OK, I am getting Synology to operate daily. With three NASes, it’s a little complicated and the documentation is there, but hard to read, so let’s go topical and talk about some key tasks.
Moving data when Volumes get full
OK, I think this is a typical problem, but the main one is how you want to manage your shares, the folders users see. How should they be allocated to volumes and how should volumes be allocated to the physical storage pools?
The main consideration is that the naive thing to do is that if you have a 12-drive NAS, then you just make it all a single volume (RAID10 please, but we’ve covered that before). The reality though is that for most prosumers, you are probably going to be driving a mishmash of devices and one problem is that over time, you will want to do partial upgrades of systems. That is right now the cheapest drives are 18 TB, but in six months they will be 22 TB).
And let’s face it 12 x $300 is $3,600 so not a trivial amount of money. And many sales limit you to say 5 drives, so getting 12 plus 2 spares is a pretty difficult endeavour. I don’t have a drive set with 12 of the same, I’ve got a bunch of 16TB, 18TB, 10TB, 6TB, and 4TB drives around. The goal is to get as much coherence as possible.
So here is my solution:
- Try to stay on a single storage pool for the whole array. On one system I’ve had up to three different storage pools, but for 8-drive arrays, I stick with a single storage pool. But for my 12-drive arrays, I’m at a pair of 6-drive arrays.
- While I would like to move to a single array as soon as possible, this is going to require a lot of copying. So instead as an intermediate step, I’m going to convert one system to two arrays each with 6x20TB and then 6x18TB and 6x16TB. This is nice because the 6x20TB at 90% utilization can store all the data that I have, so the 6x18TB and 6x16TB drives can be expansion storage.
- And right now, the market is pretty interesting, in that 20TB drives used to be $280 and are now $350 or so at Newegg, but 18TB drives are $330-$70 or so with a coupon so they are the best buy. And of course, the Seagate X22 22TB is much more expensive at $480. So the sweet spot seems to be 18TB drives, but as usual with drives, it’s all about just waiting. While they recommend the Seagate IronWolf which is supposed to be tuned for NAS, I normally just get the EXOS and they seem pretty good. I’ve had one fail on arrival, but Seagate’s Warranty is five years and they replaced it right away (although I had to pay for shipping, it was nice that they were so fast).
Shared Folders vs Volumes vs Storage Pools
The next confusing this is what is the most flexible layout. Synology has three concepts. The first is a storage pool. This is the lowest-level idea and gives you the basic RAID protection. So you can set it for SHR1, SHR2, RAID5, RAID6 and RAID10. SHR which is Synology Hybrid Raid is great if the drives are small and you are not worried about rebuild times. This is something I’ve covered before, but with 10TB and higher drives, you are reading 1e13 bits and the unrecoverable bit error rates are in the 1e14 range for SATA drives. Jim Gray (RIP) wrote a great paper on this back in December 2005 (I can still remember him towering over me telling me how important Terraserver was, and he was right!). Even back then the SCSI rates were 10E15-10E16 (for SAS in the same range) and 10E13-10E15 for SATA.
But as an aside, this is the raw disk error rate maximums and the observed rates are much lower you have to consider the whole system, but to be conservative, I’ve been using RAID10 mainly so that recovery times are lower rather than my more usual RAID6.
The most important thing about RAID10 is that it can recover very fast from a single error because it just has to duplicate the disk that was lost. Of course, it is sometimes a 1-drive fault tolerant system in the worst case (if you lose the two mirror members), but can be as high as N/2 tolerant if you lose only one mirror. Since I don’t want to tempt fate with a rebuild failing to an Unrecoverable bit error, I use this. Note that you can’t add to a RAID10 array, but you can add a new drive to SHR2 at the cost of a long rebuild time.
Now the next decision is whether you just use one volume per Storage Pool. The main reason people don’t is hard for me to figure out. So I normally just have a single volume. Volumes you can grow, but you can’t shrink. It is at the volume level that you provide a file system type. These days btrfs is a good choice. It is a copy-on-write system, so you get snapshots of your previous volume states which is really nice (at the cost of using more storage) and snapshot replication, the equivalent of rsync is very fast and doesn’t require much setup, unlike Goodsync or rsync. It happens automatically in the Synology DSM
Finally, there are shared folders, which are the most flexible, this is at the file system level and they can grow and shrink, so in most of my NAS systems, I’ll have 6x of each type and then move shared folders around to make sure I don’t get over 80% of disk capacity. The 80% for hard disks is needed to prevent lots of fragmentation and slowness.
WOL doesn’t work automatically
One feature that I was disappointed with is the Synology Wakeup-On-LAN. I thought this should work when you get network access, but it turns out you have to send a special wakeup packet to manually wake the thing up.
For instance, the DS Finder application does this, you can also load a MacOS utility called wakeonlan, but it is not very convenient say if you want a media server that turns itself off, but brew install wakeonlan will do the trick assuming you can get to port 9 on the Synology machine.
