All-In-One Single LXC Container

This guide walks you through the process of deploying OpenStack using devstack in an LXC container instead of a VM.

The primary benefits to running devstack inside a container instead of a VM is faster performance and lower memory overhead while still providing a suitable level of isolation. This can be particularly useful when you want to simulate running OpenStack on multiple nodes.


Containers do not provide the same level of isolation as a virtual machine.


Not all OpenStack features support running inside of a container. See Limitations section below for details. OpenStack in a VM is recommended for beginners.


This guide is written for Ubuntu 14.04 but should be adaptable for any modern Linux distribution.

Install the LXC package:

sudo apt-get install lxc

You can verify support for containerization features in your currently running kernel using the lxc-checkconfig command.

Container Setup


For a successful run of and to permit use of KVM to run the VMs you launch inside your container, we need to use the following additional configuration options. Place the following in a file called devstack-lxc.conf:

# Permit access to /dev/loop*
lxc.cgroup.devices.allow = b 7:* rwm

# Setup access to /dev/net/tun and /dev/kvm
lxc.mount.entry = /dev/net/tun dev/net/tun none bind,create=file 0 0
lxc.mount.entry = /dev/kvm dev/kvm none bind,create=file 0 0

# Networking = veth = up = lxcbr0

Create Container

The configuration and rootfs for LXC containers are created using the lxc-create command.

We will name our container devstack and use the ubuntu template which will use debootstrap to build a Ubuntu rootfs. It will default to the same release and architecture as the host system. We also install the additional packages bsdmainutils and git as we’ll need them to run devstack:

sudo lxc-create -n devstack -t ubuntu -f devstack-lxc.conf -- --packages=bsdmainutils,git

The first time it builds the rootfs will take a few minutes to download, unpack, and configure all the necessary packages for a minimal installation of Ubuntu. LXC will cache this and subsequent containers will only take seconds to create.


To speed up the initial rootfs creation, you can specify a mirror to download the Ubuntu packages from by appending --mirror= and then the URL of a Ubuntu mirror. To see other other template options, you can run lxc-create -t ubuntu -h.

Start Container

To start the container, run:

sudo lxc-start -n devstack

A moment later you should be presented with the login prompt for your container. You can login using the username ubuntu and password ubuntu.

You can also ssh into your container. On your host, run sudo lxc-info -n devstack to get the IP address (e.g. ssh ubuntu@$(sudo lxc-info -n devstack | awk '/IP/ { print $2 }')).

Run Devstack

You should now be logged into your container and almost ready to run devstack. The commands in this section should all be run inside your container.


You can greatly reduce the runtime of your initial devstack setup by ensuring you have your apt sources.list configured to use a fast mirror. Check and update /etc/apt/sources.list if necessary and then run apt-get update.

  1. Download DevStack

    git clone
  2. Configure

    Refer to Minimal Configuration if you wish to configure the behaviour of devstack.

  3. Start the install

    cd devstack


To stop the container:

lxc-stop -n devstack

To delete the container:

lxc-destroy -n devstack


Not all OpenStack features may function correctly or at all when ran from within a container.


Unable to create LVM backed volume

In our configuration, we have not whitelisted access to device-mapper or LVM devices. Doing so will permit your container to have access and control of LVM on the host system. To enable, add the following to your devstack-lxc.conf before running lxc-create:

lxc.cgroup.devices.allow = c 10:236 rwm
lxc.cgroup.devices.allow = b 252:* rwm

Additionally you’ll need to set udev_rules = 0 in the activation section of /etc/lvm/lvm.conf unless you mount devtmpfs in your container.

Unable to attach volume to instance

It is not possible to attach cinder volumes to nova instances due to parts of the Linux iSCSI implementation not being network namespace aware. This can be worked around by using network pass-through instead of a separate network namespace but such a setup significantly reduces the isolation of the container (e.g. a halt command issued in the container will cause the host system to shutdown).