ci(deploy): auto-deploy prod stack after image build on merge to main
Add a `deploy` job to build-images.yml that runs on the self-hosted `uom_deploy` runner and, via `needs: build`, fires only after a clean image build+push. It pulls the fresh :latest images and recreates the stack (pull → down → up -d) from /home/perry/website. Guarded on refs/heads/main so a workflow_dispatch off another branch can't deploy. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
@@ -1,11 +1,21 @@
|
||||
# Build and publish the app + bot container images to Gitea's container registry
|
||||
# on every merge to main. Production then pulls prebuilt images instead of
|
||||
# building on the host.
|
||||
# Build the app + bot container images, publish them to Gitea's container
|
||||
# registry, then roll the production stack onto the fresh images — all on every
|
||||
# merge to main. Production only ever pulls prebuilt images; it never builds.
|
||||
#
|
||||
# Two jobs run in sequence:
|
||||
# build — builds & pushes website-app / website-bot images (on ubuntu-latest)
|
||||
# deploy — `needs: build`, so it starts only after a clean build+push, and
|
||||
# pulls + recreates the stack on the production host (on uom_deploy)
|
||||
#
|
||||
# Prerequisites (one-time):
|
||||
# • An always-on Gitea runner with label `ubuntu-latest` whose jobs have the
|
||||
# host Docker socket mounted (/var/run/docker.sock), so `docker build` talks
|
||||
# to the host daemon. This also gives free layer caching between runs.
|
||||
# • A second self-hosted runner labelled `uom_deploy` ON the production host,
|
||||
# with access to the Docker daemon and to /home/perry/website (the directory
|
||||
# holding the production docker-compose.yml + .env). This is what actually
|
||||
# rolls the stack; it must be able to `docker compose pull` from the registry
|
||||
# (log in once on the host, or ensure the images are public-read).
|
||||
# • Two repo secrets (Settings → Actions → Secrets):
|
||||
# REGISTRY_USER — the Gitea username that owns the token below
|
||||
# REGISTRY_TOKEN — a Gitea access token with `write:package` (+ read:package)
|
||||
@@ -14,6 +24,7 @@
|
||||
# Produces, in gitea.whitlocktech.com/<owner>/ :
|
||||
# website-app:latest + website-app:sha-<7>
|
||||
# website-bot:latest + website-bot:sha-<7>
|
||||
# then deploys the `:latest` images (docker-compose.yml defaults IMAGE_TAG=latest).
|
||||
|
||||
name: Build container images
|
||||
|
||||
@@ -85,3 +96,25 @@ jobs:
|
||||
- name: Log out (clear cached credentials from the runner)
|
||||
if: always()
|
||||
run: docker logout "${REGISTRY}" || true
|
||||
|
||||
deploy:
|
||||
# Roll production onto the images `build` just pushed. `needs: build` makes
|
||||
# this wait for a clean build+push — if the build fails, deploy never fires,
|
||||
# so the running stack is left untouched rather than torn down for nothing.
|
||||
needs: build
|
||||
runs-on: uom_deploy
|
||||
# Guard against a workflow_dispatch fired from a non-main branch: only ever
|
||||
# deploy the main line to production.
|
||||
if: github.ref == 'refs/heads/main'
|
||||
steps:
|
||||
- name: Pull the fresh images and recreate the stack
|
||||
# `pull` grabs the new :latest images the build job published; `down`
|
||||
# then `up -d` recreates the containers on them. Compose only recreates
|
||||
# services whose image digest changed, so the DB stays put.
|
||||
run: |
|
||||
set -euo pipefail
|
||||
cd /home/perry/website
|
||||
docker compose pull
|
||||
docker compose down
|
||||
docker compose up -d
|
||||
docker compose ps
|
||||
|
||||
Reference in New Issue
Block a user