What constitutes an old drive

[tunetyme]

What constitutes an old drive. I have 6 data disks (16 TB) plus parity and cache disks. This is a simple home media server and the most exercise any one disk gets is the monthly parity check. Normally, when I add media to the server it is collected on my PC (4 TB) over several weeks (sometimes months) and added at one time.  The server is shutdown for extended periods because the fans are so loud. I added three 4 TB drives (1 parity 2 data) 18 months ago and now they are suffering from old age? I have some 2 TB drives that are 10+ years old without a single hiccup.

 

How is age determined?

 

Even when my server is on, the drives are spun down. The only time a drive ever reaches 90 degrees Fahrenheit is during a parity check otherwise they never get above 82 when spun up individually. Drive old age is one of the first things that is pointed out when seeking any support and from what I can see it is not based on read write time nor spun up time but from the time it was installed. It appears to me that there must be a better way to determine a drives age. What I try to watch is sector reallocation. When that starts on a drive I want to know about it immediately so I can replace it. I believe most of us use this as a media server and we are not managing a server farm where you are engaged in constant read write time. Is there a better way to determine when to replace a drive?

 

It has been my experience that if it isn't broke don't fix it.

[MisterWolfe]

I have never replaced a disk solely based on age either in my home or business environment. They either start to get errors that are uncorrectable and I replace them, or I replace smaller drives with larger ones when my array gets full. I've had drives in the array for 7 or 8 years before being cycled out.

 

Some of those 8 year old drives were repurposed for other projects and are still going strong.

[BRiT]

Product viability is determined by SMART reports. However, I'm proactive with array health instead of reactive. I'd rather do maintenance on my schedule than having to do emergency procedures.

 

Assuming a drive didn't cause severe issues to cause immediate replacement...  I used to replace drives around when their warranty ran out at 4 to 5 years, and before the severe issues could happen. However drives seem to have substantially shorter warranty periods now, so my approach is to replace a drive when I can double their size or close to 5 years, whichever comes first. I replaced 4TB drives with 8TB before 5 years. Added benefit was shortening parity check times.

 

My drives are around 2.5 years now, so likely will replace sometime late 2021 to early 2022 if they dont release 16TB before then.

 

 

[tunetyme]

I have been replacing drives primarily when I can double the storage capacity for less than $130 which I have been doing going from 2 to 4 TB. 

8 TB drives are within the same price range if I reduce the number of drives. I have became concerned the SMART reports knew something I didn't and I would be better off adding more 4 TB drives because the 8 TB would need replacement before their useful life (filled up) ended.  

 

It would be of great interest to me if we could study users experience in drive useful life. I think that head fly time, reallocated sectors and additional info would be a better indicator than just physical age. The oldest drive in my system has power hours of 9+ months over ten years. I think it would be helpful to all to address this based on environment (business, home media, etc.) and develop a more meaningful measure of age. I believe this is available from large server farms but I am sure they use much larger drives.    

[WashingtonMatt]

https://www.backblaze.com/b2/hard-drive-test-data.html

 

https://www.backblaze.com/blog/2018-hard-drive-failure-rates/

Edited July 4, 2019 by WashingtonMatt