Mender

NeuralPlex uses Mender to perform robust, A/B partitioned OTA updates of the rootfs. The mender-update client is included in the NeuralPlex image and is used to install, verify, and roll back updates from the command line.

Tip

NeuralPlex ships with two rootfs partitions (A/B). Updates are always written to the inactive partition, leaving the currently running partition untouched until the new update has been committed.

Checking the current state

To view the currently active artifact and its status:

Show Artifact Info

root@neuralplex:~# mender-update show-artifact
neuralplex-image-<machine>-20260615120000

To check that Mender is running and see its overall status:

Check Mender Status

root@neuralplex:~# systemctl status mender-updated

Installing an update

An update can be installed from a local .mender artifact file or from a URL served over HTTP(S).

Install from a local file

root@neuralplex:~# mender-update install /path/to/neuralplex-image-<machine>-neuralplex.mender

Install from a URL

root@neuralplex:~# mender-update install https://example.com/neuralplex-image-<machine>-neuralplex.mender

Installing an update writes the new artifact to the inactive rootfs partition. The running partition is not modified, so if the install fails or is interrupted, the device will simply continue booting the current, known-good rootfs.

Rebooting into the new partition

Once an artifact has been installed, reboot the device to switch the bootloader over to the newly written partition:

Reboot into new update

root@neuralplex:~# reboot

Caution

After rebooting into the new partition, Mender starts a rollback timer. If the update is not committed before the timer expires, the device will automatically roll back to the previous partition on the next boot.

Committing an update

After the device boots successfully into the new partition and your application has verified the update is working correctly, commit the update to make it permanent:

Commit the update

root@neuralplex:~# mender-update commit

Committing tells Mender the update was successful, cancels the automatic rollback timer, and marks the new partition as the primary boot target going forward.

Rolling back an update

If an update is found to be bad before it has been committed, you can manually trigger a rollback rather than waiting for the timer to expire:

Roll back the update

root@neuralplex:~# mender-update rollback
root@neuralplex:~# reboot

This restores the bootloader environment to point at the previous, known-good partition.

Note

Once an update has been committed with mender-update commit, it can no longer be rolled back using this command. A rollback at that point requires installing a new artifact for the previous version.

Checking for updates

To manually poll the configured Mender server for a new deployment (rather than waiting for the daemon’s poll interval):

Check for update

root@neuralplex:~# mender-update check-update

Viewing update logs

To view the log output from the most recent deployment, including any failure reason:

Show deployment logs

root@neuralplex:~# mender-update show-provides
root@neuralplex:~# journalctl -u mender-updated -f

Common workflow

1. Build or download a .mender artifact for the desired neuralplex-image.
2. Copy the artifact to the device, or host it somewhere reachable over HTTP(S).
3. Install the artifact: mender-update install <artifact>
4. Reboot the device: reboot
5. Verify the device and application are behaving as expected on the new partition.
6. Commit the update: mender-update commit

If step 5 fails, run mender-update rollback and reboot instead of committing, or simply do nothing and let the automatic rollback timer revert the device on its own.