@arichtman yeah using overlay at scale does work it just requires a lot of planning and design. Ideally with some network engineers. My last experience was contrail, years ago, and it was not great.
I’ve used flannel and cilium for one-off personal things without issue. In fact #metallb worked really really well for me with #k3s in the past as well.
Here you are: Vagrant-Setup with Ansible deploying k3s and GitLab inside k3s.
https://codeberg.org/johanneskastl/gitlab_on_k3s_vagrant_libvirt_ansible
https://github.com/johanneskastl/gitlab_on_k3s_vagrant_libvirt_ansible
Currently without the Runner, that comes next.
Self-Hosting a Cluster on Old Phones - The phones most of us carry around in our pockets every day hold a surprising amou... - https://hackaday.com/2025/04/09/self-hosting-a-cluster-on-old-phones/ #handheldshacks #postmarketos #raspberrypi #bootloader #kubernetes #smartphone #unlockable #linux #k3s
@tailscale I've been using this with #k3s on my #homelab for a while and it's super cool, worth checking out.
Very easy to take a kube service and expose as `https` on your tailnet
Finally got my #Kickstarter-backed #ComputeBlade! Replacing my last #RasperryPi 4 node in the #K3S cluster—now all worker nodes are PoE-powered.
RE: "Impromptu disaster recovery"
Amos talks about how they accidentally nuked their Kubernetes based infra with one happy little accident. As per usual, an informative and fun read!
https://kevingimbel.de/link-blog/re-https-fasterthanli-me-articles-impromptu-disaster-recovery
Finally got #nextcloud configured on #k3s using their official helm chart.. That was honestly a pain, especially getting it working with #tailscale. But it's done, and seems to be working well. Hopefully no more update issues with new versions causing the docker install to go into maintenance mode.
Even managed to get it configured on #homepage.. Success!
k3s appears to have had a major upgrade in the last few days. In v1.32.2 they have finally bumped containerd to v2 and traefik to v3
it's not shipped as a default yet, but i have played with it yesterday afternoon, and it all seem to work just fine. well at least for my use cases that is
What I did in the last weeks (part 2):
All of my k3s clusters (and I have a few of them for $REASONS) are now running either #fluxCD or #argoCD. So everything #gitops now.
In general, I like the lightweightness of fluxCD, not having to run Redis and whatnot. But having a GUI is sometimes nice, even though the flux CLI is really easy to use and very intuitive.
And of course, #renovatebot is watching all repositories and sending merge requests to update things in the clusters. Nice!
Ok, #Longhorn might have just saved me here with #k3s and my storage woes. #SMB and #NFS both caused catastrophic problems with corrupted files and databases. Using #unraid as the storage backend for the later 2 but will keep it as a backup destination and network storage for more static datasets.
Thanks to everyone who helped push me in this direction. #homeassistant is officially happy again!
Ok, think I need to do the whole #k3s shared storage thing again. As of last night I have both NFS and SMB going, but neither work for #sqlite DBs of course. I get corruptions on NFS and immediate errors on SMB (can’t get exclusive locks). Thinking of trying #longhorn for the shared storage with off-cluster backup.
My primary use case is apps like #homeassistant though, so I’m not sure if I’m overthinking this. I suppose I could configure HA to use a DB like Maria but I have some apps that don’t support this at all to my knowledge, so I’d need something with proper file locking support.
Has anyone gone down this road? Am
I framing this right - Longhorn for most app storage needs, SMB/NFS for larger storages like photos / media?
From this week's ADMIN Update newsletter, @madkiss explores production-ready mini-Kubernetes installations
https://www.admin-magazine.com/Archive/2024/83/Production-ready-mini-Kubernetes-installations
#Kubernetes #containers #K8s #k0s #K3s
The new nodes are in #k3s and running well. The amount of memory is insane.. And to be honest a couple of these machines would be enough for my full workload, so this is beyond overkill.
Got the UM890 Pros configured with Debian, ready for the #k3s cluster. Thankfully they accepted the 96GB of RAM fine 
Has anyone seen such an issue with #homeassistant? I'm running it on #k3s and was doing so for months on 1.26. Upgraded to 1.32 and now seeing constant SQLite database corruptions. I'm using NFS as the CSI. Both the Linuxserver release and the native HASS ones do this. Doesn't matter if it's a fresh install or a restore from my backup.
#hass #homelab #selfhosted
Crisis averted: Don’t upgrade your #k3s/#k8s cluster remotely when your only tunnel exists on said cluster
Decided I needed to upgrade to the latest release (1.26 to 1.32) remotely while connected over #tailscale. Besides the drop, after getting home the upgrade completed without issue and it’s all up and running happily. Was wondering if I was about to experience the “why you don’t use kubernetes at home” lesson, but it seems as though k3s was the right decision still.
