DB-유

Choosing a RAID Type on Synology: SHR vs RAID 5 vs RAID 6

· synology, raid, storage

I've unboxed enough Synology units to know the first question is always the same: which RAID type? It's also the hardest thing to undo. Change your mind later and you're copying everything off, destroying the volume and starting over. DSM 6 puts Synology Hybrid RAID (SHR and SHR-2) next to the classic levels, and the wizard's default isn't always what you want.

Below is how I pick between SHR, RAID 5 and RAID 6 on 4-bay and larger units, and how I create the volume so it's Btrfs from day one. Assumptions: DSM 6.0 or 6.1, at least three drives, and nothing on them yet, because creating a disk group wipes them.

What each level costs you

The difference is capacity versus how many drives may fail before you lose the volume.

  • RAID 5: one drive of parity. With four 4 TB drives you get roughly 12 TB usable. One drive can fail.
  • RAID 6: two drives of parity. Same four drives give roughly 8 TB usable. Two drives can fail.
  • SHR: functionally RAID 5 with one-drive redundancy, but it is built out of several mdadm arrays so that drives of different sizes are all used. With four 4 TB drives it equals RAID 5 exactly.
  • SHR-2: the two-drive redundancy version of SHR, equivalent to RAID 6 with identical drives.

Where SHR wins is mixed sizes. Two 4 TB and two 8 TB drives in RAID 5 give you 12 TB (each 8 TB drive is treated as 4 TB). SHR gives you 16 TB because it builds a second array across the leftover 4 TB on the two larger drives. You can check what a given combination yields with Synology's online RAID calculator before buying drives.

Where RAID 5/6 wins is predictability. SHR volumes cannot be moved to a non-Synology box for recovery without knowing how the nested arrays fit together, and a few high-end RS/FS models do not offer SHR at all in the wizard.

One drive of redundancy or two?

My rule of thumb: up to five drives of 4 TB or less, RAID 5 or SHR is acceptable. Six or more drives, or anything 6 TB and larger, use RAID 6 or SHR-2. Rebuild times on large drives run into days, and the array is unprotected the whole time. Losing a second drive during a rebuild isn't a theoretical event; I've seen it twice on customer hardware, both times with drives bought in the same batch. I now buy drives from mixed batches when I can.

Creating the disk group and volume

In DSM 6 the cleanest way is to create a disk group first, then a volume on it, rather than letting the wizard do both in "Quick" mode.

  1. Open Storage Manager, go to Disk Group and click Create.
  2. Select the drives, then pick the RAID type. The list shows SHR, SHR-2 (only when four or more drives are selected), Basic, JBOD, RAID 0, RAID 1, RAID 5, RAID 6 and RAID 10.
  3. Leave "Perform disk check" enabled for new drives. It takes hours but catches drives with bad sectors before you put data on them.
  4. Go to Volume, click Create, choose Custom, then "Choose an existing Disk Group", and select Btrfs as the file system if the model supports it.

Btrfs matters here because it gives you snapshot replication, shared folder quotas and checksumming, and there is no in-place conversion from ext4.

After creation you can confirm what the system built from the shell:

cat /proc/mdstat
sudo mdadm --detail /dev/md2

On an SHR volume with mixed drives you will see /dev/md2 and /dev/md3 (and possibly more), each a raid5 array; on plain RAID 5 there is a single md2 device.

Did it work?

Storage Manager → Volume should say Normal with the capacity you expected, and Storage Manager → HDD/SSD should list every drive as Normal with S.M.A.R.T. status Normal. The disk check should finish without flagging anything.

I also turn on Data Scrubbing on a schedule if the volume is Btrfs; on RAID 5/6 the scrub verifies parity too.

Things that bit me

  • SHR-2 is greyed out with three drives. It needs a minimum of four. RAID 6 also needs four.
  • Adding a larger drive to an SHR volume does not immediately add capacity. The extra space becomes usable only once at least two drives have the larger size, because the second array needs redundancy too.
  • "The system detected the following disks have bad sectors" during the disk check means exactly that. Return the drive rather than accepting it into the array.
  • If you later want to expand a RAID 5 to RAID 6, Storage Manager offers "Change RAID Type", but only in the direction of more redundancy and only when a spare bay is free. SHR to SHR-2 works the same way. Going back is not possible.

With matched drives in a home or small office box, RAID 5 and SHR are the same thing, and SHR just keeps the door open for mixed drives later. That's why I default to SHR there. Once the drive count or size grows, spend the extra drive on parity.