OK, I thought that I was nearly done. I had reorganized my machines so that I would have RAID10 partitions for half of each 12-bay Synology RP-2423+ and DS-2413+. But then I started reading and discovered that the new Seagate EXOS drives I’ve been buying at 1^15 unrecoverable bit error (UBE) drives even though they are SATA. There was a change over the last five years and the big drives are not 1E14 anymore.
Also, because I’m now using Synology btrfs instead of ext4, there is a CRC check of the files and copies. Plus, my old friend (Jim Grey RIP) published an old paper pointing out that disk bit errors are a small part of the problem. It is controllers and others that cause it.
Some more reading on Reddit and other sources and the recommendations are pretty simple:
- If you have six drives or less, then you can use Synology Hybrid Raid or RAID5. This gives you a single drive of redundancy which is good enough in modern systems. That plus making sure you have a hot backup and you are in good shape. In the old days, the rebuild times for RAID were so long and because drives are typically bought at the same time, when you get an error like this, you can easily get a second one. If you are worried about proprietary SHR, you good thing you get is that you can put any random collection of drives in and it works. RAID5 requires each drive to be larger than the ones that you currently have.
- If you have more than six, then you can use SHR2 which is two-drive redundancy, but it means compared to my old RAID10 recommendation, you get 75% of the raw capacity instead of 50% which is a big deal for prosumers.
- For twelve drive systems, you can go with SHR2 and get a whopping (10-2)/12 capacity. However, this means that with 12 drives, you can’t just take down part of the array and do things. You can take two of the drives out and run “naked” if you want but that’s a little scary. RAID10 will be very fast with all those mirrored pairs, but you give 50% of your raw capacity so great if you need performance, and RAID6 or SHR2 if you want more capacity. Rebuild times will be long, which is the main thing.
- The alternative for 12-drives is to have two 6-drive storage pools of six drives and run these as SHR so you get the same 10/12 efficiency but when you take a hit, you have only a single drive redundancy. Also if the drives are the same it’s the same as RAID6. but if you split the array into 6 drives each, then you can remove “half” an array and format things. I’m kind of torn right now what to do because I have 6×20 and 6x16TB in my old array and 6×20 and 6x18TB in my new array and it seems very neat to have a total of four RAID5 storage pools with them. But I might just do SHR2 on each and then I can swap around “sub-arrays” very easily.
I’m actually kind of torn for the 12 drive systems, these are much closer to enterprise systems where having a simple process matters. The problem is that buying 12 drives at once is a lot of money, so I’ve been cherry-picking 6 or so each time I buy and that seems to work. I’m thinking that for ordinary users, SHR2 is a good choice. And if you don’t have a very full NAS then RAID10 is good if you have the dough for lots of matched drives.
TL;dr 2 or 4-bay SHR, 8-bay SHR2, 12-bay SHR2 RAID10 or 2 arrays of SHR
For 8-bay or 4-bay, I would just use SHR as failures are going to be low and you are more likely to just chuck in some random drives.
For 12-bay systems, there are a range of choice depending on use case:
- For business uses. I’d go with RAID10 not for performance but because rebuild times are fast and you care less about money.
- For prosumers who don’t want to think so hard, go to SHR2 and throw any random drives you want
- For geek prosumers (eg me!), go to SHR for six drives in two sets.
Also, a consideration is that with SHR, you are stuck with Synology. There is some appeal to just using standard RAID5 and RAID6 so that if you have a problem, others can read your drives.
How to move things around with the Control Panel Shared Folders Location setting
OK, when you make these decisions and want to move things around inside a NAS, I’ve been going to the File Station and creating two shares, new and old, and then doing a copy. Turns out that you don’t have to do this. Inside Control Panel > Shared Folder, you can edit a Shared Folder and change the Location, just to move it to a new volume and it blasts away at 110MBps or so moving things. The disadvantage of course is that during the time it does this your shares are down, but it sure is convenient.
I’m probably going to use the Shared Folder and change the location technique if the NAS is not up all the time. You don’t have to deal with multiple share names, you can just go through all your old shares and move them to a new volume. Then when done reformat delete the old volume and the storage pool and start again.
If I need constant reliability, then you can either do a Synology Snapshot replication to get your slave going (particularly if it is on a high-speed LAN) and that works for me at 30-50MBps. Or, you can do it on the local NAS as well and you get full uptime.
