• 0 Posts
  • 9 Comments
Joined 1 year ago
cake
Cake day: September 13th, 2024

help-circle
  • If you decide to go the Kunernetes route, you can try k3sup to bootstrap your VMs k3s, it a nice half step abstraction between Ansible and running curl yourself:

    https://github.com/alexellis/k3sup

    I’ve landed on k3s as my k8s distro in my environment for a number of reasons. It seems to have the “mindshare” of selfhosters, and theres lots of k3s documentation to peruse. I also really like that you can preload manifest files if you do decide to use Ansible, which makes cluster deploys that much more organized.

    If you want to go a little off beat, you could try “Canonical K8s (not Microk8s)” as a snap. That worked REALLY well, and lets you do cool shit like “k8s enable loadbalancer” to automatically enable whole components for you, if you just want to focus on “consuming” Kubernetes instead of building it. I did notice a little overhead doing it as a snap, but my Proxmox node that runs the VM is purposely low spec (Celeron quad core if you believe it, 7 tdp tho)…so your hardware wouldn’t likely notice a difference.

    https://documentation.ubuntu.com/canonical-kubernetes/release-1.32/snap/tutorial/getting-started/

    If youre doing Proxmox already, if you don’t already have a VM template and/or Terraform/OpenTofu with Proxmox operator…it may help to tool on that too. Easier to destroy/build VMs when you get frustrated.


  • I am a container evangelist, I find excuses to convert my jobs into Kubernetes workloads, and I frequently use the likes of podman for one off apps/processes and development. I use Flatpak frequently to isolate dependencies for the likes of Steam and Heroic.

    I really wanted to like Bazzite or Bluefin, but I can’t deal with the overhead from the rpm os-tree updates. I would frequently notice hitches for my use case (sunshine streaming), and the hoops I had to do to configure Nvidia drivers (for it to then not work as good as other distros) was tiresome.

    I went back to Arch (EndeavourOS), and I improved sunshine performance and had a driver that worked with less fiddling.

    I’m saying all this because, while I’m glad to see any Linux distro grow, I hope it starts delivering what it says on the tin eventually without compromises that I experienced. Markering on it being immutable and container focused is true, but I dont see the benefit (aside from more stability which as others pointed out, is already stable is most cases)?. Right now, its a simple to configure (assuming most defaults work for your setup) distro that is finding a growing niche amongst some users (obviously by the data shown). And thats good enough for now at least.





  • Like others have stated, I go by how often I use an item. How often I’ve touched it over a few months, do I think about it, etc.

    I also look for items that may have a dual purpose. For instance, do I need a meat shredder or will my hand mixer that I use for other tasks work for it? (It does) I also use this logic to prevent acquiring new items which is the other half of this equation. You don’t have to purge what you don’t acquire.

    The last thing is being honest with myself. There are some hobbies I used to love doing, but have no desire now and will probably not have time to do anyways. So items related to those get sold/donated. It may seem financially wasteful, but it also provides a test for yourself on whether you actually will use the items. If you end up buying some of those things again, forgive yourself and keep them next time you think about it.


  • I think the issue for some people (why they may buy expensive hardware) is that their server is not “enterprise grade”, literally meaning a whole server rack with a SAN, firewall, etc. If you’re new to this hobby, please consider this unsolicited advice:

    Use whatever hardware you already have or buy only what you need to achieve your goals.

    Some people want to “cosplay as a sysadmin” like what Jeff Geerling sells on his tshirts. That can mean doing this stuff for fun or maybe self teaching for a job. For those folks, buying “enterprise” could possibly make sense. But I would argue that even the core concepts of that hardware can be learned on stuff you already have.

    Enterprise hardware is loud, inefficient, and will likely have idiosyncrasies that making them run at home kinda suck. An old laptop is perfect as a place to host stuff or play with software.

    One of the things engineers/admins have to do in a datacenter is plan for rack power efficiency. That often means planning for the capacity you are going to use, for the space you have and choosing the cheapest solution for that.

    I think its considered generally more impressive with how much you can do within the constraints you have, vs having so much capacity for a cheap price. Like, how many services can you run on a Raspberry Pi? Can you create “good enough” performance for a storage area network using just gigabit? The skills you get by limiting yourself probably out perform working with “the real stuff”, even if your purpose is trying to get a job. I’d argue the same for folks who simply want to self host. Run what you got until it stops, and then try to buy for capacity again.

    Your power bill, the environment, and your wallet will thank you.