Instructions to build the images on a Windows box using the Linux shell:

Prerequisites

• Windows Linux bash shell
• In Linux shell, install the latest Node 8.x in the Linux shell: https://askubuntu.com/a/548776/639573
• In Linux shell, install dotnet 2.0: https://www.microsoft.com/net/download/linux
• Install Docker for Windows

Enlisting and building the function runtime on Linux

• Clone https://github.com/Azure/azure-webjobs-sdk-script.git into a folder windows can access (e.g. /mnt/d/code/GitHub/azure-webjobs-sdk-script-linux)
• Check out the Core branch
• Run dotnet publish WebJobs.Script.sln

Getting the correct grpc native binary

The default build only gets 32 bit binaries, and we need the 64 bit (to match the Linux Node)

• From azure-webjobs-sdk-script repo root, cd src/WebJobs.Script.WebHost/bin/Debug/netcoreapp2.0/publish/workers/node/grpc
• npm install -g node-pre-gyp
• node-pre-gyp install

Building the docker images

Now, we leave the Linux shell and use Docker for Windows (because Linux bash on Windows can't run Docker).

Building the base image with just Core and Node (no Functions stuff yet)

From this repo's root, go into base-container and run

docker build -t core-and-node .

Building the base Functions Runtime image (no user function files)

• From this repo's root, go into base-runtime-container
• Create a junction to the runtime you created earlier. Yes, this needs to be made easier, maybe using a git submodule. e.g. (adjust for your Functions repo) junction FunctionsRuntime D:\code\GitHub\azure-webjobs-sdk-script-linux\src\WebJobs.Script.WebHost\bin\Debug\netcoreapp2.0\publish
• docker build -t azure-functions-runtime .
• docker tag azure-functions-runtime davidebbo/azure-functions-runtime
• docker push davidebbo/azure-functions-runtime

Creating an App Service using the mounted App Service storage

• Create a new Web App for container, using davidebbo/azure-functions-runtime as the image
• Set WEBSITES_ENABLE_APP_SERVICE_STORAGE=true in the App Settings so that the App Service storage gets mounted
• You can now deploy your functions files using any technique (FTP, Kudu git, msdeploy, ...). WARNING: file change notifcations don't work, so you need to restart your App Service each time you make changes!

Building a custom image with some function files baked into it

• From this repo's root, go into user-functions-container
• docker build -t davidebbo/functest .
• Testing locally: docker run -p 8000:80 davidebbo/functest
• Test the C# function: http://localhost:8000/api/HttpTriggerCSharp?name=David
• Test the Node function: http://localhost:8000/api/HttpTriggerJS?name=David
• Push the custom image to docker hub (replace with your account): docker push davidebbo/functest

Using the custom image in App Service

• Create a new Web App for container, using your test image as the image (e.g. davidebbo/functest)
• Keep The Default WEBSITES_ENABLE_APP_SERVICE_STORAGE=false, since we are not using App Service storage here
• You can now hit your test functions in Azure

Running the test on Azure Container Instances (ACI)

• Install the latest Azure CLI v2 (i.e. az)
• az group create --name FuncACI --location eastus
• az container create --name funccore --image davidebbo/functest --memory 0.8 --resource-group FuncACI --ip-address public
• Run az container show --name funccore --resource-group FuncACI until you see "provisioningState": "Succeeded"
• Get the IP address and hit it in the browser
• Now append the same test paths as above, e.g. http://52.170.33.146/api/HttpTriggerCSharp?name=David