RAID 5 failed with two drives down — what now?

RAID 5 stripes data and parity. One missing member can be calculated from the others. When that member fails, the remaining disks are already the oldest, hottest, or from the same batch. A rebuild reads every block on those remaining disks to write a new spare. If a second disk is weak, the rebuild is the load that takes it out. You then have two holes in the stripe and a controller that is still writing.

That is why “the NAS said rebuild, so we clicked OK” is the most expensive sentence in this lab. Read the RAID recovery page before you let Synology, QNAP, or a hardware card finish the job.

Power the NAS or RAID card down. Do not initialize disks in Windows. Do not run vendor “repair.” Photograph the slot order. Bring every disk to Columbia or schedule Tysons. Call (410) 937-7332.

We clone each member, including the ones marked failed, then reconstruct RAID 5 (and 6, 10, 0, 1) in software. Evaluation is free. Fixed quote before paid work. No data, no charge on the recovery fee if we cannot get the files you need.

RAID 6 can lose two disks by design. People still start rebuilds that kill a third. Same rule: clone first. If you already started a rebuild, stop it and send the members anyway. Partial writes are often still recoverable from the clones.

VMDK and VHDX files come back after the RAID is reconstructed, not by installing ESXi on the broken members. Tell us if the array held VMs so we verify those files, not just a mountable NTFS volume.

RAID 5 is designed for one missing member. Two down means you need readable sectors from both “failed” disks, or you have holes in every stripe. RAID 6 is designed for two missing members — until a rebuild writes a third. RAID 10 can survive two failures only if they are in different mirrors. Two failures in the same mirror is RAID 0. People mix these up because the NAS UI said “degraded” for all of them.

Send every member regardless of level. We will tell you which layout you actually had after we see the clones. Do not buy a new NAS and initialize the old disks in it while you wait.

SHR and RAID 5 on Synology still stripe data and parity. A “repair” in DSM is a rebuild. QNAP’s storage manager does the same. An LSI or PERC card that says “foreign configuration” is asking you to write metadata. Stop. Photograph the slots. Bring the disks to Columbia. Read RAID recovery before you click OK.

Evaluation is free. If we cannot reconstruct the files you named from the clones, you do not pay a recovery fee. Shipping and return media are separate. A two-disk RAID 1 is not priced like a six-disk RAID 6 with a half-finished rebuild. You get a fixed quote before paid work. Call (410) 937-7332.

Power down. Photograph the bays. Pull the members only if you can do it without forcing trays. Pad each disk. Write the RAID level if you know it, and whether a rebuild already started. Drop at Columbia or schedule Tysons. Do not buy a replacement NAS and initialize the old disks in it while you wait.

If the array held VMs, say so on the evaluation form. We check VMDK and VHDX after the RAID is assembled, not by installing a hypervisor on the broken members.