Add Supervisor connectivity check after DNS restart (#6005)

* Add Supervisor connectivity check after DNS restart

When the DNS plug-in got restarted, check Supervisor connectivity
in case the DNS plug-in configuration change influenced Supervisor
connectivity. This is helpful when a DHCP server gets started after
Home Assistant is up. In that case the network provided DNS server
(local DNS server) becomes available after the DNS plug-in restart.

Without this change, the Supervisor connectivity will remain false
until the a Job triggers a connectivity check, for example the
periodic update check (which causes a updater and store reload) by
Core.

* Fix pytest and add coverage for new functionality

Author: agners

supervisor/plugins/dns.py

@@ -15,7 +15,8 @@
 import jinja2
 import voluptuous as vol
 
-from ..const import ATTR_SERVERS, DNS_SUFFIX, LogLevel
+from ..bus import EventListener
+from ..const import ATTR_SERVERS, DNS_SUFFIX, BusEvent, LogLevel
 from ..coresys import CoreSys
 from ..dbus.const import MulticastProtocolEnabled
 from ..docker.const import ContainerState
@@ -82,6 +83,7 @@ def __init__(self, coresys: CoreSys):
         # Debouncing system for rapid local changes
         self._locals_changed_handle: asyncio.TimerHandle | None = None
         self._restart_after_locals_change_handle: asyncio.Task | None = None
+        self._connectivity_check_listener: EventListener | None = None
 
     @property
     def hosts(self) -> Path:
@@ -111,6 +113,15 @@ def _compute_locals(self) -> list[str]:
 
         return servers
 
+    async def _on_dns_container_running(self, event: DockerContainerStateEvent) -> None:
+        """Handle DNS container state change to running and trigger connectivity check."""
+        if event.name == self.instance.name and event.state == ContainerState.RUNNING:
+            # Wait before CoreDNS actually becomes available
+            await asyncio.sleep(5)
+
+            _LOGGER.debug("CoreDNS started, checking connectivity")
+            await self.sys_supervisor.check_connectivity()
+
     async def _restart_dns_after_locals_change(self) -> None:
         """Restart DNS after a debounced delay for local changes."""
         old_locals = self._cached_locals
@@ -236,6 +247,13 @@ async def load(self) -> None:
             _LOGGER.error("Can't read hosts.tmpl: %s", err)
 
         await self._init_hosts()
+
+        # Register Docker event listener for connectivity checks
+        if not self._connectivity_check_listener:
+            self._connectivity_check_listener = self.sys_bus.register_event(
+                BusEvent.DOCKER_CONTAINER_STATE_CHANGE, self._on_dns_container_running
+            )
+
         await super().load()
 
         # Update supervisor

supervisor/supervisor.py

@@ -46,7 +46,7 @@ def _check_connectivity_throttle_period(coresys: CoreSys, *_) -> timedelta:
     if coresys.supervisor.connectivity:
         return timedelta(minutes=10)
 
-    return timedelta(seconds=30)
+    return timedelta(seconds=5)
 
 
 class Supervisor(CoreSysAttributes):

tests/plugins/test_dns.py

@@ -406,3 +406,78 @@ async def test_stop_cancels_pending_timers_and_tasks(coresys: CoreSys):
     mock_task_handle.cancel.assert_called_once()
     assert dns_plugin._locals_changed_handle is None
     assert dns_plugin._restart_after_locals_change_handle is None
+
+
+async def test_dns_restart_triggers_connectivity_check(coresys: CoreSys):
+    """Test end-to-end that DNS container restart triggers connectivity check."""
+    dns_plugin = coresys.plugins.dns
+
+    # Load the plugin to register the event listener
+    with (
+        patch.object(type(dns_plugin.instance), "attach"),
+        patch.object(type(dns_plugin.instance), "is_running", return_value=True),
+    ):
+        await dns_plugin.load()
+
+    # Verify listener was registered (connectivity check listener should be stored)
+    assert dns_plugin._connectivity_check_listener is not None
+
+    # Create event to signal when connectivity check is called
+    connectivity_check_event = asyncio.Event()
+
+    # Mock connectivity check to set the event when called
+    async def mock_check_connectivity():
+        connectivity_check_event.set()
+
+    with (
+        patch.object(
+            coresys.supervisor,
+            "check_connectivity",
+            side_effect=mock_check_connectivity,
+        ),
+        patch("supervisor.plugins.dns.asyncio.sleep") as mock_sleep,
+    ):
+        # Fire the DNS container state change event through bus system
+        coresys.bus.fire_event(
+            BusEvent.DOCKER_CONTAINER_STATE_CHANGE,
+            DockerContainerStateEvent(
+                name="hassio_dns",
+                state=ContainerState.RUNNING,
+                id="test_id",
+                time=1234567890,
+            ),
+        )
+
+        # Wait for connectivity check to be called
+        await asyncio.wait_for(connectivity_check_event.wait(), timeout=1.0)
+
+        # Verify sleep was called with correct delay
+        mock_sleep.assert_called_once_with(5)
+
+        # Reset and test that other containers don't trigger check
+        connectivity_check_event.clear()
+        mock_sleep.reset_mock()
+
+        # Fire event for different container
+        coresys.bus.fire_event(
+            BusEvent.DOCKER_CONTAINER_STATE_CHANGE,
+            DockerContainerStateEvent(
+                name="hassio_homeassistant",
+                state=ContainerState.RUNNING,
+                id="test_id",
+                time=1234567890,
+            ),
+        )
+
+        # Wait a bit and verify connectivity check was NOT triggered
+        try:
+            await asyncio.wait_for(connectivity_check_event.wait(), timeout=0.1)
+            assert False, (
+                "Connectivity check should not have been called for other containers"
+            )
+        except TimeoutError:
+            # This is expected - connectivity check should not be called
+            pass
+
+        # Verify sleep was not called for other containers
+        mock_sleep.assert_not_called()

tests/test_supervisor.py

@@ -50,8 +50,8 @@ async def test_connectivity_check(
     [
         (None, timedelta(minutes=5), True),
         (None, timedelta(minutes=15), False),
-        (ClientError(), timedelta(seconds=20), True),
-        (ClientError(), timedelta(seconds=40), False),
+        (ClientError(), timedelta(seconds=3), True),
+        (ClientError(), timedelta(seconds=10), False),
     ],
 )
 async def test_connectivity_check_throttling(