How to deploy a site with git — Deploying Into an Account, Not Onto a Server

This runbook assumes you are doing the job on somebody else's site and billing for it. Allow 45 minutes of setup, and read the whole thing once before you open a live client account.

Where a step is safe to template across every account on your list, it says so. Where it changes what the client sees, it says that too.

An environment file that reached a remote repository is compromised by definition, and on a client project the credentials in it are usually not yours to expose. Database passwords, gateway keys and mail credentials all belong outside version control from the very first commit.

Use ignored files or environment variables, and keep the values in the same password manager that holds the account credentials. If a secret does reach a repository, rotate it rather than rewriting history, because history is not the part that matters.

Git versions code, not data. The client's database follows its own backup and migration path, and their uploads directory syncs separately or simply stays where it is. Code by Git, data by backups, media by sync — three routes, deliberately kept apart.

The daily backup on the account covers the data half, and restores run from the panel in minutes. Between the two you can describe your deployment and recovery story to a client in four sentences, which is usually four more than they have heard before.