With the recent release of the HPE CSI Driver for Kubernetes, weโre introducing new CPU architectures and running virtual machine workloads in containers for stronger isolation and application compatibility. Explore the new ecosystem expansions and get ready to deploy the new CSI driver with a long list of improvements.
With this new release the supported ecosystems have been expanded, introducing new compute platforms, several improvements to the widely popular NFS Server Provisioner feature, and a glimpse into the next release.
New paradigms in container runtimes
With the introduction of the HPE ProLiant RL300 Gen11 leveraging the Ampere Altra processors based on the ARM architecture, customers can deploy infrastructure at a never seen before CPU core density and power efficiency for private cloud. Running Kubernetes on ARM has gained popularity the last few years as itโs been readily available in managed Kubernetes service on the public cloud but also available for devices running Kubernetes at the edge. The imperative for the HPE CSI Driver team has been to deliver means to run and support the CSI driver on the ARM architecture.
With 2.4.0, introductory support includes the CSI controller and CSI node driver. As the Container Storage Providers (CSP) are separate codebases, only the HPE Alletra Storage 9000 CSP (which also includes HPE 3PAR and HPE Primera) support ARM. HPE Alletra 5000/6000 and Nimble Storage will be a separate effort released later. However, the CSI driver can be deployed on heterogeneous Kubernetes clusters with both x86 and ARM by using the advanced scheduling features of the Helm chart.
Introducing KubeVirt support
The next advancement in container runtimes is introducing official support for KubeVirt. KubeVirt is a CNCF incubating project that allows deployment of KVM-based virtual machines on Kubernetes. The project recently hit their 1.0.0 milestone. The challenge to provide persistent storage for virtual machines (VMs) โ in comparison for a container โ is that VMs works best when deployed directly on native block devices. To allow seamless migration of VMs, those block devices need to be shared among the two nodes that are part of the migration. In the case of a container, the container is simply and quickly restarted when moved. Thatโs not how VM users expect a transition to occur.
In the 2.4.0 release, the Alletra 6000 CSP (which also support HPE Nimble Storage and Alletra 5000), seamless deployment on multi-writer block devices and migration of VMs is supported by KubeVirt and other downstream packaging of KubeVirt, such as Red Hat OpenShift Virtualization, are fully supported. HPE Alletra 9000 CSP support will be announced with a later release of the CSI driver.
Improvements to the NFS Server Provisioner
A popular feature of the HPE CSI Driver for Kubernetes is the NFS Server Provisioner. It allows you to deploy multi-writer Persistent Volume Claims (PVC) for your workloads that require simultaneous access to a filesystem from containers running on multiple nodes in the cluster. New capabilities available to you with 2.4.0 include:
- Anti-affinity rules are added to the NFS Pods to ensure thereโs a more even spread across the cluster to improve performance.
- CPU and memory limits are now enforced on the NFS Pods to yield more predictable performance.
- Itโs now possible to deploy the NFS resources in the same Namespace as the requesting PVC. This allows more seamless integration with data protection software that uses CSI VolumeSnapshots to perform backups.
- Kubernetes administrators can now taint and label nodes in the cluster that are designated to run only NFS Pods. This allows better performance isolation but also allow bare-metal nodes with FC connectivity to serve workloads running on nodes without FC HBAs or VM nodes with nothing but cluster connectivity.
- Two additional labels have been added to the NFS Pods, โprovisioned-byโ (PVC name) and โprovisioned-fromโ (PVC Namespace). These labels can be used to predictably create affinity rules for workloads that need proximity to a specific NFS Pod.
Note that the new features are only available to NFS Pods deployed with 2.4.0. The NFS server itself has also got an internal software revision and it is recommended to upgrade the NFS servers individually. Please see the HPE Storage Container Orchestration Documentation (SCOD) portal for more details.
Ecosystem and solutions update
HPE supports a wide variety of ecosystem partners and applications. As many releases before 2.4.0 there a few incremental updates along with a few new platforms and partnerships.
Kubernetes
First off, Kubernetes 1.27 and 1.28 is now officially supported. While there were a couple of Special Interest Group (SIG) Storage updates in those release, no new CSI specifications have been implemented by the HPE CSI Driver.
Amazon
When evaluating running managed Kubernetes in the public cloud itโs nearly impossible not to come across Amazon Elastic Kubernetes Service (EKS). EKS is built on top of an open source Kubernetes distribution called EKS-D (distribution). EKS-D is what also is being used for private cloud EKS Anywhere. This is where the HPE CSI Driver comes in. From a developer perspective itโs very practical and convenient to use the public cloud for test and development but not all workloads are designated to run there. EKS Anywhere and EKS-D bridges the gap of uncertainty for compatibility between private and public cloud and allows a more confident lifecycle of the application. Learn more about Amazon EKS Anywhere with the HPE CSI Driver on SCOD.
Canonical
Canonical delivers two different packages of Kubernetes, MicroK8s and Charmed. MicroK8s is meant for edge use cases and developer centric workflows, such as integration with CI/CD systems. Charmed Kubernetes is a more flexible solution that integrates with infrastructure providers, private- hybrid and public cloud, that give users immense ease of management for the infrastructure and application lifecycle. Both distributions have now been tested and are formally support by HPE. Explore Charmed Kubernetes and MicroK8s on SCOD.
Red Hat
HPE CSI Operator for Kubernetes 2.4.0 will continue to support Red Hat OpenShift 4.12 with the addition of 4.13. A change in policy, as HPE now offers OpenShift on HPE GreenLake, all General Availability (GA) releases of Red Hat OpenShift will be supported by the HPE CSI Driver (instead of only Extended Update Support (EUS) releases). This release also includes introductory support for OpenShift Virtualization (see above) for HPE Alletra 5000/6000 (and HPE Nimble Storage). As always, the certification process for Red Hat OpenShift is underway. Watch the Red Hat partner page on SCOD for updates.
SUSE
SUSE recently released SLES 15 SP5 and is now a supported worker node OS.
And more
Efforts are always ongoing to support our data protection and application partners with the HPE CSI Driver for Kubernetes. The HPE Storage Solutions team recently published assets for our partners Cohesity and Splunk.
- Adapt to growing data needs using containerized Splunk with HPE Alletra Storage
- Protect your containerized applications with HPE and Cohesity
Stay tuned to Around the Storage Block as more solutions in the container space develop!
Next steps
HPE CSI Driver for Kubernetes 2.4.0 is available immediately. Ensure youโre up to date with the latest enhancements and features by updating your clusters today.
- Learn about the Compatibility and Support of 2.4.0 on SCOD
- Install the latest HPE CSI Driver for Kubernetes Helm chart from ArtifactHub
- Connect with the team on the HPE Developer Community Slack. Sign up and log in.
Stay informed about future CSI driver developments by checking back to Around the Storage Block regularly. Subscribe to the Helm chart on ArtifactHub to receive notifications when the chart is updated. New platform support and new features are already in the works for the next release.
Storage Experts
Hewlett Packard Enterprise
twitter.com/HPE_Storage
linkedin.com/showcase/hpestorage/
hpe.com/storage
MichaelMattsson
Data & Storage Nerd, Containers, DevOps, IT Automation