HDD spin down delay best practice

[jebusfreek666]

Ok, assuming that I don't really care about how much electricity I use and my number 1 goal is longevity of my drives, is it better to set the drives to never spin down vs. spin down delay of say 1 hour. The servers primary purpose is plex, and deluge. It runs 24/7. They have been running nonstop basically since I set it up, and my HDD temps have always been great. I did recently just add a 3 in 2 in the 5.25 bays on the top which have been giving me a bit of trouble with temps. But even those never exceed 38 C. All others hold at 30 all the time. I have done multiple searches on this, and it seems like people are split 50/50 on this. Again, assuming I am not concerned with conserving electricity or saving the planet.... what is best for the drives themselves. 

[JonathanM]

Anecdotally, I've very seldom seen a drive fail while running, mostly they fail to come up after a power event.

 

Always spinning means less temperature swing, so less opportunity for failure there. However, I personally don't see a need to keep a drive spun up if it's only going to be used once a week, so there's that.

 

I'd say for purposes of longevity, maybe 24 hours spin down delay would be more appropriate, but that's wild supposition based on nothing but gut instinct.

[Frank1940]

There is no real data on this.  But common sense would (almost) tell you that the worst thing you can do is to spinning them up for a short period (say, fifteen minutes) and the spinning them down for another equally short period.  But you could also make a case that loading up a drive with data and then removing from a computer, storing it on shelf for twenty years and then expecting to function when reactivated might be a big gamble. We all know that a constantly  spun-up drive will fail eventually.  (And we have some data from a big server farm company-- Backblaze --to support that!)   What we don't know is if there a happy medium between these two extremes... 

Edited February 11, 2019 by Frank1940

[hitman2158]

29 minutes ago, jonathanm said:

Anecdotally, I've very seldom seen a drive fail while running, mostly they fail to come up after a power event

Yeah, agreed.

 

Many years ago I had a terrible expirience with a failing disk at work. We made software for electric checkout for our customers in the automotive industry. So, I was sent out to one of our customer to update the software. The used PC (they are running usually 24/7) was a small industrial brick, something with a 386-CPU, a few megs of RAM and so on. The harddisk was something around 20 megs or so. OS was FlexOS, nowerdays a discontinued RTOS. So I updated the software including drivers and finally had to reboot the tiny box.  Bad surprise, no MBR found. The disk was dead. The customer was not amused and kicked me off his factory, no permission for the next 6 months. It was not my fault, it was just bad luck.

 

But you will a power event after updating/upgrading OS when you have to reboot the server. Both servers in my flat will spin down the disks after 15 minutes or so and will go to sleep after additional 15-30 minutes. If I need them I wake them up via Wakeonlan, it´s just a few seconds.

[Frank1940]

23 minutes ago, hitman2158 said:

OS was FlexOS

I can remember running Flex on a Radio Shack Color Computer back in the early 1980's.  Then, I moved on to OS9 (A UNIX type OS) for the same machine.  I had Word Processors for both of those OS's.  You had to save to do a spell check and then save it again to do the printing.  (Three separate programs!)  Of course, the Color Computer only had a max of 64KB of RAM.  (That is right--- Kilobytes!!!) 

[Julius]

My personal idea on this has always been in favor of spinning them down depending on the duration they are generally not spun up again. Basically what jonathan is saying as well, if they're offline for more than, say, 12 hours, I'd say that's a win for disk longevity. I think I also read that in a google datacenter document as well. I now have unraid set to a Spin-down delay of 45 minutes, after having started with 15 minutes I noticed that was moot, since some disks would be doing stuff intermittently right after they were then spun down, like receiving a finishing bit from a backup from a remote server, or loading or unloading files from a device on the LAN. None of the tasks last longer than 45 minutes each, so if a disk is doing nothing for that time, it can be spun down and would not be accessed soon after. We don't stream stuff from the unraid server, so you may want to consider setting it to the max duration of any video you access from it, that does make sense I guess. Depending on how many disks, how you have divided the use of them under Shares (I have some of the 10 disks in the array excluded for parts, so they can remain spun down longer), and how many people are accessing your server. I have a really large cache (1TB) so that helps as well for the reads and re-reads from SSD that don't require new access from the array.

Edited July 1, 2020 by Julius

[Baumberger]

Thank you for describing your experience with spin down hdds. I basically only need the hhds once a week to transfer the data from cache ssd (1TB). Extremely annoyed about the temperatures and noises from the hdds, although they are only in standby, optimization was pending.
It's pretty clear to say that I can safely spin down my array for 6 days or more.
As a newcomer, I wonder whether the array will start up automatically, without problems and on time for the transfer time. Maybe special settings are necessary?

[Soulflyzz]

On 12/10/2020 at 1:34 PM, Baumberger said:

Thank you for describing your experience with spin down hdds. I basically only need the hhds once a week to transfer the data from cache ssd (1TB). Extremely annoyed about the temperatures and noises from the hdds, although they are only in standby, optimization was pending.
It's pretty clear to say that I can safely spin down my array for 6 days or more.
As a newcomer, I wonder whether the array will start up automatically, without problems and on time for the transfer time. Maybe special settings are necessary?

 

Mind if I grab your final build and how its been working out for you over the last 6 months?