@agentunion/kite 1.0.0

package/core/event_hub_bench/entry.py

@@ -0,0 +1,371 @@
+"""
+Event Hub Benchmark module entry point.
+Discovers hub via Registry, runs benchmarks, saves results, exits.
+"""
+
+import asyncio
+import json
+import os
+import platform
+import sys
+import threading
+import time
+from datetime import datetime, timezone
+
+import statistics
+import uuid
+
+import httpx
+import websockets
+
+_this_dir = os.path.dirname(os.path.abspath(__file__))
+_project_root = os.path.dirname(os.path.dirname(_this_dir))
+if _project_root not in sys.path:
+    sys.path.insert(0, _project_root)
+
+RESULTS_DIR = os.path.join(_project_root, "core", "event_hub", "bench_results")
+TAG = "[event_hub_bench]"
+
+
+def _discover_hub(registry_url: str, token: str, timeout: float = 30) -> str | None:
+    """Poll registry until event_hub is found. Returns ws:// URL or None."""
+    headers = {"Authorization": f"Bearer {token}"}
+    deadline = time.time() + timeout
+    while time.time() < deadline:
+        try:
+            resp = httpx.get(
+                f"{registry_url}/lookup/event_hub",
+                headers=headers, timeout=5,
+            )
+            if resp.status_code == 200:
+                info = resp.json()
+                ws_url = (info.get("metadata") or {}).get("ws_endpoint")
+                if ws_url:
+                    return ws_url
+        except Exception:
+            pass
+        time.sleep(1)
+    return None
+
+
+def _get_hub_stats(ws_url: str) -> dict:
+    """Fetch hub /stats via HTTP (derive from ws:// URL)."""
+    http_url = ws_url.replace("ws://", "http://").rsplit("/ws", 1)[0]
+    try:
+        resp = httpx.get(f"{http_url}/stats", timeout=5)
+        if resp.status_code == 200:
+            return resp.json()
+    except Exception:
+        pass
+    return {}
+
+
+def _sample_hub_resources(ws_url: str) -> dict:
+    """Sample hub process CPU/memory via psutil. Returns {} if unavailable."""
+    try:
+        import psutil
+    except ImportError:
+        return {"error": "psutil not installed"}
+    http_url = ws_url.replace("ws://", "http://").rsplit("/ws", 1)[0]
+    port = int(http_url.rsplit(":", 1)[1].split("/")[0])
+    for conn in psutil.net_connections(kind="tcp"):
+        if conn.laddr.port == port and conn.status == "LISTEN":
+            try:
+                p = psutil.Process(conn.pid)
+                mem = p.memory_info()
+                return {
+                    "pid": conn.pid,
+                    "cpu_percent": p.cpu_percent(interval=1),
+                    "rss_mb": round(mem.rss / 1048576, 1),
+                    "vms_mb": round(mem.vms / 1048576, 1),
+                    "threads": p.num_threads(),
+                }
+            except Exception:
+                pass
+    return {"error": "hub process not found"}
+
+
+# ── WebSocket client ──
+
+class WsClient:
+    """Lightweight async WebSocket client for hub benchmarks."""
+
+    def __init__(self, ws_url: str, token: str, client_id: str):
+        self.url = f"{ws_url}?token={token}&id={client_id}"
+        self.id = client_id
+        self.ws = None
+
+    async def connect(self):
+        self.ws = await websockets.connect(self.url, max_size=None)
+
+    async def subscribe(self, patterns: list[str]):
+        await self.ws.send(json.dumps({"type": "subscribe", "events": patterns}))
+
+    async def publish(self, event_type: str, data: dict) -> str:
+        eid = str(uuid.uuid4())
+        await self.ws.send(json.dumps({
+            "type": "event", "event_id": eid, "event": event_type,
+            "source": self.id,
+            "timestamp": datetime.now(timezone.utc).isoformat(),
+            "data": data,
+        }))
+        return eid
+
+    async def recv(self, timeout=5.0):
+        try:
+            return json.loads(await asyncio.wait_for(self.ws.recv(), timeout=timeout))
+        except Exception:
+            return None
+
+    async def close(self):
+        if self.ws:
+            await self.ws.close()
+
+
+# ── Benchmark runner ──
+
+class BenchRunner:
+    def __init__(self, ws_url: str, token: str):
+        self.ws_url = ws_url
+        self.token = token
+        self.results = {}
+
+    async def _make_client(self, name: str) -> WsClient:
+        c = WsClient(self.ws_url, self.token, name)
+        await c.connect()
+        return c
+
+    # ── Throughput: burst N events, measure hub processing rate ──
+
+    async def bench_throughput(self, n: int = 10000):
+        print(f"{TAG} Throughput test: {n} events...")
+        pub = await self._make_client("bench_tp_pub")
+        sub = await self._make_client("bench_tp_sub")
+        await sub.subscribe(["bench.tp.*"])
+        await asyncio.sleep(0.2)
+
+        recvd = 0
+        stop = asyncio.Event()
+
+        async def _recv():
+            nonlocal recvd
+            while not stop.is_set():
+                try:
+                    raw = await asyncio.wait_for(sub.ws.recv(), timeout=0.5)
+                    if '"event"' in raw and '"ack"' not in raw:
+                        recvd += 1
+                except Exception:
+                    pass
+
+        async def _drain_acks():
+            while not stop.is_set():
+                try:
+                    await asyncio.wait_for(pub.ws.recv(), timeout=0.5)
+                except Exception:
+                    pass
+
+        recv_task = asyncio.create_task(_recv())
+        ack_task = asyncio.create_task(_drain_acks())
+
+        hub_before = (_get_hub_stats(self.ws_url).get("counters") or {})
+
+        t0 = time.time()
+        for i in range(n):
+            await pub.publish("bench.tp.test", {"i": i})
+        send_time = time.time() - t0
+
+        deadline = time.time() + max(15, n / 300)
+        while recvd < n and time.time() < deadline:
+            await asyncio.sleep(0.1)
+        total_time = time.time() - t0
+
+        stop.set()
+        await recv_task
+        ack_task.cancel()
+
+        hub_after = (_get_hub_stats(self.ws_url).get("counters") or {})
+
+        await pub.close()
+        await sub.close()
+
+        self.results["throughput"] = {
+            "events": n,
+            "send_rate": round(n / send_time),
+            "recv_rate": round(recvd / total_time) if total_time > 0 else 0,
+            "client_recv": recvd,
+            "hub_queued": hub_after.get("events_queued", 0) - hub_before.get("events_queued", 0),
+            "hub_routed": hub_after.get("events_routed", 0) - hub_before.get("events_routed", 0),
+            "send_time_s": round(send_time, 2),
+            "total_time_s": round(total_time, 2),
+        }
+        print(f"{TAG}   send_rate={self.results['throughput']['send_rate']} evt/s, "
+              f"recv={recvd}/{n}, time={round(total_time, 1)}s")
+
+    # ── Latency: sequential pub→recv round-trip ──
+
+    async def bench_latency(self, n: int = 200):
+        print(f"{TAG} Latency test: {n} samples...")
+        pub = await self._make_client("bench_lat_pub")
+        sub = await self._make_client("bench_lat_sub")
+        await sub.subscribe(["bench.lat.*"])
+        await asyncio.sleep(0.2)
+
+        latencies = []
+        for i in range(n):
+            t0 = time.time()
+            await pub.publish("bench.lat.test", {"i": i})
+            await pub.recv(timeout=2)  # ack
+            msg = await sub.recv(timeout=2)
+            if msg and msg.get("type") == "event":
+                latencies.append((time.time() - t0) * 1000)
+
+        await pub.close()
+        await sub.close()
+
+        if latencies:
+            latencies.sort()
+            self.results["latency"] = {
+                "samples": len(latencies),
+                "avg_ms": round(statistics.mean(latencies), 2),
+                "p50_ms": round(latencies[len(latencies) // 2], 2),
+                "p95_ms": round(latencies[int(len(latencies) * 0.95)], 2),
+                "p99_ms": round(latencies[int(len(latencies) * 0.99)], 2),
+            }
+            print(f"{TAG}   avg={self.results['latency']['avg_ms']}ms, "
+                  f"p50={self.results['latency']['p50_ms']}ms, "
+                  f"p99={self.results['latency']['p99_ms']}ms")
+
+    # ── Fan-out: 1 pub, N subs ──
+
+    async def bench_fanout(self, n_events: int = 2000):
+        for n_subs in [1, 5, 10, 50]:
+            print(f"{TAG} Fan-out x{n_subs}: {n_events} events...")
+            pub = await self._make_client("bench_fo_pub")
+            subs = []
+            for i in range(n_subs):
+                s = await self._make_client(f"bench_fo_sub_{i}")
+                await s.subscribe(["bench.fo.*"])
+                subs.append(s)
+            await asyncio.sleep(0.3)
+
+            counts = [0] * n_subs
+            stop = asyncio.Event()
+
+            async def _recv(idx, client):
+                while not stop.is_set():
+                    try:
+                        raw = await asyncio.wait_for(client.ws.recv(), timeout=0.5)
+                        if '"event"' in raw and '"ack"' not in raw:
+                            counts[idx] += 1
+                    except Exception:
+                        pass
+
+            tasks = [asyncio.create_task(_recv(i, s)) for i, s in enumerate(subs)]
+            ack_stop = asyncio.Event()
+
+            async def _drain():
+                while not ack_stop.is_set():
+                    try:
+                        await asyncio.wait_for(pub.ws.recv(), timeout=0.5)
+                    except Exception:
+                        pass
+
+            ack_task = asyncio.create_task(_drain())
+
+            t0 = time.time()
+            for i in range(n_events):
+                await pub.publish("bench.fo.test", {"i": i})
+            send_time = time.time() - t0
+
+            await asyncio.sleep(max(3, n_events * n_subs / 3000))
+            stop.set()
+            ack_stop.set()
+            for t in tasks:
+                await t
+            ack_task.cancel()
+
+            await pub.close()
+            for s in subs:
+                await s.close()
+
+            avg_recv = sum(counts) / n_subs if n_subs else 0
+            self.results[f"fanout_{n_subs}"] = {
+                "subs": n_subs,
+                "events": n_events,
+                "send_rate": round(n_events / send_time) if send_time > 0 else 0,
+                "avg_recv": round(avg_recv),
+                "min_recv": min(counts),
+            }
+            print(f"{TAG}   subs={n_subs}, avg_recv={round(avg_recv)}/{n_events}, "
+                  f"min={min(counts)}")
+
+    # ── Save results ──
+
+    def save(self, ws_url: str):
+        os.makedirs(RESULTS_DIR, exist_ok=True)
+        now = datetime.now()
+        filepath = os.path.join(RESULTS_DIR, now.strftime("%Y-%m-%d_%H-%M-%S") + ".json")
+
+        hub_stats = _get_hub_stats(ws_url)
+        resources = _sample_hub_resources(ws_url)
+
+        data = {
+            "timestamp": now.isoformat(),
+            "env": {
+                "platform": sys.platform,
+                "python": platform.python_version(),
+            },
+            "hub_resources": resources,
+            "hub_counters": hub_stats.get("counters", {}),
+            **self.results,
+        }
+        with open(filepath, "w", encoding="utf-8") as f:
+            json.dump(data, f, indent=2, ensure_ascii=False)
+        print(f"{TAG} Results saved: {filepath}")
+
+
+# ── Entry point ──
+
+async def _run(ws_url: str, token: str):
+    bench = BenchRunner(ws_url, token)
+    await bench.bench_throughput()
+    await bench.bench_latency()
+    await bench.bench_fanout()
+    bench.save(ws_url)
+
+
+def main():
+    # Read boot_info from stdin
+    token = ""
+    registry_port = 0
+    try:
+        line = sys.stdin.readline().strip()
+        if line:
+            boot = json.loads(line)
+            token = boot.get("token", "")
+            registry_port = boot.get("registry_port", 0)
+    except Exception:
+        pass
+
+    if not token or not registry_port:
+        print(f"{TAG} ERROR: Missing token or registry_port")
+        sys.exit(1)
+
+    registry_url = f"http://127.0.0.1:{registry_port}"
+    print(f"{TAG} Discovering event_hub via registry...")
+
+    ws_url = _discover_hub(registry_url, token)
+    if not ws_url:
+        print(f"{TAG} ERROR: event_hub not found")
+        sys.exit(1)
+
+    print(f"{TAG} Hub found: {ws_url}")
+    print(f"{TAG} Starting benchmarks...")
+
+    asyncio.run(_run(ws_url, token))
+
+    print(f"{TAG} Benchmarks complete, exiting.")
+
+
+if __name__ == "__main__":
+    main()

package/core/event_hub_bench/module.md

@@ -0,0 +1,25 @@
+---
+name: event_hub_bench
+display_name: Event Hub Benchmark
+version: "1.0"
+type: tool
+state: enabled
+runtime: python
+entry: entry.py
+events: []
+subscriptions: []
+---
+
+# Event Hub Benchmark
+
+事件中心性能基准测试模块。
+
+- 启用后随 Kite 启动自动运行基准测试
+- 通过 Registry 发现 Event Hub，以真实模块身份连接
+- 多线程模拟多个 publisher/subscriber 压测 hub
+- 监控 hub 进程 CPU/内存占用
+- 结果保存到 `core/event_hub/bench_results/`
+- 测试完成后自动退出
+
+启用：将 `state` 改为 `enabled`
+停用：将 `state` 改为 `disabled`