[SOLVED] Minimize backup size - how to?

[Razva]

Hey,

I'm using Proxmox 5 and local ZFS for storage.

I've created a KVM VM, used the ZFS mount for storage, allocated 100GB of space, checked "discard" and installed the OS.

Immediately after OS install I've made a backup by using vzdump. Backup size was ~5 GB. Everything OK 'till here.

Next, I've downloaded 10 + 10 + 10 + 10 + 10 GB of data. All data files were different in order to keep ZFS dedup out of the way. "df" reported ~45 GB of used space. I've made a new vzdump, backup size was ~45 GB. Everything OK 'till here.

Lastly I've removed all "data" test files. "df" reported ~5 GB of used space. But when making the vzdump...the backup size remained at ~45 GB.

Is there any way to really discard unused data without interfering with the VM? I cannot ask everybody to allow me root access in order to manually clear/discard space via OS/VM.

Thanks,
Razvan

 

[Alwin]

Are your VM disks SCSI? Only then the discard is passed through to the underlying system.

 

[Razva]

Are your VM disks SCSI? Only then the discard is passed through to the underlying system.

No, they are NVMe.

 

[Alwin]

No, they are NVMe.

NVMe is your physical storage, how are the disks of the VM configured? With SCSI/SATA?

Only then the discard is passed through to the underlying system.

 

[Razva]

NVMe is your physical storage, how are the disks of the VM configured? With SCSI/SATA?

Sorry about that, it's SATA not SCSI.

 

[Alwin]

Did you activate the discard option on the VMs disks and does the OS inside the VM also do discard (trim)? Also check if your SCSI controller type is VirtIOSCSI.

 

[Razva]

Did you activate the discard option on the VMs disks and does the OS inside the VM also do discard (trim)? Also check if your SCSI controller type is VirtIOSCSI.

As previously specified I'm using SATA, not SCSI. So I guess the "Discard" option does nothing?

 

[Alwin]

SATA & SCSI in combination with the VirtIOSCSI controller can passthrough discard.
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#qm_virtual_machines_settings

 

[Razva]

SATA & SCSI in combination with the VirtIOSCSI controller can passthrough discard.
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#qm_virtual_machines_settings

Asked my provider, we're using NVMe that are connected to the motherboard via M.2.

Here are the storage settings:

Here are the VM settings:

The OS is CentOS 7, Minimal.

The behaviour is the one specified in the initial post.

Do I need to set anything OS-specific?

 

[Alwin]

Try fstrim -v / inside the VM, this should release the blocks if everything is working correctly.

 

[Razva]

Try fstrim -v / inside the VM, this should release the blocks if everything is working correctly.

Yeah, I just read that CentOS doesn't enable TRIM or discard by default. Things look better after running trim, so I guess this issue is fixed.

For those who have the same issue, just enable fstrim and you'll be good to go.

 

[Razva]

For those looking for the solution, long story short:

• use ZFS as your storage
• use VirtSCSI as "Bus/Device" when creating your VM
• setup TRIM in the OS (it's not installed by default on CentOS / Debian / Ubuntu) or DISCARD. Note that "fstrim" needs to be manually executed so you should set a daily cron task that runs before doing the backups

I hope this helps.