Shared storage has one unpleasant property: when it goes, every system leaning on it goes at the same second, and the phone call sounds different from all the others. Offline LUNs, pool metadata in pieces, VMFS datastores refusing to mount, all of it gets rebuilt here for businesses and IT providers around Liverpool, the Wirral and the M62 corridor, whether that means one shelf or a rack full of them. Enterprise storage is regular bench trade rather than a favour squeezed in.
Every san job is diagnosed free. One fixed figure follows in writing, agreed before a screwdriver comes out of the drawer.
No fix, no fee all jobs except electronic and mechanical failures, chip level work, DVR and Forensic jobs. Every band is listed on the data recovery cost page.
Matching the symptom to the fault underneath is the first job on any san, and after twenty-odd years these thirty account for very nearly everything that arrives.
A populated shelf does not lose members politely, one at a time, in the order the design assumed. Every failed disk is repaired and imaged individually, and only then does anybody start doing arithmetic across the group.
Your hosts have lost the block device they were built on. It is rebuilt from the floor upwards: RAID groups first, pool metadata second, LUN last. Any other order produces something that looks convincing and proves nothing.
The table that maps physical blocks to volumes has failed. Everything it described is still lying on the disks, so the map is rebuilt out of the structures it was describing.
Firmware updates fail, and when one fails badly the management plane never comes back up. The data behind it never moved an inch. Work goes straight at the drives and leaves the controller out of it.
Two controllers sharing a chassis, a batch and a firmware level tend to fail within hours of one another, which is not what the redundancy was sold on. Recovery goes drive-direct instead. The disks were never shown the datasheet.
Disks fine, groups fine, datastore unmountable. The virtualisation layer collapsed on top of storage that is in perfect order. Datastores are parsed out by hand and each virtual disk lifted and repaired separately.
A few keystrokes against the wrong volume, then a fortnight of regret. How much survives depends entirely on what the array has written since, so taking the platform out of service improves the odds from the moment you do it.
Snapshots lean on their parents, so one damaged link takes down every volume hanging below it. The chain gets rebuilt from images, in sequence, a link at a time.
Promising more capacity than exists works beautifully until the day it stops, and then every thin volume drawing on the pool is damaged in the same minute. Rebuilding from images puts the accounts back into a state that survives inspection.
Writes acknowledged and never committed leave a whole group with parity that disagrees with its own data. On images that disagreement can be settled carefully. On the live array it tends to get settled destructively.
NetApp aggregates and Dell EMC pools stop knowing where their volumes live. Reconstruction starts down at the RAID groups and climbs until the volumes are visible again.
Enterprise formatting defeats ordinary tooling by design rather than by accident. The lab fabric reads those sector sizes natively, without an improvised adapter and a hopeful attitude.
HPE's small-business array arrives with its vDisk metadata torn. It is a rebuild this workshop has performed often enough to find it dull, which is the right state for a rebuild to be in.
V3700 and V5000 shelves lose extents after a power failure and the volume stops assembling. Extents are remapped from the drives, patiently, until the volume is whole again.
The box has lost the target definition. The volume sitting behind it very rarely goes with it. It is extracted and presented again over plumbing that can be relied on.
Break the index and every pointer to every shared block breaks at the same instant, which is a remarkable amount of damage from a small structure. It is rebuilt until shared blocks resolve to real data again.
A handover that went wrong leaves both controllers writing independently, so the array now holds two accounts of the same afternoon. They are reconciled from images, timestamped, with the reasoning written down.
End-of-support hardware fails with no official route left open to it. The lab route carries no renewal date, and the data was never shown the support contract.
A replacement running a different firmware level inside an elderly group misbehaves in ways that read exactly like a failing disk. Stabilise the set, image it, reconstruct afterwards. In that order and no other.
Edit the masking or the fabric and every host goes blind at once, even though the volumes below are in perfect order. The mapping gets rebuilt and returned to the people who own it.
Drives from a single manufacturing lot meeting an hours-counter bug fail within days of each other. It is a documented industry embarrassment rather than bad luck. The remaining members are imaged at speed, before they reach the same hour on the clock.
Put nearline SATA behind a SAS interposer and the interposer fails far more readily than the disk does. Those boards get bypassed and the drives are imaged bare.
A tiered volume keeps its busiest blocks on the quick tier. Lose that tier and what remains is incoherent on its own. The tiering metadata plus the surviving tiers rebuild most of what the fast layer was holding.
Present the same LUN to two servers without a clustered file system and both write as though they own it outright. The two write histories are pulled apart on images, one operation at a time.
Service processor dead, or credentials that left with a contractor nobody can reach any more. The array will not talk to anyone. The disks have never asked for a password.
Snapshot space is finite, and most platforms deal with exhaustion by silently deleting the oldest snapshots. The restore point somebody was counting on went weeks ago, without an email about it. Whatever survives is carved out of the unallocated space on the imaged disks.
A failover test, a site labelled wrongly, or a relationship restarted in reverse, and the stale copy lands on the good one at wire speed. That side is imaged, and the earlier structures are lifted back out from underneath it. Stopping the replication early is worth more than anything anybody can do later.
An array that encrypts at rest keeps its keys inside the controller, or on a key manager standing beside it. If both have gone, the disks read perfectly and mean nothing whatsoever. That verdict comes out of the free assessment and it is delivered straight, because pretending otherwise would only cost you money.
Shelves daisy-chain, so a failed cable, a dead expander or a shelf powered up out of sequence can make an entire tier of disks vanish at once. It presents as total loss and it is frequently a wiring fault. Which of the two it is gets settled before anybody rebuilds anything.
Present a new server with several volumes and somebody will eventually initialise whichever one looked unused. Very little actually gets written by a quick format, so what lies beneath survives largely intact — as long as that host is disconnected before it begins filling space it thinks is empty.
Every SAN is layers upon layers. Physical disks are gathered into RAID sets, those sets pooled together, the pool sliced into LUNs, and the file systems and hypervisor datastores people actually use sit on top of the lot. Recovery works down through that arrangement and then climbs back the way it came. Image the disks, all of them. Reconstruct the RAID sets from the images. Get the pool metadata back into a state that parses. Extract the LUNs. Only after all of that does anybody open a virtual machine or a database. None of the equipment is unusual, with Dell EMC, NetApp, HPE MSA and EVA, IBM and Fujitsu the regulars, and the confidentiality agreements, change control and internal paperwork your organisation runs on fit around that sequence without adding a day to it.
Handing a customer a pile of raw blocks helps nobody. What is wanted is the small number of systems the business genuinely cannot trade without, running again by the start of the next working day. So the reconstructed LUNs get their datastores mounted, the VMDK and VHDX files come out and are repaired where they need it, and whichever SQL or Exchange workloads were named on the phone go to the front of the line, so staff are working again while the remaining terabytes copy away in the background. That is triage rather than heroics, and it is done deliberately slowly, because rushing an enterprise job is precisely how the second mistake gets made.
Enterprise recovery needs the plumbing as much as the skill, and both are kept here in depth:
Shelves and loose drives attach to our own fabric, including the awkward sector sizes enterprise stock is formatted with before it ever leaves the factory.
Whole RAID groups are captured alongside one another, after any casualty has first been round the mechanical bench and the firmware bench.
Vendor metadata is reconstructed in the order the array originally built it: groups, then pools, then LUNs, then whatever file systems were riding above.
A restored LUN presents its datastore again, and every virtual disk on it is pulled clear and repaired on its own terms.
Say on the phone that SQL Server or Exchange is what the business is waiting on. Those files are verified early and released first, so trading resumes before the long tail has finished copying.
Every original, whether a bare disk or a whole shelf, stays write-blocked for the length of the job. There is no stage at which that gets relaxed for convenience.
An unfamiliar badge seldom means unfamiliar hardware. The enterprise vendors buy from the same small set of suppliers, and most of that stock has been apart on this bench before. A pool that has lost its map is rebuilt from the RAID groups upwards. A pair of dead controllers simply sends the work drive-direct. Odd sector sizes are read natively here instead of being argued with. Most of the enterprise work here comes from freight and forwarding offices near the docks, from the plants around Runcorn and Widnes, and from the distribution parks out at Warrington. Enterprise recovery starts at £500 + VAT, rising with the number of disks, and the figure goes in writing after a free assessment that takes 2 working days from arrival.
Before anything leaves the rack, label every disk with the shelf it sat in and the slot number it came out of, because position matters more in an enterprise layout than in almost anything else we see. Photograph the front of the unit as you found it. Then send the tagged drives only: shelves, controllers and rails stay on site where they belong. Post them tracked and insured to Manchester Data Recovery, Peter House, Oxford Street, Manchester M1 5AN, or use a courier you have booked yourself. There is no collection service and no Liverpool shop counter. If a member of your team is passing Manchester anyway, the lab reception will take the parcel in person, Monday to Friday, 9:00am to 5:30pm.
Almost every job on this bench came in by tracked, insured post. It is the calmest way to move a drive that is already struggling, and a parcel handed in anywhere on Merseyside usually reaches the bench the following working day.
Still screwed inside a laptop, tower, MacBook, iMac, server or a CCTV recorder? Take the drive out and post that on its own. We do not strip machines here, and any computer shop will do it in a few minutes. The one job nobody can take on is flash soldered straight to a logic board, as on Apple Silicon Macs and a handful of very slim laptops: if the storage will not unbolt, there is nothing to send.
↓ Print the shipping & booking-in form (PDF)
Address it to Manchester Data Recovery. It is roughly 35 miles from Liverpool along the M62 if you fancy the run, and next working day by tracked post if you do not. We ring or email the moment it is booked in, and the free diagnostic closes 2 working days after that.
Not sure what belongs in the parcel? Call 0800 689 0668 before you tape it shut, or step through the free online diagnostic first.
Diagnosis free, one figure written down, and £300 + VAT covers a single drive or SSD. Start online or ring.