Files
xenia-pos-local/local_backend/services/cloud_sync.py
T
bonaminandClaude Opus 5.5 76aac203d6 feat(backend): runtime LAN-IP detection (netinfo helper) + manual override
The backend's bridge network can't see the host's NICs, so HOST_IP from
install.sh went stale silently after a DHCP change. Now:

- services/netinfo_helper.py runs as a new `netinfo` service (same backend
  image, network_mode: host): every 60s it picks the PHYSICAL LAN address
  (main-table default-route NIC if real hardware, else first real NIC with
  IPv4; never WireGuard/Tailscale/ZeroTier/bridges/veths, ignores 169.254)
  and writes it to the shared `netinfo` volume. Stdlib only. On Docker
  Desktop (linuxkit/WSL2 kernel) it reports "unsupported" instead of the
  VM's meaningless address.
- services/lan_ip.py: one resolver used by /api/system/status (lan_ip +
  lan_ip_info), the pairing QR and the cloud heartbeat's local_ip:
  override (pos_settings network.lan_ip_override) → live detection (ignored
  when older than 5 min) → HOST_IP. Flags `mismatch` when a pinned address
  is no longer on any of the machine's NICs.
- PUT /api/system/lan-ip-override (manager): set a private IPv4 or null to
  return to automatic; public/loopback/link-local/IPv6 rejected (422).
- cloud_sync._get_local_ip uses the resolver (no more socket trick that
  returned the container IP).

Tested: helper selection on a fake sysfs/route table (8 cases incl. VPN
default routes) + real ioctl/route parsing on a Linux kernel; resolver
priority/staleness/mismatch/validation (18 cases); isolated full stack:
HOST_IP fallback on Docker Desktop, override save/validate/auth, simulated
Linux detection incl. DHCP change and dead helper, heartbeat IP.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-28 16:13:17 +03:00

527 lines
20 KiB
Python

"""
Periodic cloud check-in. Runs every 5 minutes as an asyncio background task.
Grace period: 72 hours (3 days) before marking unlicensed on connectivity failure.
Lock behaviour:
- cloud sets locked=true → set lock_pending=true in state
- lock_pending is enforced at workday-close time (see business_day router)
- while a workday is open, the site keeps running; lock applies once it closes
License expiry behaviour:
- 5 days before expiry → warning only (days_until_expiry in state)
- on expiry → 5-day grace period begins (grace_expires_at in state)
- after grace + no open workday → licensed=False enforced by business_day router
"""
import asyncio
import json
import logging
import os
from datetime import datetime, timedelta, timezone
from pathlib import Path
import httpx
from config import settings
from middleware.license_check import license_state
ORDER_POLL_INTERVAL = settings.CONNECT_SYNC_INTERVAL_SECONDS
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
SYNC_INTERVAL_SECONDS = 5 * 60 # 5 minutes
GRACE_HOURS = 72 # 3 days offline grace
EXPIRY_GRACE_DAYS = 5 # days after expiry before blocking
EXPIRY_WARNING_DAYS = 5 # days before expiry to show warning
STATE_FILE = Path(__file__).parent.parent / "license_state.json"
def _load_persisted_state():
if STATE_FILE.exists():
try:
data = json.loads(STATE_FILE.read_text())
license_state.update(data)
logger.info("Loaded persisted license state: %s", data)
except Exception as e:
logger.warning("Could not load license state file: %s", e)
def _persist_state():
try:
STATE_FILE.write_text(json.dumps(license_state))
except Exception as e:
logger.warning("Could not persist license state: %s", e)
def _compute_expiry_fields(expires_at_str: str | None) -> dict:
"""Return days_until_expiry and grace_expires_at derived from expires_at."""
if not expires_at_str:
return {"days_until_expiry": None, "grace_expires_at": None}
try:
expires_at = datetime.fromisoformat(expires_at_str)
if expires_at.tzinfo is None:
expires_at = expires_at.replace(tzinfo=timezone.utc)
except ValueError:
return {"days_until_expiry": None, "grace_expires_at": None}
now = datetime.now(timezone.utc)
days_until = (expires_at - now).days # negative once expired
grace_expires_at = None
if days_until < 0:
grace_expires_at = (expires_at + timedelta(days=EXPIRY_GRACE_DAYS)).isoformat()
return {
"days_until_expiry": days_until,
"grace_expires_at": grace_expires_at,
}
def _get_local_ip() -> str | None:
"""The server's LAN IP as phones see it — same resolver as the manager's
pairing QR (override → live detection → HOST_IP), see services/lan_ip.py.
No socket tricks: inside Docker they return the unreachable bridge IP."""
from database import SessionLocal
from services.lan_ip import resolve_lan_ip
db = SessionLocal()
try:
return resolve_lan_ip(db)["effective"]
except Exception:
return None
finally:
db.close()
async def _sync_once():
if not settings.SITE_ID or not settings.CLOUD_URL:
logger.debug("No SITE_ID/CLOUD_URL configured — skipping cloud sync")
return
try:
local_ip = _get_local_ip()
async with httpx.AsyncClient(timeout=10) as client:
resp = await client.post(
f"{settings.CLOUD_URL}/api/heartbeat/",
headers={
"X-Site-ID": settings.SITE_ID,
"X-Site-Key": settings.SITE_KEY,
},
json={"version": settings.VERSION, "uptime_seconds": 0, "local_ip": local_ip},
)
resp.raise_for_status()
data = resp.json()
licensed = data.get("licensed", True)
cloud_locked = data.get("locked", False)
expires_at = data.get("expires_at")
expiry_fields = _compute_expiry_fields(expires_at)
# If cloud says locked, check whether a workday is currently open.
# No open workday → lock immediately.
# Open workday → defer to workday close (business_day router enforces it).
if cloud_locked:
from database import SessionLocal
from models.business_day import BusinessDay
db = SessionLocal()
try:
open_day = db.query(BusinessDay).filter(BusinessDay.status == "open").first()
finally:
db.close()
if open_day:
if not license_state.get("lock_pending"):
license_state["lock_pending"] = True
logger.info("Cloud requested lock — workday open, deferring to workday close")
else:
license_state["lock_pending"] = False
license_state["locked"] = True
logger.info("Cloud requested lock — no open workday, locking immediately")
# If cloud lifts the lock, clear pending too
if not cloud_locked:
license_state["lock_pending"] = False
license_state["locked"] = False
license_state.update({
"licensed": licensed,
"expires_at": expires_at,
"latest_version": data.get("latest_version"),
"waiter_domain": data.get("waiter_domain"),
"site_numeric_id": data.get("site_numeric_id"),
"last_sync": datetime.now(timezone.utc).isoformat(),
"sync_failed": False,
**expiry_fields,
})
_persist_state()
logger.info("Cloud sync OK: licensed=%s locked=%s expires_at=%s", licensed, cloud_locked, expires_at)
except Exception as e:
logger.warning("Cloud sync failed: %s", e)
license_state["sync_failed"] = True
last_sync_str = license_state.get("last_sync")
if last_sync_str:
try:
last_sync = datetime.fromisoformat(last_sync_str)
grace_expires = last_sync + timedelta(hours=GRACE_HOURS)
if datetime.now(timezone.utc) > grace_expires:
logger.error("72-hour offline grace period expired — marking unlicensed")
license_state["licensed"] = False
except ValueError:
pass
# Recompute expiry fields from cached expires_at even when offline
expiry_fields = _compute_expiry_fields(license_state.get("expires_at"))
license_state.update(expiry_fields)
IMAGE_DIR = Path("/app/data/product_images")
async def _push_product_images(db, products):
"""Upload the local image file for each product whose picture changed since
the last successful push (tracked via Product.cloud_image_hash), so the
public QR menu can show it without needing a manually-set digital_image_url.
Skips products that already have a manual digital_image_url override."""
import hashlib
site_numeric_id = license_state.get("site_numeric_id")
if not site_numeric_id:
return
for p in products:
if p.digital_image_url or not p.image_url:
continue
filename = os.path.basename(p.image_url)
filepath = IMAGE_DIR / filename
if not filepath.exists():
continue
try:
contents = filepath.read_bytes()
image_hash = hashlib.sha256(contents).hexdigest()
if image_hash == p.cloud_image_hash:
continue
async with httpx.AsyncClient(timeout=20) as client:
resp = await client.post(
f"{settings.CLOUD_URL}/api/menu/sync-image",
headers={"X-Site-ID": settings.SITE_ID, "X-Site-Key": settings.SITE_KEY},
data={"product_id": str(p.id)},
files={"file": (filename, contents)},
)
resp.raise_for_status()
p.cloud_image_hash = image_hash
db.commit()
logger.info("Pushed image for product %d to cloud", p.id)
except Exception as e:
logger.warning("Image push failed for product %d: %s", p.id, e)
async def _push_menu_snapshot():
"""Serialize all digital-visible products+categories and POST to cloud."""
if not settings.SITE_ID or not settings.CLOUD_URL:
return
try:
from database import SessionLocal
from models.product import Category, Product
db = SessionLocal()
try:
categories = db.query(Category).filter(Category.parent_id == None).all()
payload_categories = []
all_products = []
for cat in categories:
products = (
db.query(Product)
.filter(
Product.category_id == cat.id,
Product.digital_visible == 1,
Product.lifecycle_status == "active",
)
.order_by(Product.sort_order)
.all()
)
all_products.extend(products)
product_list = []
for p in products:
product_list.append({
"id": p.id,
"name": p.name,
"digital_name": p.digital_name,
"digital_description": p.digital_description,
"digital_price": p.digital_price,
"base_price": p.base_price,
"digital_discount": p.digital_discount,
"digital_available": bool(p.digital_available),
"digital_image_url": p.digital_image_url,
"image_url": p.image_url,
"quick_options": [
{"id": o.id, "name": o.name, "price": o.price}
for o in p.quick_options
],
})
if product_list:
payload_categories.append({
"id": cat.id,
"name": cat.name,
"sort_order": cat.sort_order,
"products": product_list,
})
snapshot_json = json.dumps({"categories": payload_categories})
# Resolve numeric site_id from license state (set by heartbeat response)
site_numeric_id = license_state.get("site_numeric_id")
if not site_numeric_id:
logger.debug("Menu push skipped — site_numeric_id not yet known")
return
async with httpx.AsyncClient(timeout=15) as client:
resp = await client.post(
f"{settings.CLOUD_URL}/api/menu/sync",
headers={"X-Site-ID": settings.SITE_ID, "X-Site-Key": settings.SITE_KEY},
json={"site_id": site_numeric_id, "snapshot_json": snapshot_json},
)
resp.raise_for_status()
logger.info("Menu snapshot pushed (%d categories)", len(payload_categories))
await _push_product_images(db, all_products)
finally:
db.close()
except Exception as e:
logger.warning("Menu snapshot push failed: %s", e)
async def _pull_pending_orders():
"""Fetch online orders from cloud that haven't been synced to local yet."""
if not settings.SITE_ID or not settings.CLOUD_URL:
return
site_numeric_id = license_state.get("site_numeric_id")
if not site_numeric_id:
return
try:
async with httpx.AsyncClient(timeout=10) as client:
resp = await client.get(
f"{settings.CLOUD_URL}/api/orders/pending/{site_numeric_id}",
headers={"X-Site-ID": settings.SITE_ID, "X-Site-Key": settings.SITE_KEY},
)
resp.raise_for_status()
orders = resp.json()
if not orders:
return
from database import SessionLocal
from models.order import Order, OrderItem
from models.product import Product
from services.sse_bus import broadcast_sync
db = SessionLocal()
try:
# Use a system user id=1 (first manager/sysadmin) as the opener
from models.user import User
system_user = db.query(User).filter(User.perm_access_dashboard == True).first()
opener_id = system_user.id if system_user else 1
for cloud_order in orders:
# Create local Order row — no table_id for online orders
local_order = Order(
table_id=None,
opened_by=opener_id,
status="open",
source="online",
online_order_ref=cloud_order["public_ref"],
online_order_cloud_id=cloud_order["id"],
online_status="pending_acceptance",
online_customer_name=cloud_order.get("customer_name"),
online_customer_phone=cloud_order.get("customer_phone"),
online_customer_address=cloud_order.get("customer_address"),
online_customer_notes=cloud_order.get("customer_notes"),
online_order_type=cloud_order.get("order_type"),
)
db.add(local_order)
db.flush() # get local_order.id
# Create OrderItem rows
items = json.loads(cloud_order.get("items_json", "[]"))
for item in items:
product = db.query(Product).filter(
Product.id == item.get("product_id")
).first()
db.add(OrderItem(
order_id=local_order.id,
product_id=item.get("product_id"),
added_by=opener_id,
quantity=item.get("quantity", 1),
unit_price=item.get("unit_price", 0.0),
selected_options=json.dumps(item.get("options", [])),
status="active",
printed=False,
))
db.commit()
# Mark synced on cloud
async with httpx.AsyncClient(timeout=10) as client:
await client.post(
f"{settings.CLOUD_URL}/api/orders/{cloud_order['id']}/synced",
headers={"X-Site-ID": settings.SITE_ID, "X-Site-Key": settings.SITE_KEY},
json={"local_order_id": local_order.id},
)
logger.info("Online order %s pulled and created as local order %d",
cloud_order["public_ref"], local_order.id)
db.close()
# Broadcast SSE so the manager dashboard lights up
broadcast_sync("online_order_received", {"count": len(orders)})
except Exception as e:
db.rollback()
db.close()
logger.error("Failed to process pulled orders: %s", e)
except Exception as e:
logger.warning("Order pull failed: %s", e)
async def _sync_loop():
_load_persisted_state()
while True:
await _sync_once()
await asyncio.sleep(SYNC_INTERVAL_SECONDS)
async def _push_stats_snapshot():
"""Collect live operational stats and POST to cloud for the remote manager dashboard."""
if not settings.SITE_ID or not settings.CLOUD_URL:
return
site_numeric_id = license_state.get("site_numeric_id")
if not site_numeric_id:
return
try:
from database import SessionLocal
from models.order import Order, OrderItem
from models.business_day import BusinessDay
from models.shift import WaiterShift
from sqlalchemy import func
db = SessionLocal()
try:
now = datetime.now(timezone.utc)
today_start = now.replace(hour=0, minute=0, second=0, microsecond=0)
# Open tables = orders currently in status open or partially_paid
open_tables = db.query(Order).filter(
Order.status.in_(["open", "partially_paid"]),
Order.source == "pos",
).count()
# Today's completed POS orders (paid or closed today)
today_pos_orders = db.query(Order).filter(
Order.status.in_(["paid", "closed"]),
Order.source == "pos",
Order.closed_at >= today_start,
).all()
today_order_count = len(today_pos_orders)
# Today's revenue: sum of paid items on those orders
today_revenue = 0.0
for o in today_pos_orders:
for item in o.items:
if item.status in ("active", "paid"):
today_revenue += item.unit_price * item.quantity
# Online orders today
online_orders_today = db.query(Order).filter(
Order.source == "online",
Order.opened_at >= today_start,
).count()
online_orders_pending = db.query(Order).filter(
Order.source == "online",
Order.online_status == "pending_acceptance",
).count()
# Active business day
open_day = db.query(BusinessDay).filter(BusinessDay.status == "open").first()
current_shift = None
if open_day:
active_shift = db.query(WaiterShift).filter(
WaiterShift.business_day_id == open_day.id,
WaiterShift.ended_at == None,
).first()
if active_shift:
current_shift = {
"waiter_id": active_shift.waiter_id,
"started_at": active_shift.started_at.isoformat(),
}
finally:
db.close()
snapshot = {
"open_tables": open_tables,
"today_revenue": round(today_revenue, 2),
"today_orders": today_order_count,
"online_orders_pending": online_orders_pending,
"online_orders_today": online_orders_today,
"current_shift": current_shift,
"as_of": now.isoformat(),
}
async with httpx.AsyncClient(timeout=10) as client:
resp = await client.post(
f"{settings.CLOUD_URL}/api/remote/snapshot",
headers={"X-Site-ID": settings.SITE_ID, "X-Site-Key": settings.SITE_KEY},
json={"site_id": site_numeric_id, "snapshot_json": json.dumps(snapshot)},
)
resp.raise_for_status()
logger.info("Stats snapshot pushed: %s", snapshot)
except Exception as e:
logger.warning("Stats snapshot push failed: %s", e)
async def _connect_loop():
"""Faster loop: pulls pending online orders every CONNECT_SYNC_INTERVAL_SECONDS.
Also pushes menu snapshot and stats every 5 minutes (piggybacking on this loop).
On startup, waits for the first heartbeat to populate site_numeric_id then
immediately pushes menu + stats without waiting a full cycle."""
push_every = max(1, (5 * 60) // ORDER_POLL_INTERVAL)
# Wait for the heartbeat loop to get site_numeric_id, then do an immediate push
for _ in range(30): # wait up to 30 seconds
if license_state.get("site_numeric_id"):
break
await asyncio.sleep(1)
await _push_menu_snapshot()
await _push_stats_snapshot()
tick = 0
while True:
await asyncio.sleep(ORDER_POLL_INTERVAL)
await _pull_pending_orders()
tick += 1
if tick % push_every == 0:
await _push_menu_snapshot()
await _push_stats_snapshot()
async def start_cloud_sync() -> asyncio.Task:
heartbeat_task = asyncio.create_task(_sync_loop())
asyncio.create_task(_connect_loop())
return heartbeat_task