System Requirements

This section includes detailed system requirements for seamless deployment and optimal performance of PX-CSI in your Kubernetes environment. PX-CSI can also run on a single-node Kubernetes cluster. All nodes and back-end storage systems must meet the following requirements:

Hardware requirements

Category	Requirement
CPU architecture	x86_64 (AMD64)
CPU	1 CPU
RAM	2 GB minimum
Operating System Root Partition	Minimum: 64 GB Recommended: 128 GB
Network Connectivity	Bandwidth: Recommended: 10 Gbps Minimum: 1 Gbps
Node Type	Bare metal or virtual machine (VM)

Cluster node requirements

note

PX-CSI supports Secure Boot-enabled environments and does not require additional configuration.

FlashArray block volumes

Category	Requirement
Packages	Ensure that the latest versions of the following packages are installed on nodes where you plan to run PX-CSI: RHEL iscsi-initiator-utils device-mapper-multipath kpartx (installed as part of device-mapper-multipath) util-linux cryptsetup (optional, required only for volume encryption) Ubuntu open-iscsi multipath-tools kpartx (installed as part of device-mapper-multipath) util-linux cryptsetup (optional, required only for volume encryption) Debian open-iscsi multipath-tools kpartx (installed as part of multipath-tools) util-linux cryptsetup (optional, required only for volume encryption) SUSE Linux Enterprise Server open-iscsi multipath-tools kpartx (installed as part of multipath-tools) util-linux cryptsetup (optional, required only for volume encryption) Note: For NVMe-oF/TCP protocol with DM-multipath, you need multipath version 0.8.7 or later. Native NVMe Multipath does not require DM-multipath.
NVMe multipath	PX-CSI automatically detects and supports both Native NVMe Multipath and DM-multipath for NVMe protocols. No configuration is required. Native NVMe Multipath: Enabled when nvme_core.multipath=Y. Provides faster failover and improved performance on modern Linux distributions. DM-multipath: Used when nvme_core.multipath=N or on older kernels without native multipath support. PX-CSI automatically discovers the correct device path based on the active multipath mode. For more information about multipath support with FlashArray, see Native Multipath Configuration.
Red Hat systems	Ensure that the second action, CAPACITY_DATA_HAS_CHANGED, is uncommented in the 90-scsi-ua.rules file, and restart the udev service.
CSI snapshot feature	Install the Snapshot controller and deploy the CRDs available here in your Kubernetes cluster.
FC protocol (optional)	If you are using the FC protocol, ensure that the latest FC initiator software is installed.
NVMe CLI (optional)	If you are using the NVMe protocol, ensure that the following NVMe CLI version is installed: RHEL earlier than 9.4/Ubuntu earlier than 22.04: NVMe CLI version 1.16 RHEL 9.4 to 9.5/Ubuntu 22.04 or later: NVMe CLI version 2.6 RHEL 9.6 or later: NVMe CLI version 2.11

FlashArray File Services / FlashBlade

NFS utilities — Ensure that the latest version of NFS utilities is installed on all nodes, including the master node.

Node-level DNS resolution (PX-CSI 26.1.1 and later) — NFS mounts are performed directly in the host network and mount namespace on each Kubernetes node, not in the CSI pod namespace. Each node must be able to resolve the configured NFS server hostname or VIP using its own DNS configuration. Cluster DNS (CoreDNS) is not used for NFS mount resolution. If your NFS endpoints are configured as hostnames or VIPs rather than IP addresses, verify that node-level DNS is configured to resolve them on every node before deploying PX-CSI.

Network requirements

PX-CSI supports IPv4, IPv6, and dual-stack configuration for networking.

FlashArray

• Synchronize all cluster nodes using the Network Time Protocol (NTP). Time skew between nodes can cause unexpected behavior or service disruptions.
• Ensure the FlashArray management IP address is accessible by all nodes.
• Verify that your cluster has an operational FlashArray with an existing data plane connectivity layout.
• Use one of the following storage networking protocols supported by PX-CSI:
  • iSCSI: For block storage over IP networks.
  • NVMe-oF RoCE, NVMe-oF TCP, or NVMe-oF FC: For high-performance and low-latency storage access.
  • Fibre Channel (FC): For dedicated storage area networks.
• If you are using iSCSI:
  • Ensure that the storage node iSCSI initiators are on the same VLAN as the FlashArray iSCSI target ports.
  • If using multiple network interface cards (NICs) for iSCSI connectivity, ensure that all NICs can access the FlashArray data IP address. For multi-subnet environments with network policies that restrict cross-subnet communication, consider using subnet-aware iSCSI login. For more information, see Configure subnet-aware iSCSI login.
• If you are using Fibre Channel:
  • Ensure that the storage node Fibre Channel WWNs are correctly zoned to the FlashArray Fibre Channel WWN ports.
• If you are using NVMe-oF FC:
  • Ensure that the FC fabric is zoned to allow NVMe controller discovery from each node.
• Add the following PX-CSI telemetry endpoints on port 443 to networking allow list:
  • https://rest.cloud-support.purestorage.com
  • https://register.cloud-support.purestorage.com

FlashBlade

• Synchronize all cluster nodes using the Network Time Protocol (NTP). Time skew between nodes can cause unexpected behavior or service disruptions.
• Ensure FlashBlade is accessible as a shared resource from all cluster nodes.
• Verify that both the NFSEndPoint and MgmtEndPoint IP addresses are reachable from all nodes.
• If NFSEndPoint is configured as a hostname or VIP rather than an IP address, verify that each node can resolve it using its own DNS configuration. Starting with PX-CSI 26.1.1, NFS mounts are performed in the host network namespace and do not use cluster DNS (CoreDNS).
• Add the following PX-CSI telemetry endpoints on port 443 to networking allow list:
  • https://rest.cloud-support.purestorage.com
  • https://register.cloud-support.purestorage.com

Compatibility and interoperability matrix

For information about supported Kubernetes platforms and distributions, compatible PX-CSI and Portworx Operator versions, minimum Purity version requirements, and supported features per backend, see the Compatibility and Interoperability Matrix.