Changing Tires on a Moving Car: Our Journey to Zero-Downtime Deployments
Applications built over the years carry historical design assumptions, such as that a few hours of downtime for maintenance upgrades every six months is acceptable. Today, embracing continuous delivery practices means more frequent releases, which means more downtime. This is the problem Pierre Vincent faced and successfully overcame, going from monthly deployments with a couple of hours of downtime to zero-downtime deployments on demand. Pierre will show how, by mapping out a deployment process, it becomes possible to progressively reduce its impact on users. He will also give practical advice on how to avoid downtime, such as online database migrations or progressive application rollouts. Explore the operational improvements required to have consistent, repeatable, and observable deployments so that you can have the confidence to run them with live traffic. Zero-downtime deployments don’t mean that everything stays up, or that everything is immediately running the latest version; they simply mean users don’t notice a thing while all this is happening.