Setup

This document is primarly to document the steps needed to setup the entire docker setup again in the case of disk failure.

Installation

The base OS doesn't actually matter too much, but as of right now I'm using openmediavault so a lot of things are just taken care of for me.

1. Set up a static eth interface with 192.168.1.2 and add 1.1.1.1,8.8.8.8 as dns servers.
2. Update workbench credentials to use username and password of choice
   • This will need to be updated in infisical if they change
3. Install all necessary extensions. As of right now this includes:
   • openmediavault-apt
   • openmediavault-apttool
   • openmediavault-compose
   • openmediavault-mergerfs
   • openmediavault-snapraid

note

Some of these extensions only become available after installing omv-extras.

4. Setup the non-root user to allow majority of containers to run with it's UID and GUID
5. Mount all disk filesystems and ensure that smart monitoring is enabled for all disks
6. Also ensure these packages are installed if they are not already:

apt install update && apt install -y jq curl git

Disks

Setup the disks with mergerfs and snapraid using the following:

Disk Name	Mergerfs	Snapraid Array	Snapraid Disk
2TJMV95D	[x]	Pool	d1
2THB7KAD	[x]	Pool	d2
PN2334PCKXP9WB	[x]	Pool	d3
PN1338P4HS2STB	[x]	Pool	d4
7JJEL5KC	[x]	Pool	d5
ZXA05ZWD	[x]	Pool	d6
ZXA0HSAY	[ ]	Pool	parity
ZXA1BJ4N	[ ]	Pool	parity-2
WD-WCC4N4YV6DD3	[ ]	Backup	bd1
WD-WCC4N7DX0XVN	[ ]	Backup	bparity

Tailscale

Install tailscale and follow these steps:

• Join the node into the network
• Add the server tag to the machine
• Enable ssh for the machine using tailscale --set ssh

Docker

Ensure that docker is installed and follow the below steps:

• Confirm that all variables in komodo/variables/base.toml match the expected values for the machine
• Ensure all folders are properly setup and have the correct paths

GPU Setup

Follow installation setup for nvidia cuda toolkit. This should allow you to basically copy paste instructions. Then follow the installation for [nvidia container toolkit](https://docs.nvidia.com/datacenter/cloud-native/container- toolkit/latest/install-guide.html). After getting that all setup, install the following packages:

# If nvidia dkms is not building modules this is probably needed
sudo apt install linux-headers-amd64

sudo apt-get update sudo apt-get -y install cuda-toolkit nvidia-driver-cuda
nvidia-driver nvidia-kernel-dkms nvidia-container-toolkit

# Just in case (it usually should do this on it's own)
sudo update-initramfs -u -k all

# Setting up docker with the toolkit
sudo nvidia-ctk runtime configure --runtime=docker

# Restart docker
sudo systemctl restart docker

Deploy Komodo

Run the Deploy Komodo workflow, this should have komodo all setup and ready for initial configuration. Now follow these steps:

• Add the docker repo, following komodo/repos/docker.toml
• Add a new sync following the exact same options as komodo/resource_sync/github.toml
• Run the sync to download all configurations
• Run the sync procedure to fully deploy everything
• Add a new api key for the user in komodo and add those secrets to infisical
• Run a new deployment once the api keys are updated in infisical

All of the services should be up and running at this point!