SOLVED Drives never sleep with 22.12.RC1

[dev_guy]

I set up a test build with 22.12.RC1 and a key issue for me is the ability for drives to sleep and stay spun down. But using a SSD for the system drive and two 10TB spinning drives for the test pool, they never spin down even with no apps or VMs even installed and no active connections. I can literally unplug the ethernet cable from the server and the drives still never spin down when set to do so after 5 minutes. Is this a bug with 22.12 or does spin down just not work with TrueNAS Scale? It's a very basic feature that can save a lot of energy, heat, and wear on the drives. Something in TrueNAS seems to be accessing the drives every few seconds which is why I suspect they won't spin down. Am I doing something wrong?

 

[morganL]

Have you tested with CORE or Angelfish?
If its working there, but not in Bluefin, then it's worth reporting a bug.
There are many forum threads on setup of this.

 

[Davvo]

Where is your pool dataset? Also do note that spinning down drives, if not done properly, might actually wear them out more.

 

[dev_guy]

Where is your pool dataset? Also do note that spinning down drives, if not done properly, might actually wear them out more.

It's an old myth that spinning down hard drives is hard on them. That was true over a decade ago but all modern drives "park" their heads off to the side before spinning down so there's no damage to the disk platters. Further manufactures typically rate their drives for around half a million spin up cycles. And further still lower drive temps are associated with longer drive life and sleeping drives run much cooler and lower your electric bill. And, finally, spinning hard drives usually fail from mechanical issues. The less time their moving parts are moving, the less wear and tear on the drive.

Show me a single credible study where spinning down drives is hard on them? It's a false myth. Ten drives spinning 7x24 in a server is around $50 to $100 a year in electricity in the US and even more in many other parts of the world. File servers that can't reliably spin down their drives are expensive to own and hard on their drives.

As for the data set I only set up a single data set with a single share for testing purposes. And even when the share isn't in use, the drives still won't sleep. I'll test it with the Core version and see if they spin down properly.

 

[dev_guy]

Have you tested with CORE or Angelfish?
If its working there, but not in Bluefin, then it's worth reporting a bug.
There are many forum threads on setup of this.

The drives won't spin down with the current version of CORE either. So either I'm doing something wrong with my very basic set up or there's a problem with both versions?

 

[joeschmuck]

It's an old myth that spinning down hard drives is hard on them.

That is not true. It is hard on them, specifically it's hard on the drive motor electronics.

That was true over a decade ago but all modern drives "park" their heads off to the side before spinning down so there's no damage to the disk platters.

Yup, that was the older drives, I can't argue that.

Further manufactures typically rate their drives for around half a million spin up cycles.

Really? I know the typical head loading is up to 600,000 cycles but where is the drive spinup cycles listed? I'm not trying to argue with you, I'm actually curious to know. I'm a knowledge hound.

Show me a single credible study where spinning down drives is hard on them?

I thought it was Back Blaze that had that data but I am going to need to find it. But here on the forums we have had a lot of drive failures due to motors not spinning up and the general culprit was the drive was set to sleep often.

The problem really comes about when you have these multi platter drives that require a lot of power to get them started dur to the high mass of all those heavy platters. The small laptop size drives can manage spinups like crazy due to their low mass.

And, finally, spinning hard drives usually fail from mechanical issues.

Eh, which mechanical issues? Motor Bearings? Armature Bearings? Head Crashes?

So I have four hard drives that have been spinning for 4.1 years now (over 5 years by the warranty date), that's 2 years beyond the warranty. 1750 Load cycles, 164 Start/Stop Count, 36556 hours of spinning run time. no errors at all. The load cycles do not match the start/stop cycles because I was playing around with sleeping the drives, trying to get some testing done with the parameters for someone else. The drives were only a few months old when I did that. And the drives part when I stop the ESXi VM of TrueNAS. But my point, the drives are lasting a long time and I hear this story from many people.

I will agree that the statement that spinning the drives down might wear them out more is not quite accurate. It should have been stated that spinning your drives down frequently could cause the drive motor electronics to fail sooner. They may be robust enough to make it to the warranty period but most of us prefer to add a few more years beyond that warranty before having to purchase new drives.

So I'm willing to listen to reason so long as we learn.

The drives won't spin down with the current version of CORE either. So either I'm doing something wrong with my very basic set up or there's a problem with both versions?

They should spin down, I say should only because I personally haven't done it with the current version but I see no reason it will not. There are a few hoops to jump through to make it happen. It's been documented in the forums several times, I just don't have a link off the top of my head. One of the main things it to make sure your System Data Set is not on your Pool, it is by default. This is the main issue most face.

Good Luck.

 

[Davvo]

It's an old myth that spinning down hard drives is hard on them.

It's not if you don't do properly and they spin up and down every 20 seconds.

Where is your pool dataset? Also do note that spinning down drives, if not done properly, might actually wear them out more.

Anyway, as joe wrote I should have explained better the kind of wear I was talking about. My bad.

Move your dataset to the boot pool and your drives should spin down.

 

[dev_guy]

It's not if you don't do properly and they spin up and down every 20 seconds.

Anyway, as joe wrote I should have explained better the kind of wear I was talking about. My bad.

Move your dataset to the boot pool and your drives should spin down.

I'm not at suggesting spinning up and down every 20 seconds or even every 20 minutes. More like the drives sleep for many hours at a time, like all night long and when they're otherwise not being used, greatly reducing their total hours spinning from 24 hours a day to typically only a few hours a day.

Move what dataset to the boot pool? I don't want my share(s) on the boot pool as I ultimately want to store dozens of terabytes.

And to @joeschmuck it's trivial to design the motor, and motor electronics, to survive a huge number of spin up cycles. It's not especially stressful at all. Think about all the motors in our lives that turn on and off dozens or even hundreds of times a day like say forced air heating or cooling in HVAC systems. Or all the cars that now have auto stop systems that shut the engine off after you're stopped for more than a few seconds and re-start it when you take your foot off the brake. Those are much higher power and more "stressful" examples than just spinning up a hard drive. It's a non-issue and, again, show me a credible report? Backblaze wouldn't be a good source as they're nothing like a home file server and their customers are not willing to wait to have drives spin up.

And in all the drive failures I've had in my personal and work systems and servers, out of at least a few dozen, not one of them has been due to not spinning up. Generally they develop the "click of death" where the heads can no longer reliably find the data tracks, they develop surface errors, or they otherwise just stop working even when they're still spinning.

 

[morganL]

Forum post and script here https://www.truenas.com/community/resources/hdd-spindown-timer.122/
Not yet tested on SCALE Bluefin.

 

[Davvo]

Move what dataset to the boot pool? I don't want my share(s) on the boot pool as I ultimately want to store dozens of terabytes.

I meant the system dataset that joe wrote about and I improperly called pool dataset in my first post.
You want it out of the spinners if you want them to spin down.

 

[joeschmuck]

@dev_guy We can agree to disagree, I'm fine with that. I'm just providing my experience to help the folks out here and make the best of their investment. But I will agree with the fact that if a drive is not spinning up frequently, it possibly would not be a bad thing. Unfortunately there have been a lot of people here that have their drives spin up every 5 minutes or similar for example. Once a day or less is reasonable in my opinion. If you have a reference about the motor cycle count, I'd still like to see it, from any manufacturer. I personally have not seen that and enjoy learning new things.

Hopefully the link that @morganL provided will work for SCALE, please let us know.

 

[fastleo63]

I set up a test build with 22.12.RC1 and a key issue for me is the ability for drives to sleep and stay spun down. But using a SSD for the system drive and two 10TB spinning drives for the test pool, they never spin down even with no apps or VMs even installed and no active connections. I can literally unplug the ethernet cable from the server and the drives still never spin down when set to do so after 5 minutes. Is this a bug with 22.12 or does spin down just not work with TrueNAS Scale? It's a very basic feature that can save a lot of energy, heat, and wear on the drives. Something in TrueNAS seems to be accessing the drives every few seconds which is why I suspect they won't spin down. Am I doing something wrong?

Me, too. With 22.02 "Angelfish" no problems at all, today I upgraded my TrueNAS SCALE to 22.12.0, and HDD spindown ceased to work
EDIT: my mistake!

 

[morganL]

Me, too. With 22.02 "Angelfish" no problems at all, today I upgraded my TrueNAS SCALE to 22.12.0, and HDD spindown ceased to work.

If you can show Angelfish works and Bluefin does not, please report a bug and help us diagnose.

 

[fastleo63]

If you can show Angelfish works and Bluefin does not, please report a bug and help us diagnose.

I have to apologize...
After another check (by opening my NAS enclosure and "physically" touching the drives with my hands) I can confirm that spindown works even in Bluefin.
My drives are rather noisy, because they are spinning at 7200 RPM and their head actuators sometimes produce a strong "click", very common in many 3.5" HGST Deskstar NAS HDDs.

 

[ngandrass]

Hi,

i just wanted to let you know that I published an updated version of the Spindown Timer script that was tested on TrueNAS SCALE 22.12.0.

See: https://www.truenas.com/community/resources/hdd-spindown-timer.122/