VPS operations

SSH user and sudo

Deployments run as the configured SSH user. Nginx/Apache steps need root for /etc/nginx, systemctl reload, etc.

Option A — passwordless sudo (recommended) for user deploy:

sudo visudo -f /etc/sudoers.d/central-deploy

deploy ALL=(ALL) NOPASSWD: /usr/sbin/nginx, /bin/systemctl, /bin/cp, /bin/ln, /bin/rm, /bin/test, /usr/sbin/apachectl, /usr/sbin/a2ensite, /bin/chown, /bin/chmod, /bin/mkdir, /usr/bin/php, /usr/bin/apt-get

Remote scripts use sudo -n so a missing sudoers rule fails immediately.

Project delete runs cleanup over SSH before removing the project from Central. The shared legacy root /var/www/app is never removed wholesale.

Paths

• VPS deploy root (per server, default /var/www/server):
  • <deployRoot>/apps/<slug> — application checkout
  • <deployRoot>/configs/{nginx,apache,pm2} — generated configs
  • <deployRoot>/data/<slug>/app.db — persistent SQLite
  • <deployRoot>/logs/<slug>/ — PM2 and app logs
• Release layout: releases/{deploymentId}/ with current symlink; shared/.env persists across cutovers.

Git on the VPS

Deploys clone into releases/{deploymentId}, then repoint current. Rollback repoints current to the previous successful release.

Manual recovery:

cd /var/www/central-server
git fetch origin
git reset --hard origin/main

Do not commit on the server (next-env.d.ts changes after every build).

TLS (Let's Encrypt)

Schedule renewal on each server:

# /etc/cron.d/certbot-central (example)
0 3 * * * root certbot renew --quiet && nginx -s reload

Central stores certificate expiry on the server record; the dashboard Health card warns before SSL_EXPIRY_WARN_DAYS (default 14).

Failure modes

• Invalid nginx config marks deployment failed after rollback attempt.
• SSH connectivity uses retries for transient TCP errors.