Home / Server Data Recovery

Server Data Recovery arrays, databases and mail stores

Servers are not polite about failing. A member drops out unnoticed in March, a second follows in October, the rebuild sticks at forty per cent, and the backup that was meant to cover all of it stopped running some time in the spring. This page works through what actually breaks, why your controller is not needed to rebuild the set, how databases and virtual machines are handled once the storage stage is over, and what each band costs.

Servers and arrays begin at £500 + VAT and climb with the member count. One server disk treated on its own is £300 + VAT. Looking costs nothing, the answer lands on the second working day after your drives are logged in, and the figure after it is fixed in writing.

// what the call usually turns out to be

The failures that put a server on this bench

A server almost never fails once and neatly. By the time anybody picks up the phone, two or three things have gone wrong in a particular order, and that order explains more than any of the faults does alone.

A drive went and nothing told anybody

Alert mail leaving for a mailbox that closed when its owner did. An amber light on the back of a cabinet in a room people walk past rather than into. A monitoring agent that stopped reporting after an update and was never missed. Weeks or months of running short, and the fault only surfaces on the morning the next drive goes and takes the volume with it.

The card died, or the backplane did

Perfectly healthy disks, and the thing wiring them together has stopped. A failed card, a dead backplane, a cache module emptied by a power cut, or a replacement of the wrong family that cannot make sense of the layout its predecessor wrote. None of that calls for a drive to be opened, and all of it calls for the set to be rebuilt from what the disks themselves say.

An expansion or migration that never finished

Growing a set onto another drive, changing its level, moving a volume between platforms — each one rewrites the layout underneath people who are still working. Interrupt it with a power cut or a second failure and half the volume sits in the old shape and half in the new. That is recoverable, patiently, with both shapes reconstructed on copies.

A database or a mail store that will not open

Hardware entirely blameless, file entirely unmountable. SQL data files caught by a hard shutdown. A mail store left dirty. A guest whose virtual disk no longer agrees with itself. Application trouble sitting on healthy storage, still worked on images, and needing the log files carried across alongside the data files.

Encryption right through the shares

Every folder opens and nothing inside it reads. On a file server, decryption is almost never the route. What is realistic is the originals the malware unlinked instead of overwriting, older versions, and whatever it had not reached by the time somebody pulled the network cable. Charged as media, which puts an array at £500 + VAT and upwards.

Somebody removed the wrong object

A LUN detached from a host it did not belong to. A volume formatted during a weekend of changes. A virtual disk deleted out of a pool. A set re-created by an optimist. Nearly all of it reverses cleanly, on one condition: the machine went off quickly and nothing has been written since.

// three faults share one phone call

Dead server describes three unrelated situations

The opening sentence is nearly always the same and it covers three quite separate problems, which is why the first ten minutes matter more than they look. A box that will not power up or complete its post has a hardware fault — supply, board, memory — and the storage inside it is very often perfectly well, so the answer is to move the drives or simply read them here. A box that runs happily but cannot locate its storage has a card, backplane or array problem. A box that boots, sees the array and offers a volume nothing will mount has an application or file system problem standing on hardware that is doing its job.

Sorting out which of the three is in front of you changes every decision that follows, because the action that rescues one is the action that damages another. Pulling drives is sensible in the first case and a nuisance in the third. Pointing a repair tool at the volume is defensible in the third and destructive in the first. Ten minutes of looking is cheap next to either mistake.

Why the controller stays in your rack

Cards from Dell, HP, LSI and Adaptec each reserve a region on the member drives and write their description of the array into it, and in most families that description travels with the disks rather than living in the card. Software arrays do the equivalent — Linux md, Windows dynamic disks, Storage Spaces, ZFS, and the implementation inside every NAS sold. Either way the layout is recorded on the drives. That is why a set can be reconstructed on a bench that has never seen your controller, and why posting one achieves nothing except postage.

Level matters far more than badge. A mirror gives two independent copies and the gentlest job in the trade. A stripe keeps nothing in reserve, so one failure means every drive gets imaged before anything is assembled. Single parity tolerates one gap, double parity two, and the nested levels favoured for databases behave like a stack of mirrors underneath a stripe. Each carries its own arithmetic and its own worst day, and the examination reports the level you actually have rather than the one on the handover document from four years ago.

Hot spares help, and they also start things nobody chose

A spare sits idle in the chassis waiting to be called up the moment a member drops, and it genuinely earns its place, because it begins work instead of waiting for somebody to notice a light. It is also why a number of arrays arrive here worse off than they had to be. The automatic rebuild starts that heavy front-to-back read of every survivor with no human deciding whether now is the moment, and it may well start at three in the morning across drives that are all precisely the same age.

The other thing about a spare is that it has been getting older too. A disk that has sat unpowered in a warm cabinet for four years is not a new disk in any meaningful sense, and a fair share of them fail during the very rebuild they were bought to cover.

// databases, mail stores and guests

Getting the files back is frequently the smaller half

A SQL data file lifted off a formatted volume is worth having only if its transaction log comes with it and the pair still agree. A mail store recovered in a dirty state has to be brought to a clean one before any software will mount it. A virtual machine is one file containing a whole file system, so the guest becomes a second recovery nested inside the first. All of that is ordinary work here, and it is the honest reason a server takes longer than a laptop drive: the storage stage ends and the application stage begins.

Where a guest lives on a failed set, the two problems are stacked. The array has to come back with the correct geometry before the virtual disk is even legible, and only then can the file system inside the guest be put right. A stripe order that is slightly wrong yields a virtual disk that opens beautifully and contains nonsense, which is precisely why parameters are tested against content rather than accepted from a card. VMware and virtual machine recovery takes that layer apart in full.

Downtime, and what anybody can honestly tell you about it

Nobody in this country can price the hours of a server job down a telephone line, and a firm that offers to is gambling with your week. What is fixed: looking costs nothing, the examination closes on the second working day after the drives are logged in, and it produces a written figure beside a realistic date. Call the freephone before you post and the job is marked in advance, so it reaches the front of the queue the moment the parcel is booked rather than the back of it. A lone drive finishes two to four working days after you authorise. A set takes longer because each member is imaged on its own.

What does not exist is a hotline at midnight, an engineer at your rack, or a same-day promise made before anyone has looked at a disk. This is a bench working Monday to Friday, 9:00am to 5:30pm, with its prices printed, and saying so plainly beats letting a stopped business plan around a service that was never on offer. If the company truly cannot wait, plan around the media rather than the laboratory: move people onto other hardware, restore whatever partial backup exists onto something new, and treat the recovery as the way of closing the gap rather than the way of restoring everything.

// sending it in, and what it costs

What leaves the rack and what stays in it

Drives travel; the chassis does not. Withdraw the members, mark each with its bay reading across the front, and photograph that front panel before anything moves. Card, rails, supplies, cabling and caddies that release easily all stay with you, because none of them are used here. The sensible exception is a caddy that will not come off without tools — send the drive still in it rather than forcing anything.

Write the history down and put it in the box. Controller model, level if you know it, which member failed first, whether a rebuild was tried and how far it travelled, and what the volume was for. Two lines of that saves hours regularly, and on a multi-disk job hours decide whether you get the data on Tuesday or Thursday. Send it tracked and covered, or bring it: Oxford takes devices over the counter on weekdays from 9:00am to 5:30pm, and Coventry to Oxford is about fifty-five miles of M40, roughly an hour. Nothing is collected and there is no Coventry counter, which is worth knowing before somebody clears a morning for it.

The money

Servers and arrays begin at £500 + VAT and climb with the member count. Pull a single disk out of a server and treat it on its own and it is £300 + VAT. A recorder drive out of a CCTV system, or an encrypted volume where the key can be produced, is £400 + VAT. Cards and sticks are £250 + VAT. Forensic casework with a report written to be argued with is £800 + VAT; take the report away and the verified image with its deleted material extracted is £400 + VAT, sharing a rung with the recorder band instead of inventing one. Encrypted-by-malware media is priced as media wherever it turns up.

Logical jobs run under no fix, no fee: deleted volumes, damaged file systems, sets that will not assemble with sound drives inside them, ransomware. Excluded from that are chip-level work, DVR jobs, forensic jobs and any failure that is mechanical or electronic, and invasive work takes half the agreed figure upfront. Nothing begins without your say-so, and what you agree to is what appears on the invoice.

Who sends servers from this county

Engineering firms in Rugby with three decades of drawings on one volume. Accountants and solicitors in the city centre in the fortnight before a deadline. Distribution operations off the M6 and M69 and out at the Ansty and Prologis parks, where a warehouse system nobody dares reboot runs the entire shift. Hotels and visitor businesses around Stratford whose booking database is also their ledger. Schools and academy trusts across Warwickshire running everything from one box in a cupboard. The trade never predicts the fault. What predicts it is a server so quietly reliable for so long that nobody opened the backup report.

// related pages

Nearby on this site

// questions from IT managers and directors

Common questions

No, and it is worth resisting the urge. Members only, each marked with the bay it occupied, plus a photograph of the front panel taken before you touched anything. Card, rails and supplies are unused at this end. Four bare drives make a cheap parcel that is easy to protect properly; a rack unit is neither of those things.
Usually, and it runs as two jobs in order. The set is rebuilt correctly first, then the data file and its logs are recovered together and brought into a consistent state before anything will mount them. Say which product and which version at the start, because it changes the sequence the bench works in.
No midnight line and no engineer who travels. What does exist is worth using: ring the freephone before posting and the job is marked, so it reaches the front of the queue as soon as the parcel is booked. The bench runs Mon-Fri 9:00am-5:30pm and the free examination closes on the second working day after arrival.
Very probably not. A box that will not finish its post has generally lost a supply or a mainboard, and the storage inside is untouched by any of that. Take the drives out, keep them in bay order, and either fit them to another chassis of the same family or send them in. What must not happen is anyone re-initialising the set while hunting the power fault.
From £500 + VAT for anything holding several disks, climbing with the member count, since each drive is imaged separately before assembly is attempted. One server disk treated alone is £300 + VAT. Looking is free, the answer arrives two working days after booking in, and the price is fixed in writing before you authorise anything.

Shut it down before the next good idea gets tried.

A large share of server work arrives having been made harder by an hour of willing help. Ring first, then withdraw the drives in bay order.

// getting your media to Oxford

Posting a device in — what goes in the box

Multi-disk jobs travel as a set of bare drives sharing one padded box, and that parcel wants tracking and covering for what is written on it rather than what it weighs. Booked over a Coventry post office counter in the afternoon, it usually reaches the Oxford bench the next working morning — faster than most firms manage to spare somebody for the run down the M40.

As a rule the storage comes out and the machine stays where it is. That applies to a laptop, a tower, an iMac and to the recorder sitting under a counter. Taking equipment apart is not something this bench does, and a repair shop will free a drive in a few minutes. Two things go the other way: an external drive stays sealed inside its own case, and a NAS travels as a complete unit with its disks still in their bays. A Fusion Mac is a third case — both of its drives come out and travel together, each one labelled. The single situation nobody can work around is memory soldered flat onto a mainboard, which is how Apple Silicon Macs and a good many slim laptops are built: if the storage will not unbolt, there is no parcel to send.

  • Use a box or padded mailer with some rigidity to it, and pack around the drive until nothing shifts when the parcel is tilted. Mains adaptors, docks and leads are not wanted at this end.
  • Sending a RAID or a server? Only the member disks travel — not the chassis, not the controller — and each one wants its bay number written on it. Take a photograph of the front of the unit before anything is pulled; it costs nothing and now and again it saves a day.
  • Print the shipping and booking-in form (PDF), add a name, a number you will answer and a sentence on how the trouble began, and drop it in beside the media.
  • Most people use Special Delivery, which is tracked and covered; a courier of your own does the same job. You can also bring it: the Oxford reception takes devices over the counter, Mon–Fri 9:00am–5:30pm. Neither a Coventry counter nor a collection round exists.
// write this on the label

Oxford Data Recovery

John Eccles House
Oxford Science Park
Robert Robinson Avenue
Littlemore, Oxford, OX4 4GP

↓ Print the shipping & booking-in form (PDF)

Put Oxford Data Recovery on the label. From Coventry it is roughly fifty-five miles straight down the M40, about an hour if you would rather drive it in than post it. Either way you are told the moment it is logged, and the free diagnostic finishes two working days later.

Unsure what ought to go in the box? Ring 0800 689 0668 before you seal it, or work through the free online diagnostic and let it do the asking.