End to end deployment of R Shiny in Azure using Docker, Azure Container Registry and Azure Web Apps Also used Azure DSVM Linux box during this process. The R code for the Shiny App is leveraged from the article "How to select the best performing linear regression for univariate models".
-
In the Portal, deploy a Data Science Virtual Machine (DSVM) with Ubuntu. As the Azure DSVMs come with Docker, R, Azure CLI and other useful tools already installed, this will be used as our working environment.
-
Log into the VM, create a folder for the app - My_app
-
Grab the R code for the Shiny App from the Björn Bos' Github repo and unzip into My_app
-
Create a Dockerfile in the main folder My_app. Dockerfile given here is a template leveraged from the official source.
-
Add configuration files for the Shiny app – shiny_server.conf and shiny-server.sh from above to the folder My_app. The file structure should look like:
My_appp - app/ - server.R - ui.R - global.r - about.html - data.csv - Dockerfile - shiny-server.conf - shiny-server.sh
- Navigate to your folder My_app and run the following command
docker build -t My_app .
-
Open up the port on your DSVM to allow the traffic using Azure CLI. For this app we use port 3838.
az login
Once you have successfully logged in, type:
```
az vm open-port -g <resource group> -n <VM name> --port <port#> --priority 901
```
-
Run the Docker image
docker run -p port_num:port_num My_app
Up to this point, your app is available to anyone online via the DSVM URL (www.DSVM_IP:3838). But here we assume that this DSVM will be decommissioned so we need to host our app elsewhere.
Login to the Azure CLI
```
az login
```
-
Save Docker image as a tar file
docker save -o ~/my_shinyapp.tar my_shinyapp -
Deploy image on webserver: copy file to server and run it
docker load -i my_shinyapp.tar docker run my_shinyapp
You may need to open up the port on the webserver on your own. Up to this point then, your app would be available to anyone online via the webserver URL (www.webserver_IP).
- Add docker container to Azure Container Registry. (So now we have our docker container, let’s utilize Azure Container Registry (ACR) for private container management. Azure Container Registry is your private Docker registry in Azure.) These steps can also be completed on command line
a. Create Azure Container Registry in the Portal.
b. Log into the container registry
i. az acr login --name <my_acr_name>
c. Grab the login server for the registry from the Overview section (It should be format www.<my_acr_name>.azurecr.io)
d. Tag the docker image and push
i. Use docker images to get the image you wish
ii. docker tag my_shinyapp <my_acr_name>.azurecr.io/my_shinyapp:v1
iii. Push the image to the ACR. Before running this, check the previous step was successful by running docker images again
docker push <my_acr_name>.azurecr.io/my_shinyapp:v1
e. Verify the image has been pushed in the portal by looking under the Repositories for your app. (Or using Azure CLI commands)
e. Turning on permissions for the last step: In the Portal, navigate to the Access keys under Settings and click “Enable” under Admin user (this is necessary to log in with Docker).
- Deploy container via Azure Web App for Container. In Portal create the resource Web App for containers. The name that your choose where will be the name of your app i.e. myShiny.azurewebsites.net. Make container a Linux container. Choose the appropriate App service plan. In Configure container for image source, choose Azure Container Registry.
a. Choose the correct ACR from the dropdown box
b. Choose the correct image.
c. The tag i.e. version. (v1 in this example. If none was added, the tag is latest by default)
d. Deploy the resource
Access your app online! Navigate to www.<your_app_name>.azurewebsites.net At this point, you can tear down the DSVM if you haven't already.