I think it’s fair to say it can be used as a sort of compass. It’s not super accurate and won’t be a guarantee, but it’ll point you in the right general direction.
More importantly, I think we can safely use it to advocate against buying drives with a high outlier failure rate.
It can’t tell us which drives are the best, but I think it can tell us which are the worst.
Although, in their numbers, an outlier failure could simply be a bad batch.
I think it’s safe to say that the 12+TB disks are getting higher failure rates because of complexity and sensitivity.
Either way, the average failure rate is 1.5% on their Q1’19 report. That’s not bad at all.
Well thats where I disagree that I would use their data as a compass because my experience with HGST has proven to me that even if you get the one that is the ‘best’ you still have a chance of failure… so prepare like thats going to happen instead of picking the drives you think it wont happen for.
IMO, The data gives a false sense of chance. It makes you think that you might only have X% chance of failure because thats what backblaze had and really thats not the way I would be thinking about it at all.
Its interesting data but we cant necessarily draw a conclusion to our own use case from it because we’re not a huge company with probably insane levels of redundancy because its the companies ass if it all goes to shit. I think this is why shucking of easystores got so big. Its the cheapest price per TB for a lot of people and that means instead of focusing on having the arguably lower failure rate disks in your system, you can just have more disks at your disposal. They’re all going to fail, its just a matter of when. Does BB even have that kind of data or do they just cycle them out after 3-5 years for good measure?
Maybe im not communicating the thought effectively though. I think no matter which direction you go or what you buy… your 321 is a great rule to go by.
I’m saying prepare like it’ll happen, but try to pick drives with a low failure rate to begin with. I think we just disagree on if you can accurately predict failure rate.
So the issue is that that data is “annualized failure rate”
I’d like to see the total failure rate after 4 years of operation on a model.
@Adubs you also made an interesting point – all this time I’ve kept my rack in another room in my apartment with the AC off and basically ‘air cooled’ by whatever ambient temperature. Humidity here is definitely on the high(er) side.
Since installing the 8x10TB drives, I’ve been running the AC 24/7 on that small room. In any case though, I do like @SgtAwesomesauce’s 321 recommendation and “designing/planning” for failure makes a whole lot of sense.
During the last year, I do admit, I grew complacent in my maintenance process – and any data I’ve lost as a result is on me. Thankfully, nothing critical though.
Whenever I create a new dataset in my pool, I have to also setup an accompanying rsync task – which I missed for a couple datasets. At least I can say “Oh well” and move on.
Thankfully the data synced to the Synology box saved me, but it came really close. Way too close.
@SgtAwesomesauce You really weren’t kidding when you said my recovery was time sensitive…
Dear user,
Disk 3 on DS1815+ has degraded severely and is in failing status. Please make sure that your data has been backed up and then replace this disk.
Additional disk information:
Brand: WDC
Model: WD4001FFSX-68JNUN0
Capacity: 3.6 TB
Serial number: ****
Firmware: 81.00A81
S.M.A.R.T. status: Failing
IronWolf Health Management: -
Bad sector count: 0
Disk reconnection count: 0
Disk re-identification count: 0
SSD estimated lifespan: -
Please log into to DiskStation1815 for more information.
All done, although I have a fair bit of data on another Synology that I need to rsync over; will start on that shortly.
In any case, looks like the ‘bathtub curve’ is kicking in for my previous batch of drives – so pretty much anything on both Synology Units (1515/5 bay) or (1815/8 bay) are ready to give up the ghost. They also have some WD RMA refurbs inside.
Tad off topic – I tend to order my drives from Amazon or bhphoto, and when I register the retail drives they get classed as ‘Out of Region’. I just saw their system reject my previous RMA request with a warning saying I can only make 5 Out of Region RMA claims. So that’s ok, if I throw away 3-drives, so be it.
However, this means, every year moving forwards, I can claim on 5-drives. I reached out to support about this and asked what can they do for the remaining dead 3x drives.
They send me a request to SIGN a letter stating that they will replace the 3-drives if I agree to give up the right to make ANY FURTHER Out of Warranty claims. LOLWUT?
Guess I’ve been a WD fan as their rust has served me generally well over the years, but this sorta feels like a kick in the gut given how much business I give them.
The C2000 chips indeed have a rather interesting problem with reliability.
To be precise, the LPC clock source dies and since there is no clock, the system can’t POST.
There is a fixed Stepping ( C ) so if you get to know that it is that, you should be safe.
But if not, It could fail the next time you turn it off.
So don’t turn it off for now.
It might also stop working on its own, but i have heard people start complaining after restarting, so don’t?
This is the first time I’m playing with SuperMicro kit and right now I’ve just got the mainboard running on a desk connected over IPMI. Wow, loving it, especially the way IPMI works.
I am seriously going to have trouble looking at consumer mainboards the same way.
I’ve also received the 8x 12TB WD UltraStar HDDs and am burning in the last x4 drives.
The chassis is ready, but I will only be collecting it in April, so it’ll be a while till this project is concluded.
Apologies for the late update to this project log, although the positive side to this is that I’ve completed the entire process of setting up the backup FreeNAS server and it has been running quite well over the past couple months.
Right now the first 8-bays contain the primary backup pool consisting of 8x 12TB WD UltraStar disks.
I’ve filled the remaining 16-bays with an ‘ad-hoc’ collection of WD 4TB Red/Red Pro disks collected from my older Synology Diskstations (5 + 8-bay). I’ve called this the “dead-pool” as many of these disks are on their last legs. It’s a second pool that’s there as a “3rd” redundancy.
The primary pool performs a local-replication to “dead-pool” and interestingly enough, this feature was only just recently released in the new FreeNAS11.3