Reliable Openstack deployment should be as close to 100% availability as possible and always up to date with the latest release. To achieve this, during an upgrade services should be stopped and replaced with new one with no API downtime. In projects like keystone, glance or neutron this means keeping compatibility of DB schema between releases. The complexity of maintaining this compatibility, while keeping performance and security on a high level, is a hot topic in the community. Current approach in projects like nova or cinder requires keeping the data compatibility for 2 or 3 releases. In this presentation we want to show solutions, how to reduce the data migration time period to one release cycle. This would reduce the code complexity and performance overhead in database layer.
Attendees should learn about existing solutions for online schema upgrades in Openstack projects like keystone, glance and neutron. Also, listeners will get to know the proposed improvements in upgrade process.