External JBoD case doesn't hibernate

[daTroll]

I have built a low-power UnRAID server using a mini-PC (Zotac i5 https://www.zotac.com/de/product/mini_pcs/zbox-edge-mi646-barebone-0) and a USB 3.1 JBoD case (TerraMaster D4-300 https://www.terra-master.com/global/products/d4-300.html). The performance is good enough as my purposes are cold storage of media files and TimeMachine target of backups.

 

However, I notice that the TerraMaster enclosure never goes into hibernation mode. When all of the drives are in standby (either by inactivity timeout or manually triggered spin down), the UnRAID GUI indicators are greyed out, but the LEDs are all flashing rapidly in unison and the fans keep on spinning. The flashing LEDs indicate that the UnRAID server continues to check something on the drives even when they are in standby (but no temperature is shown in the GUI). When I connect the same drive enclosure to a Windows PC, the computer ejects the drives (I guess because the file system is not supported), the LEDs go dark and the fans stop spinning.

 

I asked TerraMaster support about this, and all they can say is that they use an "Asmedia USB SATA 5G chipset". Is there a chance that UnRAID can become more ecologically friendly and support the hibernation mode of external drive enclosures like mine?

Edited June 13, 2022 by daTroll

[daTroll]

Is there some way I can monitor what the UnRAID server is doing with hard drives that are spun down and in standby mode? The flashing LEDs show that there is activity of some sort...

[AndrewZ]

Is this drive being mounted via the Unassigned Devices plugin?  My understanding is that drives that are mounted via UD will sleep if there is no activity on them (they will follow the default disk sleep time).  If it is not mounted via UD and you are passing it through to a VM, or mounting it yourself then it indeed never sleep.  @dlandon would be able to pipe in more if it being mounted via UD

 

 

[daTroll]

All drives are assigned (parity + two data drives + cache) and the same behaviour is also present when I reboot in safe mode.

 

What I can see on the LEDs is that there is activity from the server when the drives are in standby mode (automated spin down as configured in the GUI). It’s as if the server is checking whether the drives are still there. The LEDs are then flashing at about 2 Hz frequency, all of them together. When the drives are spinning but there is no actual use, then the LEDs are just on without flashing. 

[trurl]

Do  you have any dockers or VMs?

[daTroll]

No VM is running and the only docker is Plex-Media-Server. However, stopping this docker makes no difference; the LEDs are still flashing.

[trurl]

Which disk is your system share on?

[Squid]

If the docker image (or folder) or VM's are enabled and the libvirt image are stored on that drive, it will not sleep

[daTroll]

13 minutes ago, trurl said:

Which disk is your system share on?

 

The system share is on a different cache drive, which is on an internal NVMe disk. Full details (I suppose) are in the attached diagnostics.

untroll-diagnostics-20220703-1911.zip

[daTroll]

24 minutes ago, Squid said:

If the docker image (or folder) or VM's are enabled and the libvirt image are stored on that drive, it will not sleep

 

It still doesn't sleep without any docker or VM enabled.

 

/mnt/user/system/libvert is on an internal drive, i.e. not on one of the external drives in the TerraMaster enclosure that this topic is about.

Edited July 3, 2022 by daTroll

[trurl]

I guess the enclosure doesn't allow hibernate, or at least doesn't allow Unraid to spindown unused disks. Just one of many reasons USB not recommended for array or pools.

[daTroll]

8 minutes ago, trurl said:

I guess the enclosure doesn't allow hibernate, or at least doesn't allow Unraid to spindown unused disks. Just one of many reasons USB not recommended for array or pools.

 

But the drives do spin down! This I can see both in the GUI and hear from the enclosure. It's just that the fans keep spinning and the LEDs are flashing. When I attach this enclosure and drives to a Windows PC, both the drives and the enclosure are successfully put into sleep mode (and the LEDs go dark). It's only with UnRAID that the enclosure doesn't sleep. In other words: UnRAID doesn't allow the enclosure to enter sleep mode and the flashing of the activity LEDs appears to be directly related to this.

 

A friend of mine has a similar setup, but using different brands and models of hardware. He has noticed the same thing, that the external enclosure is never put to sleep automatically when the drives in his enclosure are on standby/spun down.

Edited July 3, 2022 by daTroll

[Squid]

43 minutes ago, daTroll said:

It still doesn't sleep without any docker or VM enabled.

 

Are they running or is the service disabled?  There's a difference.  As @trurl put it, where is the system share located? (if you're using the defaults for the locations)

[trurl]

According to diagnostics, appdata, domains, system are all on nvme

[SimonF]

I think this is due the way spin ups are detected from 6.9.x onwards. All of my systems HDD LEDs blink approx every 2secs. I think reading IO stats from the controller is detected as usage, but doesn't spin up drives.

[daTroll]

22 minutes ago, Squid said:

Are they running or is the service disabled?  There's a difference.  As @trurl put it, where is the system share located? (if you're using the defaults for the locations)

 

VM & the Plex docker are definitely not running, and it is still happening. The system share is not on any of the disks in the external enclosure. 

[daTroll]

12 minutes ago, trurl said:

According to diagnostics, appdata, domains, system are all on nvme

 

Correct, and this is a drive that is mounted internally in my mini-PC.

[daTroll]

7 minutes ago, SimonF said:

I think this is due the way spin ups are detected from 6.9.x onwards. All of my systems HDD LEDs blink approx every 2secs. I think reading IO stats from the controller is detected as usage, but doesn't spin up drives.

 

This is interesting! My LEDs are flashing about twice every second (i.e. 4 times as fast as yours).

 

Maybe you are onto something, but why would UnRAID want to read the IO stats from the controller of a disk that UnRAID has put in standby mode?

[SimonF]

18 minutes ago, daTroll said:

 

This is interesting! My LEDs are flashing about twice every second (i.e. 4 times as fast as yours).

 

Maybe you are onto something, but why would UnRAID want to read the IO stats from the controller of a disk that UnRAID has put in standby mode?

It may be a little quicker than 2 secs, but my HDD LED is a single one on the tower not per disk.

 

It is so Unraid knows when drives are accessed, an once it will update disk status and get temp. When drive spinup is detected is when you see 

 

Jul  3 19:54:23 Tower emhttpd: read SMART /dev/sdd

[trurl]

27 minutes ago, daTroll said:

VM & the Plex docker are definitely not running,

Not relevant but we weren't asking about specific dockers or VMs. Even if no dockers or VMs are running, system share still has open files if Docker or VM Manager is enabled in Settings.

[daTroll]

17 minutes ago, trurl said:

Not relevant but we weren't asking about specific dockers or VMs. Even if no dockers or VMs are running, system share still has open files if Docker or VM Manager is enabled in Settings.

 

Fair enough, and perhaps I wasn't expressing myself clearly. My dockers and VM are both disabled and stopped, and I can still see the activity LEDs are flashing for drives that are in standby mode (not spinning). 

 

Side note: I suppose that UnRAID will not automatically spin down a drive that has the system share or a VM domain on it. For my external drives, the GUI will clearly show when all four drives are in standby mode, and still the LEDs are flashing.

Edited July 3, 2022 by daTroll

[daTroll]

12 minutes ago, SimonF said:

It is so Unraid knows when drives are accessed, an once it will update disk status and get temp.

Ah... but the GUI doesn't display any temp for a disk that is in standby mode, it only shows * (an asterisk) instead.

 

It seems to me that an OS could determine whether there is any actual IO on a HDD without asking the controller for stats every second, no??

[SimonF]

11 hours ago, daTroll said:

Ah... but the GUI doesn't display any temp for a disk that is in standby mode, it only shows * (an asterisk) instead.

 

It seems to me that an OS could determine whether there is any actual IO on a HDD without asking the controller for stats every second, no??

I don't know the internal processes being used so I am not sure if there is another way to do this.

[daTroll]

14 minutes ago, SimonF said:

I don't know the internal processes being used so I am not sure if there is another way to do this.

Perhaps @AndrewZ can shed some light?