Finally got hardware acceleration up and running on the cluster, thanks to this great tutorial

https://jonathangazeley.com/2025/02/11/intel-gpu-acceleration-on-kubernetes/

It was relatively easy compared to dealing with Nvidia drivers on Kubernetes, and I'm quite happy about the result

I have to see if enabling this on Immich would also bring some performance improvement

#kubernetes #k3s #nvidia #selfhosted #selfhosting #homelab #intel #driver #gpu #immich #jellyfin

I was having countless errors recently with Longhorn, keeping volumes in degraded and missing replicas, new replicas would stay on "Stopped" and never start the replication process

Today I tried bumping the simultaneous rebuild per node from 1 to 3, and everything seems to get back up progressively

I think I was stuck in a deadloop, and this over conservative setting was the trigger towards greatness

I had initially configured it to only rebuild one replica at a time, because otherwise my raspberry pis would crash with a big IOwait thread pinned to 100%

But now it's bigger x86 machines, so this setting doesn't make sense anymore

#raspberrypi #longhorn #k3s #volume #homelab #selfhosted #selfhosting #arm #x86

Weekend well spent on #homelab upgrades:

📈 3 -> 5 node #K3s cluster,
💪 single-node etcd -> 3 node etcd for #HighAvailability,
💿 updated k3s version,
🧹 tidied up some cabling on home rack.

All set for a #medialab installation !

Next up:
- find some time for that ... 😅

https://github.com/gsaslis/home-on-k8s

Anyone got any good ansible playbooks for keeping K3s up to date?

Note: I'm using Longhorn, so I'd like to upgrade no more than one node at a time.

#homelab #k3s #ansible

Iniciando o ano com mais atividades no Hackerspace: 12/01 às 19 horas! Online.

Encontro online sobre o projeto LHC/Infra

Projeto para estudos e experimentos com HomeLab, projeto aberto e colaborativo e têm como objetivo, prover um ambiente para criação de laboratórios, utilizando Kubernetes com a utilização da IaC e GitOps para gerenciar toda a sua configuração.

Podendo ser replicado em sua casa ou em seu ambiente de estudos.

Além de um ambiente dedicado para o Hackerspace e projetos dos membros.

Participação sem custo, só acessar o link: https://meet.jit.si/devops-lhc

https://eventos.lhc.net.br/event/lhcinfra-1

#hackerspace #homelab #infra #k3s #gitops #devops #iac

Sector 004 Infrastructure: Fully Documented. 🖖🚀

I've mapped my entire network into a unified, LCARS-inspired tactical ecosystem. From the "Subspace Gateway" to the "Hive," everything is accounted for.

The Grid: 📡 Core: 5-Port Gateway expansion 💻 Edge: 8-Port hub for workstations 🧠 Collective: PoE+ 8-Port switch for the Hive 🤖 The Hive: RPi K3s cluster with Netdata telemetry

Check out the archives: https://github.com/r0nn1ef/sector-004-archives

#HomeLab #Networking #K3s #Kubernetes #RaspberryPi #LCARS #DevOps

What’s the difference between Talos Linux and K3s? In short: a lot.

Talos = a Linux distribution that installs Kubernetes.
K3s = a Kubernetes distribution.

Keep reading about the differences and insights on when to use each →
https://www.siderolabs.com/blog/talos-linux-vs-k3s?utm_source=mastodon&utm_medium=social

#kubernetes #k3s

It really is surprising sometimes how #homelab services grow....and I just want to emphasize that it's silly to think that you'll be the exception.

Prior to doing my big holiday cleanup of scattered docker containers being moved over to first class, fully automated #proxmox #k3s VM nodes, and #kubernetes services being managed by #ansible and #argocd, I thought for sure I knew exactly how many services I would be running.

I also felt confident that I didn't need anything more than that.

Low and behold, the old adage holds true.

Tonight I was working on my deployment of #netbox (https://github.com/netbox-community/netbox) and I realized that it uses #Postgres under the hood.

No big deal. I've got a Postgres instance already running in the cluster.....except that I wrongly made the assumption that #paperless_ngx would be the only consumer, and I turned off the superuser.

Still no big deal....except that I'm deploying everything via GitOps as much as possible, AND I've already got live, important data in Postgres.

*long sigh*

So...now I get to go do some manual surgery on users and databases to make it be a proper cluster wide, multi-tenant service, I'll make the gitops side of it look pretty later.

And that's when I realized that keeping an instance of #pgadmin around would be super useful going forward.

I was, in no way, planning on running a PgAdmin instance, but here we are.

Well...guess what? I'm also running Mongo, InfluxDB, and Redis.

Thankfully, Influx ships with a WebUI...but if I'm already adding pgAdmin, then there's no sense skipping #Mongo Compass and #Redis Insight.

And that's officially 3 more unexpected services to deploy in a single night.