You will likely want to run the Azure IoT Edge Runtime from your own Azure Container Registry (ACR). While this will be officially supported eventually, right now it is not, but there is a workaround. It isn’t officially supported out of the box because the runtime deploys the “edgeHub” container under-the-covers from Docker Hub and there’s no obvious way to override that. I’ve been working with the Azure IoT team over the last couple of days to unblock a customer who has network restrictions that prevent them from running the Edge from anything other than the West Europe Azure Region. Here’s everything that we had to do to get it running. Hopefully this helps you out while the Azure team gets this officially implemented.
The first thing you are going to want to do is go to the azure-iot-edge-config repository and learn about the two types of Edge configuration: Runtime and Module configuration. Come back to this post after you have absorbed that info.
You are first going to pull the Azure IoT Edge containers down to your local machine, tag them and then push them to your own ACR.
Open a command prompt and execute the following statements to pull the Azure IoT Edge runtime modules down to your machine.
docker pull microsoft/azureiotedge-agent:1.0-preview
Tag each of the images with ACR URI:
Replace 'myregistry` with the name of your registry
docker tag microsoft/azureiotedge-agent:1.0-preview myregistry.azurecr.io/azureiotedge-agent:1.0-preview
You will now push your newly tagged images to your ACR. But first you need to login to your ACR.
docker login myregistry.azurecr.io -u xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx -p password
You can get the ACR credentials from the Azure Portal, under Access Keys
Now push those images to your ACR.
docker push myregistry.azurecr.io/azureiotedge-agent:1.0-preview
You will now see your images under the ‘Repositories’ section of your ACR.
You now need to change two config files: Runtime and Module Configuration.
When you run
iotedgectl setup you can pass it a custom config file, which will contain your ACR credentials. See azure-iot-edge-config for more information about IoT Edge configuration requirements.
Here’s how to get your config file setup:
1. Copy this file (runtimeconfig.json) to your local machine.
2. Open file and replace all the tokens [[enter …]] with your values for your ACR URI, password, username, connection string and hostname.
3. Save the file.
Now when you setup the Edge Runtime you pass it a
--config-file parameter to your new config file.
iotedgectl setup --config-file config.json
Then run start, and you’ll see it pull from your ACR instead of Docker Hub
View the docker logs and you should see this message, which is normal. It says it is waiting for a module config, which we will do next.
docker logs edgeAgent -f
2017-11-23 04:26:57 [ERR] - Error refreshing edge agent configuration from twin.
Now we need to apply a custom configuration to the module twin config in the cloud. See azure-iot-edge-config for more information about IoT Edge configuration requirements.
git clone https://github.com/jonbgallant/azure-iot-rest.git
5. Open command and navigate to
data-plane\devices and execute the following to apply your new configuration file:
python device-conf.py --name [iothubname] --key [iothubkey] --device-id [deviceid] --config-file [path to module config]
You will see this output:
If you go back to your docker logs:
docker logs edgeAgent -f
You will now see the runtime pull that new module configuration file and run all the appropriate containers from your ACR.
And you can verify that the events are being received by using Device Explorer or the CLI monitor-events method described on this post.
That should do it!
Trust me, this experience will get better, but for now this is what we have to deal with.
Please comment or reach out if you have any issues.