Shared storage does not fail on its own. It takes everything leaning on it down at the same second, and that is what makes these calls unlike the rest of the week. SAN estates get reconstructed here: offline LUNs, damaged pool metadata, datastores that refuse to mount. The work is done for companies directly and for the IT providers acting for them, on anything from one shelf to a room of them. Enterprise storage is core bench trade, not a favour.
Every san job is diagnosed free. The fixed figure reaches you in writing first, before any tools come out.
No fix, no fee all jobs except electronic and mechanical failures, chip level work, DVR and Forensic jobs. Full pricing is on the data recovery cost page.
First job on any san is matching the symptom to the fault. Twenty-odd years in, these twenty-five cover nearly every one that reaches this bench.
Enterprise shelves lose disks by the handful, not one at a time. Each casualty gets individual repair and an image of its own before group arithmetic resumes.
The block device your servers depended on has gone. Rebuilding it runs bottom upwards, RAID groups first and pool metadata second, strictly in that sequence, because any other order proves nothing.
What has failed is the map from blocks to volumes. NetApp aggregates and EMC pools mislay their volumes in precisely the same fashion. Reconstruction comes out of the raw on-disk structures that map spent its life describing.
Mid-flash, the management plane died, while the data behind it sat perfectly still. Work goes straight at the disks from that point and ignores the controller altogether.
Pairs fail together a great deal more readily than any datasheet implied. Recovery runs drive-direct, which carries the advantage of never having read that datasheet at all.
Collapse begins in the virtualisation layer and takes down everything visible with it. We parse the datastores out, then extract each virtual disk and repair it on its own merits.
Seconds to perform, days to regret. Survival depends entirely on what has landed since, so freezing the platform improves your arithmetic straight away and costs nothing.
Break a single link and every volume depending on it comes down. The chain gets re-forged at image level, a link at a time, until those volumes stand again.
Promise more capacity than exists and every thin volume fed by that pool corrupts the moment it runs out. Rebuilt from images, and the accounts survive inspection afterwards.
Cache batteries past their life, plus one outage, leave parity contradicting data right the way across a group. Resolving that contradiction at image level is affordable. Resolving it live is destructive.
Formats of 520, 524 and 528 bytes blunt ordinary tooling, and they were designed to. Our lab fabric reads them natively, with no improvised adapters and no hopeful conversions.
HPE's small-business staple turns up with its vDisk structures torn open. This workshop has rehearsed that reconstruction to the point of boredom, which is the right amount of practice.
After a power event, V3700 and V5000 units shed extents, and the volume above them stops making sense. Remapped off the disks, extent by patient extent, until it is whole.
Through a botched handover both controllers wrote independently, so the same storage now holds two versions of events. They get reconciled at image level, timestamped and audited.
An array past end of support fails with no official route left anywhere. Lab work carries no expiry date, because data has never once read a support contract.
Edit the masking or the fabric and every host goes blind at once, while the volumes beneath sit in perfect health. We rebuild the mapping and hand it back to whoever owns it.
Disks out of one lot hitting a firmware defect in the hours counter fail inside days of each other. That is a documented industry embarrassment, not bad luck. The survivors get imaged quickly, before they follow suit.
Kill the fast tier and every tiered volume loses whichever blocks it was using hardest, leaving cold tiers that make no sense in isolation. Those cold tiers, plus the tiering metadata, rebuild what the quick layer held.
Slip up with zoning and, absent a clustered file system, corruption interleaves from both directions at once. The two write histories get separated from images, operation by operation.
Service processor dead, credentials retired alongside a contractor. The array declines to converse. The disks hold no such reservations, so extraction proceeds with no password at all.
Snapshot reserve is finite space, and most platforms deal with exhaustion by quietly deleting the oldest snapshots. The restore point somebody was counting on has gone before anyone learns it existed. Whatever remains gets carved out of unallocated space on the imaged members.
A failover test, a site label somebody got wrong, or a relationship restarted in reverse, and the stale copy overwrites the healthy one at wire speed. We image the overwritten side and recover the older structures from beneath, which works far better the earlier it is stopped.
An array encrypting at rest keeps its keys inside the controller, or else on a key manager. Lose both and every disk is readable and entirely meaningless. That is established during the diagnostic and stated plainly, because no honest way round it exists.
Shelves daisy-chain together, so one failed cable, one failed expander, or a shelf brought up out of sequence can make an entire tier of disks disappear in a moment. It presents as total loss and it is frequently a cabling fault. Which of the two it is gets established before anybody rebuilds a thing.
Present a new server with several volumes and some well-meaning administrator initialises whichever one appeared spare. Quick formats write very little, so the volume beneath comes through largely intact, on the condition that the host is disconnected before it begins filling space it believes belongs to it.
Shared storage is layers on layers. Disks in RAID sets. Sets in pools. Pools sliced into LUNs. File systems and hypervisor datastores on top of those. Recovery descends that stack and then climbs it again in the same order. Every disk imaged. Every RAID set rebuilt. Pool metadata repaired. LUNs lifted out. Only at that point are the virtual machines and databases inside them opened. The hardware is well-trodden ground. Dell EMC and NetApp. HPE MSA and EVA. IBM and Fujitsu. Neither does the paperwork. NDAs, change control and whatever governance your organisation runs to slot into the sequence without costing it any time.
Nobody has ever wanted raw blocks back. What is wanted is a handful of systems, the ones the business actually stands on, live by morning. So datastores are raised on the reconstructed LUNs. VMDK and VHDX files come out and get repaired. Named SQL or Exchange workloads are pulled to the head of the list. Operations restart while the remainder copies quietly behind them. That is triage, not heroics, and it is done at a measured pace. Hurry an enterprise job and you make the second mistake, which is always the one that costs.
SAN recovery calls for enterprise plumbing alongside recovery skill, and this workshop stocks both in depth:
Shelves and disks connect through our own fibre channel, SAS and iSCSI plumbing, unusual enterprise sector sizes included.
Entire RAID groups captured alongside one another, once any casualty disk has first been round the mechanical and firmware benches.
Vendor metadata rebuilt in the sequence it was originally created: groups, then pools, then LUNs, then whichever file systems ride on top of them.
Restored LUNs present their datastores again, and every virtual disk gets lifted out and repaired on terms of its own.
Name your SQL Server and Exchange workloads at the outset and they get verified early and pushed up the queue, so whichever services matter most return first.
Every original, whether disk or shelf, stays read-only between the day it arrives and the day it leaves. No stage of the work relaxes that for anyone.
An unusual badge rarely means unusual hardware. Enterprise vendors rebadge out of a common parts bin, and most of its contents have crossed this bench several times over. A pool that lost its map gets rebuilt upwards from the RAID groups. Two dead controllers simply push the work drive-direct. Odd sector sizes get read natively on our lab fabric rather than argued with. Operators, engineering contractors and service companies around the harbour and the industrial estates work this class of kit harder than most, and our intake shows it. Send the tagged disks; leave the shelf its rack bolts.
Before anything leaves the rack, tag each disk with its shelf plus its slot. Enterprise layouts punish guesswork about position harder than anything else on this site. Photograph the setup as you found it, then send only the tagged disks; shelves and controllers stay on site. Post everything tracked and insured, or use a courier you book yourself. Collection is not a service offered from here. Disks can also come in to reception at our Edinburgh location, 4 Redheughs Rigg, Westpoint, South Gyle, Edinburgh EH12 9DQ, Monday to Friday, 9am to 5:30pm, if that suits you better.
Nearly everything on this bench came by tracked, insured post. It is the safest way to shift a failing drive, and a parcel handed in at a North-East post office is usually here the next working day.
Still bolted into a laptop, desktop, MacBook, iMac, server or CCTV / DVR box? Get the hard drive or SSD out first. The bare drive travels on its own. Stripping machines down is not a job this lab takes on. Flash soldered to a motherboard, as on Apple Silicon Macs and a couple of very thin laptops, is our one flat no: if it will not unbolt, it cannot be worked on.
↓ Print the shipping & booking-in form (PDF)
Address it for the attention of Edinburgh Data Recovery. From Aberdeen that is roughly 130 miles and two and a half hours down the A90, or next working day by tracked post if you would rather not make the run. You get a call the moment it is booked onto the bench.
Not sure what belongs in the box? Ring 0800 689 0668 before you tape it shut, or work through the free online diagnostic.
Free diagnosis, one number written down, no fix no fee on the bulk of jobs. Start online or phone it in.