Disk chaos with a happy ending
Two pulled hard drives, an almost lost 450 TB RAID, and a last-minute rescue.
What sounds like a nightmare for any administrator ended in success thanks to Starline's experience, Areca's precision, and a touch of luck. A lesson in what true expertise can achieve—even 10 years after the warranty has expired.
The starting point
Our customer from the Vienna area had a customer who lost hundreds of terabytes of video content in one fell swoop. A case for Chief Technology Officer Konrad Beyer. But first things first.
The patient was a RAID storage system from RAIDdeluxe, which was purchased for the customer in question from Starline in 2014. The robust 64-bay RAID system RDL-BS64S6D was equipped with 32 hard drives, each with a capacity of 22 TB, and had been running reliably until the following incident.
And then something happened that could be reconstructed retrospectively as follows:
The incident
One day in 2025, two disks were accidentally removed from the RDL-BS64S6D during operation and reinserted shortly thereafter. It was mistakenly assumed that these were the two replacement disks (hot spares). In fact, one of these disks was part of the active RAID array, as it had previously been integrated unnoticed during an automatic repair.
After reinsertion, the system initially displayed “degraded” (i.e., damaged but still readable), but did not start a rebuild. After a restart, the controller recognized all hard disks, but no longer assigned them to the RAID, instead displaying them as “free.” As a result, it only recognized three disks as belonging to the RAID, while all others were considered “missing.”
Several attempts to reconstruct the original RAID by reconnecting or restarting only resulted in the controller displaying several incomplete partial RAIDs with the same name. The entire array could no longer be activated. Hundreds of terabytes were at stake.
Although there was an LTO backup of the entire data set, restoring the 450 TB from numerous tapes would have taken weeks and incurred incredibly high costs. This is where Starline, in the person of Konrad Beyer, came into play: The goal now was to restore the original RAID set without losing any data.
The solution approach
In a brief analysis, our head of technology estimated that professional data recovery would cost in the mid five-figure range.
However, he also thought it possible to attempt recovery ourselves and promptly provided step-by-step instructions on the same day. After all, Konrad Beyer knows the Areca controller built into RAIDdeluxe like the back of his hand.
First, however, he warned explicitly that any further changes—especially restarts, reinitializations, or repeatedly plugging and unplugging disks—would make the situation worse. This was because the RAID metadata on the disks had already been severely disrupted by the interventions carried out up to that point. This manifested itself in the controller displaying several “RAID sets” with the same name.
The instructions
Konrad Beyer recommended the following procedure to restore the RAID to working order, which was to be carried out in exactly this chronological sequence.
- Switch off the system completely.
- All 30 disks belonging to the original RAID had to be plugged in, as documented in the original sequence. This meant that the two accidentally removed (“faulty”) disks also had to be included.
- The hot spare disk, on which an automatic rebuild had started, should not be inserted.
- The system had to be started and waited for until it was fully booted up. In the controller menu, all existing RAID sets with the same name should be deleted.
- Then a new RAID set had to be created with exactly these 30 disks.
- And here comes the trick: The IT specialist should create a new volume – but without initialization (“No Init”) and with the same settings as the previously existing RAID set (RAID6, stripe size, block size).
- Then shut down the system again and remove the two faulty disks.
- Restart the system. The RAID should now be recognized as “degraded” with 28 active disks.
- If the volume is not activated automatically, it can be activated manually using the “Activate Incomplete RAIDset” function.
- Finally, connect the volume to the host in read-only mode and check whether the data is visible and consistent.
Engineer Beyer concluded by pointing out that this procedure should only be carried out as a last resort. Any wrong step—especially initialization or incorrect sequence of the disks—would irretrievably delete the data.
Our customer decided to follow the expert's recommendation and use the “NoInit” method.
One attempt
After inserting the disks in their original positions, the RAID was set up without initialization. After startup, the host immediately reported the volume correctly as an active NTFS drive: complete and error-free. A checksum test confirmed that all files were intact. The system was now running stably again, and the planned migration could continue.
Our conclusion
The entire process is a remarkable example of competent, dedicated, and practical technical support—far beyond the usual standard. Despite an extremely complex initial problem, our support team was able to perform a precise analysis of the RAID situation and develop a clearly structured recovery plan.
It is worth noting the cost of this support, which was provided several years after the warranty had expired: $0! Email support for regular customers is free of charge. A matter of honor for Starline, so to speak.
