Skip to content

README.md

Latest commit

 

History

History
115 lines (93 loc) · 6 KB

README.md

File metadata and controls

115 lines (93 loc) · 6 KB
Raw

MyNode UI Dev Mode (mocked backend)

Run the real MyNode web UI in a Docker container with a fully mocked backend — no bitcoind, no LND, no systemd, no real device. For developing and testing UI changes (templates, CSS, JS, Flask routes) with instant hot reload.

Quick start

make dev-ui          # from the repo root; builds + starts the container

Open http://localhost:8888 and log in with password bolt (configurable via DEV_PASSWORD in docker-compose.yml; set DEV_AUTOLOGIN=1 to skip login entirely). Autologin applies only to requests from pages on localhost / 127.0.0.1, never to requests other sites send through your browser.

Other targets: make dev-ui-build, make dev-ui-shell, make dev-ui-clean.

Edit any file under rootfs/standard/var/www/mynode/ (templates, static assets, python) or rootfs/standard/var/pynode/ on the host — templates apply on refresh, python triggers Flask auto-reload. The source and this folder are mounted read-only, so nothing in this container can modify real code. git status stays clean.

The DEV panel

Every page gets a floating DEV button (bottom-right) that opens the state switcher:

  • Device state — flip /tmp/.mynode_status to any state (stable, drive_missing, drive_formatting, choose_network, quicksync_*, upgrading, shutting_down, ...) to render every state page.
  • Bitcoin — synced / syncing at an arbitrary progress (renders the real syncing page), or the "Launching MyNode Services" starting page.
  • Lightning — LND ready/not-ready, wallet exists/missing (wallet-create flow).
  • Warnings — undervoltage / throttled / capped / fsck / usb warning pages, and the kernel security warning.
  • System — drive-almost-full error, upgrade banner, premium vs community edition, simulated reboot/shutdown, and RESET ALL back to the golden state.
  • Apps — one-click install/uninstall (marker files only, no real install). This covers the app SDK apps in rootfs/standard/usr/share/mynode_apps too (mounted read-only); their icons and screenshots are served from there.

Everything the panel does is also available as plain HTTP endpoints under /dev/... (no auth, but only from localhost pages or non-browser clients) — see dev_blueprint.py. The real app redirects GET requests that carry no origin headers to /, so curl needs -H 'Sec-Fetch-Site: same-origin'. Examples:

curl -H 'Sec-Fetch-Site: same-origin' 'localhost:8888/dev/state?value=drive_missing'
curl -H 'Sec-Fetch-Site: same-origin' 'localhost:8888/dev/sync?synced=0&progress=0.62'
curl -H 'Sec-Fetch-Site: same-origin' 'localhost:8888/dev/app?name=specter&installed=1'
curl -H 'Sec-Fetch-Site: same-origin' 'localhost:8888/dev/service?name=electrs&status=failed'
curl -H 'Sec-Fetch-Site: same-origin' 'localhost:8888/dev/reset'

The full install flow also works end-to-end from the normal UI: installing an app from the marketplace shows the real "Installing..." page with a (fake) live install log, a simulated reboot, and the app appears installed afterward. Reboot/shutdown from settings play out the real reboot page (the mock resets a fake boot-time file; the page redirects home when it sees uptime decrease).

How it works (architecture)

Zero changes to real code. The container runs the unmodified Flask app with three mock layers around it:

  1. PYTHONPATH shim (mock_pynode/) — sits before /var/www/mynode and /var/pynode on the python path, so import bitcoin_info etc. resolve to the mocks. Each mock exec's the real module, patches ONLY the functions that touch the outside world into the real module's namespace, and re-exports everything (see _mockutil.py for why patching the real namespace matters). What gets mocked:

    • bitcoin_info — fake AuthServiceProxy (bitcoind RPC) backed by fixtures/bitcoin.json; the real update/caching code runs unchanged.
    • lightning_info — lnd_get/lnd_get_v2 return fixtures/lightning.json.
    • electrum_info — fake requests serving prometheus metrics.
    • systemctl_info — service states from /tmp/mock_services/<name>.
    • device_info — uptime/serial/temp + simulated reboot/upgrade.
    • application_info — install/uninstall simulations (marker files only).
    • pam — login accepts DEV_PASSWORD.
    • price_info, thread_functions, drive_info, transmissionrpc.

  2. Fixture filesystem (seed_fixture_fs.py) — the real code reads inline absolute paths (/mnt/hdd/..., /home/bitcoin/.mynode/..., /usr/share/mynode/..., /tmp/...), so the container recreates a provisioned device's filesystem. /mnt/hdd is a tmpfs mount because the code greps /proc/mounts for it.

  3. Fake binaries (fake_bin/) — installed into /usr/bin in the container (many calls use absolute paths). systemctl reads/writes the same /tmp/mock_services files as the python mock; journalctl emits canned logs; mynode_*.sh scripts are no-ops or tiny simulations.

dev_server.py imports the real app from mynode.py, seeds the in-memory data caches, registers the /dev blueprint + overlay injector, and runs Flask with the auto-reloader (the interactive debugger is off). A light 30s refresher keeps mocked values jittering; set DEV_REAL_THREADS=1 to run the app's real background threads instead.

Customizing sample data

Edit the JSON files in fixtures/ (block height, peers, channels, balances, invoices, prices, service states, device identity) — they reload on next refresh without restarting the container. Default-installed apps are listed in seed_fixture_fs.py (DEFAULT_INSTALLED_APPS).

Known quirks

  • Port 8888 (not 8000) because the repo's release file server uses 8000.
  • The QuickSync download page shows "Waiting on download client to start..." (no transmission daemon; the stub raises, which the real page handles).
  • /dev/* endpoints are unauthenticated; the container binds to 127.0.0.1 only, and requests with another Host or from another site get a 403.