fix(monitor): 监控班次二 P2 批——D-5 db_sample 日历不可达去 fail-open(样本可读弱检/库不可达黄)+D-6 month_bucket 损坏件 rows=None 黄(对齐 per_date)+D-8 check_disk 阈值参数化消费注册表 warn_gb/crit_gb(旧硬编码 8/20=改表不改行为校准陷阱,crit_gb 全仓无消费)+D-9 vintage/NAS ts fromisoformat 捕 TypeError+D-10 db_sample close 进 finally+AL-1 upsert 更新 severity(同日黄→红升级即时落表,旧 UPDATE 不动=次日新 id 才纠正)+purge 收口 MON-202610-04(open 黄改 last_seen 28d=活着不删,旧 first_seen 14d 硬删无声抹持续黄+断 count 升级链;resolved 统一 30d,open 红永不删)+N-1 run_monitor_nas.sh rc 先存后拼($(date) 同词先展开重置 $?=exit 恒 0,rc 隔离下 NAS 崩溃唯一现场线索)+N-2 NAS 日历豁免短期方案(last_weekday 接休市表镜像+CI 同步测钉与 monitor_registry 相等=7f9b413 镜像纪律;10-06 旧码会把 10-05 假日当昨交易日;corpus 34h 日历日节奏实证无需豁免)+P3 归因([FATAL] 级别标记不当类型归 _global,scan_signatures 注记 test-only 勿接线)+T-2 NAS red 分支(hot_rank 文件缺)测试补缺+migration_logs 保洁(data_monitor_wrapper 双班剪 14d 前旧件,13 wrapper 无限增长收口);monitoring spec §5.1/§5.2/§5.3-④/§5.4/§5.5-④/§7.6/§12 随码(C-A1 universe 异常实串/C-A2 签名表勘正补 Expecting value/C-A6 disk 阈值表述+锚点行数刷新) [vps] [nas]

scripts/data_platform/data_monitor.py

@@ -152,7 +152,8 @@ def _check_entry_impl(e, ctx):
             return done("stale", "yellow", f"<={e['max_age_days']}d", f"{age}d")
         lo, hi = e["row_band"]
         rows = chk.parquet_rows(got[1])
-        if rows is not None and not (lo <= rows <= hi):
+        if rows is None or not (lo <= rows <= hi):
+            # D-6(10-05): 损坏件 rows=None 不再静默绿——对齐 per_date 判定
             return done("rows", "yellow", f"{lo}~{hi}", rows)
         return done("stale", "green", ym, ym)
 
@@ -183,23 +184,33 @@ def _check_entry_impl(e, ctx):
     if kind == "db_sample":
         exp = chk.expected_daily(ctx["days"], ctx["now"], e.get("ready", "22:30"),
                                  e.get("lag", 0))
-        worst = None
+        worst, readable = None, False
         for sym, exc in e["samples"]:
+            got_d = None
             try:
                 conn = _ro_conn(os.path.join(ctx["data_root"], "quant_trading.db"))
-                row = conn.execute(
-                    "SELECT MAX(datetime) FROM dbbardata WHERE symbol=? AND "
-                    "exchange=? AND interval='d'", (sym, exc)).fetchone()
-                conn.close()
+                try:
+                    row = conn.execute(
+                        "SELECT MAX(datetime) FROM dbbardata WHERE symbol=? AND "
+                        "exchange=? AND interval='d'", (sym, exc)).fetchone()
+                finally:
+                    conn.close()   # D-10(10-05): close 进 finally, 异常路径不泄漏
                 got_d = (row[0] or "").replace("-", "")
             except Exception:  # noqa: BLE001  只读探测, 库坏=黄红由调用方判
                 got_d = None
+            if got_d:
+                readable = True
             if exp and (got_d is None or got_d < exp) and worst is None:
                 worst = {"symbol": f"{sym}.{exc}", "expected": exp, "actual": got_d}
         if exp and worst:
             sev = "yellow" if exp not in ctx["confirmed_days"] \
                 else e.get("severity", "red")
             return done("stale", sev, exp, worst["actual"], worst)
+        if exp is None:
+            # D-5(10-05): 日历不可达不再无条件绿(fail-open)——弱检查=样本可读
+            # (对齐 per_date 存在性弱检), 库也不可达=黄
+            return done("stale", "green" if readable else "yellow", "n/a",
+                        "sample-readable" if readable else "db-unreadable")
         return done("stale", "green", exp, "ok")
 
     if kind == "vintage_json":
@@ -213,8 +224,8 @@ def _check_entry_impl(e, ctx):
         try:
             stale = not checked or (ctx["now"] - dt.datetime.fromisoformat(
                 checked)).total_seconds() > 26 * 3600
-        except ValueError:
-            stale = True
+        except (ValueError, TypeError):
+            stale = True   # D-9(10-05): 非字符串 fromisoformat 抛 TypeError 不炸
         if stale:
             return done("verdict", "yellow", "<=26h", checked)
         # 结论无顶层字段(static_vintage_check 契约): 挖洞藏在 tables.*.holes_*
@@ -237,7 +248,10 @@ def _check_entry_impl(e, ctx):
                     "exists", exists)
 
     if kind == "disk":
-        r = chk.check_disk(ctx["data_root"])
+        # D-8(10-05): 阈值消费注册表(旧硬编码 8/20 与注册表脱钩, crit_gb 全仓
+        # 无消费=改表不改行为的校准陷阱)
+        r = chk.check_disk(ctx["data_root"], warn_gb=e.get("warn_gb", 20),
+                           crit_gb=e.get("crit_gb", 8))
         return {"key": "infra-disk-data", "status": r["status"],
                 "expected": f">={e['warn_gb']}G", "actual": f"{r['free_gb']}G",
                 "evidence": r}
@@ -530,7 +544,7 @@ def _run_shift(args, db_path):
             try:
                 age_h = (now - dt.datetime.fromisoformat(nas_ts)
                          ).total_seconds() / 3600
-            except ValueError:
+            except (ValueError, TypeError):   # D-9(10-05): TypeError 同防
                 age_h = None
             if age_h is not None and age_h > 30:
                 events.append(_event(nas_silent_key, "yellow",

scripts/data_platform/data_monitor_wrapper.ps1

@@ -10,6 +10,12 @@ $ts = Get-Date -Format 'yyyyMMdd_HHmmss'
 $logDir = 'C:\sanguo_vnpy_v2\data\migration_logs'
 if (-not (Test-Path $logDir)) { New-Item -ItemType Directory -Path $logDir -Force | Out-Null }
 $log = Join-Path $logDir "data_monitor_$ts.txt"
+# migration_logs 保洁(10-05 infra 插单): 13 个 wrapper 各自追加、无任何清理机制
+# =无限增长; 本 wrapper 双班跑必经, 顺手剪 14 天前旧件(PS 重定向副本无解析消费,
+# 监控日志层真身=static/logs/, 诊断残留而已)。
+Get-ChildItem -Path $logDir -File -ErrorAction SilentlyContinue |
+  Where-Object { $_.LastWriteTime -lt (Get-Date).AddDays(-14) } |
+  Remove-Item -ErrorAction SilentlyContinue
 C:\Python310\python.exe -X utf8 C:\sanguo_vnpy_v2\scripts\data_platform\data_monitor.py --shift $Shift *>> $log
 $rc = $LASTEXITCODE
 Add-Content -Path $log -Value "data_monitor shift=$Shift rc=$rc"

scripts/data_platform/monitor_checks.py

@@ -147,13 +147,15 @@ def dir_latest_mtime(directory):
     return latest
 
 
-def check_disk(path):
+def check_disk(path, warn_gb=20, crit_gb=8):
+    """磁盘剩余水位; 阈值参数化(D-8 10-05: 注册表 warn_gb/crit_gb 消费——旧
+    硬编码 8/20 与注册表脱钩, crit_gb 曾全仓无消费=改表不改行为的校准陷阱)."""
     target = path if os.path.exists(path) else os.path.dirname(os.path.abspath(path))
     usage = shutil.disk_usage(target)
     free_gb = usage.free / 1024 ** 3
-    if free_gb < 8:
+    if free_gb < crit_gb:
         status = "red"
-    elif free_gb < 20:
+    elif free_gb < warn_gb:
         status = "yellow"
     else:
         status = "green"
@@ -195,7 +197,10 @@ def scan_signatures(text):
     """10-05 三修: ①只扫 ERROR/[FATAL] 行——'Expecting value' 出现在重试
     WARNING 行(financial_abstract/000963 实证), 良性重试不再误中
     source_changed; ②'418' 针收紧为 '418 Client Error'——裸 418 子串误中
-    rows=41823/ok=19418 进度行(anti_crawl 假阳性实证)。"""
+    rows=41823/ok=19418 进度行(anti_crawl 假阳性实证)。
+    **注(10-05 补充审计)**: 生产路径已无调用方(签名分诊走 signatures_by_type,
+    data_monitor.log_layer_results/log_only_results)——本函数仅测试消费,
+    勿再接线; 与 signatures_by_type 的行为差异=不按 [type] 归因。"""
     lines = _error_lines(text)
     seen, out = set(), []
     for needle, sig, severity in SIGNATURES:
@@ -211,10 +216,15 @@ _TYPE_TAG = re.compile(r"\[(\w+)\]")
 def signatures_by_type(text):
     """签名按日志行 [type] 归因(10-05): 'ERROR [share_capital] ...' 行的
     签名只归 share_capital——防跨类型连坐(xueqiu 的 418 把 share_capital
-    的 failed 抬红实证)。断路器行归触发类型, data_monitor 侧全局取。"""
+    的 failed 抬红实证)。断路器行归触发类型, data_monitor 侧全局取。
+    [FATAL] 是级别标记不是类型(P3 补充审计): 裸 FATAL 行
+    (akshare_static_download.py:2721 实格式)归 _global, [type] [FATAL]
+    双标记行(:2160 实格式)归 [type]。"""
     out = {}
     for ln in _error_lines(text):
         m = _TYPE_TAG.search(ln)
+        while m and m.group(1) == "FATAL":
+            m = _TYPE_TAG.search(ln, m.end())
         t = m.group(1) if m else "_global"
         sigs = out.setdefault(t, [])
         for needle, sig, severity in SIGNATURES:

scripts/nas_sync/nas_monitor_check.py

@@ -21,13 +21,34 @@ XCHECK_REPORTS = os.path.join(CORPUS_ROOT, "xcheck", "reports")
 OUT_DIR = "/volume1/stock/monitor_nas/out"
 DISK_WARN_GB = 100  # 初值, 首周观察校准(spec §20.15.1 磁盘水位 NAS 侧)
 
+# 沪深北休市表镜像(N-2 10-05)——=monitor_registry.EXCHANGE_HOLIDAYS, CI 同步测
+# (test_holiday_table_in_sync_with_vps_registry)钉两表相等, 单边改表即红(同
+# 7f9b413 strategy 侧镜像纪律); 年更义务经 VPS 侧年更闸+同步测传导到本表。
+# 旧 last_weekday 只剥周末: 0925 中秋/国庆连休窗把假日当「昨交易日」——当前被
+# hot_rank producer 的日历日节奏掩住(files 假期照落), 语义仍错。表缺年=不扣
+# (fail-safe, 黄照旧绝不静默转绿)。
+EXCHANGE_HOLIDAYS = {
+    2026: {
+        "20260101", "20260102",  # 元旦
+        "20260216", "20260217", "20260218", "20260219", "20260220", "20260223",  # 春节
+        "20260406",  # 清明
+        "20260501", "20260504", "20260505",  # 劳动节
+        "20260619",  # 端午
+        "20260925",  # 中秋
+        "20261001", "20261002", "20261005", "20261006", "20261007",  # 国庆(1008 复市)
+    },
+}
+
 
 def last_weekday(before):
-    """strictly before 且为周一~五的最近一天(YYYYMMDD)——16:00 查时当日 17:35 班未采."""
+    """strictly before 且为交易日(周一~五且非休市表日)的最近一天(YYYYMMDD)
+    ——16:00 查时当日 17:35 班未采; N-2: 接休市表镜像, 跳过连休假日."""
     d = before - dt.timedelta(days=1)
-    while d.weekday() >= 5:
+    while True:
+        hol = EXCHANGE_HOLIDAYS.get(d.year)
+        if d.weekday() < 5 and (hol is None or d.strftime("%Y%m%d") not in hol):
+            return d.strftime("%Y%m%d")
         d -= dt.timedelta(days=1)
-    return d.strftime("%Y%m%d")
 
 
 def mtime_age_hours(path, now):

scripts/nas_sync/run_monitor_nas.sh

@@ -21,7 +21,10 @@ mkdir -p "$ROOT/out"
 {
   echo "=== $(date '+%F %T') nas monitor start ==="
   "$PY" "$ROOT/nas_monitor_check.py"
-  echo "=== $(date '+%F %T') nas monitor exit=$? ==="
+  rc_nas=$?
+  # N-1(10-05): rc 先存后拼——旧版 $(date) 在同一 echo 词内先展开把 $? 重置
+  # 为 0, exit= 恒 0; rc 隔离下这是 NAS 体检崩溃在 cron.log 的唯一现场线索
+  echo "=== $(date '+%F %T') nas monitor exit=$rc_nas ==="
   # 远端 inbox 目录幂等建好; ssh 必带 -n(vps-mirror 坑: 不带会吃 stdin)
   ssh -n "${SCP_OPTS[@]}" "$VPS" \
     "powershell -Command \"New-Item -ItemType Directory -Force -Path '$VPS_INBOX' | Out-Null\"" \