Very random question has anyone ever changed their motherboard, cpu and ram without reinstalling linux.

I currently have everything set up how i want it and have 3x1tb hdd's in a zfs.

If i changed the hardware and be able to re import the zfs pool or will it all just work

Interestingly enough, I have not done this with Linux but I did actually get Windows to work after a motherboard change (intel chipset). I was utterly amazed. If Windows can do it, I am sure Linux can, lol. Only thing I can think of is if you have /etc/fstab entries pointing to device names or UUIDs and those change with the new board, may have boot issues. Others that have done this may have better advice or examples.

Haha me to i upgraded my windows pc with new hardware and windows worked i was shocked. thanks for the tip i will check my /etc/fstab file and see what entries are in there. ALL AMD HARDWARE TOO!

Generally the reason behind using an emulation layer (mdadm, LVM, zfs, btrfs) is to isolate the filesystem from actual devices. They all provide a means to scan the devices and find appropriate filesystems based on meta-data. So your zfs pools should be ok - but I don't use it, I'm a btrfs user.

The O/S support is likely to be a bit more problematic - the new motherboard is likely to have different chipsets for things like bus support and maybe even disk/SSD. It might work it might not. In the past I've simply re-installed in need, but haven't done it for a while. Having a separate /home partition makes life easier as it keeps most customisations safe, and reinstalling packages generally doesn't over-write them. /etc will get clobbered, but that shouldn't have much in the way of user changes in it - but note my sigline.

Can we assume the pool is root is on zfs also?

There may be two parts to this. One is how linux works and how zfs works.

Clones tend to work best to exact hardware. Any deviation may leave some issues. Generally if quite similar you may find that nic naming, fstab and video tend to be most common issues.

Hmmm - I was assuming not obviously; I only keep user data in my btrfs RAIDed volumes, so "normal" partitions for the system. Root in zfs would add some extra kinks.

Hi,
Yes! Swapped out a mother board once for one with a very different chip set.
Everything just worked. Network, X11, lvm. I was shocked! Don't think I had to tweak anything. I think this would have been either Debian 7 or 8.

Sorry, don't have experience with zfs.

Cheers,

Evo2.

Me too, new motherboard, just put the hard drive in and voilá. ArchLinux that is, I'm still running that install.

No experience with zfs.

I should of been more clear. I have an ssd running all of linux and then 3x1tb drives for the zfs pool.

Yes, but far more often I've cloned a HD in one PC, so I could put the clone into a different PC and need not install all the operating systems from scratch. All my currently operable PCs are multiboot, most with upwards of 10 operating systems, most without any Windows. 4 days ago I took a HD from a dead Athlon (Socket 462/A, no SSE2) and ancient NVidia AGP GPU, and put it in a P4 (Socket 478, with SSE2) with much newer NVidia PCIe GPU. It booted right up into an openSUSE Tumbleweed last updated 30 months ago, with network and video working normally. Debian Stretch, last updated 8 weeks ago, also booted normally. Fedora 25 booted OK. openSUSE 13.1 failed to configure NIC due to mismatched NIC address in /etc/udev/rules.d/70-persistent-net.rules, but OK on subsequent boot. Mageia 5 only got as far as DrakX trying to automatically "install new hardware", but after a rescue boot to disable harddrake2, worked fine on next normal boot.

I do remember many moons ago doing a disk transplant between some kind of Athlon and some kind of Intel where the HD drivers were not in the initrd, so / could not be found.

I can't speak to the ZFS pool, but otherwise you shouldn't have a problem. Most, if not all, of the drivers, etc, are compiled into the kernel and Linux does hardware detection at every boot.

The only thing that might cause a wrinkle is when swapping to very new models for which drivers/other support requirements have not yet been incorporated into your distro of choice. I'm facing this situation waiting for the release of the AMD Ryzen 3000 series processors and accompanying 570x chipset motherboards. I'm prepared for the possibility that I may have to go "out of tree" for updates to get software support for the new parts. Also prepared for the possibility I may have to wait a couple of weeks (or longer) for that support to become available.

It's been thus with Linux since the beginning...if you absolutely need out of the box support for hardware, make sure what you buy has been available to the developers who maintain your distro of choice for about six months or so to give them time to catch up.

Having root on ssd will simplify it a bit.

Suse for example suggests that one either make references generic before they move or fix them after. That they include all drivers needed for new hardware or add in later.

Once you start moving linux to a new hardware you run a risk of having to fix stull. The ZFS pool should be OK and may just have to be managed a bit. The issue with how this new system boots and how it might see the hardware may have some impact. A very new board may not have drivers in linux so that may be an issue.


Most folks would say just try it. Be prepared with a backup plan that says reload from scratch and be sure to have a real data backup to other media.

1 members found this post helpful.

Depends on the Kernel config and the differences between hardware. If the Kernel is generic ("huge") and relatively new it should go fine, unless you have some strange new hardware.

Yes. I had a aging Pentium III m'board die and I wound up replacing it with a Core2Duo-based board + new memory (obviously). It booted and ran the 32-bit Linux quite well until I had a chance to upgrade it to 64-bit.

Would it hurt to give it a try? I suspect there may be a potential problem if you're booting from the ZFS pool. I know it's possible to import md raidsets from other systems. I would imagine it's possible to do that with a ZFS pool. Simple matter of Reading The Fine Manuals, eh? (I'm not a ZFS user so I'm probably not the best to ask about that particular technology.)

Another area where things could get sticky is the UEFI boot process.

If you're making a radical change, I'd consider adding another disk (or two to boot from RAID1) and re-installing. You can move the existing configurations files -- or at least refer to them -- to regain your existing functionality. If the UEFI boot turns out to be a problem, this might be the quickest way to getting back up and running. (Caveat: I only have a couple of systems with new enough hardware where I've had to deal with setting that boot process up---so I'm no expert. The first one took a couple of tries to get that boot partition just right.) I'm not sure where to get help with that. Maybe the Security forum?

Good luck...