Instructions last updated 29/07/2015
I’ve been attempting to install FOG in a Docker container. Mostly so I can learn about Docker, but also so I can zip FOG up into a neat little package and deploy it to some old laptops for quick imaging. My process is outlined below. Here’s some things to keep in mind:
- I haven’t created a Dockerfile yet, as I’m still learning about that, plus I don’t know if you can automate
installfog.sh
in a Dockerfile friendly way - I may have messed up some steps along the way, or not done things correctly. I had to learn all about TFTPD, NFS and many other services, while overcoming hurdles such as various
service [service name] start
calls not working and days of research about running TFTP on boot2docker in Windows (long story short, doesn’t work. Prove me wrong though!) - I’ve been able to image one machine so far (a Dell Latitude E5500 with a bare-bones Windows 7 image) and I haven’t tried anything else (e.g. deploying the FOG service) so you might need to do more to get everything else running.
- The image I was using was copied over from our old FOG server, so I haven’t tried capturing an image yet.
- I may update this as time permits, or come back to answer questions, but now that I can image machines, my allocated time on this task is up.
Let’s begin!
Prerequisites
- A Linux machine running Ubuntu 15.04
- Should work with older versions and would probably work with other distros
- Might even work with boot2docker for Mac, but due to TFTP problems, won’t work on Windows. Put simply, I can’t get TFTP packets on port 69 to LEAVE the Windows host (they arrive in Windows, arrive in the container and leave the container, but that’s it)
- Docker installed
- Run
wget -qO- http://get.docker.com | sh
- Don’t install it via
apt-get
, as it’s an older version
- Don’t install it via
- Run
Installation
How to install FOG in Docker:
- Install Docker
- boot2docker can be used, but the final container won’t work on Windows, due to port 69 (TFTP) not being accessible, even with the firewall being turned off / ports added. Investigation continues
- If you’re using boot2docker on Windows (and maybe Mac), you’ll need to allow port 80 through. To do this:
- Open VirtualBox, click on
boot2docker-vm
, click onSettings
, go toNetwork
, thenPort Forwarding
. - Click on the
+
in the top-right corner. Name will behttp
, protocol will beTCP
, host IP will be blank, host port will be80
, guest IP will be blank, guest port will be80
. Click OK, then OK when done - If you wish to try TFTP on boot2docker, do the same with port 69. You might also need to repeat these steps for every port you specify with the
-p
flag when running your Docker
- Open VirtualBox, click on
- On your HOST (* not * inside your container), run
modprobe nfs && modprobe nfsd
. NFS depends on host kernel support, and the kernel modules don’t start at boot. If using boot2docker, you don’t need to do this, as it has all the NFS stuff built in, I believe. - Run
docker pull ubuntu:12.04
- Install fails on
ubuntu:latest
for some reason (fails around the start of TFTP), so12.04
preferred
- Install fails on
- Run a container with
docker run -it -p 212:212/udp -p 9098:9098 -p 21:21 -p 80:80 -p 69:69/udp -p 8099:8099 -p 2049:2049 -p 2049:2049/udp -p 111:111/udp -p 4045:4045/udp -p 4045:4045 -p 111:111 -p 34463:34463/udp -p 34463:34463 --privileged -e WEB_HOST_PORT=80 -v ~/fog/www:/transfer ubuntu:12.04
- Replace //c/Users/graydav1/Desktop/fog/www with another folder from your PC. This is used to pass files between the container and your computer (optional, good for importing existing FOG images, config file changes, easy editing of files etc.)
- 111 is portmap, 80 is HTTP (Apache), 69 is TFTP, 2049 is NFS, 34463 is mountd (part of NFS)
- I may have opened too many (or not enough) ports, but my goal was to getting this all working for my use at work, so if there’s time, I might play around with port settings
- Run
apt-get update
- Run
apt-get install wget net-tools nano
net-tools
required for IP address detection during FOG setup,nano
for editing of/etc/bash.bashrc
andwget
to download the FOG archive
- Run
wget -P /tmp http://liquidtelecom.dl.sourceforge.net/project/freeghost/FOG/fog_1.2.0/fog_1.2.0.tar.gz
to download the FOG archive - Run
tar -C /tmp/ -xzvf /tmp/fog_1.2.0.tar.gz
to untar the archive to /tmp - Run
cd /tmp/fog_1.2.0/bin/ && bash ./installfog.sh
to start the installation - Run through the installation and wait for it to finish
- Change the IP address of the FOG server line to match the host (e.g. if the container reports 172.17.0.1 but the computer’s IP address is 10.15.0.63, set the FOG address to 10.15.0.63)
- Defaults for the rest should be fine, as Docker uses a bridge to contact the outside world. Chances are your organisation already has DHCP set up, so say no to that
- You can set a MySQL password if you like, but if you’re running the container on-demand, you could probably get away with not having one
- You won’t be able to do the schema update yet, because MySQL hasn’t started properly, so just press Enter and finish the install
- Delete the install directory and archive if you wish
- Run
nano /etc/bash.bashrc
- Go to the very bottom and add these lines:
service apache2 start
to start Apachemysqld &
To run MySQL (as it won’t start viaservice
for some reason)in.tftpd --listen --address 0.0.0.0:69 -s /tftpboot
(for the same reason as above)/usr/bin/freshclam -d --quiet
for antivirus (if you use it)rpcbind
(for NFS support)service nfs-kernel-server start
(for NFS support)
- Press Ctrl+X to quit, then Y to save
- Go to the very bottom and add these lines:
- Run
nano /etc/services
- Go to the very bottom and add these lines:
mountd 34463/tcp
(use the tab key to line up the entries, same as the other entries in the file)mountd 34463/udp
- We add these to make mountd use a static port. Otherwise it uses a dynamic port, making it hard for us to let the port through Docker
- Press Ctrl+X to quit, then Y to save
- Go to the very bottom and add these lines:
- Run
nano /tftpboot/default.ipxe
- Change the final line (that starts with
chain http://...
) so that the IP address matches your host IP (e.g.chain http://10.15.0.63/fog/service/ipxe/boot.php##params
) - Press Ctrl+X to quit, then Y to save
- default.ipxe is created with the reported IP address (e.g. 172.17.0.1) and not the IP address we specified in the startup (?)
- Change the final line (that starts with
- Type
exit
to exit from the container. This will shut it down - Run
docker ps -l
and look for the ID of the container you just exited from. Remember, that’s a lowercase L and NOT the number one - Run
docker start <container ID>
- This is like a restart. Anything in /etc/bash.bashrc will run when the container starts again
- You’ll be sent back to the command prompt (i.e. not inside the container)
- Run
docker ps
to confirm that the container has started
- Go to http://localhost/fog/management on the host PC and update the schema
- Run
docker commit <container ID> grayda/fog
(replacegrayda/fog
with your own image name)- This saves our changes to an image file, ready for packaging
- (Optional). If you need to move the image to another machine (e.g. you created this on Windows, but want to run it on Ubuntu), run
docker save -o fog.tar.gz grayda/fog
to save your image as a file. Again, replacegrayda/fog
with your image name- Copy the file to your destination machine, then run
docker load -i fog.tar.gz
on there to load the image - When you’ve loaded the image to the destination machine, log in to the FOG management UI, go to
FOG Configuration
and change the necessary IP addresses in there. Also change/tftpboot/default.ipxe
accordingly
- Copy the file to your destination machine, then run
- Set your DHCP scope options (e.g. 066 and 067) as normal. Use the IP address of the host (e.g.
10.15.0.63
), NOT the IP address that the container reports. Docker’s bridge interface will take care of getting the traffic to the right spots
When you’ve made significant changes (e.g. uploaded a new image, added a client to the list etc.), then do a docker commit <container ID> grayda/fog:vX
(where X is a number, for example, v1, v2, v3 etc.)
Some other neat things you can try are:
- Use the
-v
switch indocker run
to map your/images
directory to your host, so you can easily back up your images or effortlessly load them from a secondary drive - Map the
/var/fog/www/services/ipxe
to ahtdocs
folder on your local machine so you can do dynamic boot menus that read from your own inventory system, or authenticate against your network or whatever you like - Shove everything on a USB stick or Raspberry Pi to have a portable FOG system (which is what I plan to do, actually. Simple FOG on a stick!)