Missing notifications/actions to update qemu instances after qemu upgrades

[Billy Olsen]

Qemu packages can be upgraded in a variety of methods which may not be initially obvious to the end user. For example, an unattended upgrade or a landscape upgrade may upgrade the qemu binaries on the system. This can cause problems for instances which are started with on old binaries and then modified later (e.g. attempting to add a cinder volume to grow local storage). The problem is that the running instance code attempts to dynamically load libraries from the same version of qemu, which has been upgraded.

The instance should either be rebooted or live-migrated to get the new versions of qemu binaries applied.

This bug is to capture the fact that the charm could do better help the end user be aware of the fact that the qemu binaries have been updated and that there is at least one instance running a previous version of qemu binaries.

Ideally, the following would happen:

1. a nagios alert to notify the end-user that instances need to be migrated or rebooted in order to pick up the new binaries.

2. an action for the charm to help operationally resolve this
  - This item may be more challenging as not all instances will be eligible for a live migration (e.g. sriov, dpdk, etc)

3. update the charm status to indicate this condition exists

At the very least, the alert and/or charm status bits should be updated so that the user has awareness and can take appropriate actions.

[Billy Olsen (billy-olsen)]

I also filed https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1847361 which may resolve some of the challenges within this if older binaries are kept around on disk. That would likely only be in newer Ubuntu releases so it would still need to be addressed for those.