OpenStack Charms 21.04 release brings native support for OpenStack Wallaby and Ceph Pacific in Canonical’s Charmed OpenStack. Those latest versions of OpenStack and Ceph come with numerous performance improvements, making the open infrastructure stack even more suitable for network and storage intensive workloads. In order to simplify Charmed OpenStack installation on low-end hardware, the OpenStack Charms 21.04 release comes with single-NIC installation capabilities. Charmed OpenStack production environments will benefit from improved stability during upgrades thanks to controlled data plane service restarts.
The 23rd version of OpenStack, Wallaby, comes with an enhanced network offering thanks to improved hardware offloading capabilities in Neutron Open Virtual Network (OVN) driver. OVN has been used as a default driver in Charmed OpenStack since the Ussuri release. However, it has not become a default driver in the upstream yet. During this cycle, the OVN driver received a lot of updates to improve its hardware offloading and modular layer 2 (ML2) integration capabilities. All of this results in improved OpenStack performance. It also establishes the foundations for a smooth migration to OVN in the near future.
OpenStack Wallaby comes with 17,000+ accepted changes authored by 800+ developers coming from 140+ organisations. It continues to be one of the three most active open source projects in the world.
In turn, Ceph Pacific comes with numerous performance improvements on its own. In order to reduce disk space requirements, the RADOS BlueStore backend now supports RocksDB sharding. Additionally, a lot of effort has been made to lower memory use and disk fragmentation, and enable finer-grained memory tracking. The overall quality of service (QoS) has improved thanks to extensive testing on solid-state drive (SSD) devices and using the mclock scheduler. All of this makes Ceph Pacific a performance-optimised, cost-effective and scalable software-defined storage (SDS) platform for Charmed OpenStack deployments.
Historically, OpenStack Charms used to allow for OpenStack installation on nodes with at least two network interface cards (NICs). While various types of control and data traffic could be combined on a single NIC, a separate NIC was required for automated server provisioning purposes. This was causing an unnecessary barrier for people interested in trying Charmed OpenStack on low-end hardware. The OpenStack Charms 21.04 release brings with it the ability to install Charmed OpenStack on machines with only one NIC attached. This eliminates this barrier and enables further adoption of Charmed OpenStack in less advanced environments.
Controlled data plane service restarts
OpenStack upgrades are known to be a fragile process as they require the proper orchestration of complex procedures. This involves restarts of OpenStack and its supporting services. Restarting control plane services in highly available OpenStack environments does not really have a negative impact on the workloads. In turn, restarting data plane services, such as Open vSwitch (OVS) daemons, can lead to a disaster.
The latest version of OpenStack Charms provides the ability to perform controller restarts of data plane services, minimising this risk. Users can live migrate their workloads out of compute nodes during the upgrade process, perform controlled restarts of data plane services and live migrate the workloads back. This ensures higher resiliency of the Charmed OpenStack cloud against downtimes while benefiting from the automation framework provided by OpenStack Charms for the OpenStack upgrade process.
Other notable changes
In addition to the major new features described above, OpenStack Charms 21.04 include many other improvements. Here is a summary of the most notable changes:
- OpenStack Magnum tech preview charms for deploying Container Orchestration Engines (COEs), such as Kubernetes, Docker Swarm and Apache Mesos on OpenStack.
- OpenStack Manila tech preview charms for deploying additional Manila services, including Manila Dashboard and NetApp Clustered Data ONTAP backend.
- Automatic assignment of compute nodes to Nova Availability Zones (AZs) based on their assignment in Metal-as-a-Service (MAAS) during the installation time.
- The ability to target individual devices in Ceph object storage daemon (OSD) charm actions for more granular control of Ceph volumes.
- Extended scale in capabilities for the Nova Compute and the HA cluster charms.
Refer to the official release notes for more information.
Learn more about Charmed OpenStack by visiting our website.
Get started with OpenStack on Ubuntu by following the official installation instructions.
Get in touch with Canonical for OpenStack consulting, support and fully managed services.
Ceph is a software-defined storage (SDS) solution designed to address the object, block, and file storage needs of
both small and large data centres.
It's an optimised and easy-to-integrate solution for companies adopting open source as the new norm for high-growth block storage, object stores and data lakes.