Offline buffering — what happens when the gateway loses internet
If your gateway loses its internet connection, it keeps reading your sensors and stores the readings in its own memory. When the connection comes back, it uploads the backlog and those readings appear in your history at the time they were actually taken — not at the time they were uploaded.
You don't need to configure anything. A gateway running firmware that supports buffering does this automatically.
What you'll see
While the connection is down:
- The gateway shows as offline, and so do its devices.
- Charts and gauges stop updating — there is no live data to show.
- Nothing is lost in the background; readings are piling up on the gateway.
When the connection returns:
- The gateway comes back online and live data resumes immediately.
- The backlog uploads in the background over the next few minutes.
- Reload a dashboard and the gap fills in.
The Offline Buffer panel on the gateway's page shows how many readings are currently held and when the gateway last reported in.
Two things behave differently for recovered readings
These are deliberate, and they're the important part of this page.
Alerts and automations do not fire for recovered readings. If a value crossed one of your thresholds during the outage, you will see it in the chart afterwards, but you will not get an alert email, and no rule, macro, or device command runs for it. This is on purpose: replaying hours-old readings through your automations could switch real equipment based on a reading that is no longer true, and it would deliver a burst of alerts for problems that have already passed. Check your charts after a long outage rather than relying on alerts to have caught things.
Recovered readings never appear as the "current" value. A gauge or single-value widget keeps showing the last genuinely live reading, so an old value can never be mistaken for the present state of your equipment.
How much can it hold
Roughly 6 hours for a typical setup — five devices reporting every 30 seconds. Faster reporting intervals or more tags fill it sooner; slower intervals stretch it further.
When the buffer is full, the oldest readings are dropped to make room for new ones. So after a very long outage you keep the most recent hours and lose the beginning. The Discarded count on the gateway page tells you if this has happened.
Limits worth knowing
A power cut during an outage loses part of the buffer. Readings held in the gateway's memory are lost if it restarts; readings already written to its internal storage survive. If the gateway restarts because power was interrupted and the internet was down, expect a partial recovery rather than a complete one.
If your local Wi-Fi itself is down when the gateway starts up, nothing is recorded. The gateway needs to join your network before it can read sensors at all. Buffering covers the case where your network is fine but the internet, or our service, is unreachable.
Editing a device during an outage discards that device's backlog. If you add, remove, or rename a device's tags while its gateway is offline, the readings that gateway had buffered for that device are discarded when it reconnects. Other devices on the same gateway are unaffected. We do this rather than guess: after a tag change, a stored reading can no longer be matched to the right tag with certainty, and silently filing values under the wrong tag name would be worse than a gap. If you can, wait until the gateway is back online before editing its devices.
Readings older than your plan's data retention are discarded rather than stored and immediately expired.
Data forwarding is partial. If you forward telemetry to InfluxDB, MariaDB, or Kafka, the recovered readings are sent there too. Webhook destinations are skipped, because a webhook is a notification channel and a multi-hour backlog would flood it.
Partially-read devices
If a device answered for some of its tags but not others during the outage, only the tags that actually responded are stored, and the reading is marked uncertain rather than good. Nothing is invented to fill the gaps.
"Storage failing" on the gateway page
The gateway checks its own internal storage every time it writes to it. If those checks start failing, the gateway falls back to memory-only buffering — still useful for short outages, but lost on restart — and the gateway page shows a Storage failing warning.
This means the flash memory in that unit is wearing out or faulty. Normal use should not cause this for many years, so treat it as a hardware fault and plan to replace the gateway. Everything else keeps working normally in the meantime.