SQL Server marked the database suspect. Exchange dismounted the store and won’t discuss it. The Access file at the heart of a thirty-year business opens to gibberish. Database recovery is its own discipline — pages, logs and consistency — and it rewards one behaviour above all: stopping before the ‘repair’ that makes it worse.
$ edr connecting…
Databases die two ways. Under them: the drive or array holding the MDF, EDB or DBF files fails — a storage recovery first, followed by database validation, because a database recovered 99.9% intact can still refuse to mount on the missing 0.1%. Inside them: corruption in the page structures themselves — torn pages after power loss, checksum failures, indexes pointing nowhere — where the files are perfectly readable and perfectly broken. The bench handles both: page-level repair and record extraction for structural damage, and transaction-log work (redo, replay, or rescue of the log itself) where the LDF holds the missing history.
The most destructive tools in database recovery ship with the database: repair commands that ‘fix’ consistency by discarding whatever fails validation — SQL Server’s repair_allow_data_loss says it plainly in the name — run by administrators under pressure, on the only copy. House rule for you and for us alike: nothing runs against original files, ever. Take the server offline if you can, copy the database and log files somewhere untouched if you safely can, and bring the story: what version, what happened, what the last known-good backup is. Recovery then works on bench copies, keeps every salvageable record, and documents what — if anything — was genuinely lost.
Not yet — the aggressive repair modes purge what they can't validate, permanently, and they're the single biggest cause of avoidable database loss we see. Copy the MDF and LDF files first, or image the volume; then repair decisions can be made on copies where mistakes cost nothing.
Usually — dirty-shutdown states and EDB corruption respond to log replay and store-level repair on copies, and individual mailboxes can be extracted even when the whole store won't remount. Preserve the EDB and every log file; the logs are frequently the recovery.
No — legacy formats are familiar ground here, and Access corruption is often shallower than it looks: damaged headers or indexes over intact records. Tables and data extract to a clean file or a modern format of your choice. Send the file and its backups, however old.
A free diagnostic first, always — then a fixed written quote before any work begins. No surprises on the invoice, because there's no invoice until you've approved the number.
Getting a device onto this bench only takes two moves, and the clock on your free look starts the moment it lands.
Strip out cables, caddies and power bricks — none of it helps — wrap the device so it can't shift in transit, and either post it insured to 83 Princes Street, Edinburgh EH2 2ER or bring it in yourself. Tuck a note of what happened in with it, and enclose the shipping form if you've printed one.
Arrival gets confirmed, the diagnostic runs at no cost to you, and the call that follows names the fault and states a fixed price. The bench stays idle on your device until you say go.
Free diagnostic first. The underlying storage is imaged before any database-level repair is attempted, so a bad repair attempt never becomes the second disaster.