One job: npm ci, tsc --noEmit, npm test on Node 26; on pushes to master it builds, prunes dev dependencies and rsyncs dist/, src/ (Swagger reads it) and node_modules/ over SSH as plesk-plutodev, then restarts Passenger and smoke-tests /docs/. The server's .env, logs/, tmp/ and public/ are never touched - .env is no longer generated by CI but kept on the server. Plesk moves from Node 12.22.12 to 26 with this; the bundle built on 26 also starts on 12, so the order of deploy and Plesk switch doesn't matter. SonarQube leftovers are gone. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
5.7 KiB
Deployment
api.plutodev.de runs on the Plesk server as a Node.js app (Phusion Passenger) and is deployed by
Gitea Actions (.gitea/workflows/deploy.yml), which replaces the Jenkins job (switchover
2026-11). Same workflow as Paddy/api.pmueller.me and the Nachklang API; the reusable how-to is in
Patrick's PKM ("Jenkins to Gitea Actions for a Plesk Node.js App").
push to master ──▶ Gitea ──▶ runner (container on the server)
└─ one job: npm ci, tsc --noEmit, npm test (Node 26)
└─ (master only) build, prune dev deps ─ rsync over SSH ─▶
plesk-plutodev@pmueller.me:/var/www/vhosts/plutodev.de/api.plutodev.de
+ tmp/restart.txt
Plesk (as configured, 2026-11-16)
Websites & Domains → api.plutodev.de → Node.js:
| Setting | Value |
|---|---|
| Node.js version | 26 - switched from 12.22.12 (end of life since 2022) with the Gitea switchover, see "Node version" |
| Application root | /api.plutodev.de (= /var/www/vhosts/plutodev.de/api.plutodev.de) |
| Document root | /api.plutodev.de/public |
| Application startup file | dist/app.js |
| Application mode | production |
| Environment variable | PORT=443 (Passenger takes over the port) |
Configuration: .env lives on the server
The app reads its settings from .env in the app root (dotenv): database credentials and names,
TWITCH_HIGHLIGHTS_ACCESS_KEY, HCAPTCHA_SECRET, ... The Jenkins job wrote that file on every
deploy from its own credentials. The workflow never touches it: since the switchover the .env on
the server is the source of truth and is edited there (Plesk file manager), or the values move to
Plesk's Node.js → Environment → Secrets. Nothing secret is stored in Gitea for this app.
The code also reads CRAWLER_ACCESS_KEY, which the Jenkins-generated .env never contained
(2026-11-16) - either unused or set elsewhere; not checked.
What the deploy touches - and what it must not
| Path | How | Why it's needed at runtime |
|---|---|---|
dist/ |
mirrored (--delete inside) |
the compiled app |
src/ |
mirrored | swagger-jsdoc reads ./src/models/**/*.router.ts for /docs |
node_modules/ |
mirrored, production dependencies only (npm prune --omit=dev) |
runtime dependencies |
package.json, package-lock.json |
copied, no delete | (CommonJS - the app ran without them) |
Never touched: .env, logs/ (Winston), tmp/ (Passenger's restart.txt), public/ (Passenger's
document root; gitignored in this repo, and Jenkins never copied it either).
Node version
Until 2026-11-16 Plesk ran this app on Node 12.22.12; with the switchover it moves to 26
(.nvmrc 26, so CI builds and tests with the same version).
Checked locally (2026-11-16), as x86_64 like the runner (podman --platform linux/amd64):
npm ci on Node 26 downloads bcrypt 5.0.1's prebuilt N-API binary (napi-v3), tsc, the test,
build and prune pass, and the pruned bundle starts under Node 26 and under Node 12.22.12 (/docs/
200, bcrypt loads). So the order doesn't matter: the first deploy may land while Plesk still says
12, then switch Plesk to 26. Database queries were not exercised (no DB) - check one endpoint per
app after the switch.
Don't test this on the Mac's default (arm64) containers: bcrypt 5.0.1 has no arm64 prebuilt, so npm compiles it there, and that compile fails on Node 26 - a local artefact, not what the runner does.
Gitea settings
The repo belongs to the organisation PlutoDev, so user-level settings of Paddy don't apply.
On the organisation (PlutoDev → Settings → Actions), for all PlutoDev repos: secrets
DEPLOY_SSH_KEY (the shared private deploy key gitea-actions Paddy - PlutoDev is a private project
now), DEPLOY_KNOWN_HOSTS ([pmueller.me]:42069 ssh-ed25519 …, same server); variables
DEPLOY_HOST=pmueller.me, DEPLOY_PORT=42069, DEPLOY_USER=plesk-plutodev.
On this repository (plutoapi-v2 → Settings → Actions → Variables):
| Variable | Value |
|---|---|
DEPLOY_PATH |
/var/www/vhosts/plutodev.de/api.plutodev.de |
DEPLOY_SMOKE_URL |
https://api.plutodev.de/docs/ |
Switchover
- SSH for
plesk-plutodev(was/bin/false, set to/bin/bash2026-11-16, done): the shared public key appended to/var/www/vhosts/plutodev.de/.ssh/authorized_keys(.ssh700, file600), the user added to the key-onlyMatch User …line in/etc/ssh/sshd_config. - Pause the Jenkins job (done 2026-11-16). Its credentials (
DB_USER,DB_PASSWORD,TWITCH_HIGHLIGHTS_ACCESS_KEY,HCAPTCHA_SECRET) can be deleted from Jenkins afterwards - they're in the server's.env. - Fix file ownership (as root):
ls -ld .first (the folder keeps its group), thenfind . -mindepth 1 ! -user plesk-plutodev -exec chown -h plesk-plutodev:psacln {} +--hfor the symlinks innode_modules/.bin/. - Set the variables, open a PR → green run → merge → watch the deploy, check
/docs/. - Plesk → api.plutodev.de → Node.js: version 26, restart; check
/docs/and one endpoint per app.
Rollback: revert the commit on master; the workflow deploys the previous state.
Cleaned up with the switchover (2026-11-16)
- SonarQube is gone:
sonar-project.properties,jest-sonar-reporter, the--coverage --testResultsProcessorflags and thejestSonarconfig.
Checked locally (2026-11-16)
actionlint: no findings.node:26-bookworm(x86_64):npm ci,tsc --noEmit,npm test, build,npm prune --omit=dev; the bundle then ran under Node 26 and Node 12 (see "Node version").
Not checked: the workflow on the real runner.