HPE Primera and Alletra 9000 Data Recovery
Primera and Alletra 9000 are HPE's tier-1 arrays and the direct successors to 3PAR, carrying forward the chunklet and Common Provisioning Group model in a more highly available package. They hold the workloads organisations are least willing to lose, so cases are usually urgent and involve either a controller-level event or a deletion that outran the snapshot policy.
Platform Lineage and Naming
HPE introduced Primera in 2019 as the successor to 3PAR StoreServ, keeping the chunklet-based virtualisation model while redesigning the operating system for non-disruptive upgrades. Primera shipped as the A-series (all-flash) and C-series (converged/hybrid).
In 2021 HPE folded the platform into the Alletra brand: Alletra 9000 is the Primera lineage, and Alletra Storage MP later extended the family with a modular hardware design. Because they share architecture and recovery methodology, they are handled as one family here rather than as separate thin pages.
- HPE Primera A630 / A650 / A670 (all-flash), C630 / C650 / C670 (converged)
- HPE Alletra 9000 (9060, 9080) — the rebranded Primera lineage
- HPE Alletra Storage MP
- Predecessor: HPE 3PAR StoreServ
- Chunklets, CPGs, Virtual Volumes carried over from 3PAR
Generations and Models We Evaluate
- Primera all-flash: A630, A650, A670
- Primera converged: C630, C650, C670
- Alletra 9000: Alletra 9060, Alletra 9080
- Alletra Storage MP: Modular block storage chassis in the same lineage
- Media: SAS and NVMe SSD, NL-SAS on converged models
Architecture and Data Layout
Primera and Alletra 9000 keep the 3PAR model: drives are divided into chunklets, chunklets form logical disks under a Common Provisioning Group's RAID and availability rules, and virtual volumes are thin-provisioned on top. Systems are built from node pairs (four nodes on larger models) with mirrored cache.
The operating system was redesigned around a service-based architecture so software faults are contained rather than taking the array down, but the underlying on-disk structures — and therefore the recovery approach — remain chunklet-based.
Reconstruction means imaging the SSDs, rebuilding the chunklet and CPG mapping from array metadata, reassembling virtual volumes, and then recovering the host file systems (usually VMFS or a database's own layout) inside them.
- RAID levels are defined per CPG at chunklet level (RAID 1, RAID 5, RAID 6), with cage-level availability spreading members across enclosures.
- Deduplication and compression are standard on all-flash models, so recovered blocks must be reversed through the array's data-reduction layer.
Logical Failures
- Deleted virtual volumes, CPGs or snapshots
- Array metadata damage after a dual power event
- Failed operating system update or node rescue
- Node pair down or a cluster that will not form
- Peer Persistence / Remote Copy resynchronisation over good data
- Exhausted CPG capacity leaving thin volumes in an inconsistent state
- VMFS or guest file system corruption inside exported volumes
Hardware Failures
- Multiple SSD failures affecting the same RAID sets
- Controller node failure, or loss of both nodes in a pair
- Cache DIMM or backup power module failure preventing a clean flush
- Drive enclosure, expander or midplane faults
- Firmware faults dropping several SSDs simultaneously
Encryption and Keys
These arrays support data-at-rest encryption with self-encrypting drives and local or external (KMIP) key management. Encryption keys and array key state must be preserved with the media for any reconstruction to be possible.
Frequently Asked Questions
Is Alletra 9000 recovery different from Primera?
No. Alletra 9000 is the rebranded Primera lineage and uses the same chunklet architecture, so the evaluation and reconstruction work is the same.
The array is up but a volume was deleted. What should we do first?
Check for a snapshot or a replica on the peer array — that is the fastest route. If none exists, stop provisioning and writes to the same CPG so freed chunklets are not reallocated, then arrange an evaluation.
Do you need HPE involved?
Not for the recovery itself. If the array is under support it is worth opening a case in parallel, but no HPE authorisation is required to have media evaluated.
Related
- HPE 3PAR StoreServ — /services/enterprise-storage/hpe/3par-storeserv
- HPE Nimble Storage Data Recovery — /services/hpe-nimble-storage-data-recovery
- HPE / HP SAN Storage Data Recovery — /services/hp-hpe-san-storage-data-recovery
- SAN Data Recovery — /services/other-data-recovery-services/san-recovery