Trim in the VM

[stenzelprox]

Hello everybody,
i have some questions regarding Trim in VMs on ZFS.

In VM Config discard is enabled

root@fal-pve01:~# cat /etc/pve/qemu-server/107.conf
agent: 1
bootdisk: scsi0
cores: 6
cpu: host
memory: 12288
name: fal-mail01
net0: virtio=XXXXXXXXXXXX,bridge=vmbr1
numa: 0
onboat: 1
ostype: l26
scsi0: local-zfs:vm-107-disk-1,cache=writeback,discard=on,size=99G
scsihw: virtio-scsi-pci
smbios1: uuid=286ed534-4141-4730-bd1f-ac3a82XXXXX
sockets: 1

Does discard=on only say that it is possible ?
Do I still have to activate VM Trim ?

The system runs on two NVME SSDs (TOSHIBA KXG50ZNV512G).

 

[spirit]

discard=on allow trim/discard instruction to go from to vm to the storage.

you still need to send trim command from the vm. (fstrim command on linux, or discard option in /etc/fstab)

 

[stenzelprox]

Thanks for the info.

I have now added discard in the fstab as mount option.

root@fal-mail01 ~ # cat /etc/fstab
...
UUID=d5b559e4-4ef1-45a5-813c-ecXXXXXX1 /               ext4    discard,errors=remount-ro 0       1
...

And once entered

fstrim / -v

 

[spirit]

note that your don't need discard option in /etc/fstab to have fstrim working.

discard option in /etc/fstab discard directly when you delete a file. but be carefull, only recent linux kernel are doing this async. (>=4.7).
if not, you'll wait on each file delete that discard is finished.

 

[stenzelprox]

Thanks again for the information.
The kernel shouldn't be a problem as far as I know we have nothing older than Debian Stretch.
But is batched discard or online discard better recommended ?
And if I use batched discard how often should it run the trim ?