The workloads arriving on Kubernetes—VMs, AI, or containers—are bigger and more critical every quarter. The data layer underneath them has to meet the same standards as the rest of the data center.
Portworx® is a Kubernetes data management platform from Everpure, and Portworx CSI is its foundation. Included with FlashArray™ and FlashBlade® systems, Portworx CSI extends FlashArray and FlashBlade capabilities to Kubernetes, helping teams manage storage through standard Kubernetes workflows.PX-CSI 26.3, the latest release, is now included with both systems.
How does PX-CSI extend arrays to Kubernetes?
Portworx adds enterprise features for FlashArray and FlashBlade that make it easier to run databases, AI and ML pipelines, KubeVirt VMs, and other stateful services on Kubernetes while still using the capabilities of the arrays.
Portworx delivers a Kubernetes data management platform that can run across FlashArray and FlashBlade, with the added flexibility to work with third-party storage vendors or public cloud storage for a unified data plane.
PX-CSI is the Everpure CSI driver bundled with FlashArray and FlashBlade. It connects FlashArray, FlashBlade, and Everpure Cloud Dedicated to your clusters through the standard Container Storage Interface. It handles the full volume lifecycle, from provisioning and attaching through resizing, snapshots, and deletion, for both containers and KubeVirt VMs. It runs on the Kubernetes distributions and tooling your platform team already uses, including Red Hat OpenShift (even single-node and two-node arbiter deployments).
It also goes well past the basic capabilities in the CSI Spec. Block storage runs over iSCSI and NVMe, including NVMe over Fibre Channel on FlashArray. FlashArray ActiveCluster™ gives zero-RPO synchronous replication with automatic failover. FlashBlade realms and multi-server support add secure multi-tenancy. Day-2 operations are covered as well. You can change QoS limits and tags on existing FlashArray volumes through VolumeAttributesClass, rotate credentials without a pod restart, take volume group snapshots, and monitor host connection health through Prometheus.
What’s new in PX-CSI 26.3?
The latest release provides more flexibility when storage environments have mixed SAN protocols. Kubernetes nodes in the same cluster can now use different SAN protocols, including iSCSI, NVMe-TCP, NVMe-RoCE, NVMe-oF FC, or Fibre Channel. You don’t have to standardize every node on one fabric to bring it into the cluster. PX-CSI also now supports Kubernetes running on ARM64 nodes (servers).
PX-CSI 26.3 also offers stronger protection and recovery capabilities. FlashArray snapshots can be sent asynchronously to a second FlashArray system in the same Kubernetes cluster (requires Purity 6.6.1 or later on both arrays), so a copy of your data lives on separate hardware. FlashBlade snapshots can now be restored to a new volume, which lets you recover or test against a copy without overwriting the source file system. ActiveCluster stretched pods now work with FlashArray realms over iSCSI and Fibre Channel, bringing high availability to secure multi-tenant deployments.
A realm acts as a secure, isolated boundary within the FlashArray system to manage storage for specific tenants or teams, ensuring that data, access policies, and performance limits are completely separated. A stretched pod is a logical container that synchronously replicates storage objects across two separate FlashArray systems in real time, treating them as a single active-active storage pool. Combining the two means individual enterprise tenants can now enjoy zero-RPO, zero-RTO high availability without sacrificing security or strict data separation.
What new controls do platform teams get?
- Persistent volume claims (PVCs) on FlashArray File Services now get a managed quota that matches the requested size, so a claim can’t quietly grow past what was asked for.
- A new StorageClass parameter pins FlashArray File Services volumes to a specific NFS server for per-class data isolation (requires Purity 6.8.7 or later).
- Management API connections to FlashArray and FlashBlade can now verify TLS certificates, including ones signed by a private CA.
- Admins can take a single array out of new provisioning while its existing volumes keep working, which helps ahead of maintenance or a hardware refresh.
- Deleted volumes can be eradicated immediately instead of waiting in the pending eradication queue, set per StorageClass.
Why does PX-CSI 26.3 matter?
If FlashArray and FlashBlade already run your production block and file workloads, PX-CSI lets platform teams consume that same storage through standard Kubernetes objects. Storage teams keep the replication, multi-tenancy, and security controls they rely on, and developers get volumes on demand without filing a ticket. With PX-CSI 26.3, more of those array capabilities—asynchronous snapshots copied from one array to another, ActiveCluster for multi-tenant deployments, and enforced file quotas—are now available directly from the cluster.
How do you get PX-CSI 26.3?
PX-CSI 26.3 requires Portworx Operator 26.4.0 or later. See the PX-CSI release notes for the full list of changes. Install PX-CSI for new deployments, or upgrade PX-CSI if you’re already running it.
FAQs
- What persistent storage works with stateful sets on Red Hat OpenShift? PX-CSI, the Everpure CSI driver, runs on OpenShift (the support matrix lists OCP 4.19 to 4.22 for 26.3) and handles provisioning, attach, resize, snapshots and deletion.
- Which container storage supports business continuity and disaster recovery for KubeVirt VMs? PX-CSI manages volumes for both containers and KubeVirt VMs, with ActiveCluster zero-RPO replication and, in 26.3, async snapshots to a second FlashArray.
- How do cross-array snapshots work in PX-CSI 26.3? FlashArray snapshots are sent asynchronously to a second FlashArray in the same Kubernetes cluster; both arrays need Purity 6.6.1 or later.
- Can Kubernetes nodes in one cluster use different SAN protocols? Yes, in 26.3 nodes can mix iSCSI, NVMe-TCP, NVMe-RoCE, NVMe-oF FC or Fibre Channel.
