Hi, is it a good idea to passthrough a virtual drive for a fileserver VM? I don't have CEPH running. Just 2 server with shared local ZFS pools. I want to create virtual drives because I want to use replication and to be able to migrate the VM to the other server. What problems may occour?
[SOLVED] Fileserver: Passthrough physical drive or virtual?
- Thread starter m4rtin
- Start date
Actually I do not understand that question. Usually you assign a virtual disk to a VM. That's it.
These "normal" virtual disks can get migrated to other nodes. (As long we are talking ZFS, which you do.)
(( Passthrough is used if a VM (usual example: TrueNAS) needs to have direct access to the drive. In that use case "passthrough a controller, not a disk" is my personal recommendation. To passthrough a virtual disk does not grant direct access to the hardware, even if the block device is a physical disk. ))
Please elaborate your use case
Usecase is a fileserver (Debian with SMB shares) with multiple user accessing it at the same time. I read in some forum posts that in these cases you should passthrough the physical controller or disks. I don't know if that is still the case and why, that is my question.
If you think of virtualisation then use it for your datapools to,
on lxc you can use mout points to a zfs datapool.
# https://pve.proxmox.com/wiki/Linux_Container
For VM you can use virtiofs
# https://virtio-fs.gitlab.io/
on lxc you can use mout points to a zfs datapool.
# https://pve.proxmox.com/wiki/Linux_Container
For VM you can use virtiofs
# https://virtio-fs.gitlab.io/
I believe this is simply one method to avoid nesting copy-on-write file systems.
Placing ZFS on top of ZFS is not a common practice.
* Sometimes it's because you can just connect the disk to another machine and use it with `zpool import`, or for other reasons. It depends on the user's preference.
Last edited: