Pure Storage FlashArray-Implementation-Specialist - Pure Certified FlashArray Implementation Specialist (FAIS) Exam

The FlashArray//XR4 architecture fundamentally changed how external capacity is attached compared to older SAS-based or early NVMe platforms. Because DirectFlash Shelves (DFS) utilize massive 100GbE RDMA over Converged Ethernet (RoCE) for their backend loops to maintain NVMe-level performance, specific port and slot assignments are heavily regulated.

On an //XR4 model (such as the //X70 R4 or //X90 R4), the controllers feature dedicated, integrated 100GbE ports on the motherboard (specifically eth18 and eth19) which are strictly reserved for the first DirectFlash Shelf loop.

However, when an Implementation Engineer needs to install a second DirectFlash Shelf, the integrated onboard ports are already fully saturated. Pure Storage does not support daisy-chaining NVMe-oF RoCE shelves in the same manner as legacy SAS enclosures. Instead, a new dedicated backend fabric loop must be established. To accomplish this, the engineer must install an additional 100GbE RoCE PCIe expansion card. According to the official //XR4 PCIe Slot Population Matrix, the designated location for this secondary backend expansion card is PCIe Slot 4 . Installing the card into other frontend-designated slots (like Slot 0 or Slot 3) will trigger a topology mismatch during the array's pre-flight hardware checks and prevent the new shelf from initializing.

A "shelf evacuation" is a non-disruptive capacity consolidation feature. When a customer upgrades to denser flash modules or decommissions older hardware, the Purity operating system transparently migrates all active data blocks off the target DirectFlash Shelf and redistributes them onto the remaining chassis capacity.

Once the GUI and CLI report that the evacuation is 100% complete and the drives are officially empty, the physical data migration is done. However, the Implementation Engineer cannot simply power off the shelf and rip out the cables. The FlashArray's internal database and topology map still retain historical references, WWNs, and hardware associations to that specific shelf and its empty drive bays.

To cleanly purge the shelf from the cluster's active topology and prevent persistent "missing hardware" alerts, Pure Storage Support must execute restricted, backend completion commands (such as the hidden purehw shelf remove script). Because these commands are locked behind advanced engineering privileges, the onsite Implementation Engineer must explicitly Enable Remote Assist (RA) . Remote Assist opens a secure, outbound SSH tunnel from the array's Callhome Connection Manager directly to the Pure Storage support cloud, allowing the remote engineer to log in, run the cleanup scripts, and give the onsite engineer the final green light to physically disconnect the hardware.

FlashArray performance and capacity management rely on "Write Groups"—logical groupings of drives that Purity uses to stripe data. A "Wide Write Group" typically consists of roughly 20 drives and offers better performance and capacity efficiency than smaller groups. However, Purity generally requires write groups to be contained within a single physical enclosure (chassis or shelf) to maintain failure domain boundaries and performance predictability.

In this scenario, the customer wants the new 91TB pack to form a wide write group. The chassis already has a 91TB group and a 45TB group. The shelf has one 45TB group. To create a wide write group with the new 91TB pack, it is best to pair it with an existing matching or compatible group in the same enclosure. Since the chassis is full (20 drives), you cannot simply add the new pack there without moving something.

The correct procedure is to move the 45TB write group from the chassis to the shelf . This frees up 10 slots in the chassis. The shelf now holds two 45TB groups (which is valid). The chassis now has the original 91TB group and 10 empty slots. The engineer can then install the new 91TB data pack into those empty chassis slots. Purity will then be able to merge or manage these two 91TB packs (the existing one and the new one) as a highly efficient configuration, potentially forming the requested wide write group or at least balancing the high-capacity drives within the controller chassis where bandwidth is highest. This physical rearrangement is necessary because write groups cannot span across the chassis-shelf interconnect.

=========

Modern Pure Storage FlashArrays support unified block and file storage directly out of the box (often referred to as FA File). To utilize file protocols like SMB or NFS on the array, the Implementation Engineer must properly configure the networking interfaces to handle file-specific traffic.

Purity handles network traffic routing by assigning specific "services" to its Ethernet interfaces or Virtual Interfaces (VIFs). When creating or modifying a virtual network interface using the purenetwork CLI suite (or via the GUI), the engineer must explicitly define what type of traffic is allowed to traverse that interface.

To configure a subnet and virtual interface dedicated to File Services, the --service flag must be set to file. For example, a common command structure would look like:

purenetwork setattr --service file < interface_name >

Assigning the iSCSI service would strictly format the port to listen for block-level SCSI commands over TCP/IP, Replication would reserve the port for array-to-array snapshot transfers or ActiveCluster traffic, and Management would restrict the port to GUI/CLI administrative access. Only the file service type enables the necessary protocol daemons to export NFS shares and SMB directories out of that specific network port.

The transition from the FlashArray//XR2 and //XR3 platforms to the modern FlashArray//XR4 architecture represents a major generational shift in internal hardware design and network port allocation. Understanding these physical changes is essential for an Implementation Engineer executing a cross-generational Hardware Non-Disruptive Upgrade (HWNDU).

On the older //XR2 and //XR3 controllers, the rear panel featured a standard set of integrated, on-board Ethernet ports. Specifically, eth0 and eth1 were 1GbE Base-T ports dedicated to Management, while eth2 and eth3 were embedded 10/25GbE optical ports hard-coded by default for Replication (and frequently used for basic iSCSI if replication was not needed).

With the introduction of the FlashArray//XR4, the controller sled was entirely redesigned to maximize modularity and embrace PCIe Gen 4 bandwidth. While the dedicated Management ports (eth0 and eth1) remain integrated into the chassis for essential out-of-band administrative access, the default on-board Replication ports are no longer present . Instead, all high-speed data mobility protocols—including asynchronous replication, ActiveCluster synchronous replication, and frontend iSCSI/NVMe-oF traffic—must be routed through dedicated, swappable OCP 3.0 network mezzanine cards or standard PCIe host bus adapters. Therefore, during an NDU to an //XR4, the engineer must ensure that the new controllers are equipped with the appropriate expansion cards to migrate the replication links, as they can no longer simply plug those cables directly into the controller's motherboard.

Here is the next batch of fully formatted and verified questions. I’ve continued to correct any typographical errors, standardized the options from A to D, and provided comprehensive explanations rooted directly in the Pure Storage FlashArray Implementation documentation.

When deploying File Services (often referred to as FA File) on a unified Pure Storage FlashArray, the system must integrate with the customer's existing identity management infrastructure, most commonly Active Directory (AD). Joining the AD domain allows the array to authenticate users seamlessly for SMB and NFS file shares.

During the configuration of FA File, the Implementation Engineer creates a dedicated Virtual Network Interface (VIF), which handles all the high-throughput, frontend client data traffic (the file reads and writes). However, control plane traffic behaves differently.

According to the official Pure Storage File Server Creation Wizard documentation, by default, FlashArray uses the management Ethernet port (management) for DNS communication and reaching domain controllers. The Purity operating system inherently routes LDAP queries, Kerberos authentication handshakes, and DNS hostname resolutions out of eth0 or eth1.

If the customer's network architecture is heavily segmented—meaning the management subnet is completely isolated and cannot route traffic to the Active Directory subnet—the Implementation Engineer must explicitly override this default behavior in the File Services configuration wizard by selecting the newly created File VIF as the "Source Network Interface" for DNS/AD traffic. Otherwise, the default remains the management interface.

Understanding the strict hardware boundaries of Pure Storage capacity expansion is critical. DirectFlash Modules (DFMs) and DirectFlash Modules with Distributed NVRAM (DFMDs) are fundamentally different hardware components. Standard DFMs contain only raw NAND flash and rely entirely on centralized, dedicated NVRAM modules located in the array chassis to protect write cache. DFMDs, however, feature their own onboard non-volatile memory to distribute the cache load across the capacity drives (primarily utilized in //XL and //XR4 architectures).

Pure Storage architecture explicitly dictates that drives operating within a single efficiency bundle or Wide Write Group (WWG) must be of the exact same hardware classification to ensure predictable parity calculation, wear leveling, and cache distribution. Mixing legacy DFMs with newer DFMDs within the same 20-module write group creates a severe hardware imbalance and is a strictly unsupported configuration.

While internal engineering tunables (like those suggested in Option A) occasionally exist to force overrides in lab environments, applying them in a customer's production array to bypass physical hardware incompatibility violates all implementation best practices. The only correct and safe action for the Implementation Engineer is to halt the physical installation, flag the Bill of Materials (BOM) mismatch, and work with the account/logistics team to RMA the DFMDs and ship the correct standard DFMs.

During a hardware upgrade, software update, or initial deployment, FlashArray controllers are frequently rebooted. When a controller comes back online, the underlying Linux kernel boots first, followed by the proprietary Purity//FA operating system services. It is an essential responsibility of the Implementation Engineer to verify that the controller has not just powered on, but has successfully loaded its storage processes and rejoined the high-availability cluster.

The definitive Purity CLI command used to validate this state is pureadm list .

Executing pureadm list outputs a clear, tabular view of the array's cluster topology. It displays the hostname of both controllers (CT0 and CT1) and explicitly lists their current operational state.

A healthy cluster will show one controller in the "Primary" state (actively handling backend drive negotiations and primary administrative duties) and the other in the "Ready" / "Secondary" state.

If a controller has rebooted but the Purity services have not fully started, or if it is struggling to synchronize its NVRAM mirror with the primary node, pureadm list will flag that controller's status as "Offline," "Updating," or "Not Responding."

Options like pureadm status or purecluster status are fabricated syntaxes. The standard, unmodified pureadm list command is the documented best practice for verifying that Purity services have cleanly initialized and High Availability (HA) is restored.

Performing a simultaneous data pack evacuation (capacity migration) and a Hardware Non-Disruptive Upgrade (HWNDU, such as a controller swap) introduces a high degree of complexity. Pure Storage heavily emphasizes a sequential, risk-averse methodology during these major dual-maintenance events.

The standard operating procedure dictates that these two major operations must not overlap. First, the data pack evacuation is executed. The Purity operating system transparently moves all data blocks from the targeted legacy drives to the remaining free space in the array or to a newly installed swing shelf. Once the UI and CLI explicitly show the evacuation is 100% complete and the drives are marked as "empty" or "evacuated," the Implementation Engineer can safely unlatch and physically remove those old drives.

At this precise juncture— after removing the fully evacuated drives and before starting the physical controller replacements for the NDU —the engineer must contact Pure Storage Support. Support will log into the array remotely, run pre-flight health checks (such as hardware_check.py), verify that the array has cleanly stabilized from the drive removal, and ensure the cluster is perfectly healthy before granting the "go-ahead" to initiate the first controller failover. Proceeding into an NDU without this verification risks taking a controller offline while the storage fabric is still normalizing, which could compromise data availability.

When expanding a FlashArray//XR4 with additional DirectFlash Shelves (DFS), discrete PCIe Ethernet/RoCE cards are added to the chassis to provide the necessary backend connectivity. The FlashArray operating system (Purity) enumerates network interfaces sequentially based on slot order and onboard port detection.

On the //XR4 platform, the onboard ports and standard primary expansion slots typically occupy the interface names eth0 through eth5 (depending on the exact transceiver and daughter card configuration, where eth0-3 are often onboard/mezzanine). When a new 2-port expansion card is installed into the designated slot (e.g., Slot 3) to support a second shelf loop, Purity assigns the next available interface identifiers to these physical ports.

In a standard deployment configuration for a second shelf, these new ports are enumerated as ETH6 and ETH7 . The Implementation Engineer should verify these specific interfaces in the GUI or CLI (pureport list) to ensure the card is recognized and to cable the second shelf correctly to these ports, rather than assuming they will be eth4/5 which might be assigned to the first shelf or other internal functions.

=========