Pinned Tweet
Is this still considered a #homelab?
I just discovered FS-Cache in Linux and it sped up my NFS shares 6x even with 10 gig
youtube.com/watch?v=KVNAp7MA…
The only appropriate label for my self-installing Linux boot drive that wipes everything, installs Linux, and adds my ssh keys all without touching a key.
Decided to turn my adventure yesterday morning into a post technotim.com/posts/automati…
Local models on my own hardware are getting better, and I want to use them before sending everything to a frontier model. NVIDIA Switchyard and Nemotron 3.5 Lightning are helping me do this.
youtube.com/watch?v=qxfNTcPl…
Helping my sister with her router and access point. I now have to refer to them as trash can looking thing and the circle thing so she knows which I am talking about 😂
I now have a Spark server that builds and tests my vLLM Docker image in CI against real models.
This will be the first and last time this Spark touches grass.
Post: technotim.com/posts/vllm-gb1…
Huge thanks to @nvidia for supporting this project!
What the heck is an NPU actually for?
youtube.com/watch?v=8NFV6duM…
After a long hiatus, I finally had some time to put into my k3s-ansible repo that helps you bootstrap a Kubernetes cluster in minutes.
One of the biggest things holding me back from updating it was being able to run k3s cluster tests in CI. Well, I finally have that working again and I can merge in confidence!
k3s_version: v1.36.2+k3s1
kube_vip_tag_version: v1.2.2
metal_lb_speaker_tag_version: v0.16.0
metal_lb_controller_tag_version: v0.16.0
calico_tag: v3.32.1
cilium_tag: v1.20.0
cilium_cli_tag: v0.19.7
github.com/timothystewart6/k…
I tested a bunch of 10-inch mini-rack accessories, including cooling, power, displays, a network KVM, and even a compact UPS.
Here’s what actually worked: youtu.be/ARN0PtHHKoM
I no longer only play a network engineer on YouTube.
I helped transform a local nonprofit's network from this mess into something clean, documented, and maintainable.
Homelab skills absolutely transfer.
If you're running an NVIDIA Spark or any of the clones with a GB10 (ASUS, Lenovo, Dell, HP, MSI, Gigabyte, etc) for local AI. I built a Docker image for it with the latest VLLM and all its component so that you don't have to.
No need to wait weeks for NVIDIA to do it.
github.com/timothystewart6/v…
Just racked up the UDM BEAST. With (10) 10G links and (2) 25G aggregated to my switch, it’s pretty OP.











