11 / Services

Cloud and deploy

A move off a single server, automatic deploy, and an environment that can be stood up again.

For teams where a release means logging into the server by hand, and there is no spare system.

Cloud

Deploy

Deploy

Check

Production

Backup

Button

Release

Rollback

Previous

Diagram

Beside

01

What it consists of

What goes into the system

  1. 01

    What is running now

    First, a list: which service, which database, which keys, and what falls over if the machine is switched off.

    • Services, and what they depend on
    • Where the secrets sit
    • What cannot be stopped during the day
  2. 02

    Environments

    Production and the check environment do not share one server “for now.” The check is done before the deploy reaches people.

    • Where the new version is looked at
    • Where clients do the work
    • Which data in a check must not be taken from production
  3. 03

    Deploy

    A new version goes out by a button from the repository, not through a manual login and a copied folder.

    • What assembles on its own
    • Who is allowed to deploy
    • How to roll back the previous version
  4. 04

    Move

    We leave a single machine so the downtime is short, and named in advance.

    • What moves first
    • How we slice the traffic
    • What we check immediately after
  5. 05

    Stand it up again

    The environment description lives next to the code. A new server is built from it, not from memory.

    • Containers and a diagram
    • Watching, after the move
    • What the platform bill is made of

What we do

  1. 01What is running now, and what it depends on
  2. 02A diagram in the cloud
  3. 03A move without a long stop
  4. 04Deploy from a button
  5. 05Containers, and watching

How we proceed

  1. 01Walkthrough
  2. 02Diagram
  3. 03Move
  4. 04Automatic deploy
  5. 05The bill for infrastructure

Stack

  • Docker
  • Terraform
  • GitHub Actions
  • Cloud platforms

A guide for the timing

One service takes a few weeks. Several environments and databases take from a month.

Outcome

A new version is deployed without a night window, and the server can be stood up from the description.

What usually comes later

Several regions, complex networks, and “your own cluster for growth” are not part of moving one service.