Persistent connection is not persistent

[soeffi]

Home Assistant Android app version(s):
2025.9.2-minimal

Android version(s):
13

Device model(s):
Lenovo M9

Home Assistant version:
2025.10.2

Last working Home Assistant release (if known):
unknown

Description of problem, include YAML if issue is related to notifications:
Lenovo M9 tablet set up as a wall panel, persistent connection is set to "Always". After fresh start of the app, this works flawlessly, I can receive notifications (via notify.mydevice) immediately.

However, the connection is lost after some time. The persistent connection notification is still active, no error message is visible. Nevertheless, notifications to the device then start failing silently. Manually pushing a notify action then results in a "device not connected to local push notifications" error message.

This behavior unfortunately will consistently appear after betwenn 1 and 12 hours of running the companion app. Force close solves the issue temporarily.

Companion App Logs:

homeassistant_companion_log_9-15-2025_0-55-27_125548.txt
Log file was saved after loss of connection.

Screenshot or video of problem:
n/a

Additional Information
Battery optimization for the companion app is turned off

[ilar]

As far as I can tell, it's not only not persistent, it's incredibly brittle- it will never try to reconnect, for any reason.

Easy way to instantly trigger it is to just restart the homeassistant server, or in some cases, change network connection. The websocket never recovers, and you have to force stop the app to get notifications and a connection again.

[RedDirtDog]

I'm getting the same "Not connected to local push notifications" error, however I've been unable to fix it.

I've force closed the Companion App (Android), reopened and log shows no errors.
I then triggered a notification (manually from inside an automation) which immediately produced the “device not connected to local push notification” error on the HA frontend.
Checked the log again and this was the error which occurred when I triggered the notif:

10-20 08:24:45.459 29669 29719 E IntegrationRepositoryImpl: Unable to get notification rate limits 10-20 08:24:45.459 29669 29719 E IntegrationRepositoryImpl: java.net.UnknownHostException: Unable to resolve host “[mobile-apps.home-assistant.io](http://mobile-apps.home-assistant.io/)”: No address associated with hostname 10-20 08:24:45.459 29669 29719 E IntegrationRepositoryImpl: at java.net.Inet6AddressImpl.lookupHostByName(Inet6AddressImpl.java:156) 10-20 08:24:45.459 29669 29719 E IntegrationRepositoryImpl: at java.net.Inet6AddressImpl.lookupAllHostAddr(Inet6AddressImpl.java:103) 10-20 08:24:45.459 29669 29719 E IntegrationRepositoryImpl: at java.net.InetAddress.getAllByName(InetAddress.java:1152) … 10-20 08:24:45.459 29669 29719 D SettingsPresenterImpl$getNotificationRateLimits: Unable to get rate limits

I'm able to resolve the hostname on my phone, so not sure why it's getting that issue.

[Soukyuu]

I can only re-iterate what @ilar said. I just had a case where sending one notify.notify call with a clear_notification for one tag succeeded and the second one with a different tag, just a few seconds later results in "device not connected to local push notification". The second now being persistent.

The worst part about it is, of course, the notification about the persistent connection still being there. I'm using the minimal app, if that makes any difference.

I get that using Google services is convenient for both the users and devs, and I'm glad we have some kind of alternative available (at least for notifications). I wish it worked more stable though.

[thd6n75]

This affects me too. I think, but am not certain, that whatever previous version of the app I had installed prior to 2025-9.2-minimal didn't behave this way.

Android: 15
HA server: 2025.10.4
App: 2025-9.2-minimal installed from F-Droid

Similar behavior as already described. App is set to maintain persistent connection always.

In an attempt to gather more info, I enabled the notification for persistent connection. I setup a task in Tasker to detect when the notification disappears. When that happens, it force kills the app using root, and then launches HA again. Baked into this task is the assumption that the moment the persistent connection notification disappears is the moment that notifications break. I'm unsure if that's correct or not.

I happened to be using my phone when this task fired, so I opened the app and dumped the logs as quickly as I could. However, it looks like the log entries during the failure mode got lost when the app was killed. I guess that's not surprising. Anyways, see attached.

Tasker also has the ability to listen for specific logcat activity (whereby, presumably, I could dump logs), but I don't know what to look for. I'll see if I can dump logs when the persistent notification disappears.

homeassistant_companion_log_9-30-2025_16-49-16.txt

[samuli-as]

I think the persistent connections broke in 2025.8.7. For me the trouble started then, in addition to the issue that the app would not load properly. I was using the full version from Play Store until and up to 2025.8.7 but switched to minimal for 2025.9.2 because the updated version was not available from Play Store. I rely on the persistent connection anyway so losing FCM notifications did not matter to me.

Like mentioned before, network changes from wifi to cellular or vice versa are an easy if not trivial way to trigger this for me, though the issue does not happen every time for whatever reason. I have been using a test automation that sends a notification, and at that point an error like this appears in the log:

10-17 10:03:24.939 9368 9447 E WebSocketCoreImpl: java.net.SocketException: Software caused connection abort

The behaviour is not exactly the same every time. Sometimes the next notification comes through, but often it does not no matter how many times I spam the test notification. It's worth noting that the app can even be in the foreground while I change the network, try to trigger the notification and then the error occurs. The battery optimization is disabled for the app regardless.

I have a Pixel 6 on GrapheneOS, Android 16.

homeassistant_companion_log_9-17-2025_10-39-13.txt

[thd6n75]

at 3:47 (in the middle of the night, phone unused for hours), my persistent connection notification disappeared and my Tasker task executed and dumped several days of logcat. It's 6MB of logs. Attached is the output of $ cat logcat.txt | grep -i homeassistant , and then I deleted everything before about 15 minutes prior to the event.

logcat-grepped-homeassistant.txt

Line 22 (ConnectivityService: requestNetwork for) thru line 29 (WM-WorkerWrapper: Worker result SUCCESS for Work) seem to be regular occurrences every 15 minutes.

This stuck out to me. It happened right around the time of the event:

10-31 03:46:28.137  1774  2218 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 03:46:28.137  1774  2218 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 03:47:45.302  1774  1807 E ActivityManager: FGS (dataSync) timed out: ServiceRecord{a50db36 u0 io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemForegroundService c:io.homeassistant.companion.android.minimal}
10-31 03:47:45.326  1774  1794 D ActivityManager: Stop FGS timeout: ServiceRecord{a50db36 u0 io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemForegroundService c:io.homeassistant.companion.android.minimal}
10-31 03:47:45.339  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 03:47:45.339  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy

For privacy reasons I don't want to attach the whole logcat.txt here, but I can provide it directly to a developer.

[thd6n75]

Here's a lot more context. This is probably more useful than the above.

logcat-trimmed.txt

The last line here looks like it might be the notification disappearing:

10-31 03:46:11.137  5515 29542 D SensorWorkerBase$doWork: Updating all Sensors in foreground.
10-31 03:46:11.139  5515 19008 D ServerConnectionInfo: Using external URL
10-31 03:46:11.141  5515  5515 D WM-SystemFgDispatcher: Notifying with (id:42, workSpecId: 72d3370f-de55-46a7-b0e9-ca5b6bf8067a, notificationType :1)
10-31 03:46:11.327  5515 30910 D TrafficStats: tagSocket(82) with statsTag=0xffffffff, statsUid=-1
10-31 03:46:11.420  5515 19194 D SensorReceiverBase: Nothing to update for server 2 (Home)
10-31 03:46:11.420  5515 19194 I SensorReceiverBase: Sensor updates and sync completed
10-31 03:46:11.421  2053  2423 D LocalImageResolver: Couldn't use ImageDecoder for drawable, falling back to non-resized load.
10-31 03:46:11.422  5515  5537 I WM-WorkerWrapper: Worker result SUCCESS for Work [ id=72d3370f-de55-46a7-b0e9-ca5b6bf8067a, tags={ io.homeassistant.companion.android.sensors.SensorWorker } ]
10-31 03:46:11.424  5515  5515 D WM-Processor: Processor 72d3370f-de55-46a7-b0e9-ca5b6bf8067a executed; reschedule = false
10-31 03:46:11.424  5515  5515 D WM-GreedyScheduler: Stopping tracking for WorkGenerationalId(workSpecId=72d3370f-de55-46a7-b0e9-ca5b6bf8067a, generation=0)
10-31 03:46:11.424  5515  5515 D WM-SystemFgDispatcher: Removing Notification (id: 42, workSpecId: WorkGenerationalId(workSpecId=72d3370f-de55-46a7-b0e9-ca5b6bf8067a, generation=0), notificationType: 1

[thd6n75]

Here's another log from around the time of the persistent connection notification vanishing. FGS stuff again. In this, 1174 appears to be the PID of the homeassistant app.

10-31 21:42:38.490  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:42:47.708  1774  1990 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:42:57.216  1774  1990 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:43:16.102  1774  1990 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:43:19.249  1774  1990 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:43:27.920  1774  2218 I ActivityManager: Background started FGS: Allowed [callingPackage: io.homeassistant.companion.android.minimal; callingUid: 10200; uidState: FGS ; uidBFSL: [BFSL]; intent: Intent { act=ACTION_START_FOREGROUND cmp=io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemForegroundService }; code:PROC_STATE_FGS; tempAllowListReason:<,reasonCode:SYSTEM_ALLOW_LISTED,duration:9223372036854775807,callingUid:-1>; targetSdkVersion:36; callerTargetSdkVersion:36; startForegroundCount:1; bindFromPackage:null: isBindService:false]
10-31 21:43:28.151  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:43:28.151  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:43:28.152  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
10-31 21:43:28.152  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy


....
....
(several hours pass - included above because of FGS mentions)
.....
....

11-01 03:28:00.707 31459 10576 D ServerConnectionInfo: Using external URL
11-01 03:28:00.712 31459 10576 D ServerConnectionInfo: Using external URL
11-01 03:28:00.716 31459  6257 D WebSocketCoreImpl$sendMessage: Sending message 2940: {type=ping, id=2940}
11-01 03:28:00.717 31459  6257 D WebSocketCoreImpl$sendMessage: Message number 2940 sent awaiting answer from WebSocket
11-01 03:28:00.786 31459  3357 D WebSocketCoreImpl: Websocket: onMessage (text)
11-01 03:28:00.787 31459  3357 D WebSocketCoreImpl: Message id 2940 received
11-01 03:28:08.959  1175  1175 W TracingMuxer: type=1400 audit(0.0:39463): avc:  denied  { write } for  name="traced_producer" dev="tmpfs" ino=1257 scontext=u:r:gxp_logging:s0 tconte
xt=u:object_r:traced_producer_socket:s0 tclass=sock_file permissive=0
11-01 03:28:20.428  2053 13941 D PowerUI : can't show warning due to - plugged: true status unknown: false
11-01 03:28:20.431  1774  1774 I LineageHealth: Current time is 03:28:20 AM
11-01 03:28:20.432  1774  1774 I LineageHealth: Target time is 03:30:00 AM
11-01 03:28:20.432  1774  1774 I LineageHealth: Canceling charge deadline
11-01 03:28:28.157  1774  1807 E ActivityManager: FGS (dataSync) timed out: ServiceRecord{c219009 u0 io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemFo
regroundService c:io.homeassistant.companion.android.minimal}
11-01 03:28:28.167 31459 31459 I WM-SystemFgDispatcher: Foreground service timed out, FGS type: 1
11-01 03:28:28.168 31459 31459 D WM-SystemFgService: Shutting down.
11-01 03:28:28.169  1774 18878 D ActivityManager: Stop FGS timeout: ServiceRecord{c219009 u0 io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemForeground
Service c:io.homeassistant.companion.android.minimal}
11-01 03:28:28.171  1774 18878 W ForegroundServiceTypeLoggerModule: FGS stop call for: 10200 has no types!
11-01 03:28:28.180  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-01 03:28:28.180  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-01 03:28:28.180  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-01 03:28:28.181  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-01 03:28:28.182  1774  1797 D HintManagerService: Sent cleanup message for uid 10200
11-01 03:28:28.182  1774  1797 E HintManagerService: ADPF IS GETTING PAUSED? UID: 10200 TAG: 2

In the whole log, I note there's several occurrences of the following, and they do NOT coincide with the persistent connection notification vanishing:

11-01 02:28:44.980 31459 31459 D WM-SystemFgDispatcher: Removing Notification (id: 42, workSpecId: WorkGenerationalId(workSpecId=28e6b033-c629-48c0-a3fb-577dbaa0c89d, generation=0), 
notificationType: 1
11-01 02:28:44.983  1774  2568 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy

[thd6n75]

2 more, same story. Only posting 1 of the 2 since they're essentially the same.

11-02 02:22:55.759  1774  1807 E ActivityManager: FGS (dataSync) timed out: ServiceRecord{9b85d01 u0 io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemForegroundService c:io.homeassistant.companion.android.minimal}
11-02 02:22:55.781 18094 18094 I WM-SystemFgDispatcher: Foreground service timed out, FGS type: 1
11-02 02:22:55.784 18094 18094 D WM-SystemFgService: Shutting down.
11-02 02:22:55.785  1774  2201 D ActivityManager: Stop FGS timeout: ServiceRecord{9b85d01 u0io.homeassistant.companion.android.minimal/androidx.work.impl.foreground.SystemForegroundService c:io.homeassistant.companion.android.minimal}
11-02 02:22:55.787  1774  2201 W ForegroundServiceTypeLoggerModule: FGS stop call for: 10200 has no types!
11-02 02:22:55.797  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-02 02:22:55.797  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-02 02:22:55.798  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy
11-02 02:22:55.798  1774  1885 W JobScheduler.JobStatus: Exempted app io.homeassistant.companion.android.minimal considered buggy

I'll stop posting these since they don't add anything new anymore.

[RedDirtDog]

I updated HA Core and got a shock when a notif alerted on my phone...

Here's what I'm running, haven't touched anything on the phone or HA (apart from the updates) for a bit...
Core 2025.10.4
Supervisor 2025.10.0
Operating System 16.2
Frontend 20251001.4

[TimoPtr]

I don't have time to look at the issue right now in depth since it requires a lot of focus that I don't have.

My first thoughts on this issue is that the Foreground service that we use for Persistent connection is flagged as dataSync (as mentioned in #5987) and starting with 2025.8.7 we target SDK 36 (Android 16) which comes with some limitation on the foreground services (starting with Android 15). Our services is marked with android:foregroundServiceType="dataSync|remoteMessaging" which I though would be enough to not be killed by the system since remoteMessaging is exempted from this limitation (at least I thought).

To go a little further our websocket worker used for persistent connection is using remoteMessaging

        return try {
            val type = if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.UPSIDE_DOWN_CAKE) {
                ServiceInfo.FOREGROUND_SERVICE_TYPE_REMOTE_MESSAGING
            } else {
                0
            }

in

android/app/src/main/kotlin/io/homeassistant/companion/android/websocket/WebsocketManager.kt

Line 266 in aaa47c5

return try {

where our sensor worker is using dataSync

in

android/common/src/main/kotlin/io/homeassistant/companion/android/common/sensors/SensorWorkerBase.kt

Line 51 in aaa47c5

val foregroundInfo = ForegroundInfo(

We would need to dig deeper to understand how it works, it seems that every worker are killed and not only the SensorWorker.

@thd6n75 if you can do this test it would help, disable all the sensor in your app and see if the connection stay persistent. It would mean that if there are any dataSync worker in the foreground service and we reach the limit it would kill everything. In that case maybe we have a way to somehow restart the connection (but again it needs to be investigated).

On the full flavor you might not see this behavior since it's not always running only once every 15 minutes for few seconds so far from the 6h limit. Except if you use the persistent connection on.

I think the initial issue from @soeffi is not related to this limitation and is something else.

[Soukyuu]

Maybe it's also worth looking at Signal's/Molly.im's implementation of this? I don't seem to have any problems with the notifications there so far. They use websockets if FCM is unavailable. I also know that Threema Push actually uses websockets but I'm unsure they have an openly accessible implementation of it. I think not.