Description: When System Notifications are enabled, unassigned devices disks in some cases will spin up on each notification check if spun down.

How to reproduce: Spin down unassigned disks, then turn on System Notifications.  Disks will spin up on the next notification scan.

Expected results: Unassigned disks should stay spun down.

Actual results: Disks spin up on the notification scan.

Other information: This doesn't appear to happen in all cases.  There seems to be a disk/controller combination that spins up the disk when SMART check is performed.

 

System was booted with no plugins.  Diagnostics are attached.  This is a backup/test server and the unassigned disks are an SSD, and a eSATA attached docking station.  Let me know if you need me to run some tests.

backupserver-diagnostics-20170522-0839.zip

Edited May 22, 20179 yr by dlandon

I have the same issue. And have the following theory. First let me state my assumptions:

 

1) reading the temperature or other smart attributes with smartctl or any other tool resets the internal idle timer of a HD (there is no known way around this)

 

2) reading the temperature or other smart attributes spins up a disk if that disk was previously spun down (there is no known way arround this)

 

3) therefore all reads of smart attributes in unraid check the spin status of a hd first (this can be checked without waking a drive up and it also does NOT reset the idle timer) - if spun down, no attributes are read

 

4) if unraid poll_attributes is set to 1800 (= 30 minutes) and all disks are at their default internal setting hdparm -S 240 (= 20min before spinning down), then disks will eventually spin down, Once spun down they will not spin up because of polling, as all attribute reading in unraid (should) first check the spin status and skips reading attributes from spun-down drives

 

5) Now the problem: If a drive is spinning and the tunable poll_attributes is less than the internal hd idle timer (if for example polling takes place every 19 minutes) it will never spin down again

 

6) Array drives are managed by mdcmd which has its own independent idle counter (disk settings) and  sends them to sleep with "hdparm -y" if the specified idle time is exceeded. Therefore the array drives will properly spin down no matter when and how often you poll the smart attributes (and remember unraid skips reading from spun down drives - so 2 does not apply). BUT all other drives rely on the mechanism under 1-5

 

With this in mind I believe that the system_notifcations might not check for spin status before polling and/or poll at a different (shorter) intervall than specified in the disk-settings so that these non-array drives (e.g. unassigned devices) once spun up will never spin down or spin up when polled by system notification.

 

On a related note: If somebody here has spindown problems with non-array drives (e.g. unassinged devices) check If other scripts or plugins (e.g. autofan) poll smart attributes and/or hd temperature in intervals smaller than the internal hd idle timer and turn (for the time beeing) system notifications off.

 

Also - as unassigned devices might soon be part of the core unraid system, maybe the devs will also implement spin down for all drives with an externally managed idle timer, just as the array drives spin-down is already managed by mdcmd. Outside of unraid there are a lot of linux scripts doing this (e.g. https://zackreed.me/spin-down-idle-hard-disks-without-hdparm/). Also according to that link there are different ways to check for idle status - maybe system_notifications uses the wrong one which only works on some drives but not all.

Edited July 18, 20178 yr by Videodr0me