Written by Dávid Osztertág
Dávid Osztertág (Software Engineer, Avatao)
In this article, we will cover how to use Ansible for infrastructure automation. Here at Avatao, we are big believers in infrastructure-as-code which is a way of infrastructure automation using the practices from software development. Setup tasks, configuration, identity, and access management are coded as reproducible definitions. This dramatically reduces the chance of human error. Changes in the infrastructure are reproducible and auditable. We can also make use of software development tools such as version control or automated testing and deployment.
There are two kinds of infrastructure automation tools: orchestration and configuration management. Orchestration tools are for provisioning templates of immutable systems such as VM or container images and cloud-native resources. Configuration-management tools, on the other hand, are designed for modifying a mutable base system, for example, any Linux distribution. Since we had a mixture of VMs and traditional hosts we started with the latter, but nowadays we use a combination of both. There are plenty of configuration management tools, the right choice ultimately comes down to personal preference.
We went with Ansible for infrastructure automation. From here on, I’m going to assume you are familiar with the basic concepts of Ansible or other configuration-management tools. That being said, this blog post is relatively generic and I hope you will find it useful even if you haven’t use Ansible yet.
Ansible security: How to use Ansible for infrastructure automation challenges
Ideally, every change is done through automation and developers have no direct access to the machines. In the real world, however, machines can differ from one another: For instance, they could end up with different versions of the same package depending on when they were provisioned. There could be manual changes done by an administrator for a specific use case and so on. This is called configuration drift. It can be prevented by following the best practices and avoiding manual configuration. You should always code your playbooks to require a reproducible state; running a playbook a second time should not make any changes. Updates can be handled separately. Also, you should be running automated tests before deploying to production, for example, with Vagrant using disposable VMs.
Automation without limits can quickly spiral out of control. A denial of service attack could drain the company of resources. A potential breach is harder to track down which is why central logging, metrics collection, and alerting are critical, but this is a topic for another post. Some things are just fragile, hard to debug, and get worse over time. Regularly reprovisioning systems from scratch can alleviate some of these problems.
Ansible secret management
The hardest challenge might be secret management. Passwords, service accounts, key files, and other sensitive information have to be stored somewhere but where? Ansible’s answer to this question is Ansible Vault which makes use of standard symmetric encryption (AES-256) by committing encrypted files into the source code repository. Any file can be encrypted, even task and inventory files. But there are several drawbacks: developers might accidentally commit sensitive information without encrypting it. Using git hooks can prevent this from happening if we know what kind of files are likely to contain sensitive information. It is a good practice to always define sensitive variables in separate files, prefix their names with ‘vault_’ and include them in other variables.
A far more challenging problem is access control: keys cannot be revoked. Each time we want to revoke access to some secret, the secret has to be regenerated and re-encrypted with a new key. The vault password (key) has to be distributed amongst administrators and developers who need access and we have to keep track of who has access to it. Since Ansible 2.4, there is support for having multiple vault IDs, thus different security levels can be isolated. For instance, developers would not have access to the production secrets. As for distribution, scripts can be used as a vault password file so vault passwords can be encrypted with people’s GPG keys and securely distributed.
A better solution is using tools designed for secret management such as HashiCorp’s Vault. Of course, there is a module for reading secrets from Hashi Vault. There is a module for everything. Ansible Tower, Red Hat’s commercial offering, which is also open-source and can be self-hosted, comes with a centralized credential, secret management, and much more but it might be an overkill depending on your needs. Either way, you can and should decouple secrets from source code repositories.
Ansible is an agentless configuration-management tool that uses standard SSH to communicate with the hosts which is one of the reasons why we like it. This means it’s secure and reliable. It doesn’t require bootstrapping, there is no daemon to maintain on every host. The only requirement is Python 2 or 3 which is shipped in almost every distribution by default. However, it is harder to scale than a tool with an agent architecture yet Ansible can also work in pull mode by automatically pulling a repository and running a playbook locally when there are new changes. Tests and strict policies are especially important in this case.
One last tip for this post: Ansible can handle multiple separate inventories, which is a good way to logically isolate staging and production environments while keeping feature parity and having quality assurance. Also, as with vault password files, inventory files can be scripts that can be used to dynamically generate an inventory based on resource labels in your cloud provider.
Aspire to make everything reproducible. Keep your secrets secure and away from repositories. Check out some of our tutorials to get started.
Avatao is teaching software engineers not only to write secure code but also to create a more secure software development lifecycle. With more than 600 challenges available online covering various topics from DevSecOps including security on Git and Ansible, to web security and secure coding in Java or C/C++.
Test your secure coding skills
Reading Time: 6 minutes For most companies, security is considered a side quest, which is partly related to the daily processes. In reality, security ought to be a strong foundation of any organization. To ensure the defense of the enterprise, the relevant teams need strong security knowledge and abilities.
Reading Time: 6 minutes To build an enterprise security program, one has to go back to the well-known fundamentals of organizational change: People, Process, and Technology (originates from Harold Leavitt’s “Applied Organization Change in Industry”, 1964).
Reading Time: 10 minutes If you are working on Java projects you might have heard about other languages that run on the JVM, like Clojure, Kotlin, or Scala. Programmers like to try new things out but is it worth it to pick one of them over Java?