Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Ansible deploy scripts for Stepup Infrastructure

These are the Ansible playbooks and scripts to create, deploy and manage a step-up infrastructure and to deploy the stepup components (i.e. stepup-middleware, stepup-gateway, stepup-ra, stepup-selfservice, stepup-tiqr and oath-server-php) to this infrastructure. The playbooks are targeted to a CentOS 7 image and should be usable with any environment (i.e. not be specific to a test or a production environment).

The Ansible playbooks and the deploy script require an "environment". An "environment" is the part of the playbook that contains the configuration (e.g. passwords, certificates, urls, email addresses, hostnames, ...) of the infrastructure that is being targeted. A template environment is provided in "environments/template". This template is the starting point for creating your own new environment. When using ansible playbook the environment to use is selected by specifying the inventory file of the environment using the -i option.

What is Stepup?

Stepup authentication as-a-service, or Stepup for short, is an open source project that was started by SURFnet to create what is now called "SURF SecureID" (and "SURFconext Strong Authentication" before that). It works seamlessly with OpenConext to add Step-up authentication for (SAML) Service Providers. The Stepup system manages authentication and registration of the second factors without requiring technical integration with the identity provider, which is great if you need to support many different identity providers. For SAML service providers (SPs) an "always require stepup" policy is available that allows SPs to connect to Stepup with very little to no integration effort. For a more feature rich integration SAML Scoping with RequestedAuthnContext is supported.

Stepup is not limited to be used with OpenConext. There is nothing that precludes it from being used by itself to add Step-up authentication to:

  • an existing SAML identity provider
  • one or many SAML service providers
  • other SAML proxies or hubs

How SURFnet uses Stepup to offer strong authentication to cloud services:

More information resources can be found at the end of the readme.

Deploy process

Setting up a new Stepup infrastructure consists of 4 steps:

  1. Create an "environment" that contains the configuration of the infrastructure and the stepup applications.
  2. Deploy the Stepup infrastructure. This installs all rpms, configures services, databases, firewalls, loadbalancers etc.
  3. Deploy the Stepup components. This installs the stepup applications ("components") and writes the application configuration: stepup-gateway, stepup-middleware, stepup-selfservice, stepup-ra, steup-tiqr and oath-server-php.
  4. Post installation configuration. This includes executing the scripts on the application server that initialise or update the database and running the scripts that push the configuration to the database.

Step 1: Creating a new Environment

Using the script a new environment can be created based on a template. This new environment does not have to (and typically shouldn't) be stored in this repository. The intended use is to store the environment in a different, private, repository. The secrets (private keys, password etc) in the environment are stored in files that are encrypted with a symmetric key using python-keyczar. This keyczar key can be stored in a safe location (e.g. on a deploy host), separate from the environment. The standard Ansible vault is not used in this process.

The template contains an environment.conf file that specifies the secrets to create. The first time you run the script it is copied to the new environment and you get the change to edit/update this file. Take this chance to update this file to match your configuration, likely places to update are marked with "TODO:". You can also copy the environments/template directory to a new location to make your changes there (use the --template option). The environment can be used as-is to deploy to VMs created with the scripts in Stepup-VM.

Requirements for running the script with python 2.7:

  • python 2.7
  • openssl
  • python-keyczar. You can use pip install python-keyczar to install this tool. This makes keyczart command available.

Requirements for running the script with python3:

  • python 3
  • openssl
  • python3-keyczar. You can use pip install python3-keyczar to install this tool. This makes keyczart command available.

On the mac pip install python3-keyczar works. However it might not work on centos. See

Use <new_environment_directory> --template <template_environment_directory> to create a new environment. The script will generate passwords, secrets, SAML signing certificates and SSL/TLS server certificates for use with HTTPS for the environment. All passwords, (private) keys and secrets are encrypted with a keyczar key that is specific for the environment. To issue the server certificates a self-signed CA is created using openssl.

For any other environment than one that targets the Stepup-VM you will need to make changes to the new environment. Because the Stepup software depends on external systems, additional configuration and setup is required to be able to actually use a Stepup environment. The locations in the new environment where you may need to make changes to match the requirements of your setup are marked with "TODO". Changes to make include:

  • Set hostnames, domains, email addresses
  • Replace the SSL Server certificates (for production, the certificates work fine for test in most browsers, but with warnings)
  • Configure API keys for messagebird, yubikey
  • Configure the remote "first factor" IdP
  • Adjust firewall rules (for production)
  • Move the keyczar key out of the environment (for production)

More information on the "environment" concept can be found in ansible-tools

Step 2: Create / update infrastructure

The site.yml playbook handles the configuration of your infrastructure. This playbook requires Ansible version 2.x with python < 3.0 and the environment created in the previous step. You execute Ansible from a Deploy host (e.g. you laptop) to configure other machines. Please consult the extensive Ansible documentation for Ansible installation instructions and more.

Note: In Ansible version 2.4 the handling of the inventory_dir variable was changed in a way that breaks how Stepup-Deploy uses the inventory_dir. This means that triggering handlers from included tasks is no longer possible. This affects tasks stored in the environment (i.e. in common.yml) only.

You must adjust the Ansible inventory file that was copied over from the template to match your infrastructure. The default inventory assumes you will use two (virtual) machines for running Stepup. This is a minimal setup. The two machines are:

  1. An application server (""), running an nginx+php-fpm web stack and also the database (mariadb+galera) in the template inventory).
  2. A management server ("" in the template inventory), running ELK for log processing. Although you must configure this server in your inventory to successfully deploy you application server, you can skip actually deploying the management server and have a functional application server.

The (virtual) machine(s) must be running CentOS 7. It is very unlikely that the playbook will work with another CentOS version or with another Linux distribution. 2 GB memory with 15 GB disk is sufficient to install the app server.

Configure ssh on you deploy host (i.e. the machine on which you will execute ansible-playbook) such that you can connect to machines listed in your inventory and can become root using sudo. Note that you must specify the IP address of the server in the inventory.

Use ansible-playbook -i <your_environment_directory>/inventory site.yml -e "galera_bootstrap_node=<app>" to deploy only the application server. Where you replace with the name of the application server in your inventory. You can use the "-l" option to only deploy the app server. I.e.: ansible-playbook -i <your_environment_directory>/inventory site.yml -e "galera_bootstrap_node=<app>" -l <app>


The inventory consists of one database running on the application server. The playbook can setup a Galera cluster running on multiple dedicated machines. In a cluster, when none of the MariaDB databases is running, such as during the first deploy, the first database must be bootstrapped by setting the Ansible variable galera_bootstrap_node to the hostname of the node to bootstrap. Example: ansible-playbook site.yml -i <environment_directory>/inventory -e ""

If you are using the minimal configuration in the inventory from the template, you have one database that is running on the application server. This database is configured as a cluster consisting of one node (you could add more nodes later). In this case the most important difference between a normal mysql/mariaDB and the Galera cluster version is that you ever need to start the database you must use service mysql bootstrap instead of service mysql start.

Step 3: Deploy the Stepup components

Stepup components are the applications that together make up the Stepup service. These are:

  • Stepup-Middleware. Is used by the Selfservice component and the RA component. The middleware component is the only component that writes to the database. The other components do not communicate with the middleware. The middleware component maintains the middleware and the gateway databases. Updating the configuration of the Stepup system is performed by sending commands to the middleware.
  • Stepup-Gateway. The gateway reads its configuration from the gateway database. It is a SAML proxy and handles all authentication request in the Stepup system by interacting with external authentication providers (1st factor SAML IdP, Messagebird SMS gateway, Stepup-tiqr or the Yubico Cloud). SAML Service Provides use this gateway for authentication.
  • Stepup-Selfservice. This is the web application where end users register to get stepup token (Yubikey, SMS, tiqr or U2F), can see its status and can revoke their token.
  • Stepup-RA. This is the web application where registration authorities (RAs) approve (vet) token registrations.
  • Stepup-tiqr. This is the web application that handles tiqr registration and authentications.
  • oath-service-php. This a server for storing the secrets used by tiqr.

Stepup components are deployed on a machine that is previously prepared as described in the previous steps. The playbook used for deploying the stepup components requires a prebuild tarball of the component. Prebuild components can be downloaded from the release page of the component on GitHub.

The deploy playbook is deploy.yml. A script is provided to use this ansible-playbook to deploy a single component. This script will override the component names in the deploy.yml playbook. Usage:

`scripts/ <filename of component tarball> -i <inventory> [-t <tags>] [-l <hosts>] [-v]`

The -i, -t (tags) -l (limit) and -v (verbose) options are passed verbatim to ansible-playbook

optionally, to deploy all components in the deploy.yml playbook in one go, you can call the playbook directly and provide the path to where the component tarballs are stored. I.e. ansible-playbook deploy.yml -i <inventory> -e tarball_location=<path to tarball directory on the deply host>'

Building Components

Before a component can be deployed it must be built. This creates a tarball (tar.bz2) that can then be unpacked by the deploy playbook on the application servers. The script to do that is in the Stepup-Build repository. This script will checkout a component from git on the host, but run composer and create the gzipped tarball to be deployed in a Vagrant VM.

Prebuild components can be downloaded from the release page of the component on GitHub. Make sure to get the prebuild component tar.bz2, and not the source tarball that is automatically created by GitHub. The name of a component has the form <component-name>-<tag of branch>-<timestamp of last commit>-<git commit SHA1>.tar.bz2. For example: Stepup-RA-1.0.2-20150623082722Z-2c4b6389cdbb015ddd470a19f1c04a9feb429032.tar.bz2

Step 4: Post Installation Configuration

The fourth and last step is to perform post installation configuration. This consists of:

  • Creating database schema's for the applications
  • Writing the configuration to the database

You must do this before you can use the stepup selfservice or RA interfaces.

The databases schemas and users for the Stepup components were created by the db role in Step 2, but were not further initialised. In Step 3 each component added one or more scripts to the /root/ directory on the machine(s) where it was deployed.

To perform the post installation configuration you must execute each of these scripts once. Because some scripts are order dependent they are numbered in the order they should be executed. If two scrips have the same number, their order is not important. All the scripts except "" are idempotent, meaning they can be called multiple times without ill effect.

Bootstrap and creating SRAAs

The bootstrap script is used to create the first user(s). These users cannot use the normal process via the self service and then getting vetted in the RA interface because there is no RA yet that can vet them. The bootstrap proces creates the user, called an identity in Stepup, and registers an activatated (i.e. vetted) Yubikey token for that identity.

Configuration of which identites are SRAAs (i.e. the root users, or super admins in Stepup) and the information required to bootsrap an identity is configured in group_vars/stepup-middleware.yml.

Note that the bootsrap script only works for identities that do not yet exist in Stepup. This means that you must not login to the self service interface with a user account that you later want to bootstrap, because an identity will then be created for that user. If you later try to bootstrap that identity you will get an error stating that the identity already exists.

More information


The in this repo lists the changes of not only the deployment scripts, but also the changes in the stepup components.

Pivotal Issue tracker

Much of the development discussions take place outside github in a pivotal tracker:

Releated github repositories

Stepup Components

These are the main repositories for the Stepup components that can be deployed on the Stepup infrastructure

These in turn use many components and bundles that are stored in other repositories.

Build Server

Stepup-Build is used for building releases of the stepup components. Prebuild components can be downloaded from the release page of the component on github.

Stepup VM

Stepup-VM contains scripts for setting up a VM for testing/development

Documentation from SURFnet's SecureID service (f.k.a. SURFconext Strong Authentication)

SURFnet runs an instance of the Stepup software and offers it as a service to its members. To that end it provides documentation aimed at Identity Providers, Service Provides and users of the service in the SURF SecureID section of the Get Conexted wiki.

Animation introducing SURF SecureID

Other Documentation

The first study on the architecture and processes:

Testing does a complete deploy of an application server and a management server in a CentOS 7 docker container. Apart from running automatically through, is serves a working example of a complete Stepup deploy.

The tests create a docker container named 'ansible-test' and work with both a local docker and a remote docker-machine. To run the tests:

  • ./tests/ app --clean -- Test the deployment of the database and all the stepup components
  • ./tests/ manage --clean -- Test the deployment of the management server (ELK stack)

The '--clean' option runs the tests in a clean docker container. Omit the flag to rerun the test in an existing container.

Tests must be run from the root of the git repository (i.e. where this file is located).

Use docker exec -i -t ansible-test /bin/bash to get a shell in the running container.

Note: disabled Travis docker tests (for now)


Contributions are welcome! Please open an gitub issue or a PR in the relevant github repository. Note that besides github much of the development discussion takes place in Pivotal. If you have general questions, or are not sure where your question belongs you can always ask on the openconext-users mailinglist:

You can’t perform that action at this time.