The configuration of my home infrastructure.
My main laptop, a Lenovo Carbon X1, runs NixOS.
Its configuration is specified in x1.nix
using the experimental flakes
feature. Modify this file
and switch to the new configuration:
./scripts/switch.sh x1
By default, this configuration is stored in /etc/nixos/configuration.nix
.
For testing purposes you can build a QEMU virtual machine from the configuration:
nixos-rebuild build-vm --flake .#x1 && ./result/bin/run-*-vm
To update the lock files:
nix flake update --update-input nixpkgs --commit-lock-file
My rcfiles repository completes the configuration of my laptop. Those files live in a separate repository because I also use them on my work computer which doesn't run NixOS.
A periodic job backs up my home directory. The remote connection uses an SSH
key which I store in /root/keys
:
sudo ssh-keygen -N '' -t ed25519 -f /root/keys/id_ed25519-borg-x1
The backup server configuration references the public part of this key.
The server's configuration is in the nixosConfigurations
attribute of
flake.nix. Use this script, a thin
wrapper around nixos-rebuild
, to build and activate a server's configuration:
./scripts/switch.sh nuc # redeploy the server nuc
All server send their logs to Loki. To see all logs live:
export LOKI_ADDR=http://loki.thewagner.home
nix shell nixpkgs#grafana-loki --command \
logcli query '{job="systemd-journald"}' --tail
The query
command takes a
LogQL expression as an argument.
Each commit on the master branch is automatically deployed using Cachix Deploy. For a detailed description see this blog post.
To prepare a machine for automatic deployment:
- Add the system's derivation to the
cachix-deploy
package in flake.nix - Install
cachix-agent
by including this module - In the Cachix Deploy console follow the "Add Agent" steps
- Save the generated agent token in
/etc/cachix-agent.token
using the formatCACHIX_AGENT_TOKEN=<token>
The deployment steps are defined in this file. The Cachix Deploy documentation describe how to configure GitHub Actions. The pipeline uses the following values as action secrets:
CACHIX_AUTH_TOKEN
CACHIX_ACTIVATE_TOKEN
Installing a new system takes only a few manual steps.
Create a customized installer ISO image using the command mentioned at the top of installer/iso.nix.
Copy the ISO image to a USB stick and boot the computer from it. Connect to the installer using SSH:
ssh root@nixos -o StrictHostKeyChecking=no -o 'UserKnownHostsFile /dev/null'
Execute the relevant lines from /etc/install.sh to partition the disk and create file systems.
Use the basic configuration from /etc/configuration.nix as default and set the hostname.
Run the installer then reboot the machine. The installation of the basic system is done.
Continue the system's management using NixOps.
The configuration.nix(5) man page documents all the available options for configuring the system:
man configuration.nix
All supported options are searchable online.
Query available packages:
nix search nixpkgs wget
Install a package into the user's profile
nix profile install nixpkgs#firefox
Remove old, unreferenced packages, system-wide:
sudo nix-collect-garbage
sudo nix-collect-garbage -d # also delete old system old configurations
This is documented in the Cleaning the Nix Store section of the NixOS manual.
The builtin functions of the Nix evaulator are listed here.
See the version of this repository from which the system's configuration was built:
nixos-version --json
See which version of a given package will be installed:
$ nix eval .#nixosConfigurations.nuc.pkgs.grafana.version
"10.2.4"
Evaluate configuration parameters:
$ nix eval .#nixosConfigurations.nuc.config.networking.firewall.allowedTCPPorts
[ 22 80 1883 3000 3100 8022 8080 8081 8300 8301 8302 8500 8600 9000 9080 9090
9093 9100 9883 ]
Push a runtime closure of a locally built derivation to Cachix:
export CACHIX_AUTH_TOKEN=...
nix build -L --json .#nixosConfigurations.rp3.config.system.build.toplevel \
| nix run nixpkgs#jq -- -r '.[].outputs | to_entries[].value' \
| nix run nixpkgs#cachix -- push wagdav
Linksys WRT ACM-3200 running OpenWRT.
Connect to the router with an Ethernet cable.
Download and install the OpenWRT firmware then run:
router/setup.sh first-time
Reboot the router.
Change the settings in router/config
and run
router/setup.sh
Build the Raspberry Pi's SD card image using QEMU's aarch64 emulator.
On x230
, because nuc
is configured as a remote builder
for aarch64
packages, just run:
nix build .#packages.aarch64-linux.sdcard
On other hosts, specify nuc
explicitly as a remote builder:
nix build -L .#packages.aarch64-linux.sdcard \
--builders "ssh://root@nuc aarch64-linux $HOME/.ssh/remote-builder 4 1 - - c3NoLWVkMjU1MTkgQUFBQUMzTnphQzFsWkRJMU5URTVBQUFBSUlLYUV0YzhQTnFoeEFRMjRnWTV0MjVZLzhIVTZTdFVCNmttVTF4bVZ0YTcgcm9vdEBudWMK"
The elements of --builders
argument are described here.
Uncompress the built image and write it to an SD card:
unzstd ./result/sd-image/nixos-sd-image*.zst -o nixos-sd-image.img
sudo dd if=nixos-sd-image.img of=/dev/mmcblk0 bs=4096 conv=fsync status=progress
Insert the SD card in the Raspberry Pi and power it up. The system is configured as defined in host-rp3.nix.
If the SD card is build from scratch, change or provision the following secrets:
- Host's identity (automatically generated on first boot)
- WiFi SSID and password in
/etc/secrets/wireless.env
- Tailscale authentication token
- Cachix authentication token
If this is a complete reinstall, update the host's public key in
program.ssh.knownHosts. Run ssh-keyscan rp3
to
obtain the host key's signature.
Store the WIFI SSID and password in the file /etc/secrets/wireless.env
with
the following format:
WIFI_SSID=...
WIFI_KEY=...
Connect the host to the tailnet with tailscale login
.
To connect Cachix, follow these instructions.
This article, describes how I configured my Raspberry Pi v1 camera module on my Raspberry Pi 3 running NixOS.
I found the following links useful:
I have a couple of NodeMCU boards which can be configured using the scripts in the nodemcu directory.
Enter a Nix shell
cd nodemcu
nix-shell
In this shell the following helper functions are available.
Erase everything from the device and start from scratch:
flash_erase
: Perform Chip Erase on SPI flashflash_write
: Write the Tasmota firmware to the device
Open an interactive serial terminal:
serial_terminal
Restore the firmware's factory settings:
device_reset | commit
Configure a device
device_config <WIFI_SSID> <WIFI_KEY> | commit
These commands are defined as shell hooks in shell.nix
The best way I found to provision the ESP8266 systems with custom firmware is through MQTT because it's not always easy to get access to a serial terminal.
Use the serial console or the web interface to connect the device to the WiFi and to the MQTT broker.
Build and run any of the provisioning scripts:
nix build .#sensors && ./result/tasmota_082320.sh
This will reconfigure the specified sensor by sending commands over MQTT.
On my mobile I created a dashboard using MQTT Dash.
To update the dashboard configuration file, use the
Import/Export functionality of the app and publish the dashboard state to an
MQTT topic (the default is metrics/exchange
).
The following command listens to the published configuration and updates the dashboard configuration in this repository:
nix run .#mqtt-dash-listen > nodemcu/mqtt-dash.json
The device tasmota_0E63DE
couldn't connect to the Wi-Fi network. I believe
this is because the 2.4GHz and 5GHz networks share the same SSID. I flashed
the latest tasmota firmware (14.1.0) on the device and selected the 2.4GHz
connection with the command:
Wifi 3
which corresponds to the Wi-Fi mode 802.11b/g (2.4 GHz).