Cinder (OpenStack Block Storage project) can be found in >80% of OpenStack deployments. Configuring it to be robust, resilient and fast is important. However considering its high customizability, such as choice of deployment architecture, volume backend or backup backend, an always on Cinder is not a simple goal to achieve.
In this talk we will explain how Cinder can be deployed in ways guaranteeing increased reliability and performance, as close as possible to "always on". We will walk you through latest innovations that are letting you keep your control plane and data plane up during upgrades, increased load or when experiencing breakdowns. These features include:
- cinder-volume service in A/A mode
- rolling upgrades support
- volume replication
- volume migration
We will also share insight on the improvements in these matters that are planned for the future cycles.
Attendees should walk out from the session aware of presence of various Cinder features that increase its control and data plane availability and with knowledge how to use them correctly for different tasks.