": "200",
   "message": "success",
   "entity":[
      {
         "dateTime": "2022-09-09 09:46:45",
         "eUnit":"℃",
         "eValue": "23.87",
         "eKey": "e1",
         "eName": "Atmospheric temperature",
         "eNum": "101"
      },
      {
         "dateTime": "2022-09-09 09:46:45",
         "eUnit":"%RH",
         "eValue": "80.62",
         "eKey": "e2",
         "eName": "Atmospheric Humidity",
         "eNum": "102"
      }
   ],
   "deviceId":950920,
   "deviceName": "950920",
   "deviceRemark": "2022-09-09 09:46:45"
}

Comparing the two outputs, we can see that we have successfully collected and cleaned the data to get the data we want. The comparison chart is as follows :

Developers can focus on application development on cloud-edge products without worrying about the operation and maintenance of the underlying architecture. As Kubernetes native open source IoT development framework, Shifu can be compatible with various IoT device protocols and abstract them into a microservice software object. The two complement each other very well. In particular, Shifu can abstract devices in a way that is native to OpenYurt, greatly improving the efficiency for IoT developers when we add YurtDeviceController to OpenYurt, .

With OpenYurt and Shifu, we can transform the complex IoT, cloud-side collaborative development into simple Web development.

Introduction​

This article is a guide to use Shifu to integrate RTSP protocol cameras into OpenYurt clusters , which contains the basic operations of Shifu, Docker, Linux, Kubernetes, OpenYurt. Any developer can learn how to develop Shifu through this article.

The Shifu architecture in this article is as follows

The northbound opens up the HTTP API interface via deviceshifu-http-http and the southbound interacts with the actual device via rtsp-driver.

Objectives​

1. deploy OpenYurt on Server side and Edge side via yurtctl, and add Edge side to Server side cluster
2. deploy the digital twin of webcam on Edge side
3. realize remote automation control of webcam via HTTP

Required devices​

1. two virtual machines running Linux, the configuration of Server and Edge should be 4 cores with 16G RAM and 2 cores with 8G RAM respectively
2. a webcam with RTSP protocol, the camera model used in this article is Hikvision DS-2DE3Q140CN-W.

Software environment​

• CentOS 7.9.2009
• Go v1.17.1
• yurtctl v0.6.1
• kubectl: v1.19.8 (installed by yurtctl)

Step 1 Install and deploy the OpenYurt cluster​

This article refers to the official tutorial for OpenYurt

First let's download OpenYurt and clone the project directly from the official GitHub:

git clone https://github.com/openyurtio/openyurt.git

Next, let's download the v0.6.1 version of yurtctl:

curl -LO https://github.com/openyurtio/openyurt/releases/download/v0.6.1/yurtctl 
chmod +x yurtctl

Server-side deployment​

Create the OpenYurt cluster on the Server side:

./yurtctl init --apiserver-advertise-address <SERVER_IP> --openyurt-version latest --passwd 123 

The cluster is successfully created when you see the following message, and use --token to mark that the Edge node is added to the cluster

Next, take a look at the status of each Pod by running kubectl get pods -A:

problems​

If you encounter in kubectl logs yurt-hub-server -n kube-system:

Please try kubectl apply -f config/setup/yurt-controller-manager.yaml (method from https://github.com/openyurtio/openyurt/issues/872#issuecomment- 1148167419 )

In addition, if you encounter the following output in kubectl logs yurt-hub-server -n kube-system.

Please try kubectl apply -f config/setup/ yurthub-cfg.yaml

If similar logs are encountered in yurt-tunnel-server and yurt-tunnel-agent, fix the RBAC issue in yurt-tunnel with the following command.

kubectl apply -f config/setup/yurt-tunnel-agent.yaml 
kubectl apply -f config/setup/yurt-tunnel-server.yaml

Use untaint master node to run Shifu controller.

kubectl taint nodes server node-role.kubernetes.io/master-

At this point, the Server side deployment is succeed.

Deploy the Edge side​

First, run with the token you just initialized on the Server side.

./yurtctl join <MASTER_IP>:6443 --token <MASTER_INIT_TOKEN> --node-type=edge --discovery-token-unsafe-skip-ca-verification --v=5 

Verify Node status by kubectl get nodes.

At this point, a Server-side + an Edge-side cluster is created.

Step 2 Deploy Shifu in the cluster​

Next, let's deploy Shifu to the OpenYurt cluster

On the Server side, clone the Shifu project locally.

git clone https://github.com/Edgenesis/shifu.git 
cd shifu/

Next, install Shifu.

kubectl apply -f pkg/k8s/crd/install/shifu_install.yml

Check the Pod status with kubectl get pods -A.

Just see the Pod running in the shifu-crd-system namespace.

At this point, Shifu is successfully installed.

Step 3 Deploy the camera's digital twin deviceShifu​

OpenYurt provides a very convenient node pool (NodePool) feature that allows us to manage clusters of nodes and deploy them.

To create a beijing node pool.

export WORKER_NODEPOOL="beijing" 
export EDGE_NODE="edge" 
cat <<EOF | kubectl apply -f - 
apiVersion: apps.openyurt.io/v1alpha1 
kind: NodePool 
metadata: 
  name: $WORKER_NODEPOOL 
spec: 
  type: Edge 
EOF

The output is as follows.

Next, take the Edge server label to beijing's NodePool.

kubectl label node $EDGE_NODE apps.openyurt.io/desired-nodepool=beijing

Check the status of NodePool, there should be a READYNODES.

kubectl get nodepool

Since IoT edge nodes are usually distributed within the same scenario, here you can use OpenYurt's UnitedDeployment feature to automate the deployment based on the NodePool .

Install Yurt-app-manager:

git clone https://github.com/openyurtio/yurt-app-manager.git
cd yurt-app-manager
kubectl apply -f config/setup/all_in_one.yaml

Use UnitedDeployment to deploy a virtual Haikon camera with the following YAML file.

apiVersion: apps.openyurt.io/v1alpha1
kind: UnitedDeployment
metadata:
labels:
  controller-tools.k8s.io: "1.0"
name: deviceshifu-hikvision-camera-deployment
spec:
selector:
  matchLabels:
    app: deviceshifu-hikvision-camera-deployment
workloadTemplate:
  deploymentTemplate:
    metadata:
      labels:
        app: deviceshifu-hikvision-camera-deployment
      name: deviceshifu-hikvision-camera-deployment
      namespace: default
    spec:
      selector:
        matchLabels:
          app: deviceshifu-hikvision-camera-deployment
      template:
        metadata:
          labels:
            app: deviceshifu-hikvision-camera-deployment
        spec:
          containers:
          - image: edgehub/deviceshifu-http-http:v0.0.1
            name: deviceshifu-http
            ports:
            - containerPort: 8080
            volumeMounts:
            - name: deviceshifu-config
              mountPath: "/etc/edgedevice/config"
              readOnly: true
            env:
            - name: EDGEDEVICE_NAME
              value: "deviceshifu-hikvision-camera"
            - name: EDGEDEVICE_NAMESPACE
              value: "devices"
          - image: edgenesis/camera-python:v0.0.1
            name: camera-python
            ports:
            - containerPort: 11112
            volumeMounts:
            - name: deviceshifu-config
              mountPath: "/etc/edgedevice/config"
              readOnly: true
            env:
            - name: EDGEDEVICE_NAME
              value: "deviceshifu-hikvision-camera"
            - name: EDGEDEVICE_NAMESPACE
              value: "devices"
            - name: IP_CAMERA_ADDRESS
              value: "<CAMERA_IP>"
            - name: IP_CAMERA_USERNAME
              value: "<CAMERA_USERNAME>"
            - name: IP_CAMERA_PASSWORD
              value: "<CAMERA_PASSWORD>"
            - name: IP_CAMERA_CONTAINER_PORT
              value: "11112"
            - name: PYTHONUNBUFFERED
              value: "1"
          volumes:
          - name: deviceshifu-config
            configMap:
              name: deviceshifu-hikvision-camera-configmap-0.0.1
          serviceAccountName: edgedevice-sa
topology:
  pools:
  - name: beijing
    nodeSelectorTerm:
      matchExpressions:
      - key: apps.openyurt.io/nodepool
        operator: In
        values:
        - beijing
    replicas: 1
revisionHistoryLimit: 5

apiVersion: v1
kind: Service
metadata:
labels:
  app: deviceshifu-hikvision-camera-deployment
name: deviceshifu-hikvision-camera
namespace: default
spec:
ports:
- port: 80
  protocol: TCP
  targetPort: 8080
selector:
  app: deviceshifu-hikvision-camera-deployment
type: LoadBalancer

apiVersion: shifu.edgenesis.io/v1alpha1
kind: EdgeDevice
metadata:
name: deviceshifu-hikvision-camera
namespace: devices
spec:
sku: "HikVision Camera"
connection: Ethernet
address: 0.0.0.0:11112
protocol: HTTP

apiVersion: v1
kind: ConfigMap
metadata:
name: deviceshifu-hikvision-camera-configmap-0.0.1
namespace: default
data:
driverProperties: |
  driverSku: HikVision
  driverImage: edgenesis/camera-python:v0.0.1
instructions: |
  capture:
  info:
  stream:
  move/up:
  move/down:
  move/left:
  move/right:
telemetries: |
  device_health:
    properties:
      instruction: info
      initialDelayMs: 1000
      intervalMs: 1000

Put the four files into a directory as follows.

camera-unitedDeployment/ 
├── camera-edgedevice.yaml 
├── deviceshifu-camera-configmap.yaml 
├─ deviceshifu-camera-service.yaml 
└── deviceshifu-camera-unitedDeployment.yaml

Next, deploy:

kubectl apply -f camera-unitedDeployment/

Check the UnitedDeployment status via kubectl get ud.

Confirm that Pod is deployed in the Edge server in beijing NodePool with kubectl get pods -owide.

We can check Shifu's virtual devices in the cluster via kubectl get edgedevices -n devices at

Then use kubectl describe edgedevices -n devices to see the details of the devices such as configuration, status, etc.

At this point, the digital twin of the camera is deployed.

Running results​

Next we control the camera, here we use a Pod of nginx to represent the application.

kubectl run nginx --image=nginx

When nginx is running, go to the nginx command line with kubectl exec -it nginx -- bash.

The camera can be controlled directly with the following command.

curl deviceshifu-hikvision-camera/move/{up/down/left/right}

If we want to see what the current photo and video stream, we need to proxy the camera's service locally via kubectl port-forward service/deviceshifu-hikvision-camera 30080:80 --address='0.0.0.0'.

You can view the image/video stream directly by entering the server's IP plus the port number in the browser at

<SERVER_IP>:30080/capture
<SERVER_IP>:30080/stream

Conclusion​

In this article, we showed you how to deploy Shifu in an OpenYurt cluster to support RTSP camera.

In the future, we will also try to integrate Shifu with OpenYurt YurtDeviceController We will also try to extend the capabilities of OpenYurt to manage more IoT devices in a way that is native to OpenYurt.

EMQX is a popular MQTT Broker in the world. It has a cloud-native architecture based on Kubernetes, making it extremely suitable for the increasingly complex IoT scenarios and making the transmission of device messages more efficient. Shifu, as a Kubernetes native framework, can be perfectly combined with EMQX to provide EMQX with intelligent multi-protocol device linkage capabilities.

Introduction​

This article will show you how to deploy EMQX and Shifu in a cluster, integrate a thermometer with MQTT as the communication method and a Hikvision camera with RTSP as the transmission protocol, and add an application to interact with Shifu so that every time the thermometer detects a body temperature above 37 degrees it will ask the camera to take a photo.

The simple architecture used in this article is as follows.

Preparation​

The following services and tools are used in this article.

1. Kubernetes: 1.20.10
   • kubectl
   • kubeadm
   • kubelet
2. Golang: 1.16.10
3. Docker: 19.03.9
4. EMQX: 4.1-rc1

Step 1 Deploy Kubernetes​

For this step, you can refer to the official Kubernetes tutorial for deployment: Kubernetes.

https://kubernetes.io/docs/setup/

After the deployment is complete we should see the following message printed on the terminal.

Step 2 Deploy Shifu​

Clone the GitHub repository for Shifu locally to.

git clone https://github.com/Edgenesis/shifu.git

You can then deploy Shifu with the following command.

kubectl apply -f shifu/pkg/k8s/crd/install/shifu_install.yml

Once deployed we should see that the CRD controller for Shifu has been deployed.

Step 3 Deploy EMQX​

First you need to install EMQX Operator Controller.

$ curl -f -L "https://github.com/emqx/emqx-operator/releases/download/1.1.6/emqx-operator-controller.yaml" | kubectl apply -f -

Then we write the simplest deployment.yaml:

Then it's time to deploy an EMQX:

kubectl apply -f deployment.yaml

Step 4 Integrate into the device​

For the thermometer, we just need to adjust its MQTT settings so that it can post MQTT messages to EMQX.

(For thermometers outside the cluster, we can open up External IP for access via Kubernetes Service)

In terms of cameras, Shifu's repository already includes a configuration file for Hikvision cameras using RTSP that we can simply change IP, username, password in the config file to integrate the camera into Shifu.

At this point, our device is connected and we are ready to start linking below.

Linking Applications​

We can write a Python application to implement the following logic.

The app subscribes to EMQX temperature-shifu-mqtt messages, each message includes only a number refering the current temperature; if the current temperature is higher than 37 degrees, the camera is asked take a picture and save it locally.

The application code is as follows.

Add a capture function to encapsulate all camera actions. Then we can deploy it to the cluster and start monitoring.

python3 app.py 10.244.0.33

Summary​

This article shows how to get EMQX to empower Shifu with more efficient MQTT Broker capabilities, and make Shifu to work with MQTT to provide linkage capabilities to devices. In a real-world application scenario, we can use a cheap combination of IR thermometer and camera to replace thousands of dollars of unstable temperature cameras, saving huge costs in a large-scale deployment scenario.

As an open source IoT development framework, Shifu is also based on Kubernetes, its compatibility with multiple devices and virtualization will help KubeEdge to be applied at the edge. In fact, the two complement each other very well in terms of capabilities. Not only compatible to multiple devices, Shifu running on KubeEdge can easily manage lightweight Pod running on the edge.

With the powerful combination of KubeEdge + Shifu, we can abstract IoT devices into API and turn the traditional complex IoT development model into a simple web development model!

Let's see how to make Shifu run on KubeEdge and provide value for developers!

Introduction​

This article will briefly describe the steps to deploy Shifu on KubeEdge and integrate into Hikvision camera (using RTSP for video streaming), so that KubeEdge architecture can support Hikvision camera.

The architecture used in this article is as follows.

Preparation​

The following services and tools are used in this article.

1. Kubernetes: 1.21.5
   • kubectl
   • kubeadm
   • kubelet
2. Golang: 1.16.10
3. Docker: 19.03.9
4. KubeEdge: 1.7.2

Meanwhile, the Cloud side and Edge side of KubeEdge are running on separate Linux instances, but both on Ubuntu Server 20.04 environment.

You need to install all the above mentioned services and tools on the Cloud side , but only Docker and KubeEdge on the Edge side.

Step 1 Deploy Kubernetes on the Cloud side​

You can check official tutorial on Kubernetes to finish the deployment.

After the deployment is complete we should see the terminal print out the following message.

Step 2 Deploy Shifu on Cloud side​

Clone the Github repository of Shifu to your computer.

git clone https://github.com/Edgenesis/shifu.git

Then you can deploy Shifu with the following command:

kubectl apply -f shifu/pkg/k8s/crd/install/shifu_install.yml

Once deployed we should see that the CRD controller for Shifu has been deployed.

Step 3 Deploy KubeEdge on Cloud side​

For this step, you can refer to KubeEdge's official tutorial and use keadm for deployment.

After the deployment, we should see the following message printed on the terminal.

Step 4 Get the token on the Cloud side​

Run the following command.

keadm gettoken

Please save the token you got for the Edge side.

Now that the Cloud side configuration is completed, we switch to the Edge side machine and add it to the cluster.

Step 5 Join the cluster on the Edge side​

Run the following command on the Edge side.

keadm join --cloudcore-ipport="<Cloud-side advertise-address>:10000" --token=<token obtained in step 4>

After the deployment is complete we should see the terminal print out the following message.

At this point switch back to the Cloud side and look at the nodes.

We can see that both the Cloud side and the Edge side have been deployed.

Now we can start deploying the device.

With KubeEdge, we can perform Kubernetes operations only on the Cloud side and deploy it to the Edge side, while keeping the Edge side free of Kubernetes components to keep lightweight.

Step 6 Modify the Hikvision camera configuration file on the Cloud side​

Shifu needs a simple configuration file to generate digital twin. In Shifu, the digital twin is called deviceShifu and runs in a cluster as Pod.

Shifu provides configuration file to access to Hikvision cameras, here's the path https://github.com/Edgenesis/shifu/tree/main/examples/rtspDeviceShifu/.

Shifu deploys deviceShifu on a machine with a full Kubernetes instance by default. In a KubeEdge environment, there is no need to run full Kubernetes on the edge, so Shifu also has a lightweight deviceShifu for cloud-edge collaboration. We can change deviceshifu-camera-deployment.yaml to use the edge-side deviceShifu and add nodeName to deploy it to the edge node:

Step 7 Deploy the Hikvision Camera Pod​

On the Cloud side, run the following command.

kubectl apply -f shifu/examples/rtspDeviceShifu/camera-deployment

At this point, we can look at the Pod associated with the camera.

Final step Confirmation on the Edge side​

On the Edge side, we can see that the camera-related Docker container is already running:

We can simply call the capture/stream/info/move and a host of other HTTP APIs provided by deviceShifu to perform operations on the camera, such as the following motion picture.

Related commands.

curl edgedevice-camera/move

This completes all the steps we need to run Shifu on KubeEdge.

Create the driver​

Objective​

• Complete simple LED circuit connection
• Basic Raspberry Pi/SSH configuration
• Basic Python syntax and knowledge of GPIO libraries

Devices​

• Raspberry Pi Raspberry Pi 3B+ running 64-bit Raspberry Pi OS
• 1 breadboard
• 3 LED bulbs (red, yellow, green)
• 1 x 330 ohm resistor

Basic knowledge​

• Simple Python syntax
• Basic Linux command line operations (create files, install apps, SSH, run programs)

Step 1 Circuit design​

First let's design the circuit. We need to design a circuit that will allow the GPIO output of the Raspberry Pi to control the on/off of a single LED bulb. In this article, we have used the most straightforward approach, i.e., using GPIO directly to power the LED bulbs.

The circuit diagram is as follows.

The GPIO pins 22, 23, and 19 control the red, green, and yellow LEDs respectively. 330 ohms resistors are connected in series at the end to prevent the LEDs from burning out due to excessive current.

Step 2 Circuit Implementation​

According to the Pin Layout in the official Raspberry Pi documentation, we can see the exact location of the pins, here pins 15, 16, 35 and 39 are used.

Connect these four pins to the female port of the cable, and then connect the remaining circuits on the breadboard too.

The red, green, yellow, and gray cables in the diagram correspond to GPIO22 GPIO23 GPIO19 and ground, respectively.

This concludes the circuit design and connectivity section.

Step 3 Raspberry Pi Preparation​

First install an operating system in the Raspberry Pi, the one used in this article is Raspberry Pi OS (64-bit). download link

Insert the SD card into the reader, connect it to the USB port of your computer, and then swipe the downloaded zip file into the SD card via balenaEtcher. balenaEtcher link

Insert the SD card into the Raspberry Pi, plug in the power and monitor cable, and we can start the configuration.

First, for development/debugging we need to enable SSH, open a terminal from the desktop and enter sudo raspi-config to access the configuration screen and select Interface Options.

Select SSH:

Press enter, then press left to select Yes to enable the SSH service:

After that, press right and select Finish, then enter to exit.

At this point, the SSH service is on, but we need to know the IP of the Raspberry Pi in order to SSH, so we can check it with the built-in ip addr command.

As you can see, the IP address is 192.168.15.122.

Back in the computer, you can remotely access the Raspberry Pi from the command line via ssh pi@192.168.15.122

This concludes the preparation of the Raspberry Pi and the hardware.

Step 4 Write a Driver​

Everything is in place, now let's write the first driver!

First make sure that Python is installed on your system, if not you can run the following command.

$ sudo apt-get update && sudo apt install python3 -y

After installation, you can check the status of the installation with python -V, if it shows the version, then the installation is successful:

$ python -V
Python 3.9.2

Next let's start with an LED bulb control, we want to control the red LED on/off using the following code.

The modules used in the driver are:

• RPi.GPIO to control the GPIO of the Raspberry Pi
• argparse is used to parse command line input

First set the GPIO mode to GPIO.BCM mode, in this mode the pin numbers are the GPIO numbers, not the pin order on the Raspberry Pi board.

GPIO.setmode(GPIO.BCM)

Then turn the warning off, the Raspberry Pi will only be controlled only by this driver in this article.

GPIO.setwarnings(False)

Next process the program input, this driver will accept two inputs.

1. -p, --port, representing the GPIO pin that the program is manipulating
2. -o, --operate, represents the program's operation of the GPIO pin, on represents 1, i.e. 3.3V in the circuit, off represents 0, i.e. 0V in the circuit

The code is as follows

parser = argparse.ArgumentParser()
parser.add_argument("-p", "--pin", type=int, default=None, help="Specify the GPIO pin to operate, e.g.: '17'")
parser.add_argument("-o", "--operate", type=str, default=None, help="Specify the GPIO output, e.g.: 'on/off'")
args = parser.parse_args()

Next we need to handle errors, passing the arguments into the function turnOnLed to operate when the pin count and operation are not empty, otherwise printing a warning

if args.pin and args.operate:
    turnOnLed(args.pin, args.operator)
else:
    print("need to specify both pin and operate arguments, type --help for more information")

The main part of the program ends, so let's look at the function turnOnLed that controls the LED bulb.

The first thing is to determine whether the operate variable is on or off, otherwise it returns. The output variable gpio_out is set to GPIO.HIGH when the variable is on and to GPIO.LOW when it is off.

These two values represent on or off.

if operate == "on":
    gpio_out = GPIO.HIGH
elif operate == "off":
    gpio_out = GPIO.LOW
else:
    print("operate is neither on/off, quitting...")
    return

The last thing is to set the mode of the pin to output: GPIO.setup(pin, GPIO.OUT)

and switch the pin's output to on/off: GPIO.output(pin, gpio_out)

Outcome​

The program can be executed using python led_driver.py -p {pin #} -o {operate}

If we want to make the red bulb light up, we run python led_driver.py -p 22 -o on

And that's it, a simple control LED bulb driver for Raspberry Pi!

Integrate into Shifu​

Next, we use the driver we wrote to access the Shifu framework for interaction and management.

The Shifu architecture in this article is as follows.

The northbound opens up the HTTP API interface via deviceshifu-http-http and the southbound interacts with the actual device via rpio-gpio-driver.

Objectives​

1. install k3s cluster on Raspberry Pi and install Shifu Framework
2. package the Raspberry Pi LED driver to a container image
3. deploy the digital twin of the Raspberry Pi LED in Shifu
4. Achieve remote automation control of Raspberry Pi LEDs

Devices​

• Raspberry Pi Raspberry Pi 3B+ running 64-bit Raspberry Pi OS

Basic knowledge​

• basic operation of Docker/containerd
• basic operation of K8s/K3s

Step 1 Install k3s​

First we need to run a Kubernetes cluster in the Raspberry Pi. There is no restriction on the version users can use here, but to save resources we use k3s in this article. Installation Tutorial

After installation, run kubectl version to see the current Kubernetes version.

Use kubectl get nodes to see the status of the current cluster, showing Ready means the cluster is available:

At this point, the k3s installation is complete.

Step 2 Install Shifu​

First clone the Shifu project repository on your computer:

$ git clone https://github.com/Edgenesis/shifu.git

You can deploy Shifu to the k3s cluster with one click by using kubectl apply -f shifu/pkg/k8s/crd/install/shifu_install.yml.

Execute kubectl get pods -A again to see the Shifu Framework controller deployed to the cluster:

We can also manage device resources (currently there are no devices) via edgedevices the CRD:

At this point, Shifu is successfully installed.

Step 3 Package the driver​

We need to use a small tool provided by Shifu to realize that we can manipulate the local driver remotely, for a detailed tutorial please see.

• New version: https://github.com/Edgenesis/shifu/tree/main/cmd/httpstub/sshstub
• Older version: https://github.com/Edgenesis/shifu/tree/c0cab1ce98870eb470ddc9e01e1d99c13b611411/driver_util

This tool helps to convert HTTP requests sent by the user/program to a local command line for execution.

A sample driver is provided inside the tutorial at https://github.com/Edgenesis/shifu/blob/main/examples/driver_utils/simple-alpine/Dockerfile.sample

The content is as follows.

You can see that the Dockerfile has two parts, the first is to use the golang image to compile the http_to_ssh_stub.go provided by Shifu to convert HTTP command to SSH command. The next step is to use an empty alpine image to configure SSH for demonstration.

Next let's practice it.

Considering the limitations of Raspberry Pi, this compilation will be executed from the computer side, and the compiled image will be pushed to Docker Hub for remote invocation.

First, we create a new folder, here we use dev, and then save the created Raspberry Pi LED driver to this directory:.

dev/
└── led_driver.py

The driver content remains the same.

Copy Dockerfile.sample from the driver_util/examples/simple-alpine/ directory of the Shifu project to the dev directory.

dev/
├── Dockerfile.sample
└── led_driver.py

Change the following fields to change the second part of the image from alpine to python:alpine, install the Python library for RPi.GPIO

Finally, copy the Python driver into the run container, and the new Dockerfile will look like this, with the changes marked with comments.

FROM golang:1.17.1 as builder
 
WORKDIR /
 
ENV GOPROXY=https://goproxy.cn,direct
ENV GO111MODULE=on
ENV GOPRIVATE=github.com/Edgenesis
 
COPY driver_util driver_util
 
WORKDIR /driver_util
RUN go mod download
 
# Build the Go app
RUN CGO_ENABLED=0 GOOS=$(go env GOOS) GOARCH=$(go env GOARCH) gobuild -a -o /output/http2ssh-stub http_to_ssh_stub.go
 
FROM python:alpine # modified
 
RUN apk add --no-cache --update openrc openssh \
    && mkdir -p /run/openrc \
    && touch /run/openrc/softlevel \
    && sed -ie "s/#PubkeyAuthentication/PubkeyAuthentication/g"/etc/ssh/sshd_config \
    && sed -ie "s/#PasswordAuthenticationyes/PasswordAuthentication no/g" /etc/ssh/sshd_config \
    && sed -ie "s/AllowTcpForwardingno/AllowTcpForwarding yes/g" /etc/ssh/sshd_config \
    && echo"PubkeyAcceptedKeyTypes=+ssh-rsa" >>  /etc/ssh/ sshd_config\  # modified
    && ssh-keygen -A \
    && passwd -d root \
    && mkdir ~/.ssh \
    && while ! [ -e/etc/ssh/ssh_host_rsa_key.pub ]; do sleep 1; done \
    && cp /etc/ssh/ssh_host_rsa_key.pub~/.ssh/authorized_keys
 
RUN apk add --no-cache -Uu --virtual .build-dependencies libffi-devopenssl-dev build-base musl \
    && pip3 install --no-cache --upgrade RPi.GPIO\
    && apk del --purge .build-dependencies \
    && apk add --no-cache --purge curlca-certificates musl \
    && rm -rf /var/cache/apk/* /tmp/*  # modified
 
WORKDIR /root/
 
COPY --from=builder /output/http2ssh-stub http2ssh-stub
COPY --from=builder/driver_util/examples/simple-alpine/docker-entrypoint.sh docker-entrypoint.sh
COPY dev/led_driver.py led_driver.py  # modified
RUN chmod +x docker-entrypoint.sh
 
# Command to run the executable
ENTRYPOINT ["./docker-entrypoint.sh"]

Next we will package the Docker image, because the CPU of the Raspberry Pi is ARM64 processor, the computer used for compiling in this article is x86-64, so we need to use the buildx function of Docker to build the image, the tutorial about buildx will not be described in this article, you can move to https://docs.docker.com/buildx/working-with-buildx/.

Use docker buildx build --platform=linux/arm64 -f dev/Dockerfile.sample . -t edgehub/rpi-gpio-driver:v0.0.1 --push to build the image and push it to Docker Hub.

At this point, the image packaging part is complete.

Step 4 Deploy device twin to Raspberry Pi​

Once we have the image, we can deploy the digital twin to the cluster, so let's prepare the files we need for the deployment.

First is a Kuberenetes Deployment YAML file to run deviceShifu and the driver Pod.

apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: edgedevice-rpi-gpio-deployment
  name: edgedevice-rpi-gpio-deployment
  namespace: default
spec:
  replicas: 1
  selector:
    matchLabels:
      app: edgedevice-rpi-gpio-deployment
  template:
    metadata:
      labels:
        app: edgedevice-rpi-gpio-deployment
    spec:
      containers:
      - image: edgehub/deviceshifu-http-http:v0.0.1
        name: deviceshifu-http
        ports:
        - containerPort: 8080
        volumeMounts:
        - name: edgedevice-config
          mountPath: "/etc/edgedevice/config"
          readOnly: true
        env:
        - name: EDGEDEVICE_NAME
          value: "edgedevice-rpi-gpio"
        - name: EDGEDEVICE_NAMESPACE
          value: "devices"
      - image: edgehub/rpi-gpio-driver:v0.0.1
        name: driver
        volumeMounts:
        - mountPath: /dev/gpiomem
          name: gpiomem
        securityContext:
          privileged: true
        ports:
        - containerPort: 11112
        env:
        - name: EDGEDEVICE_DRIVER_SSH_KEY_PATH
          value: "/etc/ssh/ssh_host_rsa_key"
        - name: EDGEDEVICE_DRIVER_HTTP_PORT
          value: "11112"
        - name: EDGEDEVICE_DRIVER_EXEC_TIMEOUT_SECOND
          value: "5"
        - name: EDGEDEVICE_DRIVER_SSH_USER
          value: "root"
      volumes:
      - name: edgedevice-config
        configMap:
          name: rpi-gpio-configmap-0.0.1
      - name: gpiomem
        hostPath:
          path: /dev/gpiomem
      serviceAccountName: edgedevice-sa

Please note that in the Deployment file we need to add privileged: true to the container's securityContext, so that we can use the Raspberry Pi's GPIO in the container, then mount the Raspberry Pi's /dev/gpiomem to the container as volume.

Write a Kubernetes Service YAML file to proxy requests from the deviceShifu to the real Pod from the domain.

apiVersion: v1
kind: Service
metadata:
  labels:
    app: edgedevice-rpi-gpio-deployment
  name: edgedevice-rpi-gpio
  namespace: default
spec:
  ports:
  - port: 80
    protocol: TCP
    targetPort: 8080
  selector:
    app: edgedevice-rpi-gpio-deployment
  type: LoadBalancer

A Kubernetes ConfigMap YAML file to configure deviceShifu.

apiVersion: v1
kind: ConfigMap
metadata:
  name: rpi-gpio-configmap-0.0.1
  namespace: default
data:
  driverProperties: |
    driverSku: RaspberryPiB+
    driverImage: edgenesis/rpi-gpio-python:v0.0.1
    driverExecution: "python led_driver.py"
  instructions: |
    pin:
    operate:
    help:
# Telemetries are configurable health checks of the EdgeDevice
# Developer/user can configure certain instructions to be usedas health check # of the device.
# of the device. In this example, the device_health telemetry ismapped to
# "get_status" instruction, executed every 1000 ms
  telemetries: |
    device_health:
      properties:
        instruction: help
        initialDelayMs: 1000
        intervalMs: 1000

In ConfigMap we need to configure the execution path of the driver, because we put the Python file directly under the default path when generating the image, just fill in python led_driver.py here. If the driver is a binary file, you can directly fill in the binary directory here.

Write a Shifu EdgeDevice YAML file to generate the device twin.

apiVersion: shifu.edgenesis.io/v1alpha1
kind: EdgeDevice
metadata:
  name: edgedevice-rpi-gpio
  namespace: devices
spec:
  sku: "RaspberryPi 3B+"
  connection: Ethernet
  address: 0.0.0.0:11112
  protocol: HTTPCommandline

Put these four files into the Raspberry Pi with the following directory contents.

led-deploy/
├──deviceshifu-rpi-gpio-configmap.yaml
├──deviceshifu-rpi-gpio-deployment.yaml
├──deviceshifu-rpi-gpio-service.yaml
└──edgedevice-rpi-gpio-edgedevice.yaml

Use kubectl apply -f <dir> to deploy deviceShifu to the k3s cluster:

Then check the running status with kubectl get pods.

View all device twins in the cluster with kubectl get edgedevices -n devices.

To see the details of the digital twin via describe.

Next we can interact with the device, where we deploy an nginx container to represent the application in a real-world scenario The deployment command is kubectl run nginx --image=nginx

Then execute kubectl exec -it nginx -- bash to get to the nginx command line

Finally, using curl to send commands to the device, the driver accepts commands in the format: python led_driver --pin <x> --operate <on/off>

Using Shifu to send commands will convert from HTTP to command line, the request address is written as: http://edgedevice-rpi-gpio/pin?flags_no_parameter=<pin>,--operate,<on/off>

Outcome​

The program can control the LED bulb on/off by sending an HTTP request directly to the device's domain name.

At this point, we have successfully plugged the Raspberry Pi driver into Shifu.