Re-use work which was done for composable upgrades and adapt same procedure for updates in order to isolate errors and minimize amount of duplication in procedures.
*** Bug 1391705 has been marked as a duplicate of this bug. ***
Note I am already working on patches to enable this for the docker based architecture - it's probably possible to adjust the eventual (ansible based) model there to also handle updates on the host, but I've not looked at that yet. It's not really as simple as reusing tasks from composable upgrades though, as all those really do is stop services, and packaging already restarts them for us in most cases. IMHO having a parallel effort to rework the current yum_update.sh model into ansible probably isn't worthwhile so we'll need to work together, as I think we'll need upgrades integrated with the container deployment (just updating things on the host is no longer enough, we have to restart containers with the services in, vs have packaging handle restarts as is the case today). WIP patches here: https://review.openstack.org/#/c/454816/ https://review.openstack.org/#/c/456254/ https://review.openstack.org/#/c/456307/ https://review.openstack.org/#/c/456308 https://review.openstack.org/#/c/456309/ https://review.openstack.org/#/c/456310/ https://review.openstack.org/#/c/456311 https://review.openstack.org/#/c/459390/ The end goal is to have a playbook we can run end-to-end (all steps) on each node in a rolling fashion, and depending on how we run it it may be possible to enable operators to run it one step at a time.
https://review.openstack.org/#/c/462211 https://review.openstack.org/#/c/467581 and keeping yum_update.sh for non-containerized.
*** Bug 1419748 has been marked as a duplicate of this bug. ***
100% test execution, no blockers or serious issues with this feature.