The Portainer host keeps failing on `npm ci` inside the build stage (both Alpine+libc6-compat and Debian slim exited 1 without ever surfacing the real error to us). For a dev showcase this isn't worth chasing — the dev machine is the source of truth for the built output anyway. - Dockerfile: drop the Node build stage. Image is just nginx:1.27.3-alpine with /dist copied in. No npm inside the container. - docker-compose.yml: drop the production hardening (read_only, tmpfs, security_opt, resource caps) and the container_name. Dev-only, don't inhibit things. - .gitignore / .dockerignore: stop ignoring dist/ — it's committed now. - README: document the `npm run build && commit && push` release flow and note what to reinstate when this graduates to real production. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
26 lines
737 B
YAML
26 lines
737 B
YAML
# Portainer stack — Kaiser Natron dev showcase.
|
|
#
|
|
# Flow:
|
|
# 1. Developer runs `npm run build` locally and commits the updated dist/.
|
|
# 2. Portainer → Stacks → Pull and redeploy.
|
|
# 3. The image is a tiny nginx with the prebuilt dist copied in.
|
|
#
|
|
# This is deliberately simple for a showcase box — no container-side npm,
|
|
# no hardening, no resource ceilings. Tighten when it graduates to prod.
|
|
|
|
services:
|
|
web:
|
|
build:
|
|
context: .
|
|
dockerfile: Dockerfile
|
|
image: kaiser-natron:portainer
|
|
restart: unless-stopped
|
|
ports:
|
|
- "5555:80"
|
|
healthcheck:
|
|
test: ["CMD", "wget", "-q", "-O-", "http://127.0.0.1/health"]
|
|
interval: 30s
|
|
timeout: 5s
|
|
retries: 3
|
|
start_period: 10s
|