# Production deploy (systemd)

## Behaviour

On **every** service start (clean midnight restart, crash, or power loss recovery):

1. `git fetch` + `git reset --hard origin/<branch>`
2. `git pull --ff-only`
3. `npm install --omit=dev`
4. `git submodule update --init --recursive`
5. `npm run build:client` (concatenates FrameworkV3 → `client/frameworkv3/js/framework.js`)
6. start `node --experimental-sqlite server/src/index.js`

Pull-on-crash is intentional so a fixed commit on origin can recover a bad build.

Player-facing midnight schedule (3-hour countdown toast) still lives in the Node
process: it detects that origin is ahead, warns players, then exits cleanly so
systemd runs the update script above.

## Install on the host

```bash
cd /srv/questward/current   # or your clone
sudo ./deploy/install-service.sh
```

Optional env overrides:

| Variable | Default |
|----------|---------|
| `QUESTWARD_REPO` | parent of `deploy/` |
| `QUESTWARD_USER` | `questward` |
| `QUESTWARD_GIT_BRANCH` | `master` |
| `QUESTWARD_AUTH_CLIENT_SECRET` | (unset — stub auth) |

Production secrets: **`/etc/questward/env`** (loaded by the unit). Example:

```bash
sudo mkdir -p /etc/questward
sudo nano /etc/questward/env
# QUESTWARD_AUTH_CLIENT_SECRET=...
# QUESTWARD_AUTH_REDIRECT_URI=http://questward.timfalken.com:1234/auth/callback
sudo chmod 600 /etc/questward/env
```

See root `README.md` for OAuth / auth.timfalken.com setup.

`TILE_MMO_*` env vars are still accepted as fallbacks for older installs.

## Manual ops

```bash
sudo systemctl restart questward
journalctl -u questward -f
```

Disable the in-process schedule locally with `QUESTWARD_DEPLOY_SCHEDULER=0`.
