High Start / Stop count

[Nemesis_001]

Hi,

Any idea why the drives would have such a high start / stop count?

 

[gea]

Is disk sleep active?

 

[Nemesis_001]

Thanks. Where would I check this setting?

 

[gea]

In napp-it?
System > PowerMgmt

 

[Nemesis_001]

Thank you.
Found it.
What would be a good threshold value? It is 30s right now.

 

[gea]

Do you need or want disk sleep?

-You must disable all tasks that access disks regularly like napp-it alerts, monitoring, acceleration, smart checks or the Solaris fmd fault manager
-A wake up can last up to a minute for a pool to be ready.

The threshold depends on your usage. 30s after last access is for maximum power saving.
In your case you had 58000 start/stops in 42000 power on hours

 

[Nemesis_001]

Thanks. Im not as worried as the time it takes for the disks to get ready.
Just not sure how healthy it is for the drive to have so many start/stop cycles, around 1 per hour.
Yes, I do want disk sleep for the power savings.
You think these frequent spin ups are related to napp-it functions like monitoring or the Windows guest OS? anyway to check?
Or a spin up once an hour isn't anything to be worried about with regards to drive health?

By the way,
One specific drive has one 16 pending sectors, and one relocated sector. It's a WD drive.
The periodic scrub never repaired any errors.
Are these real errors successfully repaired on the drive level? Something phantom?
Would you dump this drive?

 

[gea]

Disks are around 5 years old. This is the suggested usage time by most vendors. Power cycles can reduce lifetime but ZFS will detect all data errors. Smart additionally count disk errors below data corruption like relocated sectors. These errors may indicate a near failure - or not.

On enterprise use I would change disks after 5 years or after such smart errors. At home you can wait until a disk fails under ZFS. A hotspare may be helpful for immediate replacement of a bad disk.

 

[Nemesis_001]

Thanks gea.
It's just a home server, where critical data is backed up to the cloud. Mostly media server usage, with a Windows Guest OS.
The pool is a RAIDZ-2. I guess I'll just leave it alone if the drive doesnt develop more errors, and maybe just move the guest OS to a NVME mirror.

 

[mr44er]

Any idea why the drives would have such a high start / stop count?

Looks like a misconfiguration right from the start, this timer was too aggressive.

For example my HGST Enterprise disk have this:
Specified cycle count over device lifetime: 50000
Specified load-unload count over device lifetime: 600000

Consumer is mostly designed for anything 'lower'. There is a difference between spindown (motor stop - this refers to the 50k) and standby (head parking - this refers to 600k). There exist also standby modes without parking.

You have home usage and raidz2, that's good...but be prepared that these disks should be already (mechanical) dead.

Current pending sector will be either fixed (x-th attempt works) or go to offline uncorrectable if a write action hits these sector area and yes, thats drive internal correction handling. It mostly works. -> I don't want to say ignore it, but that is an indicator/confirmation for you, when ZFS throws errors about this disk. So eyes always on ZFS and the scrubs, replace a disk immediately if ZFS shows the first error.

Maybe you can squeeze some more lifetime out of this disks if you set the spindown-time between 10 or 15 minutes and the standby between 3 and 5 minutes. If you have a third option with standby and parking, choose something between 6 and 9 minutes.

 

[Nemesis_001]

Thanks.
I changed the timer to 10 minutes, I will follow to see how the number increases over time.
Start / Stop count I suppose refer to a full spin down?
and Load Cycle count is head parking?

Any idea how to set their timers separately? I'm not in front of the server, but iirc there is only one relevant setting for spin down.

Either way, I ordered a couple of NVMEs to mirror for the guest OS so it will be moved off the spinners, although I dont really need the extra performance.

 

[mr44er]

Start / Stop count I suppose refer to a full spin down?
and Load Cycle count is head parking?

Yes

Any idea how to set their timers separately?

No, I've never seen napp-it or the other ones. But if you have tools like camcontrol, smartctl, hdparm, sdparm available then you could check their manual. With SATA disk it's APM (0-255, where 128 ist mostly the value of head parking) you wanna check and on SAS it's idle_a,idle_b,idle_c,standby_y,standby_z with the counters in ms.

Bonusinfo: Sometimes SATA firmware interprets APMvalues wrong or doesn't save it over a cold reboot.

 

[gea]

After you have configured all disks for a given timeout, you can select menu "edit powerconf" to modify settings per disk

 

[Nemesis_001]

Seems like changing the timeout did the trick.
Power on hours increased 216h, and start/stop count increased by 8 only.