OpenStack Hibiscus is here: What’s new in the latest release

OpenStack has released its latest version: OpenStack 2026.2, codenamed Hibiscus.

The release includes updates across several OpenStack projects, including Nova, Neutron, Glance, Manila, Ironic, and Cyborg. The main areas of focus are support for modern hardware, security, storage, and operating OpenStack at scale.

Here are some of the changes worth knowing about.

GPUs have become an increasingly important part of infrastructure, particularly for AI and other compute intensive workloads.

In Hibiscus, Nova and Cyborg have improved integration for mediated devices, including virtual GPUs. This makes it easier to allocate and manage accelerator resources within an OpenStack environment.

OpenStack is also improving its support for environments that combine virtual machines, GPUs, and physical servers. This is useful for organizations that need different types of compute resources within the same cloud environment.

Security is another area that has received attention in Hibiscus.

Nova now has additional support for confidential computing technologies including AMD SEV SNP and Intel TDX. These technologies use hardware based memory encryption to help protect workloads while they are running.

Hibiscus also adds support for attestation, allowing systems to verify that a workload is running in a trusted environment.

This can be useful for organizations running sensitive workloads where workload isolation and protection of data in use are important.

AI and other data intensive workloads also put additional demands on storage.

OpenStack Manila, the shared file system service, adds support for Weka and Lustre in Hibiscus.

Both technologies are used in high performance and data intensive environments. Supporting them gives OpenStack users more options when building infrastructure for workloads that require high throughput storage.

OpenStack is not limited to virtual machines.

With Ironic, organizations can provision and manage physical servers through OpenStack. Hibiscus includes further improvements to how bare metal resources are tracked and managed in combination with Nova.

This is useful in environments where some workloads require dedicated physical hardware while others can run on virtual machines.

A single OpenStack environment can therefore include virtual machines, bare metal servers, and accelerated hardware.

There are also several changes in Hibiscus that are mainly relevant to OpenStack operators.

One example is Neutron. The memory consumption of HA router monitoring has been reduced significantly, with OpenStack reporting a reduction of more than 15 times.

For a small OpenStack deployment, this may not make much difference. At larger scale, however, reducing the resources required by individual services can have a noticeable impact.

Hibiscus also brings changes to Glance, OpenStack’s image service.

Image imports can now run in parallel, while improvements have also been made to image caching and managing images across different stores.

For larger environments, where many images are regularly created, copied, and deployed, these changes can make image management easier and reduce the time required to deploy workloads.

That depends on your environment.

Hibiscus is not a SLURP release. The previous release, 2026.1 Gazpacho, was a SLURP release. Organizations running Gazpacho therefore have the option to skip Hibiscus and move to the next SLURP release, currently planned as OpenStack 2027.1 Indri.

For organizations that need the new confidential computing capabilities, GPU and accelerator improvements, bare metal improvements, or additional storage options, Hibiscus may be worth evaluating.

For stable environments that do not require these changes, there may be less reason to upgrade immediately.

As with any OpenStack upgrade, compatibility with the hardware, drivers, networking, storage, and other components in the environment should be checked before making the move.

Hibiscus shows how OpenStack continues to develop alongside the hardware and workloads it supports.

The platform is being used for more than traditional virtual machine based private clouds. GPUs, bare metal, high performance storage, and confidential computing are all becoming increasingly relevant to OpenStack deployments.

For organizations running or considering an OpenStack based private cloud, Hibiscus brings several practical changes worth evaluating.