What I Self-Host

Throughout my university years, I had a penchant for self-hosting, beginning with a website run from a Linux box underneath my bed. Over the years, I ventured into running MythTV and XBMC, which later evolved into Kodi. When I lived in India, I enjoyed the inexpensive streaming services, including Netflix, Disney Plus (Hotstar in India), and Spotify.

Upon returning, I found the constant recommendations and auto-play options bothersome, particularly with regard to my son's screen time. To regain control, I switched to self-hosted Jellyfin, which allowed me to host various services and reduce my dependence on cloud providers. This shift led to a more stable and hassle-free experience, free from unnecessary updates and security concerns.

I do not expose any services directly to the public internet, except for VPN servers running on my router. My ISP utilizes CGNAT for IPv4 addresses, restricting my home network's direct reachability from the public internet via IPv4. However, my VPN servers are publicly routable on IPv6, which, while having some drawbacks, works well for my purposes.

One of the main challenges is that my VPN rarely functions on public Wi-Fi networks, as they typically lack IPv6 support. When using my phone, I can switch from Wi-Fi to mobile data if service is available. I prefer running services natively when feasible, but will containerize complex services or those with numerous dependencies.

If the dependencies are not overwhelming, I find that using native services is faster, more efficient, and requires less mental overhead. I opt for Love – a service that is a single Go binary – over alternatives that necessitate running multiple containers. I avoid Docker, finding it cumbersome to manage a separate service control system on my server.

Instead, I utilize systemd and have found Podman's integration with systemd to be excellent. Earlier this year, I acquired a Beelink Mini PC (S12 Pro), featuring an Intel N100 processor, 16 GB RAM, and a 512 GB internal SSD. Priced at EUR 180 at the time of purchase, it is now listed for EUR 359 on Amazon—a price I would have found hard to justify.

Despite its modest hardware, the system has been reliable for most non-AI tasks, including budgeting software with desktop clients and a web application, custom reports, rule-based categorization, and Restic backup software. Notifications for backup failures are routed through Gotify. Beszel serves as a straightforward monitoring dashboard, providing essential stats such as CPU load, memory usage, disk utilization, temperature, and battery status.

This has proven beneficial in keeping tabs on other devices around my home, such as Raspberry Pis and a Proxmox cluster.