Azure IoT Edge order of deployment operations

Time needed: 10 minutes

How to deploy/overwrite/not mess up your IoT Edge Module configuration?

1. Runtime instantiation

   Installing your runtime will provide you with $edgeAgent and $edgeHub, but with no custom modules.
   
   You can modify these using the Azure Portal UI (not reversible, but can always be redone), an actual base/single-device deployment (irreversible and will take priority over the former), or theoretically also by generating a Single-Device deployment from VS Code (I wouldn’t really do this for any production environments, though!)
2. Base deployment

   Creating a “base deployment” initiates a “baseline” deployment for your device. You can deploy using Az CLI or the Azure Portal (in which case it’ll be a “base deployment” that you can target many devices.
   
   At the time of writing this (Jan 2022), as long as you’re using a “full” deployment manifest schema, any deployment you’ll create using Az CLI (no matter whether you’re using the –layered switch or not) will be a base deployment.
   
   You can modify $edgeAgent and $edgeHub, as well as your modules and routes.
   
   This configuration will automatically overwrite any changes to modules that you might want to do using Azure Portal UI afterward, so be mindful of what you’re deploying! You might not be able to remove it. Ever.

3. Layered deployments

   Each layered deployment (a deployment that uses the layered deployment manifest file schema AND is deployed with –layered switch) will take precedence over any earlier layered deployments with equal or lower priority, and be incremental over any base deployment or runtime configuration you have created.

4. A “Configuration” deployment

   You can create a configuration deployment from the Azure Portal UI under “IoT Hub > Device Managements > Configurations”.
   
   You can use this to overwrite configurations for both Devices and Modules. You can overwrite contents in their Digital Twins, such as “properties.desired.settings”.
5. Adding new modules using Azure Portal UI (such as using “Set Modules”)

   I know, I know – it’s getting convoluted! But you can also add the new modules from the Azure Portal UI, by using “Set Modules”.
   
   This view makes it possible for you to add a new module (well, container) details and it’ll be added to your IoT Edge host (.. Docker host, that is).
   
   
   
   Any module configurations will (probably?) be overwritten by any deployments that match them (i.e. the targetCondition matches and there’s a module or a system module (“runtime settings” in the screenshot above) configuration that differs from your settings here.

6. Modifying runtime values using Azure Portal UI (such as using the “Module Identity” editor)

   You can modify the digital twin values and environment variables using the “Set Modules” option, or from the “Module Identity” editor directly.
   
   Module twin values (and probably any environment variables) will be overwritten to whatever is provisioned in the deployments, though – so any layered or base deployment WILL take precedence (after a minute or two), and any Configuration will take precedence (possibly a bit slower – no idea why).