OpenStack Nova CellsV2 provides infrastructure scaling by grouping compute nodes together through database and message queue sharding. CellsV2 are required in any Nova deployment since the OpenStack Ocata release (a default Nova installation comes with a single Cell(V2) setup).
During the Stein development cycle important changes were introduced to improve availability and flexibility in a multi-cellV2 deployment.
If one or more cell databases in a deployment go down, some of the basic but necessary nova operations like listing of servers and counting quotas would not work as expected. These issues have been tackled by using the available information from the nova_api and placement databases.
Additionally, being able to resize across cells has been a long-desired feature for deployments running multiple cells.
This talk aims to present the development challenges faced while implementing these features and the options available to operators.
This session will provide details on important operational scenarios for those already running or looking to expand to multiple cells and features being added specifically for multi-cell deployments. The presenters will share real-world production use cases from CERN Private Cloud and Huawei Public Cloud.