Do I need to use by-id in proxmox zpools?

[el_pedr0]

I think that this post is saying that it is no longer necessary to use by-id when setting up zpools in proxmox 4.x, but being new to ZFS and proxmox, I couldn't quite follow the thread.

I've got my proxmox root installed on a ZFS mirror on two drives and I'm going to be setting up more zfs pools for my data. It's for a home server and so over time I might pull drives, re-arrange drives, etc and don't want to bork the proxmox boot.

Everywhere I look, it says that best practice is to use by-id (at least for <10 disks). But is that recommendation now out of date?

 

[ttabbal]

It doesn't matter if you do or not, you end up with /dev/sdX after rebooting. Honestly, I prefer the by-id approach. The current setup, as of the previous Proxmox anyway as I haven't updated to the one released over the weekend, is kind of annoying when you move things around. Not for boot, that seems to work fine. For the main pool. If you move devices between controllers, it fails to mount the pool. You have to manually import it. For example, I swapped an older LSI1068 controller for a H310. The pool didn't mount, and any VMs that use it fail to start. IMO, it should scan the devices, and if it locates all the required drives, mount the pool. Which controller they hang off of isn't relevant. You only have to import the pool once after the changes, but it's still annoying.