I actually needed to add an additional 32GB to accommodate the 20GB docker image for LocalAI.
Resizing was much easier than expected. Shut down the machine, run the following:
sudo qemu-img resize /data/kvm/ubuntu22.04 +16G
and reboot. The cloud-init images automatically…
Oddly, this is still firing after logs have been rotated and images have been pruned:
$ df -h
Filesystem Size Used Avail Use% Mounted on
udev 7.8G 0 7.8G 0%…
The helm chart does not offer an automated way to manage the main config file. I am editing the files manually, but this should probably be source-controlled in the future.
The T4 appears to be available within Kubernetes; however, the underlying VM is running low on storage, preventing containers from being deployed.
This deployment was straight forward, but resulted in a pod with pending status. More work is needed to utilize GPUs in K8s.