OS Drive on Consumer SSD with ZFS - Wear Triaged - Safe transition to EXT4

[sigxfsz]

I fell into the quagmire of Consoomer SSD with ZFS. I've triaged a zillion not-great options to slow down the wear but rather not leave it forever. I've customized a bevy of niceties and neccessities to prevent a blow-away rebuild (Passthrough Gaming VM, OPNSense router with VFIO, my OctoPI VM) although if I must I can make it happen scorched-earth.

Is it safe to just install Proxmox on a 2nd drive installed alongside the current running OS drive without damaging the OS/VM (ZFS RAID10 HDD's) drives if I don't hit any obvious wrong buttons? If so, I can boot back into the old OS Drive, mount the EXT4 new drive, copy all the configurations over, and be done with a live-spare instead of redundant drive until bigger fish get fried.

Gentoo since 2001; Slackware before that. No dual boot since 2004. Unfamiliar with Proxmox's installer eccentricities and new to ZFS. Rather avoid a 2nd quagmire; not my first rodeo. If Proxmox is 'dual boot safe with itself' I can solve this one without further annoyance to the community. Any affordable Enterprise SSD shopping advice would be helpfull as that is Phase2 of this conundrum. I have plenty of drives lying around including identical to the OS drive (MX500) I was going to ZFS RAID/MDRAID until I learned the caveats of both with Consoomer Drives/Proxmox; want to untweak the feature limiting wear reduction hacks someday but for now just going EXT4 with a live-spare alongside it and leaving the tweaks sounds best until then.

 

[apoc]

copy all the configurations over,

pve configs are all stored in /etc/pve
You can extract / copy them off from a running system but due the nature of this path (shared, distributed filesystem) of you just mount the disk (e.g. not boot the instance) the path is not there. So the copy destination is not available on the other install.
That is not bad as you can just pull the data off, boot the new install and then insert the files once booted. Reboot again and it will adjust configuration.
At least that is what worked for me the last time.

Regarding the VM disks. As long as you have used the defaults your VMs data is stored in zvols. ZFS block implementation. You can't just copy them over to an ext4. So you would need to research a way of doing this - I can't speak from experience but worst case is likely some "did" operation.
Maybe someone else can share experiences.

 

[Dunuin]

Regarding the VM disks. As long as you have used the defaults your VMs data is stored in zvols. ZFS block implementation. You can't just copy them over to an ext4. So you would need to research a way of doing this - I can't speak from experience but worst case is likely some "did" operation.
Maybe someone else can share experiences.

Easiest would be to backup the guests and then restore them on the new disks storage.

 

[sigxfsz]

pve configs are all stored in /etc/pve
You can extract / copy them off from a running system but due the nature of this path (shared, distributed filesystem) of you just mount the disk (e.g. not boot the instance) the path is not there. So the copy destination is not available on the other install.
That is not bad as you can just pull the data off, boot the new install and then insert the files once booted. Reboot again and it will adjust configuration.
At least that is what worked for me the last time.

Regarding the VM disks. As long as you have used the defaults your VMs data is stored in zvols. ZFS block implementation. You can't just copy them over to an ext4. So you would need to research a way of doing this - I can't speak from experience but worst case is likely some "did" operation.
Maybe someone else can share experiences.

Good to know on the configs; dealt with them a few times but didn't run into that. In this case, I'd boot up the old instance with ZFS rpool and copy them to the mount ext4 new install so should be fine with that situation.

I'm keeping the ZFS on the RAID10 Enterprise HDD's (4x4GB Seagate Helium buggers. One even failed and warrantied it. Happy with ZFS for the spinning rust honestly. Even burnt on a slower NVM as a ZIL cache until I noticed it was a bad idea until I got a purpose driven enterprise one) so I don't need to migrate the VM's. Configs should see them fine. Mostly concerned if...

A: The Proxmox install from USB onto 2nd dual booting drive (I'll BIOS select not perma-dual buat in the loader) will leave the ZFS data drives alone?
B: New Proxmox boots up and see's the ZFS volumes for use/configuration refernce just dandy?

Thanks for your time. Just hoping to get this out of the way and the whole family is dependent on the OPNSense router so trying to minimize a rodeo bull I need to tame.

 

[sigxfsz]

Easiest would be to backup the guests and then restore them on the new disks storage.

I'll have them backed up regardless, for sure. Curious if the ZFS volumes are intact after 2nd drive install then visible just fine on start then if all OS configs copied over (/etc/pve.. but also everything else) keep them intact without relying on the backup?

Plan is to copy not just PVE, but all the configs for ZFS parameters, rrdcached, VFIO (net and GPU) passthrough, hugepages (If I even customized that), and whatever else via a tar unarchive of /etc not just /etc/pve.. figured it couldn't hurt if both old and new install were newest updated 7.2 systems.. old system from network new system from USB... can't imagine any differences beeing too big.

I do have to look over my VFIO/passthrough customization as I might have to kick the bootloader. I reload grub on my Gentoo box's for that but I can't recall if I had to on Proxmox.. memory serves me it was already a builtin/module and didn't need it.. but I should dig.