WeRecoverData

HPE 3PAR StoreServ Data Recovery

3PAR StoreServ arrays are still in wide production use as the workhorse SAN behind VMware clusters, databases and file servers. Their wide-striped chunklet architecture spreads every volume across most of the drives in the array, which is why a multi-drive or cage failure rarely affects one volume in isolation.

Platform Lineage and Naming

3PAR was founded in 1999, acquired by HP in 2010, and sold as HP 3PAR StoreServ before becoming HPE 3PAR StoreServ. The line ran from the InServ T-Class and F-Class through StoreServ 7000, 10000, 8000 and 20000 generations.

3PAR's successor is HPE Primera, and later HPE Alletra 9000, which inherit much of the architecture and operating model. Many organisations still run 3PAR arrays past end of support, where spare controllers and drives are increasingly hard to source.

Generations and Models We Evaluate

Architecture and Data Layout

3PAR divides every physical drive into 1 GB chunklets. Logical Disks are built from chunklets taken across many drives and cages, Common Provisioning Groups define the RAID level and availability rules for those Logical Disks, and Virtual Volumes (the LUNs hosts see) are mapped onto them — usually thin provisioned.

That three-layer mapping is held in the array's own metadata (the TOC and system volumes on the drives). Recovery requires rebuilding the chunklet-to-LD-to-VV mapping before any host-level file system can be read, which is why standard RAID reconstruction tools do not work on 3PAR media.

Because volumes are wide-striped, a recovery normally needs drives from every cage that contributed chunklets — not just the drives that failed.

Logical Failures

Hardware Failures

Encryption and Keys

3PAR supports data-at-rest encryption using self-encrypting drives with a local key manager or external KMIP server. Where encryption is enabled the array's key state must be preserved with the drives; without it the chunklets cannot be decrypted.

Frequently Asked Questions

Can data be recovered if both nodes in a pair are dead?

Often yes. Reconstruction is done from images of the drives, so it does not depend on getting the controllers running — though controller access can help confirm configuration when the hardware is intact.

Do you need every drive in the array?

Usually all drives that hold chunklets for the affected volumes, which in a wide-striped 3PAR generally means most of the array. Sending only the failed drives is rarely enough.

Our 3PAR is past end of support. Does that change anything?

Not for the recovery itself. It usually means the array cannot be returned to service afterwards, so data is exported to replacement media supplied for the case.

Related