Synology NAS Drive failure question

I just wouldn't do *anything*. I have two Synology NAS using SHR. The bad sector count can go up *and down* based on my experience, and doesn't mean the drive is going to die anytime soon. Also, look into the Backblaze C2 service to provide catastrophe insurance for your data. It's shockingly cheap.

When you look at the SMART data associated with the drives in critical status, what are you seeing?

I've primarily used WD Red Pro drives, and they *always* die after the warranty period. Usually, drives in double the capacity are pretty cheap, so I just drill or sledgehammer the old one and recycle it. For my new NAS, I did do a couple of Seagates and a couple of WD in the same storage pool just to create some manufacturing diversity.

Does anyone else feel like drives go bad way more often when they are in a NAS vs. a desktop/laptop/whatever else? Heat/vibration, or just my imagination?


Kind of off-topic but related, I think I will move away from a NAS for my own use. I feel like they're too constraining with the fixed number of drives and matching capacities and whatnot. I *still* have not completed the project I posted about here a long time ago regarding a "hyperconverged" proxmox cluster. The delay is mostly due to my house being older and there isn't really a good place to put a server rack, so I'm shifting gears again and instead of used servers I will put together something like Project TinyMiniMicro. These boxes are pretty cheap (used) and quiet, so I'll see what kind of drives I can put in them (will probably be limited to 2.5" spinners or SSDs) unless I hang some off USB. Anyway, I love the concept of Ceph for storage since I'll be able to add a host+drive(s) at will with arbitrary capacity, basically networked JBOD. Maybe one day I will get it up and running and report back how it performs vs. a NAS.

quote:

---

If you are using desktop drives, 100% using the equivalent of a WD Blue or Black series (for desktop use) will experience increased failure rates.

---

I don't think I have ever put a desktop drive in a NAS. Always NAS drives and failures seem pretty common. Not that I deal with a lot of them, so it could just be due to small sample size or of course my imagination.

quote:

---

It's not completely clear how Project TinyMiniMicro relates to replacing a NAS, to include redundancy. Are you going to buy three of them and connect JBOD via eSATA cable or something?

---

I would run proxmox+ceph. And yes I would have at least 3 mini boxes, with the idea being I can add a 4th, 5th, on up to an arbitrary number of nodes to scale my storage and compute capacity as needed. Ceph recommends 10gbit dedicated to the storage cluster, but I'll see how it goes with gigabit (plenty say it works just fine though performance isn't exceptional). You can set redundancy either via replication or erasure coding, to whatever level of durability you want.

quote:

---

I've built my own domain controller for my house (and HTPCs, etc.), and now I realize my time is worth more than the headache of dealing with unrefined interfaces and recovery options. I don't know how you generate other excuses to stay away from your wife, but there are easier ways. Synology's DSM is a no-brainer, even at extra cost. There's support available, but it's mostly fire and forget. I don't care to deal with minutiae any more than necessary at this point.

---

This project will definitely be pretty hands-on, that's for sure, but I think it's a good trade-off for me. And since I need distributed virtualization as well, I hope to satisfy all of my requirements with a single "hyperconverged" cluster that can grow along with me adding compute and storage resources incrementally. I look at it like yes it will be a "project", but it'll be like a garden for me that I tend to from time to time.

Use only western digital red pro drives that you order directly from western digital. Not from a reseller like Amazon or Newegg. Those drives have a fiber year warranty and a much lower failure rate. The red pro is made for NAS

Tbh I don’t know what most of those numbers mean. I did check the bad sector count again this morning, and yesterday it peaked at around 17k, and this morning it dropped back to 0. I did run another smart quick test and it’s now moved from critical to failing. Unsurprisingly, when I run a “seagate iron wolf“ drive check it comes back as “healthy.” Lol

I hear you on not not doing anything just based on bad sector numbers, as that’s what I had been doing for the last several months with these two drives. They showed bad sector numbers, but were stable at that number for months…until they weren’t. Plus, I started to notice some unusual noises associated with their operation which is telling me that they are/were dying.

There’s been a lot of helpful info in here, but I’m still not sure if it’s best to leave the NAS running with the failing drive, or to remove said drive from the storage pool and Operate in a degraded status until the new drive gets here or to just power the whole thing down?

TBH, the SMART data doesn't look great. The reason I asked is that sometimes the threshold is (for instance) 44, and the recorded data lists the max recorded error rate is 45, but it shows as bad, the same as having a measurement of 500 against a threshold of 45 would.

Keep in mind that shutting the unit down and having the startup process, including drive spin up *may* be more damaging than waiting another day or two to just replace the drive.

A 3 year old drive I have reports zero Raw Read Error rate, zero seek error rate, a power cycle count of 697 (in nearly 3 years), you get the idea.

Plug in the new drive, but plug the NAS into a $50 UPS. A power cycle count at nearly 14000? That will totally kill drives, even if that's over a three or four year period. That is immediate cringe, so whatever is causing that, make it stop.

My power on hours are 21k in my old NAS on the drive I'm looking at, so we're nearly the same, but your drive has a SMART metric for failure that is *massively* higher, to a logarithmic proportion, than what I have. Cross fingers, order 2 day shipping from NewEgg next time.

For actually helpful information, if you bought all of the drives at the same time, they are likely going to be statistically subject to fail relatively closely together.
The serial numbers are probably sequential, or close to it, if you bought them together. Someone sneezed in the clean room that week, who knows. Put the replacement you want for the next drive to fail on your wishlist just in case. Even with the WD Red Pro/Plus drives, if one went, I expected the next one to report as dead with 4-6 weeks.

But with 14000 power cycles on drives that have 20000 active hours? I don't know if your cat is turning off the power strip, but that is B-A-D.

Tripp Lite, APC, CyberPower have been fine from what i've seen in home/small business...I just bought a new Tripp Lite for my rack a month or two ago...

for reference, my synology has 13,610hrs with 20 drive power cycle and 0 read errors