[Help] TOS 7.0.1341 – HDDs don't hibernate while the web UI is open (storagemanager runs "vgs" every few seconds)

[neoz]

Hi all,

After updating to TOS 7.0.1341, I noticed that my HDDs don't hibernate while the TOS web interface is open in a browser. As far as I remember (I'm almost certain), this didn't happen before this version: having the UI open didn't wake the disks, and they only spun up when I accessed Volume 2. Is anyone else seeing this?

Setup
- Model: F4-424 Pro
- TOS 7.0.1341
- Volume 1: 2 × NVMe SSD, RAID1, Btrfs (system, apps, Docker)
- Volume 2: 2 × Seagate IronWolf 6 TB (ST6000VN006), RAID1, Btrfs (data only)
- HDD hibernation enabled

What happens
- With the web UI open, one of the two HDDs stays active indefinitely. If I put it into standby manually ("hdparm -y"), it spins up again within a second.
- With the browser closed, both HDDs go into standby about 15 minutes later and stay asleep (tested for 30 minutes).
- Logging into the web UI wakes both disks right away: the attributes (ctime) of the shared folders on Volume 2 are modified at that moment.

What keeps the disk awake
I traced block requests with the block_rq_issue kernel tracepoint. Over 10 minutes on the active HDD, with the UI open:

Code: Select all

    450 smartctl N   (SMART passthrough, only while the disk is already awake)
    368 vgs R        (reads)
     80 <...> R      (reads from short-lived processes, most likely vgs too)
     10 hdparm N     (my own power-state checks)

After putting the disk into standby manually, vgs wakes it 0.3 s later and keeps reading:

Code: Select all

hdparm-201955  [001] ..... 34278.985884: block_rq_issue: 8,16 N 0 () 0 + 0 none,0,0 [hdparm]
vgs-201956  [000] ..... 34279.324896: block_rq_issue: 8,16 R 131072 () 20848640 + 256 be,0,4 [vgs]
vgs-201956  [000] ..... 34282.252023: block_rq_issue: 8,16 R 131072 () 11718747648 + 256 be,0,4 [vgs]
vgs-202025  [007] ..... 34292.372089: block_rq_issue: 8,16 R 131072 () 20848640 + 256 be,0,4 [vgs]

The parent process is /usr/sbin/storagemanager, running "vgs --noheadings -o vg_name" every few seconds (about 100 runs in 10 minutes) while the UI is open. Since Volume 2 is RAID1 and vgs only reads, each read is served by one disk, which is why one HDD stays awake while the other one sleeps.

Ruled out
- System (md9) and swap (md8) are only on the NVMe SSDs.
- No disk or link errors in dmesg.
- The smartctl calls from TOSDaemon, ter_smartfan and ResourceCollection use "-n standby", so they don't wake a sleeping disk.

Questions
1. Is anyone else on 7.0.1341 seeing this?
2. Is this a known issue? Could storagemanager poll LVM less often, or use cached information, while the UI is open?

Thanks!

[TMzethar]

We are currently looking into the issue you mentioned. Once we have completed our verification and discussed it, we will update you here.

[neoz]

We are currently looking into the issue you mentioned. Once we have completed our verification and discussed it, we will update you here.

Thank you!!