Now that 2.5 is released I’d like to call out some of the new features for
those who haven’t been following the development this cycle.
In this post I’ll be covering machine series upgrades.
Rework of machine series upgrades
With v.2.5 a new command makes its appearance: upgrade-series. By design,
the bulk of the underlying operating system is upgraded manually by the user by
way of standard tooling (e.g. do-release-upgrade).
The usefulness is that workload machines can now have their series updated
natively. In previous versions the recommended approach was to add a new unit
and remove the old one.
Here, we’ll be upgrading a machine (with an ID of ‘1’) to Ubuntu 18.04 LTS
(Bionic):
juju upgrade-series 1 prepare bionic
This causes Juju to perform a few internal changes but for the most part the
machine remains the same. We now manually upgrade the machine via SSH:
juju ssh 1
$ do-release-upgrade
<any other custom care the machine may need>
Ctrl-d
When done we inform Juju:
juju upgrade-series 1 complete
Note that the upgrade of machines hosting controllers is not supported and the
documented method of creating a new controller and migrating models is still
the recommended procedure.
The charms.reactive framework also has support for allowing the charm to manage the application during the series upgrade process, which is documented in OS Series Upgrades. In summary, the behavior is as follows:
When the prepare step is invoked, the framework will set the upgrade.series.in-progress flag, which the charm will have one (and only one) chance to react to, by at least stopping the application service for the duration of the upgrade. After this, no hooks or reactive handlers will be run until after the upgrade is complete. Once the admin has performed the series upgrade and invoked the complete step, the Python virtual environment used by the charm will be automatically rebuilt and the wheelhouse reinstalled, and then the upgrade.series.in-progress flag will be removed, indicating that the charm should perform any necessary migrations or changes and then bring the application back up.
