ZFS TRIM on Proxmox
- Thread starter norsemangrey
- Start date
The man page for
[1] https://openzfs.github.io/openzfs-docs/man/8/zpool-trim.8.html
[2] https://openzfs.github.io/openzfs-docs/man/8/zpoolprops.8.html?highlight=autotrim
zpool trim [1] provides a good explanation. Basically TRIM is used to discard unused storage blocks in the filesystem. This improves SSD performance and lifespan as it does not need to continually delete old blocks on every write. It also allows thinly provisioned storage to reclaim space that is unused by the guest operating system.There are two main differences here. The first and most obvious is that
autotrim=on allows trim operations to run automatically in the background while zpool trim must be called manually. The second is that zpool trim reclaims every free block it finds, while the autotrim feature waits until a large enough range of blocks have become free before it trims, for the sake of optimization [2].The zpoolprops man page [2] also mentions that autotrim can put significant stress on the underlying storage devices, depending on their quality. With this, it also implies that for lower end (non-enterprise grade) devices, running
zpool trim periodically could be the better option. You could, for example, set up a cronjob to run zpool trim as often as you feel appropriate for your system and at times of lower activity.[1] https://openzfs.github.io/openzfs-docs/man/8/zpool-trim.8.html
[2] https://openzfs.github.io/openzfs-docs/man/8/zpoolprops.8.html?highlight=autotrim
Last edited:
Interesting. I was not aware of zpool trim, but used fstrim -a instead, but I was not shure if it works, since the zvol are not mounted on the host.
It would trim the root filesystem, if you are using ZFS root. But it would not trim the ZVOLs that are the backing store for the VM's drives.
How is trim or autotrim used or set in proxmox, GUI or commandline? I have a PVE host w. 4x8TB RAID Z1-0 and only ONE large CT of 12TB.
rpool has 31,99TB but only 320GB free, 31,66TB allocated. I wonder where the "missing" 20TB went (o.k. ~20% for the RAID blocks, the OS, meta-data, logs, etc. but still...) There are no ZFS snapshots, no VZdumps or anything else that uses up more space then it should.
So I looked into trim and only got like:
What am I doing wrong? Was the host initially wrongly configured? How can I get autotrim to run and automatically free some space so the disks won't fill themselves up all the time?
Cheers,
~R.
rpool has 31,99TB but only 320GB free, 31,66TB allocated. I wonder where the "missing" 20TB went (o.k. ~20% for the RAID blocks, the OS, meta-data, logs, etc. but still...) There are no ZFS snapshots, no VZdumps or anything else that uses up more space then it should.
So I looked into trim and only got like:
Code:
:~# zfs get autotrim
bad property list: invalid property 'autotrim'
:~# zfs set autotrim=on rpool
cannot set property for 'rpool': invalid property 'autotrim'
:~# zfs get autotrim rpool
bad property list: invalid property 'autotrim'
:~# zpool get trim rpool
bad property list: invalid property 'trim'
:~# zfs version
zfs-2.2.8-pve1
zfs-kmod-2.2.8-pve1
:~# modinfo zfs | grep version
version: 2.2.8-pve1
srcversion: 571935691D8EEAF8FF853F9
vermagic: 6.8.12-15-pve SMP preempt mod_unload modversions
:~# zpool get all rpool | grep -i trim
rpool autotrim off default
:~# zfs set autotrim=on rpool
cannot set property for 'rpool': invalid property 'autotrim'What am I doing wrong? Was the host initially wrongly configured? How can I get autotrim to run and automatically free some space so the disks won't fill themselves up all the time?
Cheers,
~R.
RTFM.
You can setup a cron job with something like:
see:
Code:
#zpool get/set autotrim
zpool get autotrim
zpool set autotrim=[on,off]You can setup a cron job with something like:
Code:
zpool trim <zfs-pool-name>man zpool trim
Last edited:
Are you sure you need autotrim and that the behaviour you see is related to it? Also there's already a cronjob doing a zfs trim for you.
I am absolutely NOT sure, but I now did
zpool set autotrim=on rpool (cheers @news :°) and hope it will free up some space 
Where can I see this CRON job? Doenst show in the normal crontab.
What other issue could fill up the disk or why is it so full if trim is run regularly?
This is the content of
As far as I know, trimming, or lack thereof, should not affect the available space of the pool. This is before and after for one of my (non-RAID) pools
I'd recommend you create a new post for your space usage issue as I don't think it's related. If you do, please make sure to share at least this
No need to quote, my message is directly above yours.
/etc/cron.d/zfsutils-linux
Bash:
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin
# TRIM the first Sunday of every month.
24 0 1-7 * * root if [ $(date +\%w) -eq 0 ] && [ -x /usr/lib/zfs-linux/trim ]; then /usr/lib/zfs-linux/trim; fi
# Scrub the second Sunday of every month.
24 0 8-14 * * root if [ $(date +\%w) -eq 0 ] && [ -x /usr/lib/zfs-linux/scrub ]; then /usr/lib/zfs-linux/scrub; fiAs far as I know, trimming, or lack thereof, should not affect the available space of the pool. This is before and after for one of my (non-RAID) pools
Bash:
NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
bigdata 6.98T 4.20T 2.78T - - 34% 60% 1.00x ONLINE -
pve-storage: Fri Dec 12 10:24:56 2025
NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
bigdata 6.98T 4.20T 2.78T - - 21% 60% 1.00x ONLINE -I'd recommend you create a new post for your space usage issue as I don't think it's related. If you do, please make sure to share at least this
Bash:
zpool status -v
zfs list -t all -ospace,refreservation
Last edited:
trim doesn't free up space - it just helps the SSD flash controller know which flash blocks do not have data in them, so it can pre-erase those so they are ready for use when needed.
It can when called inside a VM, for example. I just wanted to mention this becuase I feel like I generalized above.
Last edited:
/usr/lib/zfs-linux/trim is for cron.
from code looks like it:
- trim all healthy pools,
- with no trim running,
- with zfs property "org.debian
eriodic-trim" (auto|on), default=auto - and auto trim only if pool has nvme disk.
timer is for systemd.
systemd service looks like this:
Code:
# cat /usr/lib/systemd/system/zfs-trim@.service
[Unit]
Description=zpool trim on %i
Documentation=man:zpool-trim(8)
Requires=zfs.target
After=zfs.target
ConditionACPower=true
ConditionPathIsDirectory=/sys/module/zfs
[Service]
EnvironmentFile=-/etc/default/zfs
ExecStart=/bin/sh -c '\
if /usr/sbin/zpool status %i | grep -q "(trimming)"; then\
exec /usr/sbin/zpool wait -t trim %i;\
else exec /usr/sbin/zpool trim -w %i; fi'
ExecStop=-/bin/sh -c '/usr/sbin/zpool trim -s %i 2>/dev/null || true'check code and you can correct me.
you can judge on your own. bud every has it purpose.
if you are activating systemd service, you know you have nvme there, and want to trim it at specified timer (weekly, monthly...).
looks like this service is not looking for health status of pool. this looks like only drawback.
Last edited: