Unraid spin down disks to save them and power or are we destroying them?

[Maticks]

My old Synology 8 Bay system with Raid 5 constantly had to read and write all disks if i did anything on the volume. That's the nature or raid volumes data is always being read and written to everything.

 

After installing Unraid around 2 months ago which is the best NAS for my specific needs being Media Storage, The Data Disk with the Media is only spinning if a file is requested and if a Write happens on the Data Disk then it will write to both Parity Drives as well. For this reason my Parity Drives are always spinning.

However i have 1-2 Data Drives ever spinning at the same time under this new system which is pretty neat in itself. It may be 18 hours between any of the Data Drives spinning up.

This provides a very interesting view of Unraid vs something like Synology/FreeNas where all drives will be read and written to for Raid setups.

 

However... 3 days of reading around on forums posts about the question will spinning down my hard drive to keep it cool, not spinning so in theory cause less wear and tear to the system be better.

I am less clear now on this subject than ever before, i would have thought with electronics if you turn something off and moving surfaces it isn't spinning to be damaged these would be good things.

Some places say yes its good, then some sources say it will kill the drive in a few months with only limited power downs but it seems drives are built for 100,000 power downs these days.

 

Mythbusters did a great video on this subject with fluro lighting was it better to turn them off or just leave them on which one made them last longer and use less power.

I was hoping Linus Tech Tips did a video on this Hard Drive subject it seems to be very much a 50% in either direction with no actual data to support it.

When you spin down 20 Hard Drives its much quieter and chewing less power plus heat so there is that but no drive exceeds 29c so i am not in any danger there leaving them all spinning.

Most of my drives are WD Red's and Seagate DM drives before i knew to buy Reds.

 

Any help will be appreciated

 

[themctipers]

spin down to save power and less power on hours.

 

less spinning = less failure rate from motors wearnig out

[Maticks]

1 minute ago, themctipers said:

spin down to save power and less power on hours.

 

less spinning = less failure rate from motors wearnig out

Yes indeed but then a few sites have stated that the first 1-3 seconds starting the disk from the stop damage the motor more than if it was just keeping the drive spinning 24x7.

Less Spinning less failure yes i do agree with you there but i am not 100% sure which is why i've posted here.

[GDRRiley]

1 minute ago, Maticks said:

Yes indeed but then a few sites have stated that the first 1-3 seconds starting the disk from the stop damage the motor more than if it was just keeping the drive spinning 24x7.

Less Spinning less failure yes i do agree with you there but i am not 100% sure which is why i've posted here.

almost every OS spins down the hard drives are designed to do it a lot. 

[themctipers]

1 minute ago, Maticks said:

Yes indeed but then a few sites have stated that the first 1-3 seconds starting the disk from the stop damage the motor more than if it was just keeping the drive spinning 24x7.

Less Spinning less failure yes i do agree with you there but i am not 100% sure which is why i've posted here.

you're not constantly hitting the NAS wiht reads/writes, so during the time when you're not (which is usually most of the time, like 90% of the time nobody is using the nas with my experience) it'll spin down the disks

 

spinning it down = no energy used,less chance of failure, less power on cycles, and it doesnt heat up, and its safe to move around and get bumped around if that happens.

[Maticks]

I have set mine back to Spin Down after 1 Hour Idle then.

[zMeul]

hard drives these days are rated in load/unload cycles - heads

plus, a motor that spins constantly will last longer than one that has to spin up

[AshleyAshes]

1 hour ago, zMeul said:

hard drives these days are rated in load/unload cycles - heads

plus, a motor that spins constantly will last longer than one that has to spin up

Well, let's do some math here.

 

The Seagate 8TB drive is rated for 300 000 Load/Unload cycles.  So, let's assume an aggressive number of one cycle every hour, which would mean your drive is spinning up not long after is spins down assuming it waits around 30mins or so of idle before it spins down, all day, every day.  It would take you 12 500 days or more than 34 years to reach 300 000 cycles.

 

Yeah, I think that's fine.  You'd basically have to build a torture script to constantly spin down and spin up the disc to use up it's rated cycle count.

 

And this is why my FlexRAID systems spins down it's drives individually when not needed by the pool.  Not to mention it saves a lot of power and heat.

[SCHISCHKA]

there is more power and wear on a drive spinning up. I only use drive spin down on my laptop to save battery. On my desktop I keep all drives spinning, except my backup drive which i only write to once a day; I keep it unmounted and powered off.

[GW2]

On 6/30/2017 at 11:42 PM, AshleyAshes said:

Well, let's do some math here.

 

The Seagate 8TB drive is rated for 300 000 Load/Unload cycles.  So, let's assume an aggressive number of one cycle every hour, which would mean your drive is spinning up not long after is spins down assuming it waits around 30mins or so of idle before it spins down, all day, every day.  It would take you 12 500 days or more than 34 years to reach 300 000 cycles.

 

Yeah, I think that's fine.  You'd basically have to build a torture script to constantly spin down and spin up the disc to use up it's rated cycle count.

 

And this is why my FlexRAID systems spins down it's drives individually when not needed by the pool.  Not to mention it saves a lot of power and heat.

Math to the rescue!

 

I've heard both sides of the argument a million times and the real problem is that neither is wrong, it's like politics. If it's just at home, my data RAID array spend 99% of it's time idle. so i have it set to spin down after 15 minutes. some days it has to spin up 3 or 4 times, but sometimes it spins up only once a day to check the drives. For my case I like saving power, they're 3TB SAS disks so they are quite power hungry. I do also understand that spinning up is (quite obviously) the most damaging time for the motor, bearings etc. This is just common sense, it's why you let your car warm up before thrashing on it, things need to heat up and fill in the tolerances for the smoothest operation. On top of that you are fighting inertia which mean the motor is working hard. The windings in the motor will work-harden faster and the motor will die sooner. In the enterprise where data is always being accessed this probably isn't even a concern, no drive will ever have the chance to spin down, even if it were set to. It's a real pickle! Id love to run some tests on this with new drives.

[AshleyAshes]

10 minutes ago, GW2 said:

Math to the rescue!

 

I've heard both sides of the argument a million times and the real problem is that neither is wrong, it's like politics. If it's just at home, my data RAID array spend 99% of it's time idle. so i have it set to spin down after 15 minutes. some days it has to spin up 3 or 4 times, but sometimes it spins up only once a day to check the drives. For my case I like saving power, they're 3TB SAS disks so they are quite power hungry. I do also understand that spinning up is (quite obviously) the most damaging time for the motor, bearings etc. This is just common sense, it's why you let your car warm up before thrashing on it, things need to heat up and fill in the tolerances for the smoothest operation. On top of that you are fighting inertia which mean the motor is working hard. The windings in the motor will work-harden faster and the motor will die sooner. In the enterprise where data is always being accessed this probably isn't even a concern, no drive will ever have the chance to spin down, even if it were set to. It's a real pickle! Id love to run some tests on this with new drives.

That's really it.  There's next to no testing since you would need a massively prolonged torture test and you'd need to execute it on a huge number of drives just to rule out random chance.  With how a drive really can 'Just up and die for no reason' without using huge number of drives in testing, you couldn't rule out one pool of drives being tested as having failed due to pure random chance.

 

The only real information available is how many desktop PCs naturally spin their discs down as the default OS configuration and these are the prime example of 'light usage' PCs.  Some will sit idle all night while the user is asleep or at work/school or see on and off usage when the user is around as they move between other tasks in the home.  There appears to be no meaningful information to suggest that the function, which is the default in Windows OS's for a very long time, causes harm to consumer users.  We're only certain of one thing: Every hard drive, no matter which hard drive, will eventually die, even if it takes years and years and years for it to happen.

[GW2]

5 minutes ago, AshleyAshes said:

That's really it.  There's next to no testing since you would need a massively prolonged torture test and you'd need to execute it on a huge number of drives just to rule out random chance.  With how a drive really can 'Just up and die for no reason' without using huge number of drives in testing, you couldn't rule out one pool of drives being tested as having failed due to pure random chance.

And even if we managed to get this magical test of a million identical hard drives put together, we know that the HDD manufacturers work in batches, and change their designs slightly on a frequent basis, who's to say this test is representative of anything more than that model of HDD?

[k0d3g3ar]

This is a very interesting subject.  I've been using Unraid now for over 10 years, and have 2 servers with about 60TB of disk in each.  They have about 12 HDDs combined in them, so I have a decent statistical case here regarding drive longevity and performance.

 

Additionally, we have 2 FreeNAS servers in our data center that have about 40TB of disk in them, about 16 drives in total, and so I can A-B compare the results of my home Unraid servers and our hosted FreeNAS servers.  Clearly our data center has better power conditioning, enterprise class servers, etc. but still hard drives fail.  That said, over 4 years we've had only 1 x 4TB drive fail in the data center in the FreeNAS servers, whereas I've had about 10 drives fail in the Unraid servers.

 

The most fickle of the drives are Seagate drives.  I've been using the 12TB Seagate Archive drives and they don't seem to last more than a couple of years.  I've had much better results with HGST data center drives for longevity, but even with that drives seem to die on me routinely.  To the point where I have to budget about $300 a month for drive replacements.  Sometimes I'm lucky to be able to claim under warranty, but most of the time I'm out of pocket and this is like sacrificing a car payment per month just to keep these drives going.

 

What could be causing this?  Well one thing that comes to mind, and what drove me to this thread, is the spinning up/down that Unraid is doing.  Not all HDDs are created equal here - you have drives specifically designed for NAS storage, drives for CCTV footage, drives for databases, etc.  I am no expert on drive design & manufacture but it seems that you do get what you pay for here.  That said, it does seem to be counter intuitive to want to try and save money on power by spinning down the drives if they are going to cost me $300 a month to replace them due to this not being the intended use-case for the drive design. 

 

Now I realize I'm probably an outlier here with my usage, but if anyone has something similar in size to this, I'd like to A-B compare how you have done it with my setup, because I'm getting really tired of retiring dead drives all the time and I can't seem to understand why I'm not doing this with FreeNAS in our data center considering those drives get more of a hammering than my home servers are subjected to.

 

K

[Jarsky]

Spinning down drives doesnt negatively impact the longevity of the drive for any reasonable period of time (5 years+). HDD spindown is part of the drives firmware itself, all the OS does is send a signal to it to sleep or wake so it isnt inherintly a specific thing to UnRAID. You can look at the manufacturers specifications on what your drives are rated for (called Start-Stop, Load/Unload, Power Cycle, etc..) in SMART its called like Load Cycle or something...a typical value for this is around 300,000-400,000. As an example of how much that is, my 2 year old UnRAID drives are up to 19,000 in 2 years.....so you're looking at around 15-20 years to reach half the rated value. 

 

I have Folder Caching enabled for if im browsing high level folders, and the drives on spin down. The folder caching helps to reduce spin up frequency. So they will only spin up when I need an actual file listing or to access a file on that drive. In the last 2 years I havent had a single failure. 

 

I'd say using Seagate Archives for active storage are probably a big part of that being a value SMR 'archive' drive.

The biggest killer to hard drives is typically heat and humidity though. It's Summer here in the middle of the day and mine are at 30-35C. Evening or Winter theyre in the 20's so I keep them relatively cool. 

 

I've been running RAID's since around 2001 in my home. In the last 20 years, ive maybe lost 5 drives (2 of them DOA in 2 weeks indicating defective drive), and ive been spinning down my media RAIDs for over a decade.  One of the biggest things is to regularly replace drives as they reach EOL. Don't wait for 1 to fail and drop completely, because to rebuild that drive you put a lot of strain on the other disks running them flat out for hours (days) at a time which increases the chance of another failure.  

 

 

[jec6613]

1 hour ago, Jarsky said:

Spinning down drives doesnt negatively impact the longevity of the drive for any reasonable period of time (5 years+).

Assuming you're replacing your drives at a regular cadence, this is exactly correct for home use.  Spin them down and save the power if your drives support it, or don't, it won't make much difference.

 

For old systems and 24/7 active systems, of the sort that many datacenters have, spindown does impact drive life in a negative manner, as does literally any power down event.  When a big datacenter loses power, there's a lot of equipment that just doesn't turn back on afterwards - mostly PSUs, HDDs, and router/switch blades.  There's some study on this effect - basically things that have been powered on and heated up for an extended period of time contract and connections break.  This is probably the source of the misinformation that consumers shouldn't spin down their drives - but consumers are not enterprise datacenters.

[Ubuntu is love]

I've heard both sides of the argument so many times and it's diffrent in every use case. It depends on the kind of disk and how they are rated. Disk do normaly break when spinning up because that's the most heavly load they will get. Does not mean they do not wear from spinning all day. There are a few things you should consider:

 

- SAS drives don't like to spin up and are also not made for that.

- How faster the drive, how harder the beating when they have to spin up.

- If the disks have to spin up more times an hour they should probably never stop spinning, for better performance and for les wear.

- Then the last point, does it matter what you do? The power cost of some disks are high enough that you should probably just let them spin down even if they get more wear from it. And if they break just buy a new one from the money you saved.

 

https://www.koreascience.or.kr/article/JAKO199911920051131.page

 

https://www.sciencedirect.com/science/article/pii/S0301679X03000720