Nice write up, but there’s just going to be a LOT of false positives.

And that’s not even letting on that literally ALL DNS queries work from cache unless you are specifically doing a live query.

None of your software is. It’s asking your OS. Your OS is asking your resolver service. Your resolver service is asking your router. Your router is 5000% caching DNS queries.

So you’re asking if there is any other way to work around physics and get a better response time to servers that are thousands of miles away?

No.

Sorry.

DLNA

Sure it’s not a proprietary binary DB, right? Probably just sqlite or something? Bet you can just dump it.

But yeah, switch to something without those problems. There’s lots of them out there.

Every part of the world is. It’s not unique to the US. Laws need to be passed to stop this.

If your requirement is a GUI, you’re not going to find anything. HA Proxy is also the most performant out of anything out there last I looked, and it’s got one of the simplest configuration setups.

• Traefik: kinda slow, mostly meant for large numbers of microservices, pretty verbose configuration
• Envoy: middle of the road, also most meant for cloud services, but should work with anything
• nginx: does have a popular 3rd party GUI, seems to be confusing for most that don’t work with it a lot
• caddy: fewer LB specific options if you’re just talking about service routing and response time, pretty easy to confirm for most, and some sort of decent 3rd party UIs, but they won’t have all the options available.

First, I think you’re attacking this from the wrong angle. You’re focused on ECC memory for some reason, but that’s not going to prevent bitrot, just potentially reduce errors in transfer, or catch issues. Your filesystem of choice has more to do with degradation in storage.

Second, you haven’t mentioned any of the boards and their storage capabilities. Do they support the correct number of drives you want to use? Do they support hot-swap, and is that even something you care about?

Last, you want more services, and but are worried about power consumption…that’s not how that works. More services means more CPU and MEM util, which means more power usage. You can either constrain your TDP at that point by using an UNDERpowered CPU and have that tradeoff, or provide a more capable CPU and take an increased TDP. There is no third option, that’s just how it works. Pick the more capable CPU and take the power hit (really, it’s going to be minor compared to a large server), and just run the things you need to run instead of coming back in a year and wanting to flip it again.

Yes, if Tailscale on your router is advertising routes, and your other devices while connected to Tailscale are picking up those advertised routes, they won’t be able to figure out how to get to your local network devices if both things are advertising the same routes.

Tailscale is a group of clients on a Tailnet which are all equal, unless you tell it otherwise. That means you need to set the client you installed on your router as a subnet router.

Even then, if you’re not familiar with networking, you’ll probably have duplicate routes if you’re not paying attention. The other option is to just install Tailscale on each server you want access to.

Your default routes are being set incorrectly. If you’re using it as an exit node, then you need to make sure it’s only being used as such for other clients on the Tailnet. You also need to make sure you’re splitting your routes correctly so that the default route on your router isn’t set for something on the Tailnet.

Generally speaking, if you’re not familiar with networking and routing, you don’t need to change the subnet settings if using a Tailscale client on your router. You also shouldn’t be advertising routes from it for your own network, or else you could end up getting issues like you’re seeing because your routing tables will be broken while Tailscale is active.

One more thing: Tailscale on your router doesn’t make it a server, it’s still a Tailscale client. You still need to setup your routing in the Tailscale server to make sure it’s not duplicating routes like this.

“Old PC” meaning what, exactly? Need some specs.

In general, a Qnap or Synology box is going to be much lower wattage than a full PC.

It’s not “standard practice”, it’s the least problematic way to distribute example templates for mass consumption without needing extra steps for the user to configure. It is assumed that anyone looking to run something in production would follow best practices and harden where needed.

Remember when everybody was making “smart” toasters and fridges and shit with cameras or WiFi for absolutely no reason.

This is that all over again.

Nobody needs “AI” in their water kettles, dryers, or dildos.