RAIDZ1 shows wrong space?

[.n3]

Hey,

I created a raidz1 with 5x 4TB disks and the availible space should be approx 16TB, because one drive is used for parity. But proxmox shows me 20TB on free disk space? Is it correct?

 

[.n3]

Hey, I think the configuration is right, because in the Overview of the disk I see the utilisation is xx from 15,84TB.

 

[Dunuin]

The "zpool" command will always show raw capacity (incl. parity) while the "zfs" command will show the size with parity already subtracted.

You might also want to search this forum for "padding overhead" because with default values your 20TB pool will only allow you to store 10TB of VM virtual disks (or only 8-9TB in case you care about performance).

A volblocksize of 32K or higher would be required with a 5 disk raidz1 using an ashift=12, if you don't want to waste space when using zvols.

 

[.n3]

Thanks for the advise. I wrote, that with a hoher blocksize the I/O will be worse and not good for DBs. I want to use a single 1TB M.2 SDD for container and vms and the RAIDZ1 will be only a data tank for nextcloud and frigate (securitys camera videos). So it would not be a probleme to setup the blocksize to 32k, right? r should I use 64k?

 

[Dunuin]

I would try to keep the volblocksize as small as possible, so 32K. And yes, shouldn't be that problematic when mostly storing big files like videos.

 

[.n3]

Thanks for the info. How can I calculate, how much padding I will have with 8K, 16k, 32k, or 64k?

I also read where you suggest how to change the blocksize, but I use directories in my RAID and it is not possible to change the blocksize:

I can only change the blocksize for ZFS but not for the directory (Verzeichnis). I don't think so, but it is possible to change the blocksize for a directory?

 

[Dunuin]

How can I calculate, how much padding I will have with 8K, 16k, 32k, or 64k?

See here:
https://web.archive.org/web/2021031...or-how-i-learned-stop-worrying-and-love-raidz

I can only change the blocksize for ZFS but not for the directory (Verzeichnis). I don't think so, but it is possible to change the blocksize for a directory?

Those Directory Storages are probably using datasets and datasets use recordsize instead of volblocksize. And padding overhead only affects zvols, not datasets.

 

[.n3]

Thanks! I'm using 5 disks so I have 25% overhead with 8KB, 16KB and 32KB. So I can use 8KB. But as long as I don't use zvols it is not important for me.

When do we use zvols?

 

[Dunuin]

Thanks! I'm using 5 disks so I have 25% overhead with 8KB, 16KB and 32KB. So I can use 8KB. But as long as I don't use zvols it is not important for me.

When do we use zvols?

PVE will create virtual disks as zvols when you create a VM on a ZFSPool storage (like the "local-zfs" or oyur "vm_container").