Files
Alexey Pavlov 3c325cbfc1 OPi: store-and-forward outbox для /api/capture (переживает обрыв сети)
- outbox.py: SQLite-очередь + JPEG на диске, at-least-once доставка
- scales.py: capture_and_send кладёт в очередь вместо прямого POST,
  heartbeat_loop каждые 30с дренирует очередь, статус публикует outbox_pending
- fw_version -> 2026-07-02-outbox
2026-07-02 16:00:21 +03:00

190 lines
8.1 KiB
Python

"""
outbox.py — надёжная доставка взвешиваний на сервер (store-and-forward).
При обрыве сети OPi↔сервер события не теряются: кладём JPEG на диск + метаданные
в SQLite, и досылаем при восстановлении связи. Сервер дедупит по event_id (upsert).
Семантика: at-least-once (повторная отправка безопасна).
Зависимости: только stdlib (sqlite3, urllib, json).
"""
import os
import time
import base64
import sqlite3
import threading
import logging
import json
import urllib.request
import urllib.error
log = logging.getLogger("outbox")
class CaptureOutbox:
def __init__(self, server_url, device_id,
db_path="/root/scales_outbox/outbox.db",
photo_dir="/root/scales_outbox/photos",
timeout=15, max_attempts=20):
self.url = server_url.rstrip("/") + "/api/capture"
self.device_id = device_id
self.timeout = timeout
self.max_attempts = max_attempts
self.photo_dir = photo_dir
os.makedirs(os.path.dirname(db_path), exist_ok=True)
os.makedirs(photo_dir, exist_ok=True)
self._lock = threading.Lock() # защита БД (heartbeat + main)
self._flush_lock = threading.Lock() # одна flush за раз
self._db = sqlite3.connect(db_path, check_same_thread=False)
self._db.execute("PRAGMA journal_mode=WAL")
self._db.execute("""
CREATE TABLE IF NOT EXISTS outbox (
event_id TEXT PRIMARY KEY,
weight REAL,
weight_raw TEXT,
captured_at TEXT,
photo_path TEXT NOT NULL,
created_at TEXT NOT NULL,
attempts INTEGER NOT NULL DEFAULT 0,
status TEXT NOT NULL DEFAULT 'pending' -- pending | failed
)""")
self._db.commit()
self._reconcile()
def _reconcile(self):
"""После рестарта убрать осиротевшие JPEG."""
with self._lock:
known = {r[0] for r in self._db.execute("SELECT photo_path FROM outbox")}
for name in os.listdir(self.photo_dir):
p = os.path.join(self.photo_dir, name)
if p not in known:
try:
os.remove(p)
except OSError:
pass
# ── публичный API ─────────────────────────────────────────────
def enqueue(self, event_id, weight, weight_raw, captured_at, jpeg_bytes):
"""Положить взвешивание в очередь. JPEG на диск, метаданные в БД."""
photo_path = os.path.join(self.photo_dir, f"{event_id}.jpg")
with open(photo_path, "wb") as f:
f.write(jpeg_bytes)
with self._lock:
self._db.execute(
"INSERT OR REPLACE INTO outbox "
"(event_id, weight, weight_raw, captured_at, photo_path, created_at, attempts, status) "
"VALUES (?,?,?,?,?,?,0,'pending')",
(event_id, weight, weight_raw, captured_at, photo_path,
time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())))
self._db.commit()
log.info("enqueued %s (%.1f kg)", event_id, weight)
def pending_count(self):
"""Количество событий в очереди."""
with self._lock:
return self._db.execute(
"SELECT COUNT(*) FROM outbox WHERE status='pending'").fetchone()[0]
def flush(self):
"""Досылает pending-события (старые первыми). Вернёт (отправлено, осталось)."""
if not self._flush_lock.acquire(blocking=False):
return (0, self.pending_count())
sent = 0
try:
while True:
with self._lock:
row = self._db.execute(
"SELECT event_id, weight, weight_raw, captured_at, photo_path, attempts "
"FROM outbox WHERE status='pending' ORDER BY rowid LIMIT 1").fetchone()
if row is None:
break
event_id, weight, weight_raw, captured_at, photo_path, attempts = row
try:
with open(photo_path, "rb") as f:
b64 = base64.b64encode(f.read()).decode()
except OSError:
self._drop(event_id)
continue
payload = json.dumps({
"event_id": event_id,
"weight": weight,
"weight_raw": weight_raw,
"device_id": self.device_id,
"captured_at": captured_at,
"photo_b64": b64,
"media_type": "image/jpeg",
}).encode()
try:
req = urllib.request.Request(
self.url,
data=payload,
headers={"Content-Type": "application/json"})
with urllib.request.urlopen(req, timeout=self.timeout) as r:
resp = r.read().decode()
if _ok_response(resp):
self._delete(event_id)
sent += 1
log.info("delivered %s", event_id)
else:
log.warning("server rejected %s: %s", event_id, resp[:100])
self._park(event_id)
except urllib.error.HTTPError as e:
if 400 <= e.code < 500:
log.error("rejected %s: %d", event_id, e.code)
self._park(event_id)
else:
self._bump(event_id, attempts)
break
except (urllib.error.URLError, OSError, TimeoutError) as e:
log.warning("network unavailable (%s), pending=%d",
type(e).__name__, self.pending_count())
break
finally:
self._flush_lock.release()
return (sent, self.pending_count())
# ── внутреннее ────────────────────────────────────────────────
def _delete(self, event_id):
with self._lock:
row = self._db.execute("SELECT photo_path FROM outbox WHERE event_id=?",
(event_id,)).fetchone()
self._db.execute("DELETE FROM outbox WHERE event_id=?", (event_id,))
self._db.commit()
if row:
try:
os.remove(row[0])
except OSError:
pass
def _drop(self, event_id):
"""Удалить событие (фото потеряно)."""
with self._lock:
self._db.execute("DELETE FROM outbox WHERE event_id=?", (event_id,))
self._db.commit()
def _park(self, event_id):
"""Паркировать событие (не отправлять больше)."""
with self._lock:
self._db.execute("UPDATE outbox SET status='failed' WHERE event_id=?", (event_id,))
self._db.commit()
def _bump(self, event_id, attempts):
"""Увеличить счётчик попыток, паркировать если превышен лимит."""
status = "failed" if attempts + 1 >= self.max_attempts else "pending"
with self._lock:
self._db.execute("UPDATE outbox SET attempts=attempts+1, status=? WHERE event_id=?",
(status, event_id))
self._db.commit()
def _ok_response(text):
"""Проверить, что в ответе есть 'ok': true."""
try:
return bool(json.loads(text).get("ok"))
except Exception:
return False