Is it OK to create zfs filesystems on the proxmox host for usage other than vm's?

[frijsdijk]

I've recently chosen for proxmox as my server to host a couple of vm's. Before this I was using Ubuntu with Qemu, so the migration is pretty straight forward. The server that I'm using for proxmox has 8 1TB drives connected to a HBA, and via proxmox I've configured those drives in a zfs raidz2, and everything works like sunshine. Proxmox boots from 2 separate identical SSD's that are directly connected to the motherboard SATA ports, and during install I've chosen to set these two disks in a zfs raidz1 for OS redundancy.

I have a few questions:

- Is it "OK" to ssh to the host itself (so proxmox OS), and via shell create some zfs (encrypted) filesystems that I'll be using for other stuff (personal backups), not for VM's.

- I'm currently building the machine at home (I have 2 of these identical servers, one at home, the other one is already hanging in a DC). When I'm done, I'll take the HBA, SAS cables, disks out of that chassis, go the the datacenter, and swap some hardware and fit these disks into the server that's already in the DC. For proxmox this should be pretty transparant, other than that the NIC will have another MAC address. I'm assuming this will just work right away. Just wondering if there are any gotcha's.

Thanks so far, and I hope to be part of this community in the future. Proxmox is pretty awesome.

 

[Dunuin]

during install I've chosen to set these two disks in a zfs raidz1 for OS redundancy.

I guess you mean a mirror. raidz1=raid5.

via proxmox I've configured those drives in a zfs raidz2

Keep in mind that storing guests needs alot of IOPS performance. HDDs are terrible at that and it gets even worse when using and raidz1/raidz2/raidz3 as the IOPS performance will only scale with the number of vdevs and not the number of disks. So your 8 disk raidz2 won't be faster than a single HDD as you just got one vdev. A 8 disk striped mirror (raid10) would give you 4 times the IOPS performance and would be way better suited for a VM/LXC storage.

- Is it "OK" to ssh to the host itself (so proxmox OS), and via shell create some zfs (encrypted) filesystems that I'll be using for other stuff (personal backups), not for VM's.

Sure, but never encrypt datasets/zvols on your rpool when using grub, or your PVE won't be able to boot anymore.

- I'm currently building the machine at home (I have 2 of these identical servers, one at home, the other one is already hanging in a DC). When I'm done, I'll take the HBA, SAS cables, disks out of that chassis, go the the datacenter, and swap some hardware and fit these disks into the server that's already in the DC. For proxmox this should be pretty transparant, other than that the NIC will have another MAC address. I'm assuming this will just work right away. Just wondering if there are any gotcha's.

Shouldn't be a big problem. Just make sure you got webKVM/physical access to that server to edit the /etc/network/interfaces in case the NIC names changes, because then the webUI and SSH won't be working. And in case you are using any passthrough, make sure to disable the autostart of the VMs using it.