balena-labs-projects / autohupr Goto Github PK

Some users are running processes in the devices that can't be interrupted when the HUP process is initiated from the dashboard. HUP's current implementation doesn't have into account the containers' locks, so the HUP process will just run when invoked, killing any running process.

This interruption has resulted in some users complaining that they would have to manually control HUP when they consider it should be done by balena.

If there's a lock in place, no update will be triggered. If there are no locks, the app will create a lock (to avoid a release update) and will then start the HUP script.

A case for failed hostOS updates is that devices attempt to update when they are not in a healthy state. Autohupr could check the device health by running and checking healthchecks before actually triggering the update.

This block should have an environment variable that will contain the target OS version. When the target version is greater (or different) than the running version.

The user will be able to trigger an update by modifying the variable, either in a fleet or in a device.