I’d considered doing something similar at some point but couldn’t quite figure out what the likely behaviour was if the workers lost connection back to the control plane. I guess containers keep running, but does kubelet restart failed containers without a controller to tell it to do so? Obviously connections to pods on other machines will fail if there is no connectivity between machines, but I’m also guessing connections between pods on the same machine will be an issue if the machine can’t reach coredns?

Cloudflare would probably meet my technical needs, but I refuse to give them any money due to how enthusiastic they are to have white supremacists on their platform

I’ve started a similar process to yours and am moving domains as they come up for renewal, with a slightly different technical approach:

• I’m using AWS Route 53 as my registrar. They aren’t the cheapest, but still work out at about half the price of Gandi and one of my key requirements was to be able to use Terraform to configure DS records for DNSSEC and NS records in the parent zone
• I run an authoritative nameserver on an OCI free tier VM using PowerDNS, and replicate the zones to https://ns-global.zone/ for redundancy. I’m investigating setting up another authoritative server on a different cloud provider in case OCI yank the free tier or something
• I use https://migadu.com/ for email

I have one .nz domain which I’ll need to find a different registrar for, cos for some reason route53 doesn’t support .nz domains, but otherwise the move is going pretty smoothly. Kinda sad where Gandi has gone - I opened a support ticket to ask how they can justify being twice the price of their competitors and got a non-answer

This is relevant to my interests, thanks. Looks like it’s pretty early stages though?

Cool - was trying to get set up with v1.94, but had real trouble getting pgvecto-rs to work properly, pgvector seems much more stable and better supported and was a breeze to get running

• An HP ML350p w/ 2x HT 8 core xeons (forget the model number) and 256GB DDR3 running Ubuntu and K3s as the primary application host
• A pair of Raspberry Pi’s (one 3, one 4) as anycast DNS resolvers
• A random minipc I got for free from work running VyOS as by border router
• A Brocade ICX 6610-48p as core switch

Hardware is total overkill. Software wise everything is running in containers, deployed into kubernetes using helmfile, Jenkins and gitea

Something like odoo (https://www.odoo.com/) might work?

You probably aren’t going to find something that works for your specific needs right out of the box, so your best bet would be finding a platform that gets you 80% of the way there and provides enough of a plugin mechanism that you can develop the remaining 20% of the functionality yourself

This is something I’m also interested in; if you find something please update us

• There has been some technical decisions over the last few years that I don’t think fit my needs terribly well; chief of these is the push for Snaps - they are a proprietary distribution format, that adds significant overhead without any real benefit, and Canonical has been pushing more and more functionality into Snap
• I previously chose Ubuntu over Debian because I needed more up to date versions of things like Python and PHP, with Docker this isn’t really a concern any more, so slower, more conservative approach Debian takes isn’t as big of an issue

Ubuntu LTS, but in the process of replacing it with Debian

From the previous issue it sounds like the developer has proper legal representation, but in his place I wouldn’t even begin talking with Haier until they formally revoke the C&D, and provide enforceable assurances that they won’t sue in the future.

Also I don’t know what their margins are like, but even if this cost them an extra $1000 in AWS fees on top of what their official app would have cost them (I seriously doubt it would be that much unless their infrastructure is absolute bananas), then it would probably only be a single-digit number of sales that they would have needed to loose to come out worse off from this.

The irony is just chefs kiss,

https://glitchtip.com/

API compatible, but lower resource consumption - is missing some of the newer features (big one for me is tracing, but just install Tempo).

Not actually tried it, but looks promising

Yeah, this is an absolute blocker for me. If its not supported upstream then it’s a no-go. I don’t want to be running whatever hacked up Ubuntu image the manufacturer put together then stopped updating in 3 months when the next iteration gets churned out

Infrastructure as code/config as code.

The configurations of all the actual machines is managed by Puppet, with all its configs in a git repo. All the actual applications are deployed on top of Kubernetes, with all the configurations managed by helmfile and also tracked in git. I don’t set anything up - I describe how I want things configured, and the tools do the actual work.

There is a “cold start” issue in my scheme - puppet requires a server component that runs on Kubernetes but I can’t deploy onto kubernetes until the host machines have had their puppet manifests applied, but at that point I can just read the code and do enough of the config by hand to bootstrap everything up from scratch if I have to

So some fun facts:

• If you buy enough licenses from Microsoft, instead of giving you a bunch of unique licence keys to keep track of, they will give you a license that you can install on a server, and a special “volume license key” that you use on every machine - then, instead of talking to Microsoft to activate themselves, they connect to your server which ensures that it is only activating as many machines as you have licenses for
• These volume license keys are public knowledge to the point that Microsoft publish them on their site because they are useless unless you have a server to validate the activations
• The server protocol is not complex, has been reverse engineered, and there are open source server implementations that forget the whole “ensure you have the right number of licenses” part

This. If there was something profitable to do with old computers besides sell them, the ewaste merchants would have figured it out by now

“GPT” describes a machine learning technique - tools like ChatGPT use this technique along with massive training sets to produce their results, but there isn’t anything stopping you doing it on a small scale so you can slap “GPT” or “AI” on your product and jump on the band wagon without actually adding anything of value

What do you mean by “increase security”? Security isn’t a thing where you get +5 points for every antivirus you have installed - it’s about risks, and how you mitigate them. A perfect antivirus isn’t going to protect you if you have a crappy password on something you forgot about, or if you are running software with a serious security vulnerability.

You can install OpenStack, but I’d probably not let any random person run code on your machine