Register for the iXsystems Community to get an ad-free experience and exclusive discounts in our eBay Store.

HDD Reliability Thread

siconic

Newbie
Joined
Oct 12, 2016
Messages
90
I haven't really seen a post on this, and I know that BackBlaze posts their quarterly results, but they are running their stuff in a DC environment, and under climate controlled circumstances. As amateur storage enthusiasts, we are still not getting much consumer data on disk reliability, or reliability reports from our own personal environments, so I thought I would suggest this as a sticky, and start collecting some data.

Data points I think would be most useful to illustrate reliability:

Conditions disk is used in (Data Center, Home office, Garage, Shed, Etc.)
Manufacturer of disk
Model or Type (SAS, SSD, SATA, SCSI)
Size
Power On Hours
Max Temp
Power Cycle Count


My setup is in my garage, in the Florida heat and humidity. With the exception of 4 BRAND NEW consumer grade Seagates in 2014, which lasted all of 4 months in my garage before all 4 died, I can say from personal experience, I will never run anything but Hitachi HGST Ultrastar (SAS or SATA commercial) drives, as I have had 4 failures out of 33 disks, in 5 years. Also bear in mind these are HIGH HOUR drives I get off ebay as storage requirements grow.

I am running everything off of an R610 server, a Rackables SE3016 16 Bay JBOD, LSI 9207-8e HBA. I started with 7 x 1TB HGST Ultrastars, each had ~32k hours, and 4 new Seagates. 2 x 1TB HGST drives died, and all 4 Seagates died. I do not have any data from my first set, except the ~ hours and failure rate.

Following are some data points:
1st Set:
1. 13 x 1TB HGST Ultrastars (SAS), 4 x 1TB Seagate Consumer (SATA)
2. 52k hours when decommissioned
3. Max Temp ~110F
4. Failures: 2 HGST, 4 Seagate

2nd Set:
1. 11 x 2TB HGST Ultrastars, Model HUA722020ALA330 (SATA); 3 x 2TB Seagate, Model ST32000SSSUN2 (SAS)
2. Replacing, All between 66k and 67k hours. 1 pending failure prompted the replacement cycle.
3. Restart Count - Between 125 and 250
4. Max Temp 109F
5. 1 pending HGST and 2 Seagate

3rd Set (incoming replacement):
1. 6 x 4TB HGST Ultrastars, HUS724040ALA640 (SATA), 5 x 3TB HGST Ultrastars, HUS723030ALS640 (SAS)
2. Starting hours, 4TB disks 25k hours, 3TB disks, awaiting arrival
3. Restart Count - Between 40 and 50 for 4TB disks
4. Max Temp 109F


I will say, I am partial to HGST, since not only have they survived my torture chamber of a setup, but BackBlaze has the lowest failure rates from that brand. I am however seeing good stats on Toshiba as well in BackBlaze reporting, so they may be able to withstand my abuse as well.

Thanks guys, keep the data flowing!
 
Last edited:

siconic

Newbie
Joined
Oct 12, 2016
Messages
90
As a side note, I will be putting my 2TB disks up for sale once my upgrade is complete. Probably $15 each if you buy all, I have 2 good backups, so thats 12 good disks.
 

tfran1990

FreeNAS Experienced
Joined
Oct 18, 2017
Messages
210
With a lot of PMR Disc still on the market and the transition to some of the SMR discs i think it is a good idea to collect data like this from the users here.
 

Adrian

FreeNAS Aware
Joined
Jun 29, 2011
Messages
85
I am not volunteering, but a lot of this is amenable to automatic capture (SMART?) and transmission (HTTPS/put?) to a central collation service. Condition could be set when initialising the system. Would need a means of manually specifying the mode of demise of decommissioned disks.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
11,837
I will never run anything but Hitachi HGST Ultrastar (SAS or SATA commercial) drives
Well that's nice, but Hitachi sold HGST to WD, and anecdotal evidence suggests that quality has fallen.

This has been discussed before, i.e. https://www.ixsystems.com/community/threads/st4000dx002.42219/post-273017

Overall the usefulness and quality of this sort of data collection is very difficult. The BackBlaze data has a bunch of flaws and isn't particularly high quality. The main redeeming value is that it is at least somewhat consistent in terms of collection environment. It would be more difficult to collect data from random hobbyists because their systems tend to lack environmental stability, which is one of the major factors in drive longevity. Trying to draw useful conclusions would be... hm. Difficult.
 

cobrakiller58

FreeNAS Experienced
Joined
Jan 18, 2017
Messages
329
I was just wishing Amazon and Alphabet would release similar data to backblaze but that is asking for too much, though I'm sure they likely track the information internally.
 

CompuGlobalHyperMegaNet

FreeNAS Experienced
Joined
Sep 13, 2014
Messages
119
I was just wishing Amazon and Alphabet would release similar data to backblaze but that is asking for too much, though I'm sure they likely track the information internally.
If you / they could filter out customers who have bought more than X number of disks at any one time and more than Y number of disks all time, I thing that'd be the best way (short of a widespread and targeted gathering of data from home users) of getting relevant data. Anything other than that, including this thread unfortunately, simply can't gather enough data to draw meaningful conclusions.
 
Top