fix(tuya): retry once on a stale persistent socket before reporting offline

installer/CHANGELOG.md

@@ -16,6 +16,7 @@
 ### Fixes
 
 - **installer:** restart service units on update and record the installed manifest (#43) (0f2aec2)
+- **installer:** make redeploy, repair and repo-built bundles install cleanly (#46) (bf1fe10)
 - **installer:** expand ~ in the repo package path (873bdda)
 - **installer:** replace remote files the SSH user cannot open for writing (eca0198)
 - **installer:** give step 20 its MQTT arguments from the node on redeploy and repair (bb4cd40)

installer/webui/CHANGELOG.md

@@ -10,6 +10,7 @@
 
 ### Fixes
 
+- **installer:** make redeploy, repair and repo-built bundles install cleanly (#46) (bf1fe10)
 - **webui:** show the error detail of a failed run (5017432)
 - **installer:** give step 20 its MQTT arguments from the node on redeploy and repair (bb4cd40)
 - **installer:** add texts for every fault code the bootstrap steps emit (23a64c3)

scripts/bootstrap/CHANGELOG.md

@@ -15,6 +15,7 @@
 
 - **bootstrap:** make steps 20, 40, 65, 70 and the diagnosis work on a real node (#37) (11eef9d)
 - **installer:** restart service units on update and record the installed manifest (#43) (0f2aec2)
+- **installer:** make redeploy, repair and repo-built bundles install cleanly (#46) (bf1fe10)
 - **bootstrap:** ignore a commented-out userspace-networking flag in step 40 (a1aeee1)
 - **bootstrap:** let step 70 keep an installed Caddy when the bundle has no Caddy pack (f8fbc8d)
 

services/tuya_mqtt/CHANGELOG.md

@@ -1,11 +1,15 @@
 # Changelog
 
-## v0.4.0 (2026-09-21)
+## v0.4.1 (2026-09-22)
 
 ### Features
 
 - **services:** give every service its own version and changelog (#45) (5bc91b8)
 
+### Fixes
+
+- **tuya:** retry once on a stale persistent socket before reporting offline (41e9f55)
+
 ## v0.3.2 (2026-09-15)
 
 ### Features

services/tuya_mqtt/manifest.json

@@ -1,6 +1,6 @@
 {
   "service_id": "tuya",
-  "version": "0.4.0",
+  "version": "0.4.1",
   "unit": "tuya.service",
   "bootstrap_step": "85",
   "schema": "config.schema.json",

services/tuya_mqtt/tests/test_tuya_poll.py

@@ -0,0 +1,73 @@
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+import pytest
+
+import tuya_mqtt
+
+
+class FakeMqttClient:
+    def __init__(self):
+        self.published = []
+
+    def publish(self, topic, payload=None, qos=0, retain=True):
+        self.published.append((topic, payload))
+
+
+def make_device(**overrides):
+    cfg = tuya_mqtt.TuyaDeviceConfig(
+        id="heizungs_ventil",
+        name="Heizungsventil",
+        device_id="dev123",
+        local_key="key123",
+        ip="192.0.2.50",
+        **overrides,
+    )
+    return tuya_mqtt.TuyaDevice(cfg)
+
+
+class FakeTuyaDevice:
+    """Zweiter status()-Aufruf auf demselben Objekt schlägt fehl, wie es das
+    TinyTuya-'Unexpected Payload'-Muster auf dem persistenten Socket zeigt."""
+
+    def __init__(self, fail_status=False):
+        self.fail_status = fail_status
+        self.status_calls = 0
+
+    def set_version(self, version):
+        pass
+
+    def set_socketPersistent(self, persistent):
+        pass
+
+    def status(self):
+        self.status_calls += 1
+        if self.fail_status:
+            return {"Error": "Unexpected Payload from Device", "Err": "904", "Payload": None}
+        return {"dps": {"1": True}}
+
+
+def test_poll_one_retries_once_after_stale_socket_error(monkeypatch):
+    device = make_device()
+    connections = [FakeTuyaDevice(fail_status=True), FakeTuyaDevice(fail_status=False)]
+    monkeypatch.setattr(tuya_mqtt, "connect_device", lambda cfg: connections.pop(0))
+    device.device = connections.pop(0)
+    client = FakeMqttClient()
+
+    tuya_mqtt.poll_one(device, client, simulation_active=False)
+
+    topics = dict(client.published)
+    assert topics["outstation/heizungs_ventil/switch"] == "ON"
+    assert topics["outstation/heizungs_ventil/status/online"] == "1"
+
+
+def test_poll_one_raises_when_retry_also_fails(monkeypatch):
+    device = make_device()
+    device.device = FakeTuyaDevice(fail_status=True)
+    monkeypatch.setattr(tuya_mqtt, "connect_device", lambda cfg: FakeTuyaDevice(fail_status=True))
+    client = FakeMqttClient()
+
+    with pytest.raises(RuntimeError):
+        tuya_mqtt.poll_one(device, client, simulation_active=False)

services/tuya_mqtt/tuya_mqtt.py

@@ -143,23 +143,37 @@ def publish_device_discovery(client: mqtt.Client, device: TuyaDevice, node_devic
     )
 
 
+def connect_device(cfg: TuyaDeviceConfig) -> Any:
+    device = tinytuya.Device(cfg.device_id, cfg.ip, cfg.local_key)
+    device.set_version(cfg.version)
+    device.set_socketPersistent(True)
+    log.info("[%s] Tuya-Gerät initialisiert: %s", cfg.id, cfg.device_id)
+    return device
+
+
 def poll_one(device: TuyaDevice, client: mqtt.Client, simulation_active: bool) -> None:
     if simulation_active:
         publish(client, device, "switch", "ON" if device.simulated_switch_state else "OFF")
         publish_online_status(client, device, True, "Simulation aktiv")
         return
 
     if device.device is None:
-        device.device = tinytuya.Device(
-            device.cfg.device_id, device.cfg.ip, device.cfg.local_key
-        )
-        device.device.set_version(device.cfg.version)
-        device.device.set_socketPersistent(True)
-        log.info("[%s] Tuya-Gerät initialisiert: %s", device.cfg.id, device.cfg.device_id)
+        device.device = connect_device(device.cfg)
 
-    status = device.device.status()
-    if "Error" in status:
-        raise RuntimeError(f"Statusfehler: {status}")
+    try:
+        status = device.device.status()
+        if "Error" in status:
+            raise RuntimeError(f"Statusfehler: {status}")
+    except Exception as exc:
+        # Der persistente Socket verwirft nach einer erfolgreichen Abfrage
+        # gelegentlich unaufgefordert nachgeschobene Payloads (TinyTuya
+        # "Unexpected Payload"/Err 904); ein frischer Verbindungsaufbau
+        # behebt das zuverlässig, siehe journalctl-Muster auf dem Node.
+        log.info("[%s] Abfrage fehlgeschlagen (%s), verbinde neu und versuche erneut", device.cfg.id, exc)
+        device.device = connect_device(device.cfg)
+        status = device.device.status()
+        if "Error" in status:
+            raise RuntimeError(f"Statusfehler: {status}")
 
     dps = status.get("dps", {})
     log.info("[%s] Roher Gerätestatus (dps): %s", device.cfg.id, dps)
@@ -182,11 +196,7 @@ def set_switch(device: TuyaDevice, client: mqtt.Client, active: bool, simulation
         device.simulated_switch_state = active
     else:
         if device.device is None:
-            device.device = tinytuya.Device(
-                device.cfg.device_id, device.cfg.ip, device.cfg.local_key
-            )
-            device.device.set_version(device.cfg.version)
-            device.device.set_socketPersistent(True)
+            device.device = connect_device(device.cfg)
         result = device.device.set_value(device.cfg.switch_dp, active)
         log.info("[%s] Tuya-Antwort: %s", device.cfg.id, result)
     publish(client, device, "switch", "ON" if active else "OFF")