v2_SNARE

Dockerized ROS 2 Galactic RViz2 environment with browser access through noVNC and a small Node.js control panel for launching isolated AWS ECS Fargate tasks.

What This Includes

• ROS 2 Galactic base image with RViz2 installed.
• Browser terminal through ttyd.
• Virtual desktop stack using Xvfb, openbox, x11vnc, websockify, and noVNC.
• rviz2-wrapper script that starts RViz2 on the virtual display.
• Express control-panel API that launches ECS Fargate tasks per student.
• GitHub Actions workflow for building the Docker image, pushing it to ECR, and rendering an ECS task definition.

Project Structure

.
|-- control-panel/
|   |-- package.json
|   `-- server.js
|-- docker/
|   |-- Dockerfile
|   |-- ros_entrypoint.sh
|   |-- rviz-novnc.html
|   |-- rviz2-wrapper.sh
|   `-- supervisord.conf
|-- .github/workflows/deploy.yml
|-- docker-compose.yml
|-- task-definition.json
`-- README.md

Prerequisites

• Docker and Docker Compose.
• Node.js 18 or newer for the control panel.
• AWS account with ECS, ECR, IAM, VPC subnet, and security group configured.
• AWS credentials with permission to run ECS tasks and push images to ECR.

Run RViz2 Locally

Build and start the container:

docker compose up --build

Open the browser services:

• noVNC desktop: http://localhost:6090
• browser terminal: http://localhost:8081
• direct VNC: localhost:5901

From the browser terminal, launch RViz2:

rviz2-wrapper

The wrapper sets DISPLAY=:98, sources ROS Galactic, and starts rviz2 inside the virtual desktop session.

Control Panel API

Install dependencies:

cd control-panel
npm install

Create a .env file in control-panel/:

AWS_ACCESS_KEY_ID=your-access-key
AWS_SECRET_ACCESS_KEY=your-secret-key
PORT=3000

Start the API server:

npm start

Launch a lab environment:

curl -X POST http://localhost:3000/api/launch-lab \
  -H "Content-Type: application/json" \
  -d "{\"studentId\":\"student-001\"}"

The API launches the ECS task definition rviz2-novnc-task in the ros-learning-cluster cluster and passes STUDENT_ID into the container.

AWS Deployment

Deployment is handled by .github/workflows/deploy.yml on pushes to main.

The workflow:

1. Authenticates to AWS.
2. Logs in to Amazon ECR.
3. Builds the Docker image from docker/.
4. Pushes the image to the rviz2-novnc ECR repository.
5. Replaces REPLACE_ME_IN_PIPELINE in task-definition.json with the pushed image URI.
6. Registers/deploys the task definition to ECS.

Required GitHub environment/secrets:

• Environment: v2_snare_light
• AWS_ACCESS_KEY_ID
• AWS_SECRET_ACCESS_KEY

Current AWS settings are configured for:

• Region: ap-northeast-1
• ECS cluster: ros-learning-cluster
• ECR repository: rviz2-novnc
• ECS task family: rviz2-novnc-task
• Container name: rviz2-container

ECS Task Ports

The task definition exposes:

• 8080: ttyd browser terminal inside the container.
• 6090: noVNC web UI.

The local Docker Compose file maps these as:

• Host 8081 to container 8080.
• Host 6090 to container 6090.
• Host 5901 to container 5901.

Configuration Notes

The ECS cluster, subnet, security group, and task definition names are currently hard-coded in control-panel/server.js. Update these values before using a different AWS account or VPC:

cluster: 'ros-learning-cluster'
taskDefinition: 'rviz2-novnc-task'
subnets: ['subnet-08f8aad16309319ec']
securityGroups: ['sg-02390fef96a5c1ea4']

The task execution role is configured in task-definition.json:

"executionRoleArn": "arn:aws:iam::221131121759:role/ecsTaskExecutionRole"

Security Notes

• Do not commit .env files or AWS credentials.
• Remove credential debug logging from control-panel/server.js before production use.
• Restrict security group ingress for ports 8080, 6090, and 5901.
• Consider using IAM roles or OIDC for GitHub Actions instead of long-lived AWS access keys.

Useful Commands

Build the RViz2 image directly:

docker build -t rviz2-novnc ./docker

Run the control panel:

cd control-panel
npm start

Check running Docker services:

docker compose ps

Stop local services:

docker compose down