Mirroring provides a copy of every piece of information, hence RAID 1 recovery is typically the easiest form of RAID recovery. However, rebuilding the RAID incorrectly can destroy the mirroring advantage and make the process significantly harder than it should be.
In general, when one member cannot read a particular sector, the other can. The goal is to create a single complete image of both members and then build up a complete file system from that. It is not to rebuild the RAID onto a disk that could also be failing.
In a RAID 1 system data is written simultaneously to two or more hard drives (or storage devices). Because of this, if a drive in the array fails the array will continue operating from the remaining good drive(s) making RAID 1 the easiest and most reliable level to implement. RAID 1 does provide protection against loss of data due to a single hard drive failure; however, it does not provide protection against the simultaneous corruption of data on all member drives in the array as a result of actions such as deletions, ransomware attacks, corrupted files or errors generated by a faulty controller. Both member drives would be corrupted equally. A mirror represents redundant data and not a backup. It may seem like semantics but it is very important on the day you need it.
One of the greatest surprises with RAID 1 systems is how quickly both drives can fail at almost exactly the same time. This happens because both disks are typically the same make and model and were purchased from the same batch of products. They were also likely installed on the same date, began their operational life handling virtually the exact same type and quantity of read/write operations, etc. So, when one disk begins developing bad sectors, the other disk is often just around the corner. Therefore, a mirror could appear to be functioning properly for many years and then within a matter of weeks both disks in the array begin showing signs of failure. The second failure mechanism in mirrors is completely quiet. A disk simply goes offline, no one notices because the volume still appears to mount correctly and the array continues to operate without any redundancy protections until the next disk fails.
Every member is cloned as read-only onto failure-tolerant hardware; if one or both of the original disks cannot be read (i.e., have unreadable sectors), each attempt is logged rather than causing the cloning process to stop. The two cloned disks are then compared sector by sector. For any given sector, either disk can be clean enough to be used as the final version. Since two identical disks will very seldom fail in the exact same location(s), it’s likely that the final usable disk will be complete even though neither original disk is. After the disks have been compared to get the best possible usable disk(s), the file system is rebuilt from it, while no changes were ever made to the original disks.
When a mirror is managed by a controller (or Windows dynamic disks) the metadata offset is factored into where the volume starts reading from; this feature has quietly defeated many DIY efforts.
First, do not force a rebuild when either side is suspect; forcing a rebuild on a suspect mirror will be the single largest reason why a recoverable mirror now cannot be recovered. Second, do not reinsert a disk that the controller has already ejected; if it resyncs in the wrong direction it will write over the good copy with the bad one. Third, do not run repair utilities such as chkdsk against the array while it is in degraded state because any writes to a mirror are writes to everything you have left. If data matters more than uptime then power down and image.
First, image both of the disks “read-only.” Then, create an image of what the combined disks would look like if they were working together. In other words, whenever there was a point where the first disk could not read data in a particular sector, the second disk would be able to supply it. Once this new merged image is created, the file system can be rebuilt from this new single image rather than trying to rebuild it from either of the two disks separately. This method leaves both original disks intact and unaltered, which gives you no restrictions on your options down the road.
Generally, if the surviving drive is genuinely healthy, it’s perfectly fine to do a rebuild on a brand new drive. The issue comes in when both drives are also degraded; rebuilding will read all data off the good drive (survivor) top to bottom, so if that drive fails at some point during the rebuild, you’re left with two bad drives. If either drive shows bad sectors or SMART warnings, always clone/ image both before starting the rebuild.
The array will be operational, however it can no longer rely on mirroring (RAID) for protection. In this case, the original mirror set is now “degraded”. Therefore, if there are errors on the last disk left in the array that was once part of the mirror set, you have lost all your information. Therefore, please copy any critical information off to another piece of media prior to replacing any drives. Once replacement is complete, do not rely on the array until the rebuild process completes without error.
Diagnostic cost: free. Cost for RAID / NAS / server recovery: from £500 plus VAT. Cost will be confirmed by way of a written quote after completion of a free 48-hour diagnostic. Where a drive member requires physical drive level repair, 50% deposit required upfront. Remaining balance due only upon successful recovery of data.
All disks, clearly labelled as to which bay they originally came from, and note the controller type or whether it was a software mirror. Do not rebuild. Do not re-insert ejected disks. Do not initialise anything. If it is in an enclosure, the disks need to be removed.