Upd logger

app.py

@@ -99,6 +99,7 @@ def upsert_mongo(docs):
         col.create_index("timestamp", unique=True)
         ops = [UpdateOne({"_id": d["timestamp"]}, {"$set": d}, upsert=True) for d in docs]
         col.bulk_write(ops, ordered=False)
+        logger.info(f"🪣 Inserted to MongoDB.")
     except Exception as e:
         logger.error(f"❌ Mongo error: {e}")
 
@@ -137,6 +138,7 @@ def fill_missing(df: pd.DataFrame) -> pd.DataFrame:
     train = df[df["consume_clean"].notna()]
     pred  = df[df["consume_clean"].isna()]
     if not train.empty and not pred.empty:
+        logger.info("Start cleaning 🧹")
         model = LinearRegression().fit(train[["voltage","current","power"]], train["consume_clean"])
         try:
             y_hat = model.predict(pred[["voltage","current","power"]])
@@ -156,6 +158,7 @@ def fill_missing(df: pd.DataFrame) -> pd.DataFrame:
             y_fb = fb_model.predict(fb_pred[["ts_sec"]])
             df.loc[fb_pred.index, "consume_clean"] = pd.Series(y_fb, index=fb_pred.index)
         df.drop(columns=["ts_sec"], inplace=True)
+        logger.info("Finish cleaning 🧹")
     # Giá trị cuối và thải giá trị thừa
     df["consume"] = df["consume_clean"]
     # Đánh dấu những bản ghi vẫn còn thiếu consume
@@ -195,6 +198,8 @@ def backfill_worker():
     while not stop_event.is_set():
         time.sleep(BACKFILL_INTERVAL)
         df_win = pd.DataFrame(window)
+        if "need_backfill" not in df_win.columns:
+            continue  # chưa có dữ liệu đã qua fill_missing()
         pending_mask = df_win["need_backfill"]
         if not pending_mask.any():
             continue