Allow ZHA startup to fail instead of raising ConfigEntryNotReady (#77417)

* Retry startup within ZHA instead of raising `ConfigEntryNotReady`

* Add unit tests

* Disable pylint warning for intentional broad except

homeassistant/components/zha/core/const.py

@@ -408,3 +408,7 @@ class Strobe(t.enum8):
 
     No_Strobe = 0x00
     Strobe = 0x01
+
+
+STARTUP_FAILURE_DELAY_S = 3
+STARTUP_RETRIES = 3

homeassistant/components/zha/core/gateway.py

@@ -13,7 +13,6 @@ import time
 import traceback
 from typing import TYPE_CHECKING, Any, NamedTuple, Union
 
-from serial import SerialException
 from zigpy.application import ControllerApplication
 from zigpy.config import CONF_DEVICE
 import zigpy.device
@@ -25,7 +24,6 @@ from homeassistant import __path__ as HOMEASSISTANT_PATH
 from homeassistant.components.system_log import LogEntry, _figure_out_source
 from homeassistant.config_entries import ConfigEntry
 from homeassistant.core import HomeAssistant, callback
-from homeassistant.exceptions import ConfigEntryNotReady
 from homeassistant.helpers import device_registry as dr, entity_registry as er
 from homeassistant.helpers.dispatcher import async_dispatcher_send
 from homeassistant.helpers.entity import DeviceInfo
@@ -62,6 +60,8 @@ from .const import (
     SIGNAL_ADD_ENTITIES,
     SIGNAL_GROUP_MEMBERSHIP_CHANGE,
     SIGNAL_REMOVE,
+    STARTUP_FAILURE_DELAY_S,
+    STARTUP_RETRIES,
     UNKNOWN_MANUFACTURER,
     UNKNOWN_MODEL,
     ZHA_GW_MSG,
@@ -166,17 +166,27 @@ class ZHAGateway:
         app_config[CONF_DEVICE] = self.config_entry.data[CONF_DEVICE]
 
         app_config = app_controller_cls.SCHEMA(app_config)
-        try:
+
-            self.application_controller = await app_controller_cls.new(
+        for attempt in range(STARTUP_RETRIES):
-                app_config, auto_form=True, start_radio=True
+            try:
-            )
+                self.application_controller = await app_controller_cls.new(
-        except (asyncio.TimeoutError, SerialException, OSError) as exception:
+                    app_config, auto_form=True, start_radio=True
-            _LOGGER.error(
+                )
-                "Couldn't start %s coordinator",
+            except Exception as exc:  # pylint: disable=broad-except
-                self.radio_description,
+                _LOGGER.warning(
-                exc_info=exception,
+                    "Couldn't start %s coordinator (attempt %s of %s)",
-            )
+                    self.radio_description,
-            raise ConfigEntryNotReady from exception
+                    attempt + 1,
+                    STARTUP_RETRIES,
+                    exc_info=exc,
+                )
+
+                if attempt == STARTUP_RETRIES - 1:
+                    raise exc
+
+                await asyncio.sleep(STARTUP_FAILURE_DELAY_S)
+            else:
+                break
 
         self.application_controller.add_listener(self)
         self.application_controller.groups.add_listener(self)

tests/components/zha/test_gateway.py

@@ -1,6 +1,6 @@
 """Test ZHA Gateway."""
 import asyncio
-from unittest.mock import patch
+from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
 import zigpy.profiles.zha as zha
@@ -211,3 +211,51 @@ async def test_gateway_create_group_with_id(hass, device_light_1, coordinator):
     assert len(zha_group.members) == 1
     assert zha_group.members[0].device is device_light_1
     assert zha_group.group_id == 0x1234
+
+
+@patch(
+    "homeassistant.components.zha.core.gateway.ZHAGateway.async_load_devices",
+    MagicMock(),
+)
+@patch(
+    "homeassistant.components.zha.core.gateway.ZHAGateway.async_load_groups",
+    MagicMock(),
+)
+@patch("homeassistant.components.zha.core.gateway.STARTUP_FAILURE_DELAY_S", 0.01)
+@pytest.mark.parametrize(
+    "startup",
+    [
+        [asyncio.TimeoutError(), FileNotFoundError(), MagicMock()],
+        [asyncio.TimeoutError(), MagicMock()],
+        [MagicMock()],
+    ],
+)
+async def test_gateway_initialize_success(startup, hass, device_light_1, coordinator):
+    """Test ZHA initializing the gateway successfully."""
+    zha_gateway = get_zha_gateway(hass)
+    assert zha_gateway is not None
+
+    zha_gateway.shutdown = AsyncMock()
+
+    with patch(
+        "bellows.zigbee.application.ControllerApplication.new", side_effect=startup
+    ) as mock_new:
+        await zha_gateway.async_initialize()
+
+    assert mock_new.call_count == len(startup)
+
+
+@patch("homeassistant.components.zha.core.gateway.STARTUP_FAILURE_DELAY_S", 0.01)
+async def test_gateway_initialize_failure(hass, device_light_1, coordinator):
+    """Test ZHA failing to initialize the gateway."""
+    zha_gateway = get_zha_gateway(hass)
+    assert zha_gateway is not None
+
+    with patch(
+        "bellows.zigbee.application.ControllerApplication.new",
+        side_effect=[asyncio.TimeoutError(), FileNotFoundError(), RuntimeError()],
+    ) as mock_new:
+        with pytest.raises(RuntimeError):
+            await zha_gateway.async_initialize()
+
+    assert mock_new.call_count == 3