How hard is it to run a 'sudo yum update' ? What about running 'sudo yum update' on your cloud? This is a talk about the work we recently did on updates in TripleO. We start with an existing deployed OpenStack cloud, with 3 controller nodes in HA configuration and running tenant VMs. Each of the nodes that make up the cloud deployment need to be updated, without disruption to the tenants. Ultimately, the goal is that the cloud administrator, from a 'Director' node can do something like:
This will update each node in sequence, bringing services down, updating packages and then bringing services back up again. For controller nodes in particular this is more complicated because we need to deal with the Pacemaker cluster, ensuring the wanted changes of the cluster configuration are in place for it to behave as expected during the update and also ensuring that the cloud services remain available.
This talk will at a high level introduce some of the tooling we rely on to achieve this goal and also highlight some of the issues we faced. No live demo but will likely include a time-lapse video showing the update process.