spindown-v25.patch: trailing newline in standby_disks prevents last disk from being recognized

[d22b]

The TrueNAS v25 spindown patch reads /run/middleware/standby_disks using:

file.read().split(',')

If the state file contains a trailing newline, for example:

sda,sdc,sde,sdf,sdg\n

the resulting list is:

['sda', 'sdc', 'sde', 'sdf', 'sdg\n']

Therefore the last disk is no longer matched correctly:

'sdg' in file.read().split(',')
# False

This causes the standby guard in DiskEntry.temp() to fail specifically for the last entry in standby_disks.

Observed behavior

On TrueNAS SCALE 25.10, five HDDs were configured with a 10-minute standby timer.

Four disks consistently entered standby while one remained active/idle.

Initially this appeared to be related to a specific HDD, /dev/sdX assignment, or SATA port. Multiple physical disk/port swaps showed that this was not the case: the affected physical disk changed.

Tracing with bpftrace finally showed recurring 512-byte device requests against the disk that would not enter standby.

Example:

14:58:08 BLOCK ... comm=python.d.plugin bytes=512 sector=0 nr_sector=0
14:59:53 BLOCK ... comm=IoThread        bytes=512 sector=0 nr_sector=0
15:03:08 BLOCK ... comm=python.d.plugin bytes=512 sector=0 nr_sector=0
15:04:53 BLOCK ... comm=IoThread        bytes=512 sector=0 nr_sector=0

The processes were identified as:

• python.d.plugin — TrueNAS/Netdata disk-temperature collector
• middlewared / IoThread — middleware disk-temperature polling

The Netdata temperature collector runs every 300 seconds.

Isolation of the device request

Individual DiskEntry properties were tested while tracing block:block_rq_issue.

The following produced no disk request:

DiskEntry.serial
DiskEntry.lunid
DiskEntry.identifier

Only:

DiskEntry.temp()

generated the 512-byte disk request.

Before the fix:

/run/middleware/standby_disks:
'sda,sdc,sde,sdf,sdg\n'

sdg present: False

DiskEntry(name="sdg", devpath="/dev/sdg").temp()
→ TempEntry(temp_c=41.0, crit=70.0)

This means the standby guard was not triggered and temp1_input was read through the drivetemp kernel driver.

That temperature read issues a command to the HDD and resets/restarts the ATA idle timer before the configured standby timeout can expire.

Root cause

Two locations in spindown-v25.patch parse the file without removing trailing whitespace:

file.read().split(',')

The final element therefore retains the newline.

This also explains why the problem appeared to move between disks after /dev/sdX assignments changed: it followed the final entry in standby_disks, not a specific physical HDD or SATA port.

Proposed fix

In middlewared/utils/disks_/disk_class.py:

- if self.name in file.read().split(','):
+ if self.name in file.read().strip().split(','):

And in middlewared/plugins/disk.py:

- disk_list = file.read().split(',')
+ disk_list = file.read().strip().split(',')

A more defensive alternative would be:

[x.strip() for x in file.read().split(',') if x.strip()]

although .strip().split(',') is sufficient for the observed state-file format.

Verification

After changing both reads to:

file.read().strip().split(',')

the same direct temperature test returned:

TempEntry(temp_c=None, crit=None)

and the block_rq_issue trace remained completely empty.

Finally, all five HDDs were tested simultaneously with a 10-minute ATA standby timer.

Result after approximately 11 minutes:

sdg    drive state is: standby
sde    drive state is: standby
sdf    drive state is: standby
sdc    drive state is: standby
sda    drive state is: standby

So the change reproducibly fixes the spindown failure.

Environment

• TrueNAS SCALE 25.10
• Native TrueNAS HDD standby: 10 minutes
• TrueNAS-Tricks v25 spindown patch
• Multiple SATA HDDs
• Verified using bpftrace on block:block_rq_issue