机架管理前后端（rack/）：FastAPI + SQLite + Vue3，经统一门户上线

- 后端 prrack：MQTT（phonerail-backend @ 127.0.0.1:1884）↔ 遥测入库 / 事件增量同步 / WebSocket 推送；命令白名单 + 重传同 seq；审计记门户 X-Auth-User（只绑 127.0.0.1，main.py 拒绝非回环）。 - 自动恢复：只看电流（平线 / 过低，服务器够不到手机 adb），先长按电源再断电重启，每小时限次后停手；默认逐台关闭。adb 探活接口 recovery.Prober 预留。 - 前端：总览 / 板子页（五路操作、电流电压曲线）/ 事件 / 操作记录 / 自动恢复；手机宽度可用。 - 部署 deploy/deploy.sh：服务器 Python 3.8 上先跑测试再替换；systemd phonerail-rack，每日备份。 - 门户应用 id 16、key rack（phonerail 已被需求文档站占用），授权 admin + staff。测试 26 例（本机 3.14 / 服务器 3.8 均过）。线上板 b81f3fc9c8bc 在线，WS 经门户 101 + 1 条/s 遥测。 Co-Authored-By: Claude Opus 5.5 <[email protected]>
2026-09-27 12:33:28 +08:00
@@ -0,0 +1,302 @@
+"""协议镜像 + Hub（经进程内假 broker 对接模拟板子）+ 自动恢复状态机。
+
+和固件测试同一个口径：断言打在「板子那头真的收到 / 执行了什么」上（SimBoard.received、
+它的 rails 状态），不看 Hub 自己回了什么 —— 重传换了 seq 的现象恰恰是 Hub 这边一切正常、
+手机被按了两遍。"""
+import asyncio
+import json
+
+import pytest
+
+from prrack import protocol
+from prrack.bus import MemBroker, MemBus, topic_matches
+from prrack.db import Db
+from prrack.hub import CmdTimeout, Hub
+from prrack.recovery import GAVE_UP, HARD, IDLE, SOFT, Params, PhoneRecovery, RecoveryManager
+from prrack.simboard import HUNG, NORMAL, OFF, SimBoard
+
+B = "b81f3fc9c8bc"
+
+
+def run(coro):
+    loop = asyncio.new_event_loop()
+    try:
+        return loop.run_until_complete(coro)
+    finally:  # 收掉 Hub 的后台任务（flusher / maint），免得退出时刷一屏 pending 警告
+        tasks = asyncio.all_tasks(loop)
+        for t in tasks:
+            t.cancel()
+        loop.run_until_complete(asyncio.gather(*tasks, return_exceptions=True))
+        loop.close()
+
+
+async def settle():
+    for _ in range(5):
+        await asyncio.sleep(0)
+
+
+def rig(tmp_path, present=(True,) * 5, timeout=0.05):
+    broker = MemBroker()
+    db = Db(str(tmp_path / "t.db"))
+    hub = Hub(db, MemBus(broker), attempt_timeout=timeout)
+    sim = SimBoard(MemBus(broker), B, present=present)
+    return hub, sim, db
+
+
+# ---- 协议镜像 -----------------------------------------------------------------
+
+def test_build_command_whitelist_and_ranges():
+    assert protocol.build_command({"cmd": "key", "rail": 3, "keys": "vdn+pwr", "ms": 2000}) == \
+        {"cmd": "key", "rail": 3, "keys": "vdn+pwr", "ms": 2000}
+    bad = [
+        {"cmd": "wifi", "ssid": "x"},                       # 只收 USB 的命令不给界面
+        {"cmd": "events"},                                   # 服务器自己同步
+        {"cmd": "set_voltage", "rail": 1, "mv": 4011},       # SR-13 之上
+        {"cmd": "set_voltage", "rail": 1, "mv": "3900"},
+        {"cmd": "key", "rail": 1, "keys": "pwr+home", "ms": 100},
+        {"cmd": "key", "rail": 1, "keys": "pwr", "ms": 15001},
+        {"cmd": "power", "rail": 6, "on": True},
+        {"cmd": "power", "rail": True, "on": True},         # bool 不是 int
+        {"cmd": "vbus", "rail": 1, "on": 1},
+        {"cmd": "fan", "duty": 1001},
+    ]
+    for b in bad:
+        with pytest.raises(protocol.CmdError):
+            protocol.build_command(b)
+
+
+def test_seq_is_strictly_increasing_and_in_range():
+    t = [1700000000.0]
+    g = protocol.SeqGen(clock=lambda: t[0])
+    a = [g.next() for _ in range(5)]
+    assert a == sorted(set(a)) and all(1 <= x <= 0xFFFFFFFF for x in a)
+    t[0] -= 100  # 时钟往回跳也不许重复
+    assert g.next() > a[-1] or g.next() == 1
+
+
+def test_event_table_matches_firmware_numbers():
+    # firmware/main/evlog.h 冻结的数字（抽几个钉住）
+    assert protocol.EVENT_TYPES[3][0] == "OVER_VOLT"
+    assert protocol.EVENT_TYPES[12][0] == "REBOOT"
+    assert protocol.EVENT_TYPES[13][0] == "CMD_REJECT"
+    assert protocol.EVENT_TYPES[16][0] == "PD_LIMIT"
+    assert protocol.CMD_NAMES[9] == "power" and protocol.CMD_NAMES[12] == "net"
+
+
+def test_topic_matching():
+    assert topic_matches("phonerail/+/rsp", "phonerail/%s/rsp" % B)
+    assert not topic_matches("phonerail/+/rsp", "phonerail/%s/tele" % B)
+    assert topic_matches("phonerail/#", "phonerail/x/y")
+    assert protocol.parse_topic("phonerail/%s/tele" % B) == (B, "tele")
+    assert protocol.parse_topic("evse/123/state") is None
+
+
+# ---- Hub ---------------------------------------------------------------------
+
+def test_tele_is_ingested_and_board_discovered(tmp_path):
+    hub, sim, db = rig(tmp_path)
+
+    async def go():
+        await hub.start()
+        sim.start()
+        sim.tick(1000)
+        await settle()
+        hub.flush()
+
+    run(go())
+    snap = hub.snapshot(B)
+    assert snap["online"] and snap["fw"] == "0.1.0-sim"
+    assert snap["status"]["rails"][0]["en"] is True
+    assert db.q("SELECT COUNT(*) c FROM tele")[0]["c"] == 5
+    assert len(snap["phones"]) == 5 and snap["phones"][0]["auto_recover"] == 0
+
+
+def test_command_round_trip_and_audit(tmp_path):
+    hub, sim, db = rig(tmp_path)
+
+    async def go():
+        await hub.start()
+        sim.start()
+        return await hub.call(B, {"cmd": "power", "rail": 2, "on": False}, "alice")
+
+    rsp = run(go())
+    assert rsp["ok"] is True
+    assert sim.rails[1].en is False
+    row = db.q("SELECT * FROM audit")[0]
+    assert row["user"] == "alice" and row["rail"] == 2 and row["ok"] == 1
+
+
+def test_retransmit_reuses_the_same_seq_and_executes_once(tmp_path):
+    """⭐ 回应丢了一次：Hub 用**同一个 seq** 重发，板子判重，只执行一次。"""
+    hub, sim, db = rig(tmp_path)
+    sim.drop_next_replies = 1
+
+    async def go():
+        await hub.start()
+        sim.start()
+        return await hub.call(B, {"cmd": "power_cycle", "rail": 1}, "bob")
+
+    rsp = run(go())
+    assert rsp["ok"] is True and rsp.get("dup") is True
+    sent = [r for r in sim.received if r.get("cmd") == "power_cycle"]
+    assert len(sent) == 2 and sent[0]["seq"] == sent[1]["seq"]
+    assert [e[3] for e in sim.events].count(12) == 1  # 重启事件只记了一次
+
+
+def test_timeout_is_audited_and_raised(tmp_path):
+    hub, sim, db = rig(tmp_path)
+    sim.drop_next_replies = 99
+
+    async def go():
+        await hub.start()
+        sim.start()
+        with pytest.raises(CmdTimeout):
+            await hub.call(B, {"cmd": "power", "rail": 1, "on": True}, "carol")
+
+    run(go())
+    row = db.q("SELECT * FROM audit")[0]
+    assert row["ok"] == 0 and json.loads(row["response"])["err"] == "timeout"
+    assert len({r["seq"] for r in sim.received}) == 1  # 三次都是同一个 seq
+
+
+def test_events_are_synced_incrementally(tmp_path):
+    hub, sim, db = rig(tmp_path)
+    for i in range(19):
+        sim._ev(1, 8, i)  # 加上 BOOT 共 20 条：要分 3 页
+
+    async def go():
+        await hub.start()
+        sim.start()
+        sim.tick(1000)
+        for _ in range(20):
+            await settle()
+        sim._ev(2, 3, 0, 4200, 1)  # 又来一条过压
+        sim.tick(2000)
+        for _ in range(20):
+            await settle()
+
+    run(go())
+    seqs = [r["seq"] for r in db.q("SELECT seq FROM events ORDER BY seq")]
+    assert seqs == list(range(21))
+    assert db.q("SELECT ev_cursor FROM boards")[0]["ev_cursor"] == 21
+
+
+def test_empty_slot_reply_passes_through(tmp_path):
+    hub, sim, db = rig(tmp_path, present=(True, True, True, False, True))
+
+    async def go():
+        await hub.start()
+        sim.start()
+        return await hub.call(B, {"cmd": "power", "rail": 4, "on": True}, "dave")
+
+    assert run(go()) == {"seq": sim.received[-1]["seq"], "ok": False, "err": "empty"}
+
+
+# ---- 自动恢复：状态机 ------------------------------------------------------------
+
+P = Params(window_s=60, flat_ua=5000, low_ua=30000, settle_s=30, max_per_hour=3)
+
+
+def feed(ph, t0, secs, ua_fn):
+    out = []
+    for s in range(secs):
+        st, act = ph.feed(t0 + s * 1000, ua_fn(s), 3)
+        if st or act:
+            out.append((t0 + s * 1000, st, act))
+    return out
+
+
+def test_flat_current_escalates_soft_then_hard_then_gives_up():
+    ph = PhoneRecovery(P)
+    r = feed(ph, 0, 61, lambda s: 412000)
+    assert r[-1][1] == SOFT and r[-1][2][1] == {"cmd": "key", "rail": 3, "keys": "pwr", "ms": 10000}
+    t = r[-1][0]
+    # settle 期间的样本一概不看（哪怕是平线）
+    assert feed(ph, t + 1000, 29, lambda s: 412000) == []
+    r = feed(ph, t + 30000, 62, lambda s: 412000)
+    assert r[-1][1] == HARD and r[-1][2][1] == {"cmd": "power_cycle", "rail": 3}
+    t = r[-1][0]
+    r = feed(ph, t + 30000, 62, lambda s: 412000)
+    assert r[-1][1] == GAVE_UP and r[-1][2] is None
+    assert feed(ph, t + 200000, 100, lambda s: 412000) == []  # 放弃之后再也不动
+
+
+def test_recovers_after_soft_when_current_moves_again():
+    ph = PhoneRecovery(P)
+    t = feed(ph, 0, 61, lambda s: 412000)[-1][0]
+    r = feed(ph, t + 30000, 70, lambda s: 300000 + (s % 7) * 40000)
+    assert r[-1][1] == IDLE and ph.reason == "recovered"
+
+
+def test_low_current_is_detected_and_normal_is_not():
+    ph = PhoneRecovery(P)
+    assert feed(ph, 0, 200, lambda s: 300000 + (s % 5) * 50000) == []
+    ph2 = PhoneRecovery(P)
+    assert feed(ph2, 0, 61, lambda s: 800 + s % 3)[-1][2][0] == "low"
+
+
+def test_rate_limit_gives_up():
+    ph = PhoneRecovery(Params(window_s=60, settle_s=30, max_per_hour=2))
+    t = 0
+    for _ in range(2):  # 两次软重启都「恢复」了，但每次很快又卡
+        assert feed(ph, t, 65, lambda s: 412000)[-1][1] == SOFT
+        t += 65000  # 下一段接着上一段喂完的时刻往后走（时间不倒退）
+        # 前 30 s 还在 settle 里不算数，之后要攒满一个 60 s 窗口才判「恢复」
+        assert feed(ph, t, 100, lambda s: 300000 + (s % 7) * 40000)[-1][1] == IDLE
+        t += 100000
+    r = feed(ph, t, 65, lambda s: 412000)
+    assert r[-1][1] == GAVE_UP and ph.reason == "rate_limit"
+
+
+# ---- 自动恢复：接到 Hub + 模拟板上 -------------------------------------------------
+
+def test_auto_recovery_end_to_end_with_sim(tmp_path):
+    """⭐ 模拟板上第 3 台卡死：开了自动恢复 → 5 分钟后下长按电源 → 板子执行 → 恢复。
+    没开自动恢复的第 2 台（也卡死）一个命令都不收到。"""
+    hub, sim, db = rig(tmp_path)
+    clock = [1_000_000]
+    rec = RecoveryManager(hub, lambda: clock[0])
+    sim.set_phone(3, HUNG)
+    sim.set_phone(2, HUNG)
+
+    async def go():
+        await hub.start()
+        sim.start()
+        await settle()  # 等 Hub 处理完上线消息、建好这块板的 phones 行
+        db.x("UPDATE phones SET auto_recover=1 WHERE rail=3")
+        rec.reload()
+        for s in range(300 + 130 + 320):
+            clock[0] += 1000
+            sim.tick(clock[0])
+            await settle()
+
+    run(go())
+    keys = [r for r in sim.received if r.get("cmd") == "key"]
+    assert len(keys) == 1 and keys[0]["rail"] == 3 and keys[0]["ms"] == 10000
+    assert not [r for r in sim.received if r.get("rail") == 2]
+    assert sim.rails[2].phone == NORMAL
+    states = [r["state"] for r in db.q("SELECT state FROM recovery_log ORDER BY id")]
+    assert states == [SOFT, IDLE]
+    assert db.q("SELECT user FROM audit")[0]["user"] == "auto-recovery"
+
+
+def test_auto_recovery_ignores_rails_that_are_off(tmp_path):
+    hub, sim, db = rig(tmp_path)
+    clock = [0]
+    rec = RecoveryManager(hub, lambda: clock[0])
+    sim.set_phone(1, OFF)
+    sim.rails[0].en = False  # 这一路本来就关着：电流 0 不是故障
+
+    async def go():
+        await hub.start()
+        sim.start()
+        await settle()
+        db.x("UPDATE phones SET auto_recover=1")
+        rec.reload()
+        for s in range(400):
+            clock[0] += 1000
+            sim.tick(clock[0])
+            await settle()
+
+    run(go())
+    assert not [r for r in sim.received if r.get("cmd") in ("key", "power_cycle")]