After updating to 25.10.0, I noticed that my disks wake up every 90 minutes. This was not the case in the previous version 25.04.2.1. To illustrate the behavior, I have attached a screenshot of the power consumption.
I used these settings:
HDD Standby: 30
Advanced Power Management: Level 1
In order to narrow down the problem source, I have stopped all apps and disabled all shares. I don’t run any VM’s. So I suspect there is some activity of TrueNAS itself that accesses the discs, but I have no idea where to look or what to do, and so I would appreciate any help.
EDIT: I should mention that the boot pool is on different drives.
I think this is more than just a bug. I think it is a question of what TrueNAS is supposed to be used for. It seems that TrueNAS is always optimized for performance.
I run two TrueNAS servers: One main server and a backup server with a replica of the main server. Both have spinning drives for capacity and SSD’s.
Needless to say that the backup server sees very little activity. The only thing that happens on the server is that it periodically gathers backups from other computers and runs a TrueNAS replication task. To ensure that drives will sleep, I have disabled SMART data collection.
Ideally, it would almost always sleep and thus use minimal power. It would only wake when the admin connects to the GUI and when there was a task to run. That has not been possible to configure easily. Until now has consumed about 150 USD/EUR of power each year. With the change to SMART data collection in 25.10, it will jump to about 400 USD/EUR (based on my measurements).
I really like TrueNAS, but right now I am considering dropping TrueNAS on the backup box and just run bare Linux with OpenZFS. That way I would have far more control over power usage. Obviously, that would come with added complexity. Ideally, TrueNAS would have a general setting where one could choose its role. Application server, file server, backup server, etc.
That way decisions affecting power usage could be made with that in mind though it would require more testing.
Just FYI, I have no reverted back to version 25.04.2.1, and everything is back to normal. I do hope this will be fixed soon (the bug has been reported October 4, 2025).
FWIW: I have done this and it was not difficult at all if you know your way around Linux & ZFS.
Use Case:
Wanted a universally recoverable solution eliminating TrueNAS from my offsite recovery and be able to recover anywhere on anything I could get my hands on that runs Linux. TrueNAS is not always friendly here.
Ability to export and import a rotating set of 4 physical hard drives to offsite storage on a quarterly basis.
TrueNAS remains primary and backup replica targets for clients and mounts.
How:
Using Ubuntu 24.04 LTS, Webmin and Poolsman ZFS management interface. CLI is cool but much easier in a GUI to deal with ZFS…
Built an Ubuntu server, configured OpenZFS, established SSL logins & credentials, replicated a seed set of snapshots via mbuffer and zfs send/receive streams.
Daily a TrueNAS cron job executes zfsbud which handles the replication tasks of existing TrueNAS snapshots. TrueNAS jobs are ok but I didn’t want to have a dependency on TrueNAS schedulling, UID GID or be bound to my TrueNAS snapshot pool. Instead, I selectively grab weekly and monthly snapshots for the quarter and push them across to the offsite box.
Just watched the latest episode of TrueNAS Tech Talk. They just say “don’t do spindown” and that the new behaviour is “intentional design”. The UI still exposes the disk setting “HDD Standby” which will certainly not work anymore. I do hope they fix this issue, but I fear their “solution” will just be to remove the setting from the UI.
I saw it too. I understand that the people who pay the bills are enterprise customers and that we are just along for the ride.
I understand that they want to hide SMART because they think results are sometimes wrong and generally not trustworthy. That seems reasonable.
What I do not understand is that they will not allow us to change the frequency of polling or disable it entirely as it causes major problems for low power environments. Especially if SMART is often wrong.
I do think they are wrong when they say that the system needs to write often to disks. While that is true for some systems, other systems spend lots of time during the day with no users accessing it and no transactions at all to the some of the ZFS file systems. In particular I can say that my backup server only see action every three days. I can see very clearly on the attached power meter whenever the disks are spinning.
For me, the best thing would be if they allowed the system to essentially sleep keeping only the boot pool and a bit more running and allow everything else to power down. I suspect small companies would like something similar as it would allow them to place a backup server outside an expensive server room; in a place without special cooling and where noise could be an issue. That way it could be geographically separated from the main server without much cost.
Well we will see what happens. I will stay on 25.04 for now. I will then investigate what it takes to make a bare linux backup server that is able to sleep. I am considering a setup where the system unmounts the spinning drives and only mount them on demand or when it needs to do a backup. I think that would be possible. Power in this part of Europe is far too expensive to just waste it like this. With a little luck I would be able to make the backup server in question use something like 10W while “sleeping”. If all it does is to wait for an incoming connection then that should be doable.
I saw that as well, and it sounded more like the bug for them isn’t that you set your disk to standby and then it still wakes up every 90 minutes — but that you can set your disk to standby at all. The “we have to make that call for the safety of your data” argument doesn’t really sit with me.
I only partly understand the arguments being made. First, there’s the data‑integrity/safety aspect: my pool is at risk if ZFS can’t write a few bytes every five seconds even when there isn’t any activity on that pool for hours. Really? I kind of doubt it — when the HDD spin down worked fine on 25.04.
Then there’s the business argument. Sure, there may be liability or warranty issues for the systems sold, and if they decide that no spin‑down should be allowed, they could disable the ability to set disks to standby. But I’m wondering because power efficiency and savings are still a big deal for large businesses in many parts of the world. Many companies still work on the traditional five‑day week with up to 12 hours of activity per day—there’s a lot of energy that can be saved by spinning down disks during the rest of the week.
For the community edition of TrueNAS I don’t get that at all. Please give me the option to decide if and when I want to spin my drives down. If you think it’s that dangerous, remove the option from the GUI but write an article in your wiki with proper warnings on how to disable checks and enable HDD standby via the shell. To be fair, QTS also has no GUI option for enabling HDD spin‑down you have to tinker with the system in the shell, but it works great and flawless for me over the past ten years.
I totally get the energy‑saving argument. I just migrated from a QNAP that I used for about five years: four Intel cores, 16 GB RAM, two NVMe drives, and six 6 TB HDDs. With HDD standby and running apps/containers on the NVMe mirror, the whole system drew about 25 W. When the drives were active it drew 55 W; the spinning drives used more energy than the system itself. It’s not only the €50+ I save per year by spinning down — it’s also an absolute waste of energy to keep them spinning 24/7 für my home application when the disks are only used for maybe eight hours a day probably less on average over the year.
That said, when I migrated to TrueNAS I switched to three refurbished 16 TB drives and now only save about 10 W when spinning down (≈40 → 30 W), so the concern is smaller. Still, I’d like to spin the drives down.
Yeah, issue is the noise for me here… I’m not gonna sleep next to an array of enterprise hard drives a metre or two from my bed.
This not working/intentionally not working now means I have to screw around with booting up and shutting down the machine for the weekly archival jobs instead of just letting it run and only spin up when it needs..
