Snapshots spinning up pools not related to snapshot task

[renaX]

Hi,

I am running TrueNAS Scale 25.04.2.4, on a system with 2 pools, main(2x6TB HDD in a mirror) and secondary(a 2TB NVMe SSD), main is used for backups of secondary and as a the main mass storage space, secondary as a drive for VM’s, docker containers etc. This has been done to make the server power consumption as low as possible (main being spun down for the majority of the day).

Recently I made a snapshot task that takes snapshots every hour (I’ve got reasons) of a dataset on the pool secondary, but I’ve noticed that every time this specific snapshot task starts, it also spins up the main pool even though the snapshot is being done on a different pool, and there isn’t an immidiate replication task after it. This is further supported by the fact that there’s no data being written/read off of main during these spin-ups (accoring to Disk I/O reports).

Putting away the fact that spinning up drives for 5 minutes 24x times a day is not healthy (both for my sanity and my drives), can I somehow prevent spinning up drives of pool main while taking a snaphot of a dataset on the pool secondary?

Also since I was asked here’s:

System info

- CPU: i5-6500 - MB: ASUS B150M-C - HDD's: HGST-HUS726T6TALE6L4 (6TB-SATA) - SSD: WD Red SN700 2000GB - Boot SSD: INTEL SSDPEKKF256G8L - Power consumption: ~28W (0.8kW/h a day) *(noone cares anyway)*

[SmallBarky]

Where do you have the system dataset stored?
Menus System, Advance Settings, Storage?

[renaX]

My system dataset is in boot-pool which is located on a dedicated drive.

Its a 256GB Intel drive.

[renaX]

Okay so I’ve done some digging, and a similar issue discussed elsewhere made me realize that it could be caused by a bad configuration of a replication task.

To give context, the dataset Im trying to take snapshots of hourly, is a sub-dataset of dataset apps, let’s call it mc-server. I’ve also got a daily snapshot task that takes snaphots of all the datasets inside apps. Here’s the config for the task.

And here’s the configuration of the connected replication task

Originaly I had only Run Automatically ticked, which ment that the task would start

immediately after the linked periodic snapshot task completes

as the tooltip states. This made me think that, for some reason, the task would get triggered by the hourly snapshot task. That theory was wrong, because when I also ticked the Schedule option, the periodic snapshot task still spun the drives up.

Then I thought, maybe its because this snapshot task doesn’t have a linked replication task, which as far as I know shouldn’t make much of a difference, but I figured it could be worth the try. I setup a linked replication task that was set to run daily, and wouldn’t you know it, absolutly no change whatsoever.

I also set the hourly snapshot task to not allow taking empty snapshots, and the drives spun-up despite the fact that there where no shapshots taken, which further proves my point of something waking the drives up, just because there’s a snapshot task running.

Just to prove that Im not just making this up, this is the server’s power consumption today. I enabled the hourly snapshot task at 11:00, that was a non empty snapshot, and the rest are empty snapshots. (under normal circumstances this would be a relatively smooth line)

What was also weird was that at each of these drive wakeups I’ve got some I/O activity on the main pool. This could of course be just the system reading some data at the start.

At this point I’m starting to think that this might just be a “feature”.

[Jaywalker]

I think this is the same issue that I reported here almost 1.5 years ago:
https://ixsystems.atlassian.net/browse/NAS-128946

It has been closed by the Truenas team without a fix since they do not officially support HDD Standby

[renaX]

Well that’s disappointing to say the least.

Having the drives spin-up 24x a day is not something that I’ll allow happening, and not spinning the drives down is a 33% power consumption increase of my machine so that’s a no-go as well.
With 30W thats about 65EUR yearly, while with 40W that’s 87EUR yearly. Call me cheapskate.

I guess I’ll have to find a different tool for making more frequent backups.

[urbanvaughn15]

I noticed the same issue where snapshot operations randomly spin up unrelated pools. It seems tied to how the middleware queries dataset metadata during replication or snapshot listing. Switching from recursive to targeted snapshots reduced unnecessary pool activity. Hopefully, a future TrueNAS update optimizes this behavior since idle disks shouldn’t spin up just for catalog reads.