Shiny App - Dockerized for CapRover

This repository is a starter template for R Shiny applications, packaged with Docker and ready to deploy on CapRover.

Quick Start

We suggest building and running the example app as-is first. Then once you've got that far, go back and start adding your own app content (and building any R package as needed in the Dockerfile).

1. Build the Docker Image

The shiny user inside the container needs to read files you mount from your host machine. To make this work, build the image with your user's UID/GID:

docker build --build-arg SHINY_UID=$(id -u) --build-arg SHINY_GID=$(id -g) -t my-shiny-app .

If deploying to a server where the data files are owned by a different user, use that user's UID/GID instead. If you installed CapRover on a fresh VM, the correct UID and GID to use there are most likely 1000.

2. Run Locally with Docker

docker run -p 3838:3838 -v "$(pwd)/data_mount:/data_mount" my-shiny-app

Then open http://localhost:3838

Without Docker (for development): Open app/shiny-app.Rproj in RStudio and click "Run App". The app will read from data_mount/ in the repo root.

3. Deploy to CapRover

1. Push the Docker image to a container registry that's hooked up in your CapRover deployment.

2. Create a new app in CapRover, making sure to check Has Persistent Data.

3. In HTTP Settings, set the Container HTTP Port to 3838.

4. In App Configs > Environment Variables, add:

   APP_DATA_PATH=/data_mount
   

5. In App Configs > Persistent Directories, add a volume mount:

   • Path in App: /data_mount

• Map to a host path or named volume containing your data

6. If you need password protection, in HTTP Settings > Edit HTTP Basic Auth assign a username and password. This approach does not allow multiple different usernames.

7. Under the Deployment tab, use "Method 6: Deploy via ImageName"

Project Structure

.
├── app/
│   ├── app.R              # Your Shiny application
│   └── shiny-app.Rproj    # RStudio project file
├── data_mount/            # Put data files here for LOCAL development
├── Dockerfile
├── shiny-server.conf
└── README.md

Working with Data

Your app reads data from APP_DATA_PATH:

• Local development: Defaults to ../data_mount (relative to app/)
• Docker: Set to /data_mount via the .Renviron file in the image

In your R code:

APP_DATA_PATH <- Sys.getenv("APP_DATA_PATH", unset = "../data_mount")

# Helper to build paths
data_path <- function(...) file.path(APP_DATA_PATH, ...)

# Use it
my_data <- read.csv(data_path("my_data.csv"))

Adding R Packages

Install packages in the Dockerfile, not at runtime. Edit the Dockerfile:

RUN R -e "install.packages(c('dplyr', 'ggplot2', 'plotly'), repos='https://cloud.r-project.org')"

This keeps container startup fast and ensures reproducible builds.

Get building! Your next steps are:

1. Edit app/app.R to build your application
2. Add data files to data_mount/ for local testing
3. Update the Dockerfile to install any packages you need
4. Rename shiny-app.Rproj if you like

Troubleshooting

Application exits during initialization

If you encounter this error message on startup:

[INFO] shiny-server - Error getting worker: Error: The application exited during initialization.

That is an indicator that you might be missing an R package, or that something in the R code is not working as expected. Turn on logging by uncommenting preserve_logs true; in shiny-server.conf and check the logs on the container in /var/log/shiny-server/ for more information.

See rstudio/shiny-server#353 for more information.

Shared library / system library version errors

If your app crashes with errors like:

unable to load shared object '/usr/local/lib/R/site-library/stringi/libs/stringi.so':
  libicui18n.so.70: cannot open shared object file: No such file or directory

Or similar errors mentioning .so files (shared objects), this means an R package was compiled against a different version of system libraries than what's available in the container.

Solution: Install those packages from source in the Dockerfile so they compile against the correct system library versions.

How to identify which packages need source compilation:

1. Check the error logs in /var/log/shiny-server/ - the error will tell you which package failed to load
2. Look for the library name in the error (e.g., libicui18n.so = ICU libraries, used by stringi)
3. Compile that package and any packages that depend on it from source

Common examples:

ICU library issues (stringi/janitor):

# Install ICU development libraries
RUN apt-get update && apt-get install -y libicu-dev && rm -rf /var/lib/apt/lists/*

# Compile stringi from source
RUN R -e "install.packages('stringi', repos='https://cloud.r-project.org', type='source')"

# Compile packages that depend on stringi from source too
RUN R -e "install.packages(c('snakecase', 'janitor'), repos='https://cloud.r-project.org', type='source', Ncpus=2)"

ImageMagick library issues (magick package):

# Install ImageMagick development libraries
RUN apt-get update && apt-get install -y libmagick++-dev && rm -rf /var/lib/apt/lists/*

# Compile magick from source
RUN R -e "install.packages('magick', repos='https://cloud.r-project.org', type='source')"

General pattern:

1. Install system library (the -dev package)
2. Compile the R package from source using type='source'
3. Install remaining packages as binaries for speed