Aqara Hub Keeps Going Offline at Night: Why and How to Stop It

Aqara hubs that drop offline overnight follow a predictable pattern caused by cloud reconnect logic. Here is what is happening and how to make it stop.

Aqara Hub Keeps Going Offline at Night: Why and How to Stop It
Difficulty Beginner
Estimated time 15 minutes
Last tested May 2026
Verified on HomeKit, Alexa, Google Home +2

You set up your Aqara hub, paired your sensors, built your automations, and it ran perfectly for a few days. Then you started noticing that around three or four in the morning, your phone shows the hub as offline. By the time you wake up, it is back online and pretending nothing happened. Your motion sensor automations missed their morning triggers, and you can never quite tell if the hub is going to be online when you actually need it. This pattern is so common with Aqara hubs that it has a recognizable cause.

What is happening between 3am and 5am

Aqara’s cloud infrastructure performs maintenance during the early morning hours in their primary timezone, which is China Standard Time. During those windows, the cloud rotates the long-lived authentication token your hub uses to stay connected. The hub is supposed to negotiate a new token transparently. In firmware versions before approximately 4.1.0, the token rotation handler had a race condition that left the hub in a state where its cached token was rejected by the cloud and its new token request was queued but not actually sent. The hub would sit offline until something triggered a reconnect: a manual restart, a network blip, or the hub’s own internal heartbeat after a long delay.

This explains the symptom precisely. The hub goes offline around the same time every night. It comes back on its own after several hours. There is no visible reason. Your local Wi-Fi never lost connectivity.

How to confirm this is your issue

Look at the Aqara app’s hub log. Go to the hub, tap the gear icon, scroll to “Logs” or “Diagnostic Info” depending on app version. You are looking for entries that look like:

  • Cloud connection lost: token expired
  • Reconnect attempt failed: auth error
  • Cloud connection restored several hours later

If you see this pattern, the cause is confirmed and the fix is simple. If you see different errors (DNS failures, network unreachable, MQTT errors), the cause is different and you should treat this as a network problem instead.

The actual fix

Update the hub firmware. The fix shipped with version 4.1.0 and has been refined in every release since. To force a firmware check, open the Aqara app, go to the hub, tap the gear, scroll to “Firmware Upgrade”. If you have automatic updates enabled but the hub is on an older version, the auto-update may be stuck behind another pending update. Manual triggering will push it through.

The update takes around six to eight minutes. The hub will be offline during the update. Do not unplug it. After the update completes, the hub will reboot and reconnect to the cloud automatically. Check the log again the next morning to confirm the issue is gone.

If the update does not fix it

A small number of users continue to see the issue even on current firmware. In those cases, the root cause is usually that the hub’s MAC address is on Aqara’s cloud blacklist due to too many failed authentication attempts during the buggy firmware period. The fix is to remove the hub from the Aqara app entirely and re-add it, which resets the cloud-side authentication record.

Before doing this, export your device list and automation list. The Aqara app does not have a clean export, so screenshot the device list and write down each automation. When you re-add the hub, you will need to re-pair each Zigbee device, but the pairing process is fast (typically thirty seconds per device once the hub is in pairing mode).

Why your HomeKit accessories also disappear

If you use the Aqara hub as a HomeKit bridge, its HomeKit accessory advertisement stops the moment its cloud connection drops, because the Aqara firmware ties HAP availability to cloud presence. This is an Aqara-specific quirk that no other major HomeKit bridge does. The result: when the Aqara hub goes offline at 3am, all of its accessories also disappear from HomeKit until the cloud reconnects. This means HomeKit automations that rely on Aqara sensors will silently skip their morning triggers, even though HomeKit itself is perfectly healthy.

This is one of the strongest arguments for keeping safety-critical sensors on a HomeKit-native bridge that does not need a cloud connection to expose accessories locally.

What to do tonight if you cannot wait for the update

If you need a working hub before bed and you cannot update the firmware right now, set a smart plug on a schedule to power-cycle the hub at 5am. The hub will boot fresh, get a clean token, and stay online through the morning. This is a workaround, not a fix, but it gets you through the night without missed automations.

Preventing this on future hubs

If you are buying a new Aqara hub, look for the M3 or later. The earlier M1S and M2 hubs are reliable now but had longer histories of cloud reconnect issues. The newer hubs also support Matter, which lets you bypass the Aqara cloud entirely for HomeKit or Google Home integration by using Matter bridging instead of the legacy cloud bridge.

The pattern to log if support gets involved

If you escalate to Aqara support, they will ask for the exact times of disconnect and reconnect events. The app log has these but they are easy to miss. Take a screenshot of the log section showing at least three disconnect cycles before you write the ticket. Note the firmware version visible at the top of the hub settings page in the same screenshot.

If the disconnects involve HomeKit accessories specifically going offline at the same moment, that is the Aqara-specific behavior described earlier and is a separate bug from the cloud reconnect issue. Our HomePod Thread border router guide is the right next read if the affected accessories are Thread-based.

The exposure to Matter as the longer-term answer

Aqara has been adding Matter bridging to newer hubs. If your hub supports it and you can expose Zigbee devices to your main ecosystem via Matter, the nightly cloud-reconnect issue is bypassed entirely because Matter does not depend on the Aqara cloud. The migration is invasive enough that it is only worth it on hubs where the offline issue is chronic. The wider pattern of cross-ecosystem device exposure is in our routine truncation guide.

Related reading: display time-recovery guide.

Works with
HomeKitAlexaGoogle HomeMatterZigbee

Frequently asked questions

Why does my Wi-Fi not show any outage at 3am?

Your Wi-Fi is fine. The disconnect is at the application layer between the hub and the Aqara cloud, not at the network layer. Your router will not log it because no network event occurred.

If I disable cloud sync, will my hub stay online locally?

Yes for local automations, but no for HomeKit and other ecosystem integrations on this hub, because Aqara's firmware ties HAP availability to cloud presence. Newer Matter-capable Aqara hubs decouple this.

How often should the firmware actually update?

Aqara typically pushes a firmware update every two to three months. If you have not seen an update in six months and you are on a version older than 4.1.0, the auto-update is stuck and you should trigger it manually.

My hub came back online after I rebooted my router. Does that mean my router was the problem?

Not necessarily. Any network event (DHCP renewal, router reboot, even a brief packet loss) can trigger the hub to abandon its stuck reconnect state and retry. Your router was not the underlying cause, but it was the thing that broke the hub out of the bad state.

Should I downgrade if a new firmware is worse?

Aqara hubs do not allow downgrade. If a new firmware introduces a regression, you have to wait for the next fix. This is one reason to wait one to two weeks after a release before manually triggering it on hubs that are working fine.

Why does the issue not affect my SmartThings setup?

SmartThings uses a different cloud architecture with shorter-lived tokens that are designed to fail closed (the hub keeps trying) rather than fail open (the hub gives up). Different design tradeoffs lead to different failure modes.