So the MTU of Tailscale is actually 1280, but is the connection even going through the VPN or rather through my VPS, when Uptime-Kuma is trying to connect to my local domain?

Thanks, since I access my home network and server through the public IPv4 of a VPS via Tailscale this could actually be the issue. I’ll look into it, when I find the time.

Not sure how this helps, but here you go.

Yeah, I suspect it’s simply an issue on the side of DuckDNS. :/

Yeah, it works fine through my browser. Sometimes the websites load a little longer. I feel like it’s an issue with DuckDNS as it’s seemingly random when it works and when not.

IPv6 doesn’t work:

docker exec -it Uptime-Kuma curl -6 proxmox.datenprolet.duckdns.org
curl: (6) Could not resolve host: proxmox.datenprolet.duckdns.org

Besides that the issue has disappeares since last night. I automatically restart all containers at night and moved from uptime-kuma:1 to uptime-kuma:latest. That shouldn’t make a difference, but maybe it did?

And it’s not a typo in my config, but in my post. But good catch. ;)

It’s HTTPS, what else should it be, when I monitor a domain?

What do you mean by tracker? I’m monitoring local domains, that point to local services and their respective web interfaces like Proxmox or Nextcloud. The local domains have a wildcard SSL certificate via DuckDNS.

Well, I’m monitoring the GUI of Proxmox on which I run a Debian VM which itself runs Uptime-Kuma and Nextcloud in Docker, so yes that’s on the same hardware.

Yep

Yes, Uptime-Kuma is running on the same domain as the other services, except the Nginx-Proxy-Manager, which runs on a VPS which I access via WireGuard. And yes, I’m using Docker. I tried curl’ing one of the domains from the Uptime-Kuma container and got the folllowing error: curl: (35) OpenSSL SSL_connect: SSL_ERROR_SYSCALL in connection to service.datenprolet.duckdns.org:443. So thanks, now I have an idea about what I should investigate.

Isn’t this using a lot of computing power?

What’s costs do you mean? It’s free and open source.

Immich has image and facial recognition by default and a very neat Android app. Also it’s running in my home server, which has more power if Immich needs it. In that case I’d say software should serve one purpose and serve that good. Immich is just for picture management and does that very good. Nextcloud is a cloud and the Photos app is just a small extra that can’t compete with a full-fledged software. Nextcloud runs fine on my Raspberry Pi 4, but it’s only used by me and three friends. It’s mainly limited by your network speed and disk speed I’d say. And I’m using an external hard drive without issues.

I use Nextcloud and Immich and would recommend both. Immich might be a bit overkill, but it’s also well maintained, feature-rich and has a large community. It’s super easy to set up and works great.

Borg

It’s easy to use, there are CLI-wrapper and GUIs, it’s crossplattform, deduplicates, compresses, encrypt and based on rsync. I use it for alle backups between machines and networks.

But it’s proprietary, unfortunately.

Thanks, looks promising. I’ll give it a try.

I don’t want to configure a whole Dashboard for at least CPU, RAM, Storage and Network for up to 5 hosts.

I used the following dashboard now, but it’s not really satisfying and also doesn’t really fit more than 4 nodes. https://grafana.com/grafana/dashboards/11756-hpc-node-exporter-server-metrics-v2/

Interesting, because Tailacale doesn’t use any special ports. How would that be detected? And could you maybe use Headscale on a dynamic port to circumvent that?