Charm upgrade of neutron-gateway

Hi good folks of juju and Openstack,

I have recently undergone an charm upgrade of our Openstack environment. Initially in a test/staging environment. The reason for this is that we need to upgrade the juju charms to the latest before we deploy Octavia into the cluster.

However, when undergoing the procedure in our test/staging environment i noticed some downtime (around 250ms when pinging an instance’s floating ip) when initiating the upgrade of neutron-gateway. I assume that juju initiates the charm upgrade on all units at the same time. During the upgrade it seems that some updates with apt is applied, i therefor also assume that some services are restarted.

What i am wondering is if there is a way to stage the charm-upgrade?
What i would like to accomplish is for juju to only upgrade the most sensitive charms under my command.

Best regards,
Emil

1 Like

Hi Emil,

Currently Juju doesn’t have support for a rolling upgrade of units in an application. We do have the feature on roadmap, however, in the form of “branches,” which will allow you to perform upgrades and other actions on a branch (subset) of your model.

We’ll post more info on branches when they become available.

~ PeteVG

Hey @emil-jacero, we have run into similar challenges as you when upgrading Openstack and Ceph charms.

A workaround for this is to stop all the juju unit agents for the specific application, upgrade the charm, then start the juju unit agents one by one which we just did for ceph-osd upgrade (bouncing all the OSDs of an active cluster at the same time is not fun).
Haven’t tried for neutron-gateway, but give it a go in test and let us know.