iSCSI ZFS ZVOL VolBlocksize for Windows & Linux VMs

[fahadshery]

Hi,

I am creating z ZVOL to pass it as an iSCSI device to proxmox so that I could create windows/linux VMs on it.
As we know that ZVOL volblocksize (if you get it wrong) can cause padding overheads as well as performance hits!

I know, if you wish to run a VM with EXT4 or NTFS, It doesn't matter what the workload is - Ext4 will always write 4k chunks. As will the NTFS…
and
VMware (VDI) is typically around 24 - 32K but that is accessed via iSCSI/FC.

I was wondering what could be the ZVOL volblocksize for Proxmox?

thanks

 

[fahadshery]

so what you guys are volblocksize for ZVOL passed to Proxmox using iSCSI to create VMs on?
Do you guys use NFS for it? If yes then what's the ZFS dataset record size you're using for it?
thanks

 

[bbgeek17]

The answer is a usual "it depends". The recommendation should be sought from your storage vendor/support. Proxmox does not generate IO on its own, its the application in your VM that does. There could be many layers in between: App>VM Kernel>VM File system>VM LVM>Qemu>Hypervisor LVM>iSCSI>Storage ZFS. Things could be combined/merged/optimized at various layers.
You'll need to understand what your IO pattern is and then apply that to your Storage Vendor recommendations.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox