Periodic reads (30min) preventing disk spindowns

[Pycrs]

Hello everyone,
I recently built a system and got TrueNAS SCALE up and running, been very happy with it too. One thing that has been bugging me is that I can't keep my drives in standby mode because there is a periodic read on my ZFS pool. System dataset is on the boot drive, tried stopping all the containers, turned SMB off. SMART tests, snapshots, etc are scheduled so far apart that I'm pretty certain that those are not the culprits. Every read is 0.15KB in size, every 30 minutes on the dot, which is oddly specific. At this point I'm pretty much out of options so I've decided to try my luck here.
System specs:
Intel Core i5-8400T
2*8GB DDR4 RAM (non ECC)
500GB SATA M.2. SSD (apps)
100GB SATA Intel SSD (boot drive)
3*4TB IronWolfs

Thanks in advance for any reponses.

 

[Pycrs]

Forgot to mention that those periodic reads wake the drives up one at a time, when It's me waking them up, the all start up at the same time, don't know if it means anything.

 

[nddj]

I was about to put up the exact same thread just now. In my case, it seems to go crazy because loading occurs every 5 minutes.

I am keeping an eye on this thread.

 

[Davvo]

I was about to put up the exact same thread just now. In my case, it seems to go crazy because loading occurs every 5 minutes.

I am keeping an eye on this thread.

That's the system dataset, move it into an SSD.

Every read is 0.15KB in size, every 30 minutes on the dot, which is oddly specific.

Looks like temperature readings.

 

[nddj]

That's the system dataset, move it into an SSD.


Looks like temperature readings.

The very first build moved the system dataset to the boot-pool(SSD, of course). Because I knew about this problem from the beginning. However, Read is happening once every 5 minutes. what is happening...

 

[Davvo]

The very first build moved the system dataset to the boot-pool(SSD, of course). Because I knew about this problem from the beginning. However, Read is happening once every 5 minutes. what is happening...

Verify, the 5 minutes mark is the system dataset's signature.

 

[nddj]

Verify, the 5 minutes mark is the system dataset's signature.

This is exactly what I saw this morning. Is there a way to confirm that this mark is from the signature of the system dataset?

 

[Pycrs]

Looks like temperature readings.

I've read somewhere that if you set up spindowns, i.e. there is Always On in the standby period field, the temperature will not be recorded for those disks, furthermore, in the reporting tab, only the two SSDs - boot and apps - have their corresponding temperature graphs, the HDDs only have IOs graphs.

 

[Pycrs]

I've read somewhere that if you set up spindowns, i.e. there is Always On in the standby period field, the temperature will not be recorded for those disks, furthermore, in the reporting tab, only the two SSDs - boot and apps - have their corresponding temperature graphs, the HDDs only have IOs graphs.

*there is anything other than Always On, you get the point

 

[Davvo]

Well 30 minutes is the default value for the temperature readings, henceforth my guess.

 

[Pycrs]

Well 30 minutes is the default value for the temperature readings, henceforth my guess.

Is every temp reading being logged by any chance maybe? So that I can explore this possibility a bit more?

 

[Pycrs]

UPDATE: I disabled SMART checks on all the drives in the pool and so far, about an hour, no reads - thus the pool is still sleeping, yay. Not sure if temperature reads are a part of SMART checks but it seems like they are. Now, I have SMART tests defined in the Data Protection tab, does this mean that when those tests go off, none of the disks with the disabled SMART checks will be checked? Ideally I would want my weekly SMARTs to check all the drives and waking the drives up once a week is no big deal. Any way I can achieve this? Not sure if I got my point across.

 

[Davvo]

Take a look at this script.

HDD Spindown Timer 2.2.0

This resource was originally created by user: ngandrass on the TrueNAS Community Forums Archive. Please DM this account or comment in this thread to claim it. Disk spindown has always been an issue for various TrueNAS / FreeNAS users. This script utilizes iostat to detect I/O operations (reads...

www.truenas.com

 

[Pycrs]

So I followed the instructions on the github page, disabled TrueNAS spindowns. That caused periodic temperature reads every 5mins and temperature graphs are now being displayed alongside the IO ones. I tried multiple poll rate/idle time combinations with the script, the disks never spun down. So now it looks like either I get my power and noise savings or SMART tests.

 

[Pycrs]

UPDATE: Considering the option of telling SMART to ignore sleeping disks and schedule a script that wakes the pool before every SMART test - weekly SHORT and bimonthly LONG.

 

[Davvo]

FWIW I run daily short and weekly long smart tests. I do not spin down my drives.