talos VS Flatcar

I think this is underrated as a design flaw for how Linux tends to be used in 2024. At its most benign it's an anachronism and potential source of complexity, as it's worst it's a major source of security flaws and unintended behavior (eg linux multitenancy was designed for two people in the same lab sharing a server, not for running completely untrusted workloads at huge scale).

I haven't had a chance to try it out but this is why I think Talos linux (https://www.talos.dev/) is a step in the right direction for Linux as it is used for cloud/servers. Though personally I think multitenancy esp. regarding containerized applications/cgroups is a bigger problem and I don't know if they're addressing that.

How to create a VM with the Talos image is beyond the scope of this article. Please refer to the official documentation for guidance. After bootstrapping the control plane, the next step is to deploy the Talos CCM along with a CNI plugin.

Maybe in the future I will try others systems, like Talos which is designed for Kubernetes - secure, immutable, and minimal.

I've used Talos Linux[1] on a production infrastructure. To keep a Maintainability. (Because there are no person to maintain a infrastructure 24/7)

All the configurations are made and came from YAML. So I can manage and share on Git. And able to spin a new node (or cluster) ASAP.

For my own, I'm using a NixOS as a daily driver. It's pretty great to spin up machine and environment ASAP. (I don't know why I keep saying `ASAP`, but time is a money.)

However the downside is require a strong knowledge of Nix Language. Sometime the installer crashses.

Without that, it's pretty great.

---

[1]: https://www.talos.dev/

Log aggregation: https://reclaim-the-stack.com/docs/platform-components/log-a...

Observability is on the whole better than what we had at Heroku since we now have direct access to realtime resource consumption of all infrastructure parts. We also have infinite log retention which would have been prohibitively expensive using Heroku logging addons (though we cap retention at 12 months for GDPR reasons).

> Who/What is going to be doing that on this new platform and how much does that cost?

Me and my colleague who created the tool together manage infrastructure / OS upgrades and look into issues etc. So far we've been in production 1.5 years on this platform. On average we spent perhaps 3 days per month doing platform related work (mostly software upgrades). The rest we spend on full stack application development.

The hypothesis for migrating to Kubernetes was that the available database operators would be robust enough to automate all common high availability / backup / disaster recovery issues. This has proven to be true, apart from the Redis operator which has been our only pain point from a software point of view so far. We are currently rolling out a replacement approach using our own Kubernetes templates instead of relying on an operator at all for Redis.

> Now you need to maintain k8s, postgresql, elasticsearch, redis, secret managements, OSs, storage... These are complex systems that require people understanding how they internally work

Thanks to Talos Linux (https://www.talos.dev/), maintaining K8s has been a non issue.

TIL about Talos (https://github.com/siderolabs/talos, via your github/onedr0p/cluster-template link). I'd been previously running k3s cluster on a mixture of x86 and ARM (RPi) nodes, and frankly it was a bit of a PiTA to maintain.

About a month ago I setup a Kubernetes cluster using Talos to handle my container load at home.

I considered deploying Talos a few weeks ago, and I ran into this:

https://github.com/siderolabs/talos/issues/8367

Unless I’ve missed something, this isn’t a big deal in an AWS-style cloud where extra storage volumes (EBS, etc) have essentially no incremental cost, and maybe it’s okay on bare metal if the bare metal is explicitly designed with a completely separate boot disk (this includes Raspberry Pi using SD for boot and some other device for actual storage), but it seemed like a mostly showstopping issue for an average server that was specced with the intent to boot off a partition.

I suppose one could fudge it with NVMe namespaces if the hardware cooperates. (I’ve never personally tried setting up a nontrivial namespace setup.)

I assume https://www.talos.dev/

Basically a small OS that will prop itself up and allow you to create/adopt into a Kubernetes cluster. Seems to work well from my experience and pretty easy to get set up on.

First, obligatory xkcd [0].

> This challenge/solution consumed my entire interest for that day. My dopamine hit was because I wouldn't have to do the BigBoringTask ever again.

Yep. Occasionally I have to stop and remind myself that all I'm trying to do is rename 10 files (for example), and by the time I remember the {ba,z}sh-ism for parameter substitution, I could have probably manually renamed them. I usually tell myself that it's not nearly as fun, though.

This does occasionally present detrimental facets, though. I have a homelab, and as most people with one, its primary purpose is storing and serving media files (I promise I do other things too, but let's be honest – Plex is what people care about). I run apps in K3OS, which has been dead for quite some time. The NAS is in a VM under Proxmox, and I build images with Packer + Ansible. I've been wanting to shift K3OS over to Talos [1] for some time, but I had convinced myself that it was only worthwhile if all of it was in IaC, starting from PXE. I got most of the way there, and then stopped due to work taking more of my life than I wanted. Unfortunately, around this time the NAS broke (as in a hardware failure, not a software issue), and I was refusing to bring it back until the entire homelab was up to my absurd self-imposed standards. Eventually I convinced myself this was a ridiculous punishment, replaced the dead hardware, and brought it back.

[0]: https://xkcd.com/1319/

[1]: https://www.talos.dev/

Like everything, it's context dependent, but wowzers my life has improved so much since I got on board the Flatcar or Bottlerocket train of immutable OS. Flatcar (née CoreOS) does ship with docker but is still mostly a general purpose OS but Bottlerocket is about as "cloud native" as it comes, shipping with kubelet and even the host processes run in containers. For my purposes (being a k8s fanboy) that's just perfect since it's one less bootstrapping step I need to take on my own

Both are Apache 2 and the Flatcar folks are excellent to work with

https://github.com/flatcar/Flatcar#readme

https://github.com/bottlerocket-os#bottlerocket

Shoutout to my favorites, Flatcar https://github.com/flatcar/Flatcar#flatcar-container-linux (Apache 2) and Bottlerocket https://github.com/bottlerocket-os#bottlerocket (Apache 2)

Flatcar grinds my gears in that they have their own cutesy (and ragingly stupidly named) Ignition/Butane/whatever instance provisioning file format when they deprecated the damn-near-standard cloud-init. Bottlerocket also has their own thingy, but at least they go wholesale toward static Kubernetes Pod manifests which are much easier to reason about

Sounds like you're looking for the CoreOS Linux successor FlatCar https://www.flatcar.org/

It's actually based on some ChromeOS update tools under the hood but is a regular Linux distro, just super minimal and designed to run containers.

I guess if you found my comment to be "comically hyperbolic" then replying to mine with a "comically reductionist" is fair game

So, anyway, I actually did dig up a concrete example of my experience with it, and I cannot link to the "Additional information" section but that is both why I think the thing was a mess and also why the Miroservices YT joke resonated: https://github.com/flatcar/Flatcar/issues/220

I think the CoreOS boot strategy was decomposed into a bunch of different executables, each responsible for doing their own little slice of the world. Maybe it drew inspiration from systemd in that way. But, just like my real life experience with microservices, it requires keeping a bunch of different projects and their upgrade paths in ones head, knowing their disparate config formats, and when one of them inevitably has a bug, understanding how to troubleshoot what went wrong with the system as a whole

And, again in trying to be reasonable in this discussion[1] I do also understand why one would opt for the data URI, given how much of the rest of Ignition loads content from URLs. I don't believe cloud-init has that remote content paradigm baked into in nearly the same way, so I hear you about that.

And yes, my belief is that JSON is a data-exchange format from _computer to computer_ and making people write them is a poor DX choice, IN MY OPINION. And, to reiterate, I know that CoreOS's perspective is that it is a computer-to-computer transmission from the transpiler-project-o-the-day to the Ignition binary, but that is predicated on one having access to that transpiler binary in all cases, which is quite different from the problem that cloud-init is trying to solve

fn-1: I'm sorry you got hurt by my "tire fire" outburst, and that evidently derailed this whole interaction, but it was my experience