Adding Observability to My VPS: Uptime Kuma and Umami

After setting up my personal VPS infrastructure on AWS, I wanted to dive deeper into Site Reliability Engineering (SRE) concepts. Since I was already paying for an EC2 instance (smol-ivan) to host my main Astro website, it made sense to utilize the remaining capacity to explore basic service monitoring and analytics.

I decided to self-host two lightweight tools:

Uptime Kuma: For uptime monitoring and service health status pages.
Umami: For privacy-friendly, simple web analytics without the footprint of traditional tracking platforms.

1. Keeping Environments Isolated

To keep operations clean, I separated the deployment lifecycle of these tools from my main website pipeline. My personal site goes through automated CI/CD runs via GitHub Actions whenever I push updates. Monitoring tools, on the other hand, follow a “deploy once and let run” pattern.

I placed Uptime Kuma and Umami in a dedicated docker-compose.yml stack on the VPS:

# Infrastructure services stack (smol-ivan)

services:
    uptime-kuma:
        image: louislam/uptime-kuma:1
        container_name: uptime-kuma
        restart: always
        volumes:
            - kuma-data:/app/data

    umami:
        image: ghcr.io/umami-software/umami:postgresql-latest
        container_name: umami
        restart: always
        environment:
            DATABASE_URL: ${DATABASE_URL}
        ports:
            - "3000:3000"

volumes:
    kuma-data:

2. Nginx Reverse Proxy & TLS Extension

Since I already had Certbot configured for TLS certificate management on my domain, adding subdomains for Uptime Kuma and Umami was straightforward.

flowchart TD
    N["Nginx Proxy"]

    N --> A["Astro Site<br/>(Port 8080)"]
    N --> U["Uptime Kuma<br/>(Port 3001)"]
    N --> M["Umami<br/>(Port 3000)"]

Configured Nginx server blocks for each subdomain.
Ran certbot –nginx to issue expanded certificates.
Routed internal traffic directly to the Docker containers.

3. Takeaways

Given that my personal website doesn’t get high volumes of traffic, adding these tools didn’t drastically change day-to-day operations. However, taking monitoring and analytics out of the “black box” category and configuring them directly on my own server made the underlying infrastructure concepts much clearer.