Chef Server startup notes

10 Nov 2016 · Eight minute read · on Gianluca's blog

News! Tech related notes are NOW published to ShippingBytes. See you there! I always felt this was not the right place for me to write consistently about tech and tools. So if you want to read more about that see you at the other side

I worked with different provisioning tools and configuration managers in the last couple of years: Chef, Saltstack, Puppet, Shell, Python, Terraform. Everything that was allowing me to make automation and describe my infrastructure as a code.

I really think that this is the correct street and every companies need to stop to persist random commands in a server:

Chef is my first configuration manager, I started to use it with Vagrant few years ago but I never had change to deep drive into it and into the full chef-server configuration from scratch.

I had this change few days ago and I am here to share some notes. I used digitalocean to start 1 Chef Server and two nodes, during this post I am not focused about the recipe and cookbook syntax but I will share some commands and notes that I took during my test to start and configure a Chef Server.

First of all doctl is the command line application provided by digitalocean to manage droplets and everything, I used that tool to start my droplets.

The Chef Server doesn’t run in little box, we need 2gb RAM, I tried with small size but nothing was working, the installation process gone out of memory very soon. Thanks Ruby.

$ doctl compute droplet create chef-server \
  --region ams2 --size 2gb --image 20385558 \
  --access-token $DO --ssh-keys $DO_SSH

$ doctl compute droplet create n1 \
  --region ams2 --size 512mb --image 20385558 \
  --access-token $DO --ssh-keys $DO_SSH

$DO contains my digitalocean access key and $DO_SSH the id of the ssh key to log into the servers. You can leave the last one empty and you will receive an email with the password.

When the process is gone you will be able to copy the ip of the chef-server and go into it.

$ doctl compute droplet ls

ID              Name            Public IPv4     Public IPv6     Memory  VCPUs   Disk    Region  Image           Status  Tags
30cw4230        chef-server                                     2gb     1       20      ams2    Debian 8.6 x64  new
q0514230        n1                                              512     1       20      ams2    Debian 8.6 x64  new

This provisioning script installs Chef-Server from the official deb package and also install chef-manage. Chef-manage provides a nice web interface to manage users, cookbooks and everything is stored into the server.

cd /tmp
sudo apt-get update
sudo apt-get install -y unzip curl
curl -LS -o chef-server-core_12.9.1-1_amd64.deb
sudo dpkg -i chef-server-core_12.9.1-1_amd64.deb
sudo chef-server-ctl reconfigure

Our server is up an running reachable over HTTPS (port 443). This configuration is just for testing purpose. It’s not a good practice leave a Chef Server public as we are doing. It’s a better idea to close it under a VPN for example.

Chef supports an authentication and authorization level based on users and companies. We are creating a new user called Gianluca Arbezzano with username and as password hellociaobye. We are also create an organization and we are associating user with org.

chef-server-ctl user-create gianarb Gianluca Arbezzano 'hellociaobye' --filename /root/gianarb_test.pem
chef-server-ctl org-create tf 'ThumpFlow' --association_user gianarb --filename /root/tf-validator.pem
chef-server-ctl org-user-add tf gianarb

At this point we can configure a nice UI for our Chef Server with this simple commands:

sudo chef-server-ctl install chef-manage
sudo chef-server-ctl reconfigure
sudo chef-manage-ctl reconfigure --accept-license

Chef-Server works with the concept of Organization and User. The organization is a group of users that share cookbooks, rules and so on. Users can update cookbooks and there is also a set of permission to manage access on particular resources like:

At this point we have one user with its own key and credential. You can come back into the UI and use username (gianarb) and password (hellociaobye) to login in. The key (–filename) is used to configure knife and encrypt communication between client and server. There are 3 main actors and this point that we need to know:

Chef Server has a HTTP api and knife is a CLI that provide an easy integration for your node and workstation. With this command we are installing knife. You can do it in your local environment, to become a workstation and into the server. (it’s usually a god practice create a user, we are doing everything as root right know but it’s BAD! don’t be bad!).

We have two certificate one is gianarb_test.pem and it’s identify a specific user, we need to generate our for every workstation/member of the team and the validation_client represent the organization, it could be the same across multiple users.

curl -O -L
bash ./

You can copy paste the 2 keys into the local machine and run this command that will drive your into the process to create a ~/.chef/knife.rb file that your cli uses to communicate with the chef server.

knife configure

This is an example of generated knife configuration file that I did in my server. I lose times to understand the chef_server_url it contains the hostname of the server but also the `/organization/' be careful about this or knife will come back with an HTML response in your terminal.

log_level                :info
log_location             STDOUT
node_name                'gianarb'
client_key               '/root/gianarb_test.pem'
validation_client_name   'tf-validator'
validation_key           '/root/tf-validator.pem'
chef_server_url          'https://chef-server:443/organizations/tf'
syntax_check_cache_path  '/root/.chef/syntax_check_cache'
cookbook_path            ["/home/gianarb/git/chef-pluto/cookbooks"]
ssl_verify_mode          :verify_none

The last 2 commands download and validate the SSH certificate because in the default configuration the CA is unofficial and we need to force our client to trust the cert.

knife ssl fetch
knife ssl check

Know that we did that in our server and also in our local environment we can clone chef-pluto a repository that contains recipes, rules and cookbooks to configure our node, we need to syncronize it into the server.

git clone /home/gianarb/git/chef-pluto/chef-pluto
cd /home/gianarb/git/chef-pluto/chef-pluto
knife update /

The last command update all our repository into the chef server. You can log in into the web ui and see the micro cookbook and the power rule.

micro is an application that I wrote in go and it just expose the ip of the machine. It’s a binary and the cookbook downloads and starts it, pretty straightforward.

At this point we need to make a provisioning of our first node, usually is the server that install and start the Chef Client into the node, what we can do it’s store a private key into the server to allow chef to connect to the node. I copied the digitalocean private key into the server (~/do), from security point of view you can create a dedicate one. You can also use the -P option if you are not using an ssh-key to run this example.

knife bootstrap <ip-node> -N node1 --ssh-user root -r 'role[power]' -i ~/do

If everything it’s good you can reach the application from port 8000 into the browser. The log is something like:

$ knife bootstrap -N testNode --ssh-user root -r 'role[power]' -i ~/do
Doing old-style registration with the validation key at /root/tf-validator.pem...
Delete your validation key in order to use your user credentials instead

Connecting to -----> Existing Chef installation detected Starting the first Chef Client run... Starting Chef Client, version 12.15.19 resolving cookbooks for run list: ["micro"] Synchronizing Cookbooks:   - micro (0.1.0) Installing Cookbook Gems: Compiling Cookbooks... Converging 2 resources Recipe: micro::default   * remote_file[Download micro] action create_if_missing (up to date)   * service[Start micro] action start     - start service service[Start micro] Running handlers: Running handlers complete Chef Client finished, 1/2 resources updated in 02 seconds

knife started the client, syncronized cookbooks, it assigned the power role at the node and run the correct recipes. Your server is ready and you can create and delete nodes to make your infrastructure complex how much you like.

Chef is quite old and it’s in ruby (the first one could be a plus but the second one no really) but it continue to be a good way to make a provisioning o your infrastructure. Lots of people moved to Ansible but the agent that they reject offer a very good orchestration feature that it’s something that I usually search.

I worked with StalStack and it’s very nice, the syntax is easy and it seems less expensive in terms of configuration, resources and setup but I am not really sure about the YAML specification. I am not a ruby developer and I don’t love the ruby syntax but in the end is a programming languages and I am doing infrastructure as a code.

Something weird with this website? Let me know.