Disk compression feature
- Thread starter r59461900
- Start date
I am using the default filesystem: ZFS.
It gives be both a) sparse files = only blocks actually containing data will occupy space on disk and b) transparent compression = everything a VM writes gets compressed on the fly, usually (not guaranteed) by a factor of 1.5:1 or better. (And the compression can be fine-tuned for every dataset/ZVOL to get a higher ratio or less CPU load.)
Choose your tool-set, choose wisely - changing it after the fact is... difficult.
You may possibly want to place a feature request over there, in the official tracker: https://bugzilla.proxmox.com/ (...and link it here, in both directions.)
Disclaimer: I am a ZFS fan and there are other storage solutions / filesystems which may be a better choice for some situations / tasks.
It gives be both a) sparse files = only blocks actually containing data will occupy space on disk and b) transparent compression = everything a VM writes gets compressed on the fly, usually (not guaranteed) by a factor of 1.5:1 or better. (And the compression can be fine-tuned for every dataset/ZVOL to get a higher ratio or less CPU load.)
Choose your tool-set, choose wisely - changing it after the fact is... difficult.
You may possibly want to place a feature request over there, in the official tracker: https://bugzilla.proxmox.com/ (...and link it here, in both directions.)
Disclaimer: I am a ZFS fan and there are other storage solutions / filesystems which may be a better choice for some situations / tasks.
Last edited:
Last edited:
- run "zpool set autotrim=on YOURPOOLNAMEHERE" to enable a continuous, automatic TRIM. Without this, you only have the default PVE schedule where TRIM is run shortly after midnight on the first Sunday of each month. Use "zpool trim" to run it on the pool immediately.
- edit /etc/pve/storage.cfg and add "sparse" or "sparse 1" for any ZFS pool(s) created after install to enable thin provisioning.
- edit /etc/pve/storage.cfg and add "sparse" or "sparse 1" for any ZFS pool(s) created after install to enable thin provisioning.
zpool trim isn't related to the virtual disk's size though. I have some ZFS specific trim docs here as well.Yes, I mentioned it partially for completeness but also for VMs running ZFS, such as PBS.
The arch and debian wiki recommend against continous trim, sadly with no further explanation except freezes on old sata-hardware: https://wiki.archlinux.org/title/Solid_state_drive#TRIM
Does anybody here has an idea what woukd be a safe default?
Does anybody here has an idea what woukd be a safe default?
I mount the hard drive to a directory via /etc/fstab, and the disk format is EXT4. This way, changing the order of the hard drive data cables will not affect the directory mounting of the hard drive. The hard drive contains only test virtual machines, so data loss is acceptable. I back up all virtual machines every six months, and when the SSD health drops below 90%, I replace it with a new SSD.
Attachments
Last edited: