We already wrote about a RAID 0 stripe when one drive dies, about a RAID 5 that loses a second drive, about the Rebuild button in general, about Force Online after the RAID battery dies, about a Synology volume that has already crashed, and about an Unraid or TrueNAS pool that will not mount. This page is the mirror that stayed up. One member failed. The array kept serving. The warning sat there for weeks or months. Then the second failure, or the rebuild, spent the only copy.
What a RAID 1 mirror is
RAID 1 keeps the same volume on two disks. Sometimes on three. Usable space is one member, not the sum. Read one disk and you have the files. That is the whole protection. There is no parity block somewhere else to recompute a missing sector from. The other disk is the copy.
The usual places that mirror shows up:
- A two-bay NAS or enclosure set to RAID 1. Synology and QNAP both offer it. A WD My Book Duo is the two-disk mirror called out on the RAID service page. The share stays online when one bay goes red.
- Motherboard RAID and Intel RST. A mirror pair in the BIOS, or in the Rapid Storage app, on two SATA drives or two NVMes. The RST status line says degraded. The volume letter is still there.
- Windows. A mirrored dynamic disk, or a Storage Spaces two-way mirror. Spaces says the virtual disk is degraded and still attached. Simple (no resiliency) is a stripe, which is the RAID 0 post. Two-way mirror is this one.
- A hardware RAID card — PERC, Smart Array, MegaRAID, Adaptec — with a two-disk RAID 1 virtual disk. The card beeps. One LED is amber. The operating system keeps writing.
- One pair inside RAID 10. RAID 10 is mirrored pairs, then a stripe across those pairs. Losing both disks of one pair is this failure inside the stripe. The other pair being healthy does not rebuild the pair you lost.
A two-member mirror can lose one disk and keep running. That state has a name: degraded. The controller has failed a member, marked it offline, or stopped reading it. The survivor answers every request. From the desk, the only change is a beep, a light, an email, or a banner. The files open, so the banner gets closed.
A three-way mirror can lose two disks and still have a member. The same page applies the moment you are down to the last disk. Fault tolerance left is zero. The next failure is the volume.
The first failure is the window
Degraded is not a suggestion to buy a disk eventually. It is the controller telling you the second copy is already gone from the set. Whatever that failed disk still holds is frozen at the hour it dropped. Every write after that — the spreadsheet saved on Thursday, the photos copied Friday, the database that checkpointed overnight — lands on the survivor alone. The two disks stop being copies of each other the day the first one falls out, and they drift further every day you wait.
The warning is one of these, and people treat all of them as noise:
- A controller beep that starts at boot, or a NAS that chirps until someone unplugs the buzzer. That alarm is the card or the NAS. A hard drive that beeps because the spindle will not come up is a different failure — the motor, covered under drive beeping. If the survivor itself is making that sound, power the set down.
- A bay LED that goes amber or red and stays there. The other bay stays green. The green bay is not a spare. It is the disk doing all of the work.
- An email from DSM, QTS, a RAID management agent, or the server. It lands in a folder. The folder is the log of the day the window opened.
- A Degraded banner: Synology Storage Manager, QNAP Storage & Snapshots, Windows Storage Spaces, Intel RST, or the RAID card's BIOS. The word means one member is out and the volume is still mounted. It does not mean "healthy enough to ignore."
The share stays up, so the first failure sits there for weeks or months. During that time the survivor takes every read and every write. It is often the same model, bought the same day, with the same hours. The disk that failed first is evidence about the one you have left, not a sign that the remaining disk is the good one.
While the volume is still mounted and that disk is still quiet, the files can be copied to a different physical disk. That copy is the thing the mirror never was. A mirror tolerates one dead member. It is not a backup you can restore after you have built a new array on top of these drives. The rebuild page already says redundancy is not a backup. This is what that looks like on a two-disk set.
A degraded RAID 1 that has been left that way is one disk away from offline. If the survivor is clicking or a rebuild is already grinding, power it down and get the set evaluated. Free evaluation, no data, no fee.
What not to click
The buttons show up when you finally open the screen, or when the volume drops and the firmware offers a way to "get it back." Each one writes. On a mirror that has been degraded for weeks, the write lands on the only current copy, or it throws that copy out in favor of the stale disk.
Rebuild, Repair, or Resync onto a replacement, while the source is the disk you have been ignoring. A RAID 1 rebuild reads the survivor from end to end and writes a new member. That is ordinary maintenance when the survivor is healthy and the failure is today's news — the condition in the rebuild vs. recovery post. It is a different job after a month of degraded running. You do not have a health confirmation. You have a disk that has been the entire array. If it has bad sectors, the mirror has nothing to reconstruct them from. The controller can abort, skip the sector, or write a block that was never the file. The read itself is what finishes a marginal drive. Synology's Repair on a volume that is already Crashed is that other page. Repair on a mirror that merely says Degraded, after the warning sat for weeks, is this one.
Force Online, Bring Online, or Mark as Good on the member that failed first. That disk missed every write since it dropped. Forcing it back tells the card to trust a stale member. The resync can run the wrong direction and overwrite the survivor with the older copy. This is not the Force Online in the battery post. That screen is a power cut, a failed or missing BBU, and write-back cache the controller may still be holding. This screen is a mirror member that has been failed for weeks while the volume kept serving. Do not treat them as the same button.
Clear Configuration, Clear Foreign, Delete Array, or Intel RST "Reset disks to non-RAID." Those drop the metadata that records which disk is current and that the two disks were a mirror. The file bytes may still sit on the platters. The controller no longer has a volume to hand you, and the next screen is Initialize or a new array. Clearing the configuration does not extract the files.
Initialize. On a RAID card, Initialize (fast or full) creates a new virtual disk and writes it. Full init is a wipe. Fast init still writes the map a later rebuild trusts. In Windows, Initialize Disk is the GPT or MBR dialog on a member you pulled and attached to a PC. People do that to "see if the good drive is OK," confirm the dialog, and slide the disk back into the bay. The mirror metadata at the start of the disk is what they just replaced.
Delete Volume, then a new volume. Delete Volume on the card, in Disk Management, or in DSM/QTS drops the volume the share was using. Creating a new one formats. Storage Spaces Delete Pool is the same family on a mirror pool, and it has its own page: don't delete the pool. A degraded two-way mirror that still mounts is not that failed-pool banner. Do not delete it so the warning goes away.
"Recreate the mirror and restore from backup" when the backup is the mirror, or when there is no backup. The wizard will build a new RAID 1 on these disks or on a fresh pair you do not have. Initialize runs. Restore does not, because the files were never anywhere else. If a real backup exists on a different disk, the recreate belongs on empty disks after that backup has been checked — not on the members that still hold the only copy.
chkdsk /f, Error Checking, or a filesystem scrub to clear a degraded banner. Degraded is a member, not a dirty NTFS bit. Those tools write the filesystem on the survivor. What /f does is the CHKDSK post. Run it here and you have a filesystem repair on top of a one-disk mirror.
Why a rebuild after weeks is worse
A rebuild on the day a disk fails, with a quiet survivor and a copy of the files already sitting on something else, is the maintenance job the controller was built for. A rebuild after the degraded state was ignored is a different risk, for mechanical reasons.
The survivor is the source, and it is tired. It has taken the full load since the partner dropped. Pending sectors and slow reads show up under a sequential read of the whole surface, which is what a rebuild is. On RAID 5 a failed read during rebuild can sometimes be the parity problem described in the two-drives-down post. On RAID 1 there is no other stripe to recompute from. The sector the survivor cannot return is a hole, unless the disk that failed weeks ago still has an older copy of that block. The controller will not quietly go read the failed member for you. It will hammer the survivor.
The two disks disagree, and the wrong one can win. Force the old member online, or let a resync start without knowing which disk the card thinks is current, and the stale copy can overwrite the weeks of writes that exist only on the survivor. That direction is not undoable from a button labeled Rebuild again.
The second disk dies in the middle. This is the sentence people arrive with. The rebuild had been running for hours or days. The survivor clicked, dropped offline, or the volume went read-only and then gone. The new disk holds a prefix of the volume. The survivor holds the current copy in whatever sectors it can still give up. The original failed disk holds the mirror as of the day it dropped out. Three partial stories. Letting the controller "finish" after the source has left is how it writes a guess across the new disk.
Same batch, same hours. Drives bought together fail close together. The first failure was the schedule. Waiting until the second one follows is how a redundant pair becomes two bad disks and no current copy. Large disks make the rebuild long enough that the survivor has time to fail during the read that was supposed to save it.
This screen is not the other RAID pages
RAID 0, one drive dies is a stripe. Nothing was mirrored. The surviving SSD holds fragments, and a rebuild does not have a second copy to read. If your pair was a mirror and one disk failed while the share stayed up, you are on this page. If the volume died the moment one stick died, and the firmware says stripe or RAID 0, you are on that one.
RAID 5 with two drives down is the parity version of an ignored first failure. Three or more disks, one parity block, the array runs degraded after the first loss, and the second loss takes it offline. Read that page when the set is RAID 5 or SHR-1. The "do not rebuild on the originals" rule is the same. The reconstruction is not. A mirror does not have parity to reassemble, and a RAID 5 does not have a full second copy of every file on one disk.
Rebuild vs. recovery is the general button: a rebuild writes, a lab images and reconstructs from the copies. This page is the mirror-shaped way that button gets pressed too late. Use the other post for the definition. Use this one when the first member failed, the volume stayed online, and the warning was ignored until the survivor or the rebuild became the problem.
The RAID battery is a power cut. The controller reports a failed, missing, or learning BBU, and Force Online can commit or drop dirty cache. If your screen is that battery line, stop and use that post. If the array has been degraded for weeks because a disk died and nobody replaced it, the cache is not the story. Forcing the old disk online is.
Synology Volume Crashed means the volume is already offline. Repair is a write in that state. That post also says a still-mounted Degraded volume can be a normal replace-and-rebuild when every remaining disk is healthy and you are following Synology's procedure for one known-dead member. This page does not overrule that. It is the RAID 1 case where Degraded was left until nobody can call the survivor healthy, or until the second disk is already failing. The NVMe write-back cache post is a different device again: an M.2 cache SSD, where a two-SSD cache can itself be a small RAID 1. That crash is the cache page. Do not pull the cache module and follow this article, and do not Repair a crashed volume because a mirror article said the share was still up.
QNAP Volume Not Active is the offline banner. Recover is the write. That post's Degraded line — share still up, one known-dead member, the rest healthy — is their maintenance case, including the warning that Degraded — Readonly can go Not Active if you replace the disk blindly. This page is that Degraded mirror after the email was ignored. Unraid and TrueNAS are the unmountable import. Unraid's main array is not a RAID 1 mirror. A TrueNAS mirror that will not import is that post. A mirror that is imported, degraded, and still shared, with the first failure weeks behind you, is this one.
Storage Spaces covers Delete Pool, Reset, and Repair-VirtualDisk when the pool is already Failed. It also says a degraded virtual disk that still attaches can be a normal replace-and-repair when the remaining disks are healthy. This page is the two-way mirror left on that Degraded banner until the last disk is the problem. Repair-VirtualDisk then rebuilds from the survivor. Do not delete the pool because the banner is annoying.
What to do right now
If the volume is online, one member is failed, the survivor is quiet, and you are looking at the warning today: copy the files to a disk that is not in the set before anyone rebuilds. If the copy starts erroring, or the disk clicks, stop and power down. If the degraded state is already weeks old, if you have no other copy, if the volume has gone offline, or if a rebuild is running and the source is struggling: do not start or finish a rebuild to find out.
- Stop. Cancel the rebuild prompt. If a rebuild, a resync, or an initialize is already running and a disk is clicking, grinding, or dropping out, power the machine down. Finishing the pass is more writes. If the volume is up and the disks are quiet, stop clicking. Do not start a rebuild as a test.
- Photograph the screen before the next reboot changes the words. Degraded versus offline versus failed, which bay, RAID 1 versus RAID 10, the time the first disk dropped if the log still shows it, and the exact button the firmware is offering. DSM, QTS, Storage Spaces, Intel RST, or the RAID BIOS — whichever you have. Screenshots help. They are not a diagnosis.
- Label every member before anything is unplugged. Bay, slot, or port. Keep the disk that failed weeks ago, the survivor, and any spare you already inserted for the rebuild. Member identity is which copy is current. A pile of unlabeled disks is how the stale one gets forced online later.
- Write down what you remember: when the beep or the email started, whether anyone muted it, whether anyone already rebuilt, forced a disk online, cleared the configuration, initialized, deleted the volume, or recreated the mirror. And whether a backup exists on media that is not this array. "The NAS is the backup" means it does not.
- Leave the originals alone. Do not save recovered files back onto a member. Do not format the survivor "so I can at least use the good drive." Do not move a member to a USB dock so Disk Management can initialize it. Copying off is reasonable only while the volume is mounted and the disk stays quiet. If it is offline, clicking, or mid-rebuild, that is an evaluation, not a copy.
Pack loose disks the way you would any failed drive — how to ship a failed drive safely. If a member is clicking, start with power it off. A server or NAS whose disks you cannot pull cleanly can come in as a chassis. Do not open a drive on a table to see which one failed.
How a lab handles it
We do not rebuild the mirror on your controller to see if the share returns. We do not Force Online the stale member, clear the configuration, initialize, delete the volume, or create a new RAID 1 on the same disks.
Each member is imaged write-blocked first. That includes the survivor, the disk that failed weeks ago, and any replacement a rebuild had already started writing. A drive that only reads in pieces is still imaged as far as it will go. Imaging uses the lab tools already named on this site, including the ACE Lab PC-3000 and DeepSpar Disk Imager. Work happens on the copies.
On those images the question is which member is current, where the two copies diverged after the first dropout, and whether a partial rebuild wrote a prefix onto the new disk. A sector the survivor can no longer read may still exist on the disk that dropped out, as whatever was on that block before it left the mirror. That is an older sector, not a current copy of the volume. Deciding which blocks to keep is work on the images. Forcing the old disk online so the controller can "fill in" the holes is how the originals get overwritten. A wrong guess on a copy can be thrown away. The same guess on the disks cannot.
What imaging cannot do: put back sectors a resync already overwrote, a configuration that was cleared and then initialized, or a new mirror that was created across these disks. We will tell you if the current copy is not there.
The public case log has no RAID 1 job, and no case where a first mirror member was left failed until the second disk died. We will not invent one.
The closest published array work is parity, and one of those write-ups is the "left degraded for months" fact people are actually living. 21607: twelve HPE 1.8 TB SAS drives on one SmartArray, a five-drive RAID 5 and a seven-drive RAID 6. Three RAID 6 members had failed, and some had quietly fallen out months earlier without being replaced. Nothing was rebuilt on the original controller. Every member was imaged, the stale member was identified from the images, and the arrays were reconstructed virtually. Full recovery on that set. It is RAID 6 and RAID 5, not a two-disk mirror. DDR-2026-0003: a Synology four-drive RAID 5, two members failed, nothing rebuilt in the NAS, each drive imaged, the array reconstructed from the images. Full recovery. RAID 5 tolerates one failure. That set had lost two, so it was offline rather than degraded. A third published array, DDR-2026-0014, is a two-SSD RAID 0 stripe. No mirror member to ignore. Those outcomes belong to those drives. They are not a rate, and they are not a prediction about the degraded pair on your bench. We don't advertise a success rate.
If the firmware says stripe, start with RAID 0. If it says RAID 5 and a second disk is already down, start with two drives down. If the button is Rebuild and you are not sure the rest of the set is healthy, start with rebuild vs. recovery. For the work itself, see RAID data recovery.
RAID 1 ignored first failure FAQ
The share still opens. Can I leave a degraded RAID 1 until I have a spare?
Does Degraded mean files are already missing?
I bought a replacement. Should I start the rebuild today?
What do Force Online, Clear Configuration, Initialize, and Delete Volume actually do?
Someone said to recreate the mirror and restore from backup. The backup is the NAS.
The second drive died while the rebuild was running. Let it finish?
Is a Storage Spaces two-way mirror, an Intel RST mirror, or a two-bay NAS the same screen?
How is this different from RAID 0, a RAID 5 with two disks down, or a RAID battery Force Online?
Is there a RAID 1 ignored-first-failure case in the public log?
I already rebuilt, forced a disk online, or initialized. Is it too late?
A RAID 1 that is still serving files after one disk failed is not fine. The degraded warning is the window. Stop, and start with a free evaluation before anyone rebuilds the only copy.
Request free evaluation →Free evaluation · No data, no fee · Talk directly with a technician.