Unlocking New Potential with Mirantis OpenStack for Kubernetes 24.2

We’re excited to announce the release of Mirantis OpenStack for Kubernetes (MOSK) 24.2, which delivers significant improvements to optimize performance and simplify management for cloud operators, while also improving flexibility and control for developers. Highlights of the release include the new Dynamic Resource Balancer service,  a system for Day 2 bare metal configuration, and updates to major components, including Open Virtual Network, OpenStack Caracal, Ubuntu 22.04, and Ceph Reef.

Major Updates and Enhancements

The latest Mirantis OpenStack for Kubernetes (MOSK) 24.2 release brings several significant updates to major components, setting the stage for advanced functionalities and enhanced performance. Here’s a closer look at what’s new:

The MOSK 24.2 release introduces Open Virtual Network (OVN) as a next-generation software-defined networking (SDN) solution, available as a tech preview for new deployments. OVN supersedes Open vSwitch as the backend for OpenStack, marking a pivotal shift in the community's development focus. The transition to OVN is expected to drive future production deployments, offering improved network management and scalability. An upgrade path from Open vSwitch to OVN will become available later this year, allowing existing users to seamlessly transition to this advanced networking solution.

OpenStack Caracal (2024.1) is now available as a technical preview for greenfield deployments, providing users a glimpse into the future of OpenStack. This release includes numerous new features and improvements that enhance the overall functionality and user experience. Detailed highlights of these enhancements can be found in the MOSK release notes. The upgrade path from the current Antelope release to Caracal is expected to be available later this year, ensuring a smooth transition for existing deployments.

The gradual introduction of Ubuntu 22.04 support begins with MOSK 24.2. The management cluster will get automatically upgraded to Ubuntu 22.04 during the update to Mirantis Container Cloud 2.27 bundled with MOSK. OpenStack clusters will see comprehensive support for this host OS version later this year. This upgrade promises improved stability, security, and performance, as well as official support for the latest server hardware, aligning with the latest Ubuntu long-term support (LTS) version.

Ceph Reef is set to replace the now outdated Quincy version in the MOSK 24.2 release, bringing substantial improvements to storage cluster performance and functionality. Existing clusters will be upgraded to Ceph Reef automatically during the MOSK 24.2 update, ensuring users benefit from the latest advancements in storage technology without the need for manual intervention.

Achieving Operational Excellence for OpenStack

The MOSK 24.2 release brings a suite of enhancements aimed at elevating operational excellence, providing cloud operators with advanced tools and features to optimize performance, streamline management, and ensure robust monitoring. 

Similar to VMware’s Dynamic Resource Scheduler, the new Mirantis Dynamic Resource Balancer (DRB) service enables cloud operators to maintain optimal workload placement in their OpenStack cloud continuously and in a completely automated manner. This feature helps mitigate the “noisy neighbor” problem and prevents compute nodes in heavily used clouds from becoming “hot spots.” By dynamically moving virtual machines around, DRB ensures balanced performance across the cloud environment. Out of the box, it implements generic logic for making decisions about optimal resource allocation, however the service’s pluggable architecture allows operators to build a solution that truly works the best possible way in their specific environment.

Learn more in our deep dive blog about the Dynamic Resource Balancer.

The new Day 2 bare metal configuration system is now ready for wide adoption. This feature allows cloud operators to configure host OS settings on the cluster servers, without having to reprovision them. Built on a concept of reusable parameterized modules, the new system completely eliminates the need for shadow configuration management (e.g., Salt, Puppet, or Ansible), providing cloud operators with centralized control over all cluster settings, top-to-bottom, via a single API. Out of the box, MOSK 24.2 includes standard modules to manage third-party packages and repositories, the irqbalance daemon, sysctl settings, with more to come later. However, cloud operators can always implement their own custom modules specific to their needs. 

The new network port availability measurement tool (Portprober) continuously probes all virtual network ports in an OpenStack cloud, reporting response statistics to our logging, monitoring, and alerting stack, Mirantis StackLight. This helps cloud operators assess how specific events and actions affect the network connectivity of the cloud workloads, enabling better planning of maintenance operations based on the specifics of the environment and workload. By providing real-time visibility into network port availability, Portprober enhances overall cloud reliability and operational planning. Portprober will be available as a technical preview in MOSK 24.2.

Improved Update Experience

We’ve significantly improved the experience of updating MOSK clusters by introducing the following new features:

• ClusterUpdatePlan API: This API serves as a single point of entry for planning and executing MOSK cluster updates. Cloud operators can oversee the update status and progress, and pause between steps if necessary, ensuring a controlled and smooth update process.

• Node Update Groups API: This feature allows operators to organize compute and storage nodes into groups based on failure domains within their cluster. The system will use this information to apply host OS changes to cluster machines in a strict order, preventing uncontrolled impacts on virtual machines and data. 

Enhanced Monitoring and Alerts

Operational excellence is further supported by new monitoring and alerting features:

• “I/O Error” messages in the kernel log of the cluster servers will now trigger an alert in Mirantis StackLight. This feature helps cloud operators to detect storage device problems early, allowing for timely interventions and preventing potential outages.

• Collection of common SMART metrics is now also a part of Mirantis StackLight. Cloud operators can use the SMART (Self-Monitoring, Analysis, and Reporting Technology) data collected from the cluster’s storage devices to define alerts indicating hardware health problems. This proactive monitoring enables operators to address issues before they escalate, ensuring sustained performance and reliability.

Tenant-Facing Features 

The MOSK 24.2 release brings a range of new features designed to enhance performance, flexibility, and control for cloud users. Here’s a detailed look at these new capabilities:

RAW local images for root and ephemeral storage provide an effective compromise between the simplicity of QCoW2 and the high performance of LVM backends for virtual storage devices used by the guests to host the operating system. This feature allows workloads to achieve a significant boost in IO operations. However, it requires full (uncompressed) disk images to be stored on compute nodes. This trade-off offers improved performance for data-intensive applications, making it a valuable option for users seeking enhanced storage efficiency.

The ability to disable spoof checks on SR-IOV interfaces managed by Tungsten Fabric is now available to cloud users, simplifying the deployment of advanced Virtualized Network Functions (VNFs) in MOSK clouds designed for telecommunications. For virtual router appliances, disabling spoof checks is essential to enable certain routing functions. In Deep Packet Inspection (DPI) solutions, spoof checks need to be disabled to ensure proper network traffic flow. This feature provides the necessary flexibility for deploying complex network functions without compromising performance or functionality.

Cloud operators can now specify target nodes for hosting Tungsten Fabric’s SNAT and load balancer namespaces. This capability allows for better resource management by preserving compute node resources where highly sensitive workloads are running. By restricting these network functions to specific nodes, cloud operators can ensure optimal performance and resource allocation for their most critical applications.

Learn more

For a full list of updates in MOSK 24.2, see the release notes.

To request a private demo with one of our cloud architects, please contact us.