Percona Live Online

Learning about Vitess ("A database clustering system for horizontal scaling of MySQL" - https://vitess.io) is straightforward enough, as is running the Get Started demo on your computer. But once you want to start scaling out a sandbox cluster, or want to run realistic benchmarks against your schema design (both of which are hard to do on a personal computer), setting up a full cluster in a pinch seems daunting... or is it?

I'm here to show you, with a live demo/tutorial, that deploying and evaluating a Vitess sandbox cluster, into a public cloud environment, can be done super easily. In fact, my aim is to bootstrap a fully functioning cluster within 10 minutes of starting the demo.

With the remaining demo time, I will demonstrate other Vitess operations, such as:
* Scaling up and down the cluster
* Increasing and decreasing the number of shards without losing data
* Configure zonal SSDs for MySQL
* Backup and restore (so you can shut down the cluster to save money or discard an experiment, then bring it back up again with the original data)
* Deploy the experimental Vitess orchestrator component
* Planned and unplanned failovers
* Automatic rolling upgrades, and controlled rolling upgrades
* Metrics, dashboards

Even with the best possible documentation (and the Vitess documentation is quite good!), getting a fully working cluster, experimenting with it, and getting everything configured the way you want can involve a bunch of trial and error. I hope that my demo can help you bypass some of the more boring trial-and-error, and get running more quickly with your Vitess evaluation.

For this demo, I will be using the excellent open-source Vitess-operator for Kubernetes, provided by PlanetScale. Even if you aren't considering deploying Vitess on Kubernetes in production, I still highly recommend it for sandbox use. Deploying an arbitrary number of components is super trivial with the operator, and everything auto-wires automatically. No need to delay your evaluation by needing to manually bootstrap a cluster one node at a time, or write your own deployment tools.