dev_1.0.0_yuyr_2：重新提交 PR，增加 master/agent 以及系统集成测试 (#17)

Reviewed-on: #17
Reviewed-by: sundapeng <sundp@mail.zgclab.edu.cn>
Reviewed-by: xuxt <xuxt@zgclab.edu.cn>

src/master/app/scheduler.py

@@ -0,0 +1,90 @@
+from __future__ import annotations
+
+import logging
+import threading
+from typing import Optional
+
+from .config import AppConfig
+from .storage import Storage
+from .util import atomic_write_json, parse_iso, to_iso, utcnow
+
+
+class StatusScheduler:
+    def __init__(self, storage: Storage, config: AppConfig, logger: Optional[logging.Logger] = None) -> None:
+        self._storage = storage
+        self._config = config
+        self._logger = logger or logging.getLogger("argus.master.scheduler")
+        self._stop_event = threading.Event()
+        self._thread = threading.Thread(target=self._run, name="status-scheduler", daemon=True)
+        self._nodes_json_lock = threading.Lock()
+        self._pending_nodes_json = threading.Event()
+
+    def start(self) -> None:
+        """启动后台线程，定期刷新节点状态与 nodes.json。"""
+        if not self._thread.is_alive():
+            self._logger.info("Starting scheduler thread")
+            self._thread.start()
+
+    def stop(self) -> None:
+        self._stop_event.set()
+        self._pending_nodes_json.set()
+        self._thread.join(timeout=5)
+
+    def trigger_nodes_json_refresh(self) -> None:
+        self._pending_nodes_json.set()
+
+    def generate_nodes_json(self) -> None:
+        with self._nodes_json_lock:
+            online_nodes = self._storage.get_online_nodes()
+            atomic_write_json(self._config.metric_nodes_json_path, online_nodes)
+            self._logger.info("nodes.json updated", extra={"count": len(online_nodes)})
+
+    # ------------------------------------------------------------------
+    # internal loop
+    # ------------------------------------------------------------------
+
+    def _run(self) -> None:
+        # 确保启动时 nodes.json 会立即生成
+        self._pending_nodes_json.set()
+        while not self._stop_event.is_set():
+            changed = self._reconcile_statuses()
+            if changed or self._pending_nodes_json.is_set():
+                try:
+                    self.generate_nodes_json()
+                finally:
+                    self._pending_nodes_json.clear()
+            self._stop_event.wait(self._config.scheduler_interval_seconds)
+
+    def _reconcile_statuses(self) -> bool:
+        """根据 last_report 与当前时间对比，决定是否切换状态。"""
+        any_status_changed = False
+        now = utcnow()
+        rows = self._storage.fetch_nodes_for_scheduler()
+        for row in rows:
+            node_id = row["id"]
+            last_report_iso = row["last_report"]
+            current_status = row["status"]
+            last_report_dt = parse_iso(last_report_iso)
+            if last_report_dt is None:
+                # No report yet; treat as initialized until report arrives
+                continue
+            delta_seconds = (now - last_report_dt).total_seconds()
+            new_status = current_status
+            if delta_seconds > self._config.offline_threshold_seconds:
+                new_status = "offline"
+            elif delta_seconds <= self._config.online_threshold_seconds:
+                new_status = "online"
+            # Between thresholds: keep current status (sticky)
+            if new_status != current_status:
+                any_status_changed = True
+                self._logger.info(
+                    "Updating node status",
+                    extra={
+                        "node_id": node_id,
+                        "previous": current_status,
+                        "new": new_status,
+                        "delta_seconds": delta_seconds,
+                    },
+                )
+                self._storage.update_status(node_id, new_status, last_updated_iso=to_iso(now))
+        return any_status_changed