nanjj

In this guide we will talk about CloudInit Setup on FreeBSD, not bsd-cloudinit. It's because bsd-cloudinit is based on cloudbaseinit, which has no support on ConfigDrive, while our OpenStack cloud supports only ConfigDrive. So for freebsd image we need to enable CloudInit.

We need packages qemu-utils, wget:

pkg install qemu-utils wget

And download freebsd base image:

base=FreeBSD-11.1-RELEASE-amd64
release=releases/VM-IMAGES/11.1-RELEASE/amd64/Latest
wget https://download.freebsd.org/ftp/$release/${base}.qcow2.xz

Unzip and convert to raw:

unxz ${base}.qcow2.xz
qemu-img convert -p -f qcow2 -O raw ${base}.qcow2 ${base}.raw

Attach to device:

mdconfig -a -t vnode -u 0 -f ${base}.raw

Mount last partition:

mount /dev/md0p3 /mnt
mount -t devfs devfs /mnt/dev
cp /etc/resolv.conf /mnt/etc
chroot /mnt

Install cloud-init:

pkg install py27-cloud-init

Enable services:

cat >> /etc/rc.conf << EOF
cloudinit_enable="YES"
sshd_enable="YES"
EOF

Cloudinit call blkid to select config drive. But blkid doesn't work for FreeBSD filesytem. Now work around blkid as below:

cat > /usr/bin/blkid << EOF
#!/bin/sh
[ -e /dev/iso9660/config-2 ] || exit 1
echo /dev/iso9660/config-2
EOF
chmod a+x /usr/bin/blkid

Mount for cd9660 on FreeBSD does not support -o sync. Patch cloudinit mount:

cd /usr/local/lib/python2.7/site-packages/cloudinit
patch -p1 << EOF
--- a/util.py   2017-10-27 07:14:55.968737000 +0000
+++ b/util.py   2017-10-27 07:15:03.481245000 +0000
@@ -1364,6 +1364,7 @@
         if mtypes is None:
             mtypes = ["auto"]
     elif platsys.endswith("bsd"):
+        sync = False
         if mtypes is None:
             mtypes = ['ufs', 'cd9660', 'vfat']
         for index, mtype in enumerate(mtypes):
EOF

Cleanup and exit:

set history = 0
exit

umount /mnt/dev
rm /mnt/etc/resolv.conf
umount /mnt
mdconfig -d -u 0
qemu-img convert -p -f raw -O qcow2 ${base}.raw ${base}.qcow2

Once you boot an instance, you can access it via ssh, please notice the default user is beastie.

First to list the geoms:

gpart list
=>
Geom name: vtbd0
modified: false
state: CORRUPT

Show the Geom:

gpart show vtbd0
=>       3  44040315  vtbd0  GPT  (40G) [CORRUPT]
         3       118      1  freebsd-boot  (59K)
       121   2097152      2  freebsd-swap  (1.0G)
   2097273  41943040      3  freebsd-ufs  (20G)
  44040313         5         - free -  (2.5K)

If the state shows CORRUPT, we need to recover it:

gpart recover vtbd0
vtbd0 recovered

Show again and the CORRUPT mark disappeared:

gpart show vtbd0
=>       3  83886069  vtbd0  GPT  (40G)
         3       118      1  freebsd-boot  (59K)
       121   2097152      2  freebsd-swap  (1.0G)
   2097273  41943040      3  freebsd-ufs  (20G)
  44040313  39845759         - free -  (19G)

Apparently the partition 3 should be resized.

Run resize on partition 3:

gpart resize -i 3 -a 4k vtbd0
vtbd0p3 resized
gpart show vtbd0
=>       3  83886069  vtbd0  GPT  (40G)
         3       118      1  freebsd-boot  (59K)
       121   2097152      2  freebsd-swap  (1.0G)
   2097273  81788799      3  freebsd-ufs  (39G)

According FreeBSD Doc, growfs should be run as below:

growfs /dev/vtbd0p3

But it doesn't work with below error reported:

growfs: /dev/vtbd0p3: Operation not permitted

A workaround is to run service growfs onestart:

service growfs onestart
Growing root partition to fill device
vtbd0 recovering is not needed
vtbd0p3 resized
super-block backups (for fsck_ffs -b #) at:
 42314112, 43596352, 44878592, 46160832, 47443072, ...
 56418752, 57700992, 58983232, 60265472, 61547712, ...
 70523392, 71805632, 73087872, 74370112, 75652352, ...

After that I run shutdown -r now, the VM hang at:

run_interrupt_driven_hooks: still waiting after 60 seconds for xpt_config
run_interrupt_driven_hooks: still waiting after 120 seconds for xpt_config

After force restarted it's bootable again.

Looks FreeBSD image is not polished well for cloud using.