Portworx CSI Release Notes

Portworx CSI (PX-CSI) is a lightweight storage orchestration solution that uses the Kubernetes Container Storage Interface (CSI) framework to integrate the Everpure platforms.

PX-CSI integrates with Kubernetes to directly access FlashArray Direct Access volumes (FADA) and FlashBlade Direct Access volumes (FBDA).

If you're setting up PX-CSI for the first time, see Install Portworx CSI. If you're already using PX-CSI, upgrade to the latest version. For upgrade instructions, see Upgrade Portworx CSI.

For information about supported versions, release numbering, and support phases, refer to the PX-CSI Life Cycle Support Policy.

PX-CSI 26.1.2

April 15, 2026

This release also addresses security vulnerabilities.

Supported Operator version: 25.6.1 or later

Resolved Issues

Issue Number	Issue Description	Severity
PWX-52561	For FlashBlade PVCs, PX-CSI created SMB export objects that are not required for NFS. Starting with PX-CSI 26.1.2, it does not create SMB export objects for new PVCs.	Minor
PWX-53224	PX-CSI pods enter a CrashLoopBackOff state if nodes do not have the iscsiadm utilities installed, even when using only FlashArray file services or FlashBlade file systems.	Minor

PX-CSI 26.1.1

March 31, 2026

Supported Operator version: 25.6.1 or later

important

Starting with PX-CSI 26.1.1, you must install the util-linux package on all cluster nodes. For detailed package requirements by distribution, see Cluster node requirements.

What's New

• Configurable maximum volumes per node
  

  PX-CSI now allows you to configure the maximum number of volumes that can be attached to a single node by using the MAX_VOLUMES_PER_NODE environment variable in your StorageCluster specification. You can also set different limits for specific nodes by using the spec.nodes[].env section to override the value. For more information, see StorageCluster CRD reference.

  Note that if you don't specify a value, Kubernetes doesn't enforce a limit. Make sure you don't create more volumes than what PX-CSI supports. For more information, see Volume attachment limits.

• Array ID–based volume placement
  

  PX-CSI supports specifying the target FlashArray or FlashBlade by using array IDs for volume provisioning. Use the portworx.io/pure-array-id parameter in a StorageClass or as a persistent volume claim (PVC) annotation to provision volumes on a specific array when multiple arrays are connected to the cluster. This feature provides flexible volume placement strategies for FlashArray and FlashBlade. For more information, see Control volume placement using array IDs.

Resolved Issues

Issue Number	Issue Description	Severity
PWX-52125	For ActiveCluster volumes NodePublishVolume failed when two multipath devices appeared on a node for the same volume, which caused VM migrations to become stuck. This issue occurred when stale SCSI devices and multipath devices from a second FlashArray became active during volume attachment, which created ambiguity during the publish operation.	Minor

PX-CSI 26.1.0.1

March 3, 2026

Supported Operator version: 25.6.1 or later

Resolved Issues

Issue Number	Issue Description	Severity
PWX-51779	Pods using FlashBlade volumes with pure_nfs_endpoint specified in the StorageClass got stuck in ContainerCreating state with the error mount.nfs: remote share not in 'host:dir' format.	Minor

PX-CSI 26.1.0

February 26, 2026

Supported Operator version: 25.6.1 or later

What's New

• FlashArray ActiveCluster support for high availability
  

  PX-CSI now supports FlashArray ActiveCluster for synchronous replication between two FlashArrays, enabling high availability and disaster recovery. Configure ActiveCluster by setting the pure_fa_pod_name parameter in the StorageClass to the stretched pod name. PX-CSI establishes multipath connections to both arrays, enabling automatic failover if one becomes unavailable. For more information, see Use FlashArray ActiveCluster for high availability.

• FlashBlade//EXA support
  PX-CSI now supports FlashBlade//EXA, Everpure high-performance solution for AI and HPC workloads. FlashBlade//EXA uses node group–based provisioning instead of traditional size-based placement. To provision volumes on FlashBlade//EXA, specify the pure_fb_node_group parameter in your StorageClass. For more information, see Create and use FlashBlade PVCs.

• Subnet-aware iSCSI login for multi-subnet environments
  PX-CSI now supports subnet-aware iSCSI login, so that iSCSI initiator interfaces on Kubernetes nodes connect only to FlashArray target interfaces within the same subnet. This is useful when clusters span multiple network subnets and network policies restrict cross-subnet communication. To enable this feature, set the ENABLE_SUBNET_AWARE_ISCSI_LOGIN environment variable to true in your StorageCluster specification. For more information, see Configure subnet-aware iSCSI login.

• Static provisioning for importing existing volumes
  PX-CSI supports static provisioning, so you can import existing FlashArray and FlashBlade volumes into Kubernetes without provisioning new ones. This helps migrate existing data into Kubernetes and use volumes created outside of it. For more information, see Static Provisioning of Volumes.

• New FlashBlade StorageClass parameters
  PX-CSI now supports new StorageClass parameters for FlashBlade file systems, providing better control over file system configuration:

  • pure_fb_node_group: Specifies the node group name for FlashBlade//EXA deployments.
  • pure_fb_hard_limit_enabled: Enables or disables hard quota limits on the file system (default: true).
  • pure_fb_snapshot_directory_enabled: Enables or disables the snapshot directory (default: false).
  • pure_fb_nfsv3_enabled: Enables or disables the NFSv3 protocol (default: true). For FlashBlade//EXA, dafault is "false".
  • pure_fb_nfsv4_1_enabled: Enables or disables the NFSv4.1 protocol (default: true).

  For more information, see StorageClass reference.

• Nvidia GPU Direct Storage support for FlashBlade
  PX-CSI now supports Nvidia GPU Direct Storage (GDS) with FlashBlade by mounting NFSv3 file share with RDMA. This support enables high performance, low latency storage access for GPU workloads, AI/ML training, and HPC applications. To use GDS, specify proto=rdma in the StorageClass mount options.

Resolved Issues

Issue Number	Issue Description	Severity
PWX-50861	Telemetry registration with Pure1 failed because a required header was missing when a proxy was configured using the PX_HTTP_PROXY or PX_HTTPS_PROXY environment variable.	Minor
PWX-50855	During NodeUnstageVolume, multipath device flush failures could result in incomplete device cleanup, leaving stale multipath devices on the node.	Minor
PWX-50391	When using the NVMe protocol, pods might get stuck in the ContainerCreating state with the multipath device not found error after restarting on the same node.	Minor
PWX-51355	Telemetry pods crashed when call-home requests failed with some network errors, such as ERR_TOO_MANY_REDIRECTS, instead of connection refused errors.	Minor
PWX-49846	Volumes failed to unmount after upgrading to PX-CSI 25.8.0 when using containerd as the container runtime.	Minor
PWX-49579	When many PVCs were created simultaneously, they could get stuck in the pending state due to FlashArray API rate limiting. The CSI provisioner sent excessive parallel requests to the backend array, causing provisioning failures. The provisioner now includes throttling to prevent API rate limit errors.	Minor
PWX-48903	Kubelet volume metrics (kubelet_volume_stats_*) were not available for PX-CSI volumes, preventing monitoring of volume capacity and usage through standard Kubernetes metrics endpoints. This affected integration with Kubernetes monitoring tools and dashboards. For more information about available metrics, see Monitor PX-CSI.	Minor
PWX-48887	Telemetry pods failed to collect controller logs when PX-CSI was installed in a custom namespace.	Minor
PWX-48518	Node-level environment variable overrides specified in the StorageCluster specification were not applied, preventing per-node customization for features like concurrency settings and timeouts. For more information about node-specific configuration, see StorageCluster CRD reference.	Minor
PWX-51697	When you add custom annotations to the StorageClasses that the Portworx Operator creates, it removes them during reconciliation loops. Integrations that rely on these annotations (for example, external tool adding annotations or GitOps system managed StorageClass annotations) might not detect the intended StorageClasses. Note: This issue is fixed in Portworx Operator version 25.6.1.	Minor
PWX-49579	PVCs could remain in the pending state when you create many PVCs simultaneously. This issue is resolved by reducing the number of PX-CSI provisioner worker threads to prevent bottlenecks.	Minor

Known Issues

• PWX-52361: Custom kubelet directory is not supported.

  Workaround: Contact Everpure Support for assistance.

  Affected Version: All