Files
git.stella-ops.org/ops/devops/console/README.md
2025-12-07 15:12:34 +02:00

45 lines
3.5 KiB
Markdown

# Console CI runner (offline-friendly)
Status: runner spec + CI now wired to PRs; runner image scaffold + CI build workflow now available with baked npm + Playwright cache.
## Runner profile
- OS: Ubuntu 22.04 LTS (x86_64) with Docker available for Playwright deps if needed.
- Node: 20.x (LTS). Enable corepack; prefer npm (default) to avoid extra downloads.
- Caches:
- npm: `~/.npm` keyed by `src/Web/package-lock.json` hash.
- Playwright: `~/.cache/ms-playwright` pre-seeded with Chromium so `npm test -- --browsers=ChromeHeadless` can run offline. Seed once using `npx playwright install chromium` on a connected runner, then snapshot the directory into the runner image.
- Angular build cache: optional `~/.cache/angular` if using angular.json cache; safe to keep.
- Artifacts retention: keep lint/test/build outputs 14 days; limit to 500 MB per run (coverage + dist + test reports). Artifacts path: `artifacts/` (dist, coverage, junit/trx if produced).
## Pipeline steps (expected)
1) Checkout
2) Node 20 setup with npm cache restore (package-lock at `src/Web/package-lock.json`).
3) Install: `npm ci --prefer-offline --no-audit --progress=false` in `src/Web`.
4) Lint: `npm run lint -- --no-progress`.
5) Unit: `npm test -- --watch=false --browsers=ChromeHeadless --no-progress` (headless Chromium from pre-seeded cache).
6) Build: `npm run build -- --configuration=production --progress=false`.
7) Artifact collect: `dist/`, `coverage/`, any `test-results/**`.
## Offline/airgap notes
- Do not hit external registries during CI; rely on pre-seeded npm mirror or cached tarballs. Runner image should contain npm cache prime. If mirror is used, set `NPM_CONFIG_REGISTRY=https://registry.npmjs.org` equivalent mirror URL inside the runner; default pipeline does not hard-code it.
- Playwright browsers must be pre-baked; the workflow will not download them.
### Runner image (with baked caches)
- Dockerfile: `ops/devops/console/Dockerfile.runner` (Node 20, npm cache, Playwright Chromium cache). Builds with `npm ci` + `playwright install chromium --with-deps` during the image build.
- Build locally: `IMAGE_TAG=stellaops/console-runner:offline OUTPUT_TAR=ops/devops/artifacts/console-runner/console-runner.tar ops/devops/console/build-runner-image.sh`
- `OUTPUT_TAR` optional; when set, the script saves the image for airgap transport.
- Runner expectations: `NPM_CONFIG_CACHE=~/.npm`, `PLAYWRIGHT_BROWSERS_PATH=~/.cache/ms-playwright` (paths already baked). Register the runner with a label (e.g., `console-ci`) and point `.gitea/workflows/console-ci.yml` at that runner pool.
- CI build helper: `ops/devops/console/build-runner-image-ci.sh` wraps the build, sets a run-scoped tag, emits metadata JSON, and saves a tarball under `ops/devops/artifacts/console-runner/`.
- CI workflow: `.gitea/workflows/console-runner-image.yml` (manual + path-trigger) builds the runner image and uploads the tarball + metadata as an artifact named `console-runner-image-<run_id>`.
### Seeding Playwright cache (one-time per runner image, host-based option)
```bash
ops/devops/console/seed_playwright.sh
# then bake ~/.cache/ms-playwright into the runner image or mount it on the agent
```
## How to run
- PR-triggered via `.gitea/workflows/console-ci.yml`; restrict runners to images with baked Playwright cache.
- Manual `workflow_dispatch` remains available for dry runs or cache updates.
- To refresh the runner image, run the `console-runner-image` workflow or execute `ops/devops/console/build-runner-image-ci.sh` locally to generate a tarball and metadata for distribution.