Upgrade from 11.1-RELEASE-p15 to 11.2-RELEASE failure

P

Paul Floyd

I just tried to upgrade to 11.2-RELEASE and it did not go well.

I ran freebsd-update -r 11.2-RELEASE upgrade, then init 1 and then freebsd-update install. All seemed to go well, and I got the message saying to reboot and run freebsd-update install again to complete the process.

On reboot I got a load of data error messages, too fast to read, just before the grub boot screen. I tried booting to single user, but got a console with /bin/sh and a read-only mount. freebsd-version says that I'm on 11.2-RELEASE-p4 kernel and 11.1-RELEASE-p15 userland. I remounted read-write and ran 'freebsd-update install' and it ran OK, and I now have 11.2-RELEASE-p4 for both but still booting to read-only console.

I've also tried booting multiuser but I just get a white blob (I use KDE, nVidia drivers for a GeForce card).

A few other details:
I'm using a ZFS mirror. zpool status says the 2 disks are clean, and zpool scrub found no errors. I haven't done a zpool upgrade.
The PC is triple boot, Windows 10 Fedora 24 (each with a disk each) and FreeBSD, chain booting from Fedora grub. Windows and Linux are still OK.

/boot/loader.conf contains
Code: 
kern.geom.label.disk_ident.enable="0"
kern.geom.label.gptid.enable="0"
zfs_load="YES"
#nvidia_load="YES"
boot_verbose="YES"
I tried the solution mentioned here
https://forums.freebsd.org/threads/...ls-to-boot-11-2-kernel-no-vb-no-nvidia.66538/
namely setting /boot/loader.conf:kern.vty="sc" seems to have no effect.

dmesg comntains some ACPI warnings and errors but not much else.

Any suggestions as to what to try next?
 
What is the error? You don't "just" get dropped into single user mode; the system must have mentioned something as to why it's doing that.
 
ShelLuser said:
What is the error? You don't "just" get dropped into single user mode; the system must have mentioned something as to why it's doing that.
Click to expand...

I can't see any error. At the point where it seems to be giving up the output is

Trying to mount root from zfs:zroot/ROOT/default [] ...
Root mount waiting for: usbus7 [repeats down to] usbus0
Root mount waiting for:usbus7 usbus3
Root mount waiting for:usbus3

If there are any error messages, they aren't in the console output/output of dmesg.
 
A quick update:
I've managed to run most of the 'pkg upgrade's, though I did get a lot of 'missing dependence xproto' messages on the way. There are something like 11 packages not upgraded, mostly Linux and nvidia 340. There error message that I'm getting with them is that Linux support is not enabled in the kernel.

I then checked the doc for Linux support and added
linux_load="YES"
to loader.conf.

After that I was able to 'pkg upgrade' everything. I uncommented nvidia_load also in loader.conf rebooted and am now back to the KDE desktop.
 
This sounds like the NVidia driver didn't match the new kernel and hung or crashed. The fix would have been to reinstall that, as you have done. However personally I would never do a major version upgrade without disabling things like graphics drivers and automatic loading of X via a window manager etc.

If you make it so it's back to basics and just drops you at a console login prompt then the upgrade is likely to go much more smoothly.

The other thing that can cause issues is that it's recommended to reinstall the boot code on the disks when you do a major upgrade just in case anything has changed or else this can cause your pool to not be recognised.
 
I hadn't realized that single user mode was so spartan. I now have the hang of remounting zfs and starting the network.

Do you recommend doing zfs upgrade first then grub-install or the other way round?
 
xtaz said:
This sounds like the NVidia driver didn't match the new kernel and hung or crashed. The fix would have been to reinstall that, as you have done. However personally I would never do a major version upgrade without disabling things like graphics drivers and automatic loading of X via a window manager etc.

If you make it so it's back to basics and just drops you at a console login prompt then the upgrade is likely to go much more smoothly.

The other thing that can cause issues is that it's recommended to reinstall the boot code on the disks when you do a major upgrade just in case anything has changed or else this can cause your pool to not be recognised.
Click to expand...

My next attempts were even more unsuccessful. I ran zpool upgrade and then the gpart bootcode commands to update the boot sectors on the two drives. This seemed to cause even more error messages on boot, but I was able to reboot a couple of times. zpool scrub ran OK. As it was still booting I proceeded to run pkg autoremove and then pkg install kde5.

When I rebooted after this I got the following error

lots of gptzfsboot errors then

ZFS: i/o error - all block copies unavailable
ZFS: can't read MOS of pool zroot
gptzfsboot: failed to mount default pool zroot

Downloading bootonly disk, more to follow...

[edit 1]
... OK booted 11.2 bootonly.

Escaping to the grub command line and running 'lsdev' gives reasonable looking output.

When booted from the DVD, I can do a zpool import OK.

[edit 2]
Well sort of, I can only see var tmp and usr, not things like boot etc or usr/local - not sure if this is normal.

On rebooting I get to the boot: prompt. No idea where to go from there.
 
Paul Floyd said:
Well sort of, I can only see var tmp and usr, not things like boot etc or usr/local - not sure if this is normal.
Click to expand...
You may need to do zfs mount -a, as some things may be set to "noauto" (namely /!). Sorry I can't be of more help otherwise though, as I use neither freebsd-update nor nVIDIA cards under FreeBSD.
 
I'll try that. I see that there are plenty of other people that have had similar issues with mirrored zfs systems. ATM it looks like I'm going to have to do a fresh install.

[edit 1]
zfs mount -a
-> still no sign of /

[edit 2]
zfs mount zroot/ROOT/default did the trick, I can mount the root partition
 
Last edited:
You must log in or register to reply here.
Back
Top