Chef Trial Install
Important: The chef install only supports OpenStack Icehouse and we’ve heard reports that it doesn’t work smoothly outside our test lab. We recommend using one of our packaged installs.
If you’re interested in trying out Calico but you don’t have much familiarity with installing OpenStack, you can install an OpenStack-with-Calico deployment using Chef. If you’re familiar with Chef, this is the fastest way to get a quick test environment for playing around with Calico.
This install procedure has only been tested on Ubuntu 14.04. Other versions of Ubuntu may also work successfully. It is unlikely that this install will work with Red Hat Enterprise Linux or its variants.
This procedure requires a Chef server: it is not sufficient to use chef-solo. The installation uses knowledge about the other nodes in the deployment to determine how to configure a node.
What You Get
This install guide creates a deployment that contains a single control node and at least two compute nodes. All the compute nodes are connected in a BGP mesh, no BGP route reflector is required. This is the simplest possible configuration for Calico.
Note that this is not a production install of Calico or of OpenStack. It is intended for trial and testing purposes only.
How To Use It: For Chef Newbies
If you are unfamiliar with Chef you may want to use the hosted Chef as a service, as mentioned on the Getting Started with Chef website. The following instructions provide a step-by-step guide on installing Chef and using the hosted Chef service to install a test OpenStack+Calico environment.
- At least four servers running Ubuntu 14.04. These are for:
- One OpenStack control node
- Two OpenStack compute nodes.
- One Chef bootstrap server for running the node bootstrapping.
You can in principle get away with using only three machines and performing the bootstrapping from one of the OpenStack nodes, however this has never been tested. If you try this out, please let us know how it went!
Prepare Your OpenStack Nodes
The default Ubuntu 14.04 installation may have hosts configuration that causes problems for the Chef installation. We recommend ensuring the following preparation is performed on each OpenStack node prior to starting the bootstrap process:
- Ensure your hostname is correctly set. Edit the
/etc/hostnamefile and set the correct name for each node (names must be unique).
- Ensure your loopback IP and OpenStack DNS entries are configured.
- Set the loopback IP address to 127.0.0.1.
- Set your hostname IP address to your static IP.
- Configure the hostnames / IPs of the other OpenStack nodes.
If you are using VMware and VMs for each of these machines (for testing), ensure the VM setting allows the VM to expose hardware assisted virtualization to the guest OS (setting under the CPU configuration).
- If you are intending to use the hosted Chef service, follow the instructions on Getting Started with Chef to create a new user and to create an organization for that user.
- On your Chef bootstrap server, follow the instructions on Getting Started with Chef to install Chef.
Clone the calico-chef repo on the Chef bootstrap server (or somewhere sensible):
git clone https://github.com/projectcalico/calico-chef.git
- Create a directory for the chef installation configuration. The
following instructions assume you created a
.chefdirectory under the
calico-chefdirectory that was created by the git clone.
- Generate the knife config for the organization on the hosted chef
site, download the
- Download both the organization and user keys from the website into
Load the roles from file. From the
knife role from file ../roles/*.rb
Bootstrap your OpenStack nodes
Setting up your test deployment nodes is a two-stage process. The first is to bootstrap each node with the appropriate role. Once all roles are assigned to the nodes, the chef client is re-run to update the configuration on each node. This is necessary to set up the BGP mesh, which needs to have knowledge of each compute node in the deployment.
To install the controller node, from the chef bootstrap server run:
knife bootstrap --sudo --verbose --run-list "role[controller]" --ssh-user <username> --ssh-password <password> <controller node hostname>
To install each compute node, from the chef bootstrap server run the following for each compute node:
knife bootstrap --sudo --verbose --run-list "role[compute]" --ssh-user <username> --ssh-password <password> <compute node hostname>
Check that all nodes appear in the host Chef Web UI.
Now, log on to each node in the OpenStack deployment (controller and compute nodes) and run:
This will fix up the BGP mesh between the compute nodes.
You should now be ready to go!
How To Use It: For Experienced Chef Users
We provide a single
calico cookbook in this
repository. This cookbook
can be cloned in git and then added to the Chef server in the usual way.
The cookbook defines two roles:
compute. To set up
your deployment, perform the following steps:
- Bootstrap one machine with the
controllerrole. This will install OpenStack’s control components on to that machine. When using OpenStack you will mostly interact with the UI on that machine.
Bootstrap at least two further machines with the
Note that this procedure works best when you assign all your compute machines the
computerole before executing their run lists. This way you’ll only need to execute the run list for each compute machine once.
If you execute the run list for a compute machine before all the compute machines have been assigned their role, you’ll need to re-run the run-list once all compute machines are present. The
computerole builds up config that relies on knowing all the other
- Play with Calico!