OTA from the Dashboard
Assign an approved edge release or staged rollout and confirm it landed, without touching the Pi.
This is the everyday path from the Update Model page: assign or trigger an approved release from the Cloud Dashboard, then watch the device confirm it on its own. Use staged rollout controls where available; single-device legacy updates are still supported during the git-v1 transition.
Open the device
In the Cloud Dashboard, go to the Locations detail page for this store and open the Maintenance tab.
Start the update
Under Edge Software Update, select or trigger the approved release or staged rollout that applies to this device. On legacy single-device screens, click Update Edge Software.
Confirm
Confirm the release or rollout target shown in the dialog. For production, that target should be a release record, release tag, or staged rollout, not the tip of main. Click Confirm Update. The command is queued immediately — you'll see "Update command queued — device will apply it on the next heartbeat cycle."
Watch it land
The device picks up the command on its next heartbeat, applies the assigned release, and restarts. Give it one heartbeat cycle, then check the Edge Software version shown at the top of the store page. For a legacy single-device update, use Command History to see the actual Git ref or result the device applied.
Field tip
The device re-execs itself after updating, so a brief offline blip right after the update is normal — that's not a failure. The dashboard shows the new version within one heartbeat cycle; if it's still on the old version after a couple of cycles, check the Command History on the Maintenance tab for an ERROR status before assuming something's wrong.
There's no manual Rollback button in the dashboard yet for every single-device legacy flow. If an update misbehaves on one device, use the Manual SSH fallback as break-glass and roll it back to an explicit known-good release. Staged rollouts to a canary group of stores are different: those are watched automatically, and a bad rollout can be rolled back without waiting for manual detection (see below).
Last updated on