This weekend was one of thoses weekend … Sunday morning while trying to destroy an unused jail i got a zfs warning: “pool is degraded”.

As it turn out i was missing a drive in my zfs mirror setup, so heck .. no fun.

After a few email with the hosting company they recommended i migrate over a new server, which is running ok if you’re reading this ;).

I needed to make a backup of the old system, not wanting to reinstall everything from scratch, i dig a bit and made two “files” from the original server:

1/ a backup of the geometry:

[email protected]:~# gpart backup ada0 > /var/tmp/backup.geom

which gave me something like that:

  GPT 152
  1   freebsd-boot         40       1024
  2   freebsd-swap       2048    4194304 swap0
  3   freebsd-zfs    4196352 1949327360 zfs0

2/ a backup of the filesystem:

[email protected]:~# zfs snapshot -r [email protected]_20171029
[email protected]:~# zfs send -R [email protected]_20171029 > /var/tmp/backup.zfs

And voila ! i just had to put this two file over an ftp server.

(well actually the first time i tried using the -D option of zfs send to further shrink the stream, as it turn out it wasn’t a great idea, the recv method didn’t like it).

The new serveur was made available at around 20:00 sunday evening, thanks to some limitation in the cloud tools provided i had to start an automated pre-scripted install before being able to “hack over” the hardware. So i did started the process just before going over to the movie (latest Blade Runner, fine by the way but so slow … ).

So after the movie, the server was up and running and i only had to … reboot it in rescue mode. The rescue mode is a PXE-booted environment that come in multiple “flavor”, including three FreeBSD version, that allow an ssh acces and will allow the reset everything easily.

When the freebsd/rescue is booted /tmp is mounted as a tmpfs, lucky me the memory available on the server is greated that the size of the zfs backup stream so i downloaded everything over it.

The drives where re-partitionned using:

[email protected]:~# gpart restore -F -l ada0 < backup.geom
[email protected]:~# gpart restore -F ada1 < backup.geom
[email protected]:~# gpart modify -i 2 -l swap1 ada1
[email protected]:~# gpart modify -i 3 -l zfs1 ada1

Boot code was installed using:

[email protected]:~# gpart bootcode -b /boot/pmbr -p /boot/gptzfsboot -i 1 ada0
[email protected]:~# gpart bootcode -b /boot/pmbr -p /boot/gptzfsboot -i 1 ada1

Apparently the swap took care of itself ;)

The only thing left was .. the file system, this get a bit tricky and it took quite a bit to get it done right but the process should be only:

[email protected]:~# zpool create zroot mirror gpt/zfs0 gpt/zfs1
[email protected]:~# zfs recv zroot < backup.zfs
[email protected]:~# zpool set bootfs=zroot/ROOT/default zroot

If, like me, you had to restart the server in rescue mode because the first boot didn’t get you there, there is a few interesting command to consider:

[email protected]:~# zpool import -R /mnt zroot
[email protected]:~# mount -t zfs zroot/ROOT/default /altroot

anyway, short things short the new server was up and running after only a 15 minutes service interruption !