PXE is unnecessary unless you’re going to be creating a reusable boot image. Just faster to use LiveUSB.

What did you getaid off, and what are you trying to apply to? Maybe help to understand on what you’re trying to learn.

Just for your own sanity, just install Talos on the 3 machines, understand how to join them to a cluster, then deploy some stuff around the cluster. Get a feel for the basics before you get into the mess of trying to do it all in VMs.

I’d also check some comparisons on the various flavors of different lube stacks: k3s, microk8s, kubedge…etc. Theres so many now it’s hard to track.

Because you’re being rate limited. Don’t let these tools constantly hammer Github’s API in mass fits and starts or you may get a back off from GH.

Neh

And I got downvoted into oblivion for bringing it up 🤣

Then why if you aren’t familiar would you make a comment you didn’t see anything?

Do you randomly walk into other people’s jobs with zero proficiency and speak to how they’re doing at it?

Here’s a very simple list of issues that any Node dev would immediately say is generated and has not been cleaned up:

• package.json
• knip (check that tab usage)
• WILDLY fucking stupid Dockerfile
• What directory am I in here??? Note the dupe code because whatever shit tool lost track
• What in the absolute fuck?

I mean I can keep going, but if you even glanced at this and didn’t IMMEDIATELY get it, you are bad at your job.

Edit: This one is fucking hilarous

Lolwut??? Did you check the GitHub at all?

Check the code

This is so vibecoded 🤣 Nawthx

Logs

And literally everything else can do it better.

But at least you got that.

😘

No. Just…no.

The failure rate is going to be absolute INSANE as well.

I didn’t miss it, but didn’t loop back. Apologies.

I disregarded that as a solution in my response to that, because it’s not really a solution to OP’s request. Yes, they are cheaper. No, they are not functional for this need due to lack of PCIe. Running SSDs on these devices is not a feature because of the bus speed and connection limitations.

Sure it’s possible. No, it’s not functional for the needs requested here, or even a good suggestion. If somebody wanted a RELIABLE backup target using SSDs, this is the last possible scenario I would even suggest, and only if working from a box of scraps.

I’m not discounting your point that it’s cheaper at all, but it’s like…okay…if someone asked me where to get steak, because they need steak for a recipe they are cooking for dinner, my response shouldn’t be “Well, you could get steak right there, but it costs $X, and you can get Chicken wayyyyyyyy over there. It’s not beef, and it’s not what your recipe calls for, but it’s cheaper and possible to get.”

You’re asserting a position into justification for an argument that doesn’t exist. OP isn’t asking what they could theoretically run backups to. That could be an esp32 board for even cheaper. It’s also an even worse solution than an RPi. It’s just not what they’re asking for is my point.

This ONLY works at an insane scale. This will never hit the consumer market.

Again…I think you’re just missing the point here and trying to justify a worse solution without cause. A cheaper, more functional solution exists already. Trying to assert “you don’t need this or that” is not useful.

If I went to a car dealership and they told me a newer model of a car I wanted was in stock for $X, I’m not going to say “Okay,bsure, but I want the shitter version, and I’ll pay more for it.”

Makes zero sense.

Then something has changed about the local deployment and concentration of the network near you. Don’t know what to tell ya 🤷

As long as the provider is the same, and your instances are using properly using DoH or DoT, you have nothing to worry about.

If you’re super concerned though, I’d be using Mullvad over Cloudflare though. Just saying.

See my other response. This is quite normal.

Yes, that’s called Round-Robin Load Balancing.

To get more specific, your DNS provider spins up a large number of DNS resolvers out in the world on a CDN network that resolves clients to the most geographically convenient server(s) at any point in time based on the GeoIP info of your public IP.

Once you resolve one set of addresses at any given time, it caches your request, so the next time you ask these DNS servers for something you’ll get a response right back from them as fast as possible.

You constantly checking is just going to show this. It’s quite normal.

You’re going out of your way to prove some unnecessary point with this solution though.

Only the RPi5 has PCIe, first of all, and the older boards would need a slow USB interface for any type of larger storage. Then you have longevity and reliability questions because of the age of the boards…it’s just worth it.

OP wants a simple solution. RPi of any kind just ain’t it when you get down a simple list of Pros v Cons list.