ci-actions/README.md
Christopher Mühl 01d6a3e779
Switch to S3 binary cache with isolated store paths
- Build in isolated Nix container
- Push to S3 binary cache (no host /nix/store access)
- Pull specific store paths to alvin
- Mount only specific /nix/store/hash to /var/www (read-only)
- Generate signing keys for cache authentication
- Update documentation with binary cache setup

Security improvements:
- Build container has no access to host /nix/store
- Web server only mounts its specific store path
- Proper isolation at every layer

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2026-02-16 14:53:11 +01:00

370 lines
8.5 KiB
Markdown

# CI Actions
Reusable Forgejo/Gitea actions for toph's infrastructure.
## Available Actions
### `deploy-nix-site` (Recommended)
Deploy a static site built with Nix flake to S3 and Nomad. Provides proper isolation and reproducibility.
**Requirements:**
- Repository must have a `flake.nix` with a default package output
- Runner label: `nix` (uses `docker://nixos/nix:latest`)
**Usage:**
```yaml
name: Deploy
on:
push:
branches: [main]
jobs:
deploy:
runs-on: nix
steps:
- uses: actions/checkout@v4
- uses: https://git.toph.so/toph/ci-actions/deploy-nix-site@main
with:
site-name: mysite
traefik-rule: Host(`mysite.example.com`)
env:
S3_ACCESS_KEY: ${{ secrets.S3_ACCESS_KEY }}
S3_SECRET_KEY: ${{ secrets.S3_SECRET_KEY }}
NIX_SIGNING_KEY: ${{ secrets.NIX_SIGNING_KEY }}
```
**Inputs:**
- `site-name` (required): Site identifier
- `traefik-rule` (required): Traefik routing rule
- `flake-output` (optional): Nix flake output, defaults to `.#`
- `s3-endpoint` (optional): S3 endpoint, defaults to `https://s3.toph.so`
**Example flake.nix:**
```nix
{
description = "My static site";
inputs.nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
outputs = { self, nixpkgs }:
let
system = "x86_64-linux";
pkgs = nixpkgs.legacyPackages.${system};
in {
packages.${system}.default = pkgs.stdenv.mkDerivation {
name = "my-site";
src = ./.;
buildInputs = [ pkgs.nodejs_20 ];
buildPhase = ''
npm ci
npm run build
'';
installPhase = ''
cp -r dist $out
'';
};
devShells.${system}.default = pkgs.mkShell {
packages = [ pkgs.nodejs_20 pkgs.vite ];
shellHook = "echo 'Run: vite'";
};
};
}
```
---
### `deploy-site`
Deploy a static site to production via S3 and Nomad (non-Nix).
**Usage:**
```yaml
name: Deploy
on:
push:
branches: [main]
jobs:
deploy:
runs-on: self-hosted
steps:
- uses: actions/checkout@v4
- name: Build site (if needed)
run: npm run build
- name: Deploy
uses: https://git.toph.so/toph/ci-actions/deploy-site@main
with:
site-name: mysite
traefik-rule: Host(`mysite.example.com`)
source-dir: dist # optional, defaults to current dir
env:
S3_ACCESS_KEY: ${{ secrets.S3_ACCESS_KEY }}
S3_SECRET_KEY: ${{ secrets.S3_SECRET_KEY }}
```
**Inputs:**
- `site-name` (required): Site identifier used as Nomad service name
- `traefik-rule` (required): Traefik routing rule (see examples below)
- `source-dir` (optional): Directory with built files, defaults to `.`
- `s3-endpoint` (optional): S3 endpoint, defaults to `https://s3.toph.so`
**Environment variables:**
- `S3_ACCESS_KEY`: S3 access key (set via Forgejo secrets)
- `S3_SECRET_KEY`: S3 secret key (set via Forgejo secrets)
**What it does:**
1. Packages the site directory as a tarball
2. Uploads to S3 at `s3://artifacts/<commit-sha>.tar.gz`
3. Sets public-read ACL on the artifact
4. Dispatches Nomad job to deploy the site
5. Site becomes available via the specified Traefik rule with Let's Encrypt SSL
**Infrastructure handles:**
- Docker containers (static-web-server)
- Resource limits (100 CPU, 64MB RAM)
- Traefik routing & Let's Encrypt SSL
- Automatic restarts
## Traefik Rule Examples
**Single domain:**
```yaml
traefik-rule: Host(`example.com`)
```
**Multiple domains (with www):**
```yaml
traefik-rule: Host(`example.com`) || Host(`www.example.com`)
```
**Subdomain:**
```yaml
traefik-rule: Host(`blog.example.com`)
```
**toph.so domain:**
```yaml
traefik-rule: Host(`mysite.toph.so`)
```
**Path-based routing:**
```yaml
traefik-rule: Host(`example.com`) && PathPrefix(`/docs`)
```
## Setup
### 1. Deploy S3 service (SeaweedFS)
```bash
cd /srv/infra
nix run .#bosun -- run s3
```
### 2. Generate binary cache signing keys
```bash
nix-store --generate-binary-cache-key cache.toph.so cache-priv-key.pem cache-pub-key.pem
# Public key (add to nix.conf trusted-public-keys on hosts that will fetch):
cat cache-pub-key.pem
# Example: cache.toph.so:9zFo64TPnxaQeyFM6NS9ou2Fd8OQv4Ia+MuLMjLBYjY=
# Private key (store in Forgejo secrets):
cat cache-priv-key.pem
# Keep this secret!
```
### 3. Create S3 buckets
```bash
# Configure AWS CLI
export AWS_ACCESS_KEY_ID=<your-access-key>
export AWS_SECRET_ACCESS_KEY=<your-secret-key>
export AWS_ENDPOINT_URL=https://s3.toph.so
export AWS_EC2_METADATA_DISABLED=true
# Create binary cache bucket
aws s3 mb s3://nix-cache
# (Optional) Create artifacts bucket for non-Nix deployments
aws s3 mb s3://artifacts
aws s3api put-bucket-acl --bucket artifacts --acl public-read
```
### 4. Configure alvin to trust the binary cache
Add to `/srv/infra/hosts/alvin/default.nix`:
```nix
nix.settings = {
substituters = [
"https://cache.nixos.org"
"s3://nix-cache?endpoint=s3.toph.so&scheme=https"
];
trusted-public-keys = [
"cache.nixos.org-1:6NCHdD59X431o0gWypbMrAURkbJ16ZPMQFGspcDShjY="
"cache.toph.so:9zFo64TPnxaQeyFM6NS9ou2Fd8OQv4Ia+MuLMjLBYjY=" # Your public key
];
};
```
Then deploy: `sudo nixos-rebuild switch --flake .#alvin`
### 5. Add Forgejo secrets
In your repository settings (or organization settings for global secrets):
- `S3_ACCESS_KEY`: S3 access key
- `S3_SECRET_KEY`: S3 secret key
- `NIX_SIGNING_KEY`: Contents of `cache-priv-key.pem`
### 6. Configure SSH access from runner to alvin
The runner needs to pull store paths to alvin's `/nix/store`. Add the runner's SSH key to alvin or use an agent socket mount.
## Examples
### Simple HTML site
```yaml
name: Deploy
on:
push:
branches: [main]
jobs:
deploy:
runs-on: self-hosted
steps:
- uses: actions/checkout@v4
- uses: https://git.toph.so/toph/ci-actions/deploy-site@main
with:
site-name: mysite
traefik-rule: Host(`mysite.toph.so`)
source-dir: . # HTML files in repo root
env:
S3_ACCESS_KEY: ${{ secrets.S3_ACCESS_KEY }}
S3_SECRET_KEY: ${{ secrets.S3_SECRET_KEY }}
```
### Node.js/Vite site with custom domain
```yaml
name: Deploy
on:
push:
branches: [main]
jobs:
deploy:
runs-on: self-hosted
steps:
- uses: actions/checkout@v4
- name: Setup Node
uses: actions/setup-node@v4
with:
node-version: 20
- name: Install & build
run: |
npm ci
npm run build
- uses: https://git.toph.so/toph/ci-actions/deploy-site@main
with:
site-name: myapp
traefik-rule: Host(`app.example.com`) || Host(`www.app.example.com`)
source-dir: dist
env:
S3_ACCESS_KEY: ${{ secrets.S3_ACCESS_KEY }}
S3_SECRET_KEY: ${{ secrets.S3_SECRET_KEY }}
```
### Hugo site
```yaml
name: Deploy
on:
push:
branches: [main]
jobs:
deploy:
runs-on: self-hosted
steps:
- uses: actions/checkout@v4
with:
submodules: true # For Hugo themes
- name: Setup Hugo
run: |
wget https://github.com/gohugoio/hugo/releases/download/v0.121.0/hugo_extended_0.121.0_linux-amd64.tar.gz
tar xzf hugo_extended_0.121.0_linux-amd64.tar.gz
sudo mv hugo /usr/local/bin/
- name: Build
run: hugo --minify
- uses: https://git.toph.so/toph/ci-actions/deploy-site@main
with:
site-name: myblog
traefik-rule: Host(`blog.example.com`)
source-dir: public
env:
S3_ACCESS_KEY: ${{ secrets.S3_ACCESS_KEY }}
S3_SECRET_KEY: ${{ secrets.S3_SECRET_KEY }}
```
## S3 Access
- **API endpoint**: https://s3.toph.so
- **Console**: https://s3-console.toph.so (restricted to Tailscale)
- **Credentials**: Stored in Nomad variables at `nomad/jobs/s3`
- **Backend**: SeaweedFS (S3-compatible)
## Troubleshooting
### AWS CLI not found
The action automatically installs AWS CLI if not present on the runner.
### Access denied during upload
Check that:
1. `S3_ACCESS_KEY` and `S3_SECRET_KEY` are set in Forgejo secrets
2. The credentials match those in the S3 Nomad job config
3. S3 service is running: `nomad job status s3`
### Site not accessible
Check:
1. Nomad allocation is running: `nomad job status static-site`
2. Traefik has picked up the service: check Traefik dashboard
3. DNS resolves: `dig <site-name>.toph.so`
4. Certificate is valid: `curl -v https://<site-name>.toph.so`
## Development
To add a new action:
```bash
mkdir -p new-action
cd new-action
vim action.yaml
```
Then commit and push to `main`.
## License
MIT