Infrastructure automation tools rely on data about the infrastructure being automated. This data includes device type, OS and version, configurations, IP address, and so on. This information must be accurate and up to date for automations to succeed. Many automation tools will connect to sources of truth or systems of record to gather device information needed for automation tasks.
The startup Opsmill has released an open beta of Infrahub, a datastore/source of truth that provides a flexible schema that models infrastructure elements and their relationships. It’s meant to serve as a central repository for information about infrastructure including networks, compute, security, and cloud infrastructure.
One challenge for a source of truth is how to represent all the information about devices and their attributes, as well as their relationships to other devices. “We don’t have an automation problem, we have a data problem,” said Damien Garros, a co-founder of OpsMill in an interview. “That’s what we want to build with OpsMill and Infrahub: a platform for all the data in infrastructure.”
Infrahub integrates with automation platforms including Ansible, Nornir, and Terraform for automation deployments.
Pick Your Source of Truth
OpsMill says Infrahub can work with all kinds of infrastructure, but it’s anticipating uptake in the network automation sector. There are already several popular sources of truth for network automation, including NetBox and Nautobot. How does Infrahub differ?
First, Infrahub is built on a graph database (in this case Neo4j). The OpsMill team chose a graph database as opposed to a SQL-style database because they believe it provides for a robust data schema while also natively mapping relationships between and among objects in the database.
Infrahub Components
Source: OpsMill
Infrahub’s schema is described in YAML. It’s built around nodes, attributes, relationships, and other descriptors. OpsMill emphasizes the flexibility of its data schema. This flexibility lets customers define infrastructure attributes, as well as show relationships between objects and describe the business logic that links objects.
Speaking of a flexible schema, one thing to know is that the open beta of Infrahub, the only version currently available, doesn’t have a default data schema. It’s up to the user to define one out of the box. If that flexibility appeals rather than intimidates, you may be a good candidate for this platform. A forthcoming Enterprise version will include a default schema to help get customers up and running.
Infrahub also includes what Garros calls data lineage – that is, the ability to tag objects with metadata such as the owner of the object, when the object was last updated, or the original source of the object. Given that infrastructure data often lives in multiple repositories (for example, on the device itself or in ServiceNow), Garros says the data lineage capability can help engineers ensure that data is correctly synchronized across sources.
Finally, OpsMill’s Infrahub differentiates from other sources of truth because it includes Git-like features such as branching, diffing, and merging directly in the database. It also includes version control to manage changes to the data model.
Note that OpsMill isn’t positioning Infrahub as a replacement for Git. “We have integration with Git for things like Jinja templates,” says Garros.
Population Control
A source of truth isn’t useful if it doesn’t have any data in it. OpsMill says it will partner with discovery tools to populate Infrahub. At present it integrates with IP Fabric. The company says it’s in discussion with Slurpit and SuzieQ. ServiceNow is also on its roadmap.
More Details
You can read the Infrahub documentation at https://docs.infrahub.app/ You can access an open beta at https://github.com/opsmill/infrahub. OpsMill hopes to have a version 1 release of Infrahub available by the end of 2024.
